Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Trade Liberalisation and Labour Market Adjustment in Botswana

Trade Liberalisation and Labour Market Adjustment in Botswana Abstract We study the effects of domestic trade liberalisation on labour markets in Botswana. South Africa is the dominant member of the Southern Africa Customs Union. As such, when South Africa liberalised trade in the 1990s, this induced large and plausibly exogenous tariff reductions for the other customs union members, including Botswana. Using labour force surveys from Botswana spanning a decade, we find that trade liberalisation did not affect the relative size of industries in terms of employment. However, trade liberalisation had effects within industries. We find an increase in the prevalence of working in an informal firm and self-employment, but mixed evidence of effects on unemployment. Hours worked decreased in response to trade liberalisation, partially driven by the movement of workers to informal firms. Despite large increases in aggregate income, trade liberalisation is associated with a reduction in monthly income, but the results are imprecise. Our results also suggest that a positive export demand shock, the 2000 African Growth and Opportunities Act, is associated with a reduction in employment in informal firms in the clothing industry. 1 Introduction To fully realise the gains from trade, workers must reallocate across industries and across firms within industries. However, reallocation is costly for workers (Artuc et al., 2015, Dix-Carneiro, 2014). Evidence from various developing countries suggests that reallocation across sectors in response to trade liberalisation is uncommon.1 More recently, evidence from developing countries has emerged about additional margins of labour market adjustments, such as between formal and informal firms, into non-traded sectors and out of the workforce.2 These studies suggest that earlier attempts to look for the reallocation of workers across industries in response to trade liberalisation in developing countries may have been missing additional margins of adjustment.3 Very little evidence, however, exists on labour market adjustments to changes in trade policy in Sub-Saharan African countries, with the exception of South Africa (Erten et al., 2019). This is largely due to the lack of high quality data. Credible causal estimates of labour market adjustments to trade policy require representative and comparable data before and after a significant change in trade policy along with a viable identification strategy. These ingredients are generally missing within Sub-Saharan Africa.4 In this paper, we overcome these problems by using repeated nationally representative labour force data from Botswana that spans a period of large trade liberalisation. Botswana is a member of the Southern Africa Customs Union (SACU). SACU’s external trade policy has historically been determined by South Africa (Southern Africa Customs Union, 2017). Thus, Botswana experienced a large and externally imposed trade liberalisation when South Africa embarked upon dramatic reductions in tariffs beginning in the middle of the 1990s. The tariffs faced by Botswana have been plausibly exogenous to events within the country, making it an unusual case in which to study the causal effects of trade liberalisation. Botswana has been politically stable for many decades. It has experienced rapid growth and structural change since independence (McCaig et al., 2017). We focus on the period of 1995/96 through 2005/06 due to the availability of nationally representative labour force survey data at the beginning and end of this period. There are no other labour force surveys. During this period, real PPP adjusted GDP per capita grew by 3.4% per annum.5 This period of rapid growth was accompanied by a large increase in the size of the working age population, from 743,000 to 923,000 individuals, stubbornly high levels of unemployment and a significant increase in the rate of employment in informal firms within the manufacturing sector. However, there have been no studies exploring whether the large SACU trade liberalisation influenced labour markets in Botswana during this period. Our research exploits both the exogeneity of the tariff reductions as well as the large variation in tariff cuts across industries to estimate the effects of trade policy on the labour market in Botswana. From a theoretical standpoint, the impact of the tariff reductions on industry employment, informality and unemployment is ambiguous. Tariff reductions by South Africa are expected to lower the price of imports and increase the demand for imports in both South Africa and Botswana. From Botswana’s perspective, the increase in imports from third parties has benefits and costs.6 Cheaper intermediate inputs are likely to raise labour productivity, providing a boon to both the traded and non-traded goods sectors. In addition, to the extent that businesses in the exportables sector were competing for scarce resources with the highly protected importables sector, the reduction in import tariffs could boost the production of exportables. At the same time, cheaper imports may also displace workers and shut down businesses, potentially affecting productivity, employment and the diversity of goods produced in Botswana. Finally, because South Africa has been, and still is, the primary destination for most of Botswana’s non-mineral exports, opening up to trade is likely to erode Botswana’s market share in South Africa, possibly reducing the number of products Botswana exports to South Africa. The net effect of these competing forces will determine the impact of trade liberalisation on industry employment, unemployment, informality, hours worked and income in Botswana. We use nationally representative labour force data from the 1995/96 and 2005/06 labour force surveys. These data cover urban and rural areas, the formal and informal sectors and all industries. In this period, Botswana experienced an increase in the prevalence of informal work in aggregate and within manufacturing, the most exposed sector to the SACU trade liberalisation.7 Informality stood at 27% of the manufacturing sector in 1995/96 and rose to 31% by 2005/06. Additionally, the labour force surveys contain questions regarding the previous industry of employment for the unemployed, allowing us to examine the impacts of tariff cuts on unemployment. Unemployment is very high in Botswana, around 20% during this period. We match workers to SACU tariff cuts by industry to study the reallocation of workers across and within industries. Between 1995/96 and 2005/06, SACU tariffs fell considerably. The average tariff within manufacturing fell from 25.0 to 11.7 percentage points. Importantly, the size of the tariff cuts varied across industries, a feature we exploit for our identification strategy. Clothing experienced the largest tariff reduction, almost 60 percentage points, whereas industries such as office, accounting and computing machinery and printing and publishing experienced negligible tariff cuts. The variation in tariff cuts across industries allows us to compare differential changes in labour market outcomes across industries in relation to trade liberalisation. We find that SACU trade liberalisation was associated with little to no net movement of workers across industries. This is consistent with a large number of studies examining the impacts of domestic trade liberalisation on industry employment in developing countries.8 These studies generally find negligible changes in the allocation of industry employment within the formal sector or overall. We find mixed evidence of impacts on unemployment. We use two definitions of unemployment. Narrow unemployment includes individuals that are available to work and sought work. Broad unemployment further includes individuals that were available to work but did not seek work. The labour force surveys ask unemployed individuals about the industry they were most recently affiliated with, allowing us to match the previous industry of employment with the industry tariff cuts. Our results suggest a decrease in unemployment within manufacturing and an increase in unemployment once mining industries are added to the analysis. Although the implied magnitudes are large, the estimates are not statistically different from 0. Among workers, we find that SACU liberalisation was associated with an increase in the prevalence of working in an informal firm and in self-employment. Furthermore, workers experienced a reduction in hours worked, partially caused by the movement of workers into informal firms, and a decrease in income, although it is imprecisely estimated. Within developing countries, labour force or household survey data that tracks the previous industry of employment is rare. Thus, most of the literature focusing on trade impacts on unemployment (or non-employment, which also includes not participating in the labour market) focuses on regional exposure to trade liberalisation as unemployment cannot be directly linked to a specific industry.9 Studying India’s trade liberalisation in the 1990s, Hasan et al. (2012) are able to link the previous industry of employment to industry trade exposure. They report no increase in unemployment in Indian industries in response to domestic trade liberalisation in the 1990s. The difference in industry-level unemployment responses between India and Botswana may be due to differences in overall labour market policies and institutions. In general, unemployment is much higher in Botswana, around 20%, than in India, suggesting that this is a more common labour market outcome and thus possibly a more likely labour market adjustment mechanism to trade liberalisation. Our definition of informal employment, which is based on the registration status and size of the business, differs from that used in most of the trade and informality literature. That literature has predominantly focused on workers in formally registered firms and defined informal based on failure to comply with labour legislation.10 This approach has the drawback of implicitly ignoring workers in informal firms. Our definition more closely resembles that of McCaig and Pavcnik (2018), which defines informal employment based on the registration status of a business in Vietnam. They find a decrease in informal employment within industries in response to changes in trade policy, but unlike our study, the trade policy changes are a reduction in tariffs in a foreign market (i.e., reduced exporting costs). Hence, our study complements the existing trade and informality literature by considering a definition of informality that is based on the status of the firm (i.e., does not focus only on workers in formally registered firms) in the context of domestic trade liberalisation. This definition of informal employment is appealing due to the direct link to the high prevalence of informal firms in developing countries such as Botswana. Growing evidence suggests that there are large differences in productivity between informal and formal firms (La Porta and Shleifer (2008,2014), Nataraj (2011), McCaig and Pavcnik (2018), Ulyssea (2018)). Hence, the SACU-induced movement of workers away from formal firms to informal firms in Botswana may have lowered aggregate labour productivity, particularly within manufacturing as it was subject to the largest tariff reductions. Our results are robust to focusing on self-employment, which is strongly positively correlated with working in an informal firm. In some ways Botswana’s labour market resembles that of South Africa. Understanding these characteristics is useful for anticipating the margins of adjustment to trade liberalisation in Botswana (Erten et al., 2019). Unemployment is stubbornly high at around 20% and has been for decades. When we include discouraged workers, the unemployment rate rises to almost 30%. The reasons for high unemployment are varied but a combination of relatively high reservation wages, an underdeveloped private sector and generous social safety nets certainly contribute (World Bank, 2015). These relatively high levels of unemployment have stimulated entry into the informal sector (Siphambe, 2004). In 2005 roughly 18% of the non-agricultural labour force was employed in the informal sector. The rate of informality in Botswana is somewhat higher than it is in South Africa but significantly lower than it is in middle income Latin American countries where urban rates of informality reach as high as 70% (Maloney, 2004). By contrast, unemployment in Botswana is considerably higher than it is in middle income countries in Latin America but somewhat lower than in South Africa. Thus, apart from sectoral reallocations, we expect adjustment to trade liberalisation in Botswana to potentially impact both unemployment and informality. The remainder of this paper is organised as follows. In Section 2, we describe SACU trade liberalisation. In Section 3, we introduce the labour force surveys and present descriptive evidence on Botswana’s labour markets. In Section 4, we explore the causal effects of the trade liberalisation on industry employment size, unemployment, informality, hours worked and monthly income. Section 5 concludes. 2 Trade liberalisation in the SACU Botswana has been a member of the SACU since 1910 and SACU’s external trade policy has been dominated by South Africa (Southern Africa Customs Union, 2017). Prior to the end of apartheid in 1994, South Africa pursued a vigorous policy of import substitution (Edwards, 2005). In 1994, the process of trade liberalisation gained momentum due to South Africa’s commitment to the GATT Uruguay Round. Between 1995 and 2006, import tariffs on all traded goods fell significantly. Levels of initial tariffs varied widely across industries, reaching over 90% in clothing. Subsequent liberalisation saw the largest tariff cuts in the industries that had the highest initial tariffs. Tariffs on trade with non-SACU members have been typically set by South Africa, with little or no input from Botswana. Thus, we use South Africa’s tariff structure to determine the level of trade protection for Botswana. We measure trade protection using tariffs (including ad valorem equivalents) plus surcharges for South Africa. Our data on trade protection, provided by Lawrence Edwards, spans the period 1990 to 2008 and is described in detail in Edwards (2005). This dataset comprises tariff rates (including ad valorem equivalents) and surcharges at the 8-digit HS level. To construct the industry tariffs that correspond to the industry classification in the labour force surveys, we construct a weighted average tariff within industries based on concordances available through the World Integrated Trade Solution. The tariff declines experienced by Botswana are advantageous for causal analysis for a number of reasons. First, the tariff reductions were large as the mean tariff within manufacturing fell from 25.0% in 1995 to 11.7% in 2005. Second, the tariff reductions varied significantly across industries. Figure 1 shows the change in industry tariff between 1995 and 2005 versus the level in 1995 across manufacturing and mining. Tariff reductions ranged from 0 to almost 60 percentage points and there is a strong negative relationship between the initial tariff level and the change. Third, the tariff changes are plausibly exogenous to Botswana. Botswana has been a member of the SACU since 1910. However, according to SACU’s website, Botswana did not have a role in setting external tariffs as ‘South Africa retained the sole decision-making power over customs and excise policies’ and set the common external tariffs to benefit South African manufacturers following an import substitution strategy. Botswana continued to lack formal influence in setting common external tariffs until the 2002 SACU Agreement, which established an independent Secretariat and headquarters in Namibia. Hence, the level of tariffs prior to the end of Apartheid and the liberalisation ushered in by South Africa are not likely to have been influenced by economic conditions within Botswana. In Section 4.1 we use regression analysis to explore whether the tariff reductions are related to initial industry conditions in Botswana. Figure 1 Open in new tabDownload slide Change in Industry Tariff Versus Initial Tariff Figure 1 Open in new tabDownload slide Change in Industry Tariff Versus Initial Tariff 3 Labour force surveys and aggregate trends 3.1 Labour force survey data We rely on two labour force surveys conducted in 1995/96 and 2005/06. The labour force surveys were conducted by the Central Statistics Office of Botswana. These surveys aim to be a source of nationally representative information on the size, structure, and main characteristics of the labour force. They include information on employment in both formal and informal firms, as well as detailed categories of work status for both workers and those not working. Data for these surveys was collected throughout the 12 months of the duration of the survey. Both the 1995/96 and 2005/06 surveys asked virtually the same questions, with the 2005/06 survey adding some questions on child employment, so data from both surveys is comparable.11 There are two main definitions of employment in the labour force surveys, each with its own time frame. The surveys ask about the main type of work the person has been doing in the past 12 months (‘usual employment’) and the type of work the person did in the past 7 days (‘current employment’). We focus on current employment during the past 7 days due to the more detailed questions asked about the current job as opposed to the usual job. Furthermore, we focus on employment outside of agriculture due to comparability issues introduced by seasonal agricultural work when using the past 7 days employment. If the respondent worked during the past 7 days or was temporarily absent, she was asked about the occupation, industry, whether the work was paid or self-employment, and a series of questions to determine if the organisation the individual worked for is formal or informal.12 We follow the conventional definition of an informal sector business in Botswana. An informal business is identified by multiple characteristics, including ‘non registration of a company with the Registrar of Companies or legal professionals; informal accounts or none; 5 or less paid employees; expenditure not easily distinguishable from household ones; enterprise often temporary or mobile or in owner’s home’ (Central Statistics Office, 2009). Under this definition, all workers in central and local government, parastatals, non-government organisations, private households and private firms with more than 10 workers are defined as working in formal organisations. Workers in private firms with 10 or fewer workers are classified by the enumerator as either being in a formal or informal business based on four questions: location (permanent; on a footpath, street or open space; market; in someone’s home; or no fixed location), how many workers are paid versus unpaid within the firm, whether the firm is a registered company or is registered with a professional organisation and whether the firm keeps a complete set of accounts. This leaves some discretion for the enumerator as a firm may not keep complete accounts, but could be a registered company or the firm could not be registered but keeps complete accounts and most of it (10 or fewer workers) are paid employees. For individuals that did not work or were not temporarily absent during the past 7 days, they were asked whether they were available for work during the past 7 days. We exclude non-working individuals that reported being unavailable to work since the questionnaires did not ask these individuals about their previous industry of work and hence we have no industry tariff to match with these individuals. Among individuals that were available to work, but did not, we create two definitions of unemployment. The first definition, narrow unemployment, is based on being available to work and trying to find work in the past 30 days. Our second definition, broad unemployment, includes individuals that were available to work, but did not report trying to find work during the past 30 days. Among individuals that were available to work, but did not seek work, the most commonly reported reason was that they thought there was no work available. We focus on unemployed individuals that had previously worked and report the industry of their most recent job. Table A1 in Appendix A provides summary statistics for employed individuals. We have 977 and 631 workers in manufacturing, 1,614 and 935 workers in mining and manufacturing and excluding workers in agriculture, we have 8,895 and 6,587 observations in 1995/96 and 2005/06 respectively. Table A2 provides summary statistics for unemployed individuals based on the industry of their most recent job. Within manufacturing, unemployed individuals are younger, on average, than employed individuals. They are also more likely to be female. We report industry summary statistics in Table A3. We have 5 industries within mining and quarrying and 27 industries in manufacturing for a total of 32 traded industries. The table reports the share of economy-wide employment within each traded industry and for non-traded industries aggregated together. Within traded industries, the largest are clothing, diamond mining and copper or nickel mining. The share of workers in traded industries fell between 1995/96 and 2005/06. The timing of the initial labour force survey is after SACU trade liberalisation had already started in 1994 (Cassim et al. (2004), Edwards (2005)). As such, any adjustment that had already taken place by the time of the 1995/96 LFS will be missed in our estimates. We thus view our estimates as a lower bound of the effect by the time of the 2005/06 LFS as we match the 1995 and 2005 tariffs to the 1995/96 and 2005/06 LFSs by industry. It would be ideal to have an additional labour force survey or population census prior to SACU liberalisation as a way to check whether pre-existing trends within Botswana’s labour markets are correlated with subsequent industry tariff reductions. Unfortunately, such data is not available. There is no labour force survey prior to the 1995/96 LFS. Additionally, the 1991 population census used more aggregated industry codes, leaving only 11 manufacturing industries, and the 1981 population census grouped all manufacturing industries together. Hence, data availability is a constraint, as is common in most sub-Saharan countries (Center for Global Development, 2014). 3.2 Changes in employment, unemployment and informality We begin by examining the reported activity of working age individuals. Table 1 shows the number and share of individuals according to whether they were working; seeking work; not available to work; or available to work, but did not look for a job during the past 30 days. The estimates are based on the individual’s activity during the past 7 days, according to the 1995/96 and 2005/06 labour force surveys, respectively. We focus on individuals between the ages of 15 and 60 inclusive. Over the sample period, the labour force participation rate increased by 8.5 percentage points as the number of working age individuals increased by 24% and the number of individuals not in the labour force remained roughly constant. Among labour force participants, the share of those actively seeking work fell from 22% to 19%. However, the share of the working age population who reported that they were available to work but did not work during the past 30 days (discouraged workers) increased from 10.9% to 13.9%. If we count both those actively seeking work and discouraged workers, the share of the working age population that is unemployed is very high in both periods at 23.3% and 26.0%, respectively. Table 1 Economic Activity of the Working Age Population Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab Table 1 Economic Activity of the Working Age Population Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab HIV prevalence is extremely high in Botswana. Using Botswana’s AIDS survey and correcting for selection bias, Levinsohn and McCrary (2010) estimate prevalence rates to be around 17%. According to UNAIDS, the prevalence among working age individuals is much higher, at close to 40%. Contrary to our expectations, illness does not seem to be a major determinant of labour force participation. The share of the population reporting that they were not in the labour force due to illness increased only marginally from 3% in 1995/96 to 3.4% in 2005/06. This may be a testament to the government’s aggressive campaign to treat individuals who are HIV positive. With our data, it is impossible to tell. However, given the importance of HIV in Botswana, in our econometric work we control for the individual characteristics that Levinsohn and McCrary found to be important correlates of HIV status: age, location (urban or rural) and education. Next, in Table 2 we examine the distribution of workers across sectors. The estimates are based primarily on the current job of the individual, but for agriculture we also report estimates based on the usual job over the past 12 months because of the seasonality of agricultural employment (recall that the surveys were conducted across 12 months). Additionally, for the workers usually working in agriculture, we report the number of seasonal and non-seasonal workers. An agricultural worker is defined as seasonal if the worker reported working less than 52 weeks and reported working on their own/family lands/cattlepost (hereafter referred to as own lands) for their usual job. In contrast, a non-seasonal agricultural worker reported working all 52 weeks in the sector. The distinction between the individual’s usual and current job is crucially important in agriculture. It is immediately obvious that the two definitions yield very different trends in estimates of the number of workers in agriculture. The number of current workers in agriculture grew by 182%, while the number of usual agricultural workers grew by only 19%. The most likely explanation for these differences has to do with the timing of interviews. More households were surveyed during the lean season in 1995/96 than in 2005/06. As a result, in Section 4 we focus our empirical analysis on non-agricultural workers. Table 2 Distribution of Workers Across Major Sectors Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The estimates reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The number of agricultural workers is based on the individual’s usual job and only includes individuals who reported doing some work on their own land/farm/cattlepost as usual job. The total number of workers in 1995/96 differs from Table 4 (on labour force) since there were 476 individuals with undefined industry. Non-seasonal agricultural workers are those who reported working all 52 weeks of the year and reported the sector as their usual sector of employment. Seasonal agricultural workers are those who reported not working all 52 weeks of the year and reported the sector as their usual sector of employment. Open in new tab Table 2 Distribution of Workers Across Major Sectors Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The estimates reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The number of agricultural workers is based on the individual’s usual job and only includes individuals who reported doing some work on their own land/farm/cattlepost as usual job. The total number of workers in 1995/96 differs from Table 4 (on labour force) since there were 476 individuals with undefined industry. Non-seasonal agricultural workers are those who reported working all 52 weeks of the year and reported the sector as their usual sector of employment. Seasonal agricultural workers are those who reported not working all 52 weeks of the year and reported the sector as their usual sector of employment. Open in new tab Outside of agriculture, as reported in Panel A, there were some important changes in the distribution of workers across sectors. The share of workers in public administration and construction fell appreciably, while the number of workers in wholesale and retail trade, hotels, restaurants, etc.; and finance, insurance, real estate and business services increased. The remaining sectors experienced only marginal changes in their shares of the non-agricultural workforce. In the context of the SACU tariff cuts, it is interesting to note that the share of workers in manufacturing fell, but only by 0.6 percentage points, and the share of workers in mining fell by 1.4 percentage points or 27%. In Table 3, we return to the issue of unemployment and labour force participation by various segments of the working age population. We report these rates for males and females, urban and rural workers, by age (30 or younger and older than 30) and by education levels. We find significant differences across groups. For example, males are less likely to be unemployed and more likely to be in the labour force than females in either year, although both genders significantly increased their labour force participation rates between 1995/96 and 2005/06. Urban individuals have a higher rate of narrow unemployment than rural individuals and the gap has grown to 5.4 percentage points by 2005/06. Urban individuals are also more likely to be participating in the labour force, although the gap has shrunk considerably over time. There is not much difference in either unemployment or labour force participation rates across individuals with primary or secondary education. However, having some education, as compared to no formal education, is an important determinant of employment status. Lastly, older workers have a higher rate of labour force participation and lower rates of both narrow and broad unemployment than younger workers. Table 3 Unemployment and Labour Force Participation Rates Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The narrow unemployment rate is defined as the ratio of the number of individuals seeking work relative to the sum of the number of workers and individuals seeking work. The broad unemployment rate is defined as the ratio of the number of individuals available to work relative to the sum of the number of workers and individuals available to work. Open in new tab Table 3 Unemployment and Labour Force Participation Rates Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The narrow unemployment rate is defined as the ratio of the number of individuals seeking work relative to the sum of the number of workers and individuals seeking work. The broad unemployment rate is defined as the ratio of the number of individuals available to work relative to the sum of the number of workers and individuals available to work. Open in new tab Table 4 Percentage of Workers in Informal Firms by Major Sector Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: Sample is persons aged 15 to 60. Informal workers are all those workers in the private sector, employed at firms with less than |$10$| employees and defined as informal in the survey by question 30 in the 1995/96 LFS and question 39 in the 2005/06 LFS. Sample excludes agricultural workers.The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab Table 4 Percentage of Workers in Informal Firms by Major Sector Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: Sample is persons aged 15 to 60. Informal workers are all those workers in the private sector, employed at firms with less than |$10$| employees and defined as informal in the survey by question 30 in the 1995/96 LFS and question 39 in the 2005/06 LFS. Sample excludes agricultural workers.The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab An additional concern about the SACU tariff cuts, aside from displacement and unemployment, is the movement of workers into informal firms. The labour force surveys define informality according to a series of questions related to the ownership sector (e.g., government, parastatal, NGO or private), the number of workers in the business, the location of the business, whether the business is registered and whether the business keeps a complete set of accounts. The enumerator evaluated whether the worker’s place of work was formal or informal based on these questions.13 Note that the questions about location, business registration and accounts were only asked for workers in businesses with 10 or fewer workers in the private sector. Consequently, for all workers that were not asked the detailed questions related to formality, we classify them as formally employed as they either worked in large private firms or in sectors more likely to be formal (e.g., government, parastatal or NGOs). This definition of informality is thus based on the status of the business, not on whether the business confirms to existing labour legislation. Furthermore, the definition of an informal business, based on a combination of ownership type, employment size, registration status and account keeping, is consistent with those from other countries. For example, La Porta and Shleifer (2008, 2014) define a business as informal based on its registration status. In Vietnam, an informal firm is one that is not registered with the central government as an enterprise under Vietnam’s Enterprise Law. Vietnamese domestic private businesses are not required to register with the government if they have less than 10 workers and do not operate in more than one location (McCaig and Pavcnik (2018)). And in India, manufacturing firms are legally required to register with the government if they have 10 or more workers and use electricity or if they have 20 or more workers, but don’t use electricity (Nataraj (2011)).14 We use this employer-based definition of informality to report on the incidence of informal work by major sector in Table 4. Overall, the percentage of work in informal firms has increased by 4.8 percentage points or 36%, a very significant increase. Importantly, the incidence of informality in the manufacturing sector increased by 4.3 percentage points, which is slightly below the overall increase, but may be partially related to the SACU tariff cuts. We explore this possibility in Section 4.3. Many sectors experienced an increase in the share of workers in informal firms. In fact, the only sector to experience a decrease was community, social, household and personal services. In summary, the share of individuals working increased between 1995/96 and 2005/06; however, the share of discouraged workers also increased. At an aggregate level, employment in manufacturing fell slightly, by 0.6 percentage points, but the share of informal workers in manufacturing increased significantly, by 4.3 percentage points from 27% in 1995/96. In Section 4, we explore whether these changes are related to SACU tariffs reductions. 4 SACU trade liberalisation and labour market adjustment in Botswana 4.1 Endogeneity of SACU tariff cuts Before estimating the impacts of SACU trade liberalisation on labour markets in Botswana, we address possible concerns about the endogeneity of the tariff reductions. As shown in Figure 1, the size of the tariff reductions was closely related to the initial tariff level. Furthermore, as previously discussed, Botswana had no ability to influence the size of the tariff reductions due to South Africa dominating trade policy within SACU. Nonetheless, it is still possible that the size of the tariff reductions may vary with conditions across industries in Botswana. To explore this possibility, we estimate the following equation: $$ \begin{equation} \Delta ln(1+\textrm{tariff}_j)= \alpha+\beta X_j + u_j \end{equation} $$ (1) where |$\Delta ln(1+\textrm{tariff}_j)$| is the change in ln tariff in industry j and |$X_j$| is a vector of initial conditions in industry j. We calculate the following initial conditions from the 1995/96 labour force survey: the share of workers within the industry that work in informal firms, the share of young workers (30 years and less), the share of urban workers, the share of highly educated workers (completed at least junior secondary) and the industry’s share of total employment. We report the results in Table 5 for manufacturing and traded (mining and manufacturing together), and for both sets of industries, but omitting clothing. We find that some of the initial conditions have a statistically significant relationship with the tariff reductions. Specifically, industries with a higher share of workers in informal firms and industries with a greater share of individuals attached to it (either currently working or most recently employed in the industry) experienced greater tariff reductions. Although the coefficient is not statistically different from 0, industries with a greater share of attached individuals that were unemployed also experienced larger tariff decreases. The employment share is a particularly strong predictor of the size of the tariff cut within manufacturing. The finding that initial conditions are correlated with subsequent tariff cuts is common in the literature. For example, Goldberg and Pavcnik (2005) find that the initial share of skilled workers is correlated with subsequent tariff reductions in Columbia. Topalova and Khandelwal (2011) find that the initial share of nonproduction workers is correlated with the size of industry tariff reductions in India.15 Table 5 SACU Tariff Cuts and Initial Industry Conditions Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Note: The dependent variable is the change in ln(1 + tariff). Heteroskedasticity robust standard errors are reported in parentheses. *** p <0.01, ** p <0.05, * p <0.1. Open in new tab Table 5 SACU Tariff Cuts and Initial Industry Conditions Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Note: The dependent variable is the change in ln(1 + tariff). Heteroskedasticity robust standard errors are reported in parentheses. *** p <0.01, ** p <0.05, * p <0.1. Open in new tab In our context, the results are significantly weaker when clothing is excluded from the analysis, as shown in columns 3 and 4. Clothing is a very important industry in this analysis for two reasons. First, it is a very large industry in terms of employment, representing 25% of manufacturing employment and 15% of mining and manufacturing employment in 1995/96. Second, it is the industry that experienced the largest tariff reduction (Figure 1). These two features make it a very influential observation. Furthermore, clothing is an important industry for a second reason. The 2000 US African Growth and Opportunities Act (AGOA) led to duty free US imports from numerous Sub-Saharan African countries, including Botswana.16 Botswana’s clothing exports to the US rose by almost 700% between 2000 and 2005, from 8.9 to 61.7 million USD.17 Hence, in our labour market analysis we will check the robustness of our results controlling for the influence of AGOA on the clothing industry. 4.2 SACU trade liberalisation and industry employment Traditional trade theory predicts that workers should reallocate across industries in response to changes in tariffs. In particular, the industries that received the largest tariff cuts should contract. We begin our formal analysis of the labour market consequences of SACU’s trade liberalisation by looking for evidence of worker reallocation across industries in response to tariff changes. We estimate the following regression model: $$ \begin{equation} s_{jt}=\beta ln(1+\textrm{tariff}_{jt}) + \theta D_{2005} X_{j1995} +\lambda_j+\eta_t+u_{jt} \end{equation} $$ (2) where |$j$| indexes industries, |$t$| indexes time (i.e., 1995/96 and 2005/06), |$s_{jt}$| is the share of the workforce working in industry |$j$| at time |$t$|⁠, |$\textrm{tariff}_{jt}$| is the SACU tariff in industry |$j$| at time |$t$|⁠, |$D_{2005} X_{j1995}$| is a dummy for 2005 interacted with a vector of initial industry conditions, |$\eta _t$| is a time fixed effect and |$\lambda _j$| is an industry fixed effect. The vector of initial industry conditions is the same as in Table 5 and includes the share of workers within the industry in informal firms, the share of individuals affiliated with the industry that are male, 30 and younger, urban, have finished junior secondary, are (broad) unemployed, and the share of individuals affiliated with the industry. We are interested in the coefficient on tariff. Our identifying variation is changes in tariffs over time within an industry. Thus, we are identifying whether industries that experienced larger tariff cuts either grew or shrunk in comparison to industries that received smaller tariff cuts, conditional on initial industry conditions. Conventional trade theory predicts that the industries that experienced the largest tariff cuts should experience a reduction in the share of workers and thus that the coefficient on tariff, |$\beta $|⁠, should be positive. We estimate equation (2) for manufacturing, traded, and for all industries (excluding agriculture). For non-traded industries, we assign a tariff of 0 in both 1995/96 and 2005/06. With the inclusion of industry fixed effects, this implies that identification for non-traded industries is based on no change in the tariff. Table 6, Panel A, presents the results from estimating equation (2). For manufacturing, the coefficient is negative, suggesting an increase in employment shares in response to SACU liberalisation, but the implied magnitude is very small and the coefficient is not statistically different from 0. In column 2, the coefficient is also negative, suggesting an increase in employment share in response to the SACU tariff reductions, but again the implied magnitude is very small and the coefficient is not statistically different from 0. Lastly, in column 3 for all industries (excluding agriculture), the coefficient is positive and statistically different from 0, suggesting that employment in traded industries shrunk relative to non-traded industries in response to SACU liberalisation. Table 6 SACU Tariffs and Industry Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Note: The dependent variable is the share of workers within an industry in a given year. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. The employment shares are calculated relative to total employment within each year in the indicated industries in each column. All regressions have industry and year fixed effects and initial industry conditions (see Table 5) interacted with a 2005 indicator. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 6 SACU Tariffs and Industry Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Note: The dependent variable is the share of workers within an industry in a given year. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. The employment shares are calculated relative to total employment within each year in the indicated industries in each column. All regressions have industry and year fixed effects and initial industry conditions (see Table 5) interacted with a 2005 indicator. The industry tariff is ln(1 + tariff + surcharges). Open in new tab In Panels B and C of Table 6 we explore how sensitive our results are to the clothing industry. In Panel B we add an AGOA indicator variable as an additional control variable. This variable is 1 for the clothing industry in 2005/06 and 0 for all other observations. In Panel C we remove the clothing industry from the analysis. Recall that the clothing industry is the largest manufacturing industry in terms of employment, accounting for 25% of manufacturing workers in 1995/96. Additionally, as previously discussed, the clothing industry was subject to other large trade policy changes during this time period, specifically the AGOA. Given the size of the clothing industry and the clothing-specific effect of AGOA, it makes sense to explore how sensitive our results are to the inclusion of clothing. In both Panels B and C, the coefficients are smaller in magnitude than in Panel A for manufacturing and mining. For all industries, the estimates in Panels B and C are very similar to that in Panel A, again suggesting a relative expansion of the non-tradable sector in response to SACU tariff cuts. Lastly, note that the AGOA indicator is positive in columns 1 and 2, suggesting an expansion in the relative size of the clothing industry, as one would expect from AGOA. In summary, there is little evidence of a systematic reallocation of workers across traded industries in response to the SACU tariff cuts, but some evidence of relocation to non-traded industries.18 These results are consistent with a lack of reallocation across industries in response to domestic trade liberalisation in developing countries surveyed by Goldberg and Pavcnik (2007) and more recently found by Topalova (2010) in the case of India’s trade liberalisation in the early 1990s, Dix-Carneiro (2014) for Brazil’s trade liberalisation in the late 1980s and early 1990s and McCaig and Pavcnik (2018) in Vietnam in the early 2000s in response to reductions in a trading partner’s tariffs (i.e., export market liberalisation). Table 7 SACU Tariffs and Narrow Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Note: The dependent variable is an indicator for being available to work and seeking as work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 7 SACU Tariffs and Narrow Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Note: The dependent variable is an indicator for being available to work and seeking as work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab 4.3 SACU trade liberalisation and workers Although the tariff changes do not appear to have substantially affected the distribution of workers across industries, it may be the case that there have been important within-industry changes. For example, workers could have become unemployed, moved to informal firms, changed hours worked or experienced changes in pay. We explore whether there is evidence of changes in worker outcomes in response to the SACU tariff cuts using the following regression model: $$ \begin{equation} Y_{ijt}=\beta ln(1+\textrm{tariff}_{jt}) +\delta Z_{ijt} + \theta D_{2005} X_{j1995} + \lambda_j +\eta_t +u_{ijt} \end{equation} $$ (3) where |$j$| indexes industries, |$t$| indexes time (i.e., 1995/96 and 2005/06), |$Y_{ijt}$| is an outcome variable for worker or individual |$i$|⁠, |$tariff_{jt}$| is the SACU tariff in industry |$j$| at time |$t$|⁠, |$Z_{ijt}$| is a vector of individual characteristics (age, male indicator, urban indicator, finished primary, finished junior secondary, finished senior secondary), |$D_{2005} X_{j1995}$| is a dummy for 2005 interacted with a vector of initial industry conditions (the same as used previously in Tables 5 and 6), |$\lambda _j$| is an industry fixed effect and |$\eta _t$| is a time fixed effect. We are interested in the coefficient on tariffs. Our identifying variation is changes in tariffs over time within an industry, conditional on controlling for observable individual characteristics and initial industry conditions. Controlling for initial industry conditions is important in this context due to their correlation with SACU tariff reductions. Failing to directly control for them would then imply that the SACU tariff cuts are partially picking up any effects related to initial differences in industry size and composition. Unemployment We begin by examining two definitions of unemployment. The first is based on workers that were available to work and sought work. The second is broader and includes individuals that were available for work but did not seek employment. In both cases, our sample includes individuals that are actively working and individuals that are defined as unemployed according to the respective definition. For individuals that are currently working, their industry is that in which they are currently working, but for unemployed individuals, their industry is that in which they most recently worked. Table 8 SACU Tariffs and Broad Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 8 SACU Tariffs and Broad Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab In Table 7, we report regression results from estimating (3) for manufacturing, traded (mining and manufacturing) and all industries (excluding agriculture). As before, we explore the sensitivity of our results to the clothing industry. Within manufacturing, column 1, we consistently find a small, statistically insignificant effect on unemployment. Adding mining to the sample (column 2) causes the estimate to become negative and to increase in magnitude, but it remains statistically insignificant in all three panels. Using the estimate from Panel A, the magnitude of the coefficient for the manufacturing and mining sample is relatively large and suggests that the average reduction in manufacturing tariffs, |$-0.107$|⁠, is associated with a 1.6 percentage point increase in the likelihood of narrow unemployment, but the coefficient is estimated imprecisely, limiting our ability to draw firm conclusions. For all industries (column 3), the coefficient is also negative, but is smaller in magnitude. In Table 8, we report results for estimating equation (3) with an expanded definition of unemployment. We include unemployed individuals that sought work as well as unemployed individuals that did not seek work (i.e., discouraged). Across all specifications, the effects are noticeably larger than in Table 7, but are typically not statistically different from 0, especially once we control for the effects of AGOA (Panel B) or remove clothing (Panel C). As in Table 7, the results within manufacturing suggest a reduction in unemployment in association with SACU tariff reductions. Using the estimate from Panel B, the magnitude of the coefficient for manufacturing suggests a 2.4 percentage point decrease in unemployment, broadly defined, due to the mean tariff cut (⁠|$-0.107$|⁠), although the estimate is imprecisely estimated and not statistically different from 0. The results in columns 2 and 3 are qualitatively similar to those for narrow unemployment.19 In summary, our results examining the effects of SACU liberalisation on unemployment are suggestive of a decrease in unemployment within manufacturing and an increase in unemployment in traded industries. However, in both cases the results are imprecisely estimated and not generally statistically different from 0. Our results of a possible increase in unemployment are consistent with those reported for Brazil. Menezes-Filho and Muendler (2011), Gaddis and Pieters (2017) and Dix-Carneiro and Kovak (2019) report that Brazil’s domestic trade liberalisation was associated with a rise in non-formal employment (i.e., a rise in informal employment, unemployment or non-participation). In South Africa, Erten et al. (2019) find evidence of an increase in unemployment in local labour markets in response to SACU trade liberalisation. However, our results differ from Hasan et al. (2012) who report no increase in unemployment in India in response to domestic trade liberalisation in the 1990s. The difference between contexts likely has to do with overall labour market policies and institutions. Botswana, South Africa and Brazil have much higher rates of unemployment than India. For example, Dix-Carneiro and Kovak (2019) report mean non-employment (unemployed plus not participating in the labour force) rates of 36% to 40% across Brazilian regions and Erten et al. (2019) report broad unemployment rates of 23% for South Africa, while Hasan et al. (2012) report unemployment rates between 4.1% and 5.6% across Indian states. Botswana’s unemployment and non-employment rates are more similar in magnitude to those in Brazil and South Africa, suggesting that unemployment is, more generally, a common outcome and thus more likely to be an adjustment margin in response to domestic trade liberalisation. Work in informal firms and self-employment We next explore margins of adjustment among workers. Recent work has highlighted the role of trade in the reallocation of workers between informal and formal firms (McCaig and Pavcnik (2018) and Dix-Carneiro and Kovak (2019)). In Table 9, Panel A, we report the estimation of equation (3) using an indicator for working in an informal firm as the dependent variable. We find evidence of an increase in the prevalence of work in informal firms within manufacturing in response to the tariff reductions. The implied magnitude is large, a 7.8 percentage point increase in the prevalence of working in an informal firm in response to the average tariff reduction within manufacturing. However, the result is imprecisely estimated and not statistically different from 0. In columns 2 and 3 of Panel A, the estimate is very close to 0. Table 9 SACU Tariffs and Work in Informal Firms Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Note: The dependent variable is an indicator variable for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 9 SACU Tariffs and Work in Informal Firms Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Note: The dependent variable is an indicator variable for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab In Panel B, we add an indicator for AGOA. We find that AGOA is associated with a reduction in informality within clothing. Controlling for this influence leads to an increase in the magnitude of the coefficient on SACU tariffs relative to Panel A. The coefficient for manufacturing suggests a 8.0 percentage point increase in the prevalence of working for an informal firm in response to the mean tariff reduction within manufacturing (⁠|$-0.107$|⁠). The inclusion of mining in column 2 leads to a smaller effect but it is still statistically different from 0. Lastly, in column 3, the effect is again smaller than within manufacturing, but is statistically different from 0. The smaller effects in mining and non-traded industries are likely due to the lack of informal businesses in mining and the lower overall prevalence of such work in some large non-traded industries, such as public administration (Table 4). We check the sensitivity of our results to removing clothing from the specification and report the results in Panel C of Table 9. The estimated coefficients are very similar to those in Panel B, both in magnitude and statistical significance.20 The estimates of an increase in informality due to SACU tariff reductions are consistent with evidence from Colombia and Brazil where domestic trade liberalisation was associated with an increase in informality (Goldberg and Pavcnik (2003), Bosch et al. (2012), Paz (2014) and Dix-Carneiro and Kovak (2019)).21 However, those studies primarily focus on a definition of informality that is based on employment conditions for workers within formal firms and focus only on workers in formal firms. Our definition of informality is based on the registration and size of the business and more closely aligns with the definition used by McCaig and Pavcnik (2018). They find a reduction in informal employment in response to new export opportunities. The difference in effects of changes in trade policy are likely due to the positive nature of the export demand shock studied by McCaig and Pavcnik (2018) as compared to domestic trade liberalisation in Botswana. This is consistent with a recent survey by Goldberg and Pavcnik (2016) that highlights how the nature of trade policy changes matters. Table 10 SACU Tariffs and Self-Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Note: The dependent variable is an indicator variable for being self-employed. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 10 SACU Tariffs and Self-Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Note: The dependent variable is an indicator variable for being self-employed. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Since the definition of an informal firm varies across countries, we provide a robustness check based on self-employment. In particular, we focus on the distinction between paid employment and self-employment. Information on whether a worker is self-employed or working in wage employment is more widely available in household and labour force surveys in developing countries and McCaig and Pavcnik (2018) show that employment in an informal business is highly correlated with self-employment in Vietnam. We find that working in an informal business is highly positively correlated with self-employment in Botswana as well. The correlations are 0.82, 0.84 and 0.73 in manufacturing, traded and all industries, respectively. Thus, using an indicator for self-employment is both a useful check on the sensitivity of our results to the particular definition of an informal firm in Botswana and facilitates comparisons with other low- and middle-income countries. In Table 10, we report results from estimating equation 3 using an indicator for self-employment as the dependent variable. The specifications mimic those in Table 9. Within manufacturing, we consistently find that the SACU tariff cuts are associated with an increase in the prevalence of self-employment within an industry. The estimates are larger than those reported in Table 9 in terms of magnitude. The results in Panel C suggest an increase in self-employment of 11.7 percentage points within an industry that experienced the mean tariff reduction in manufacturing as compared to no reduction. The effects are largest within manufacturing. Self-employment within mining is unobserved in our data, which is likely why the magnitude is lower once mining industries are added to the regression. Table 11 SACU Tariffs and Weekly Hours Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 11 SACU Tariffs and Weekly Hours Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab We can further disaggregate self-employment into self-employment with employees, without employees and as an unpaid family helper. These results are reported in Tables A7 through Table A9 in Appendix A. We find that more than half of the increase in self-employment due to SACU trade liberalisation occurred within unpaid family help. The combination of self-employment with employees and unpaid family help accounts for over three quarters of the SACU-induced increase in self-employment. These results are remarkably consistent with the informality results, suggesting that the particular definition of an informal business in Botswana is not driving the results. Hours worked The analysis thus far has focused on extensive margins of adjustment: changes in unemployment and changes in the type of job (working for an informal business and self-employment). In this subsection we explore whether hours worked were affected by SACU trade liberalisation. In Table 11 we report estimates of equation (3) with ln of actual hours worked during the past week as the dependent variable. The results for manufacturing consistently suggest a decrease in weekly hours worked. The estimate in Panel A suggests a decease in weekly hours of 8.7% in response to the mean tariff reduction within manufacturing. The same pattern emerges once mining is included, particularly once clothing is controlled for. Once non-traded industries are added to the analysis (column 3), the estimate is much smaller. The reduction in weekly hours worked is partially driven by the shift of workers into informal firms as workers in informal firms work fewer hours on average. For example, within manufacturing in 1995/96, the mean number of weekly hours was 35.9 and 43.6 hours in informal and formal businesses, respectively. This is confirmed in Table 12, which repeats the previous specifications, adding an indicator for working in an informal firm as a control variable. The coefficient on informal is consistently negative, suggesting a decrease in hours of between 17% and 21% across the various specifications. Moreover, the coefficient on tariffs falls by about 0.2 ln points when the informal control is added relative to Table 11. This suggests the shift of workers into informal firms is partly responsible for the decrease in weekly hours in response to the SACU tariff cuts. Table 12 SACU Tariffs and Weekly Hours Controlling for Informal Work Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 12 SACU Tariffs and Weekly Hours Controlling for Informal Work Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Monthly income Lastly, we report results using ln monthly gross income from paid employment as the dependent variable. This includes both cash and the value of in kind payments. We present the results in Table 13. The specifications follow the same pattern as in previous tables. The results for manufacturing suggest a decrease in monthly income, although the coefficient is not very precisely estimated and is not statistically different from 0. Nonetheless, the suggested magnitude of the effect remains large. For example, the coefficient on tariffs in panel A for manufacturing suggests a reduction in monthly income of 9% in response to the mean tariff reduction. However, when we add workers in mining to the sample, the coefficient switches sign. Recall that mining industries, on average, experienced very small tariff reductions (Table A3, Figure 1). Hence, the change in the coefficient from column 1 to 2 implies that workers in mining experienced slower wage growth than workers in manufacturing industries that experienced small tariff reductions. Table 13 SACU Tariffs and ln Monthly Income Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Note: The dependent variable is the ln of monthly income from paid employment, including both cash and in kind payments. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60 in paid employment. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Income is the sum of paid employment in their primary or secondary job. Open in new tab Table 13 SACU Tariffs and ln Monthly Income Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Note: The dependent variable is the ln of monthly income from paid employment, including both cash and in kind payments. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60 in paid employment. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Income is the sum of paid employment in their primary or secondary job. Open in new tab 5 Conclusion We study the impact of trade liberalisation on workers in Botswana. A unique feature of Botswana’s trade liberalisation is that it was driven by South Africa’s control of the external trade policy of the SACU. This provides an excellent context for exploring causal effects since it decreases the likelihood that tariff reductions were influenced by factors within Botswana. The evidence presented in this paper indicates that SACU’s trade liberalisation had little impact on the allocation of employment across industries in Botswana. This is consistent with existing trade and labour market adjustment studies that generally find little or no changes in industry size in developing countries in response to domestic trade liberalisation. We find mixed evidence that the SACU tariff reductions impacted the prevalence of unemployment within industries. The results suggest a decrease in unemployment within manufacturing, but an increase once mining industries are added to the analysis. However, although the implied increase in unemployment is large, the results are imprecisely estimated and not statistically significant. The prevalence of working in an informal business increased within manufacturing due to SACU liberalisation, as did the prevalence of self-employment. We further find that weekly hours decreased and that this is driven in part by the increase in work in informal businesses, as these workers report working fewer hours on average. Lastly, we find evidence of a decrease in monthly income in manufacturing due to SACU liberalisation, but despite large coefficients, the results are imprecise and not statistically different from 0. The lack of statistically significant results is commonly the outcome of large standard errors, not of estimates that are quantitatively close to 0. This highlights the need for either larger labour force surveys in terms of sample sizes or more frequent labour force surveys to help derive more definitive results. This echoes concerns about the lack of data for a large number of Sub-Saharan African countries. The reallocation of workers out of formal firms and into informal firms and unemployment may have important consequences for labour productivity, particularly within manufacturing, the most impacted sector. A growing literature documents that informal firms are typically much less productive than formal firms. Thus, the movement of workers away from formal firms to informal firms is likely associated with a decrease in aggregate labour productivity. The results on unemployment, working in an informal business and self-employment add to the more recent literature on labour market adjustment to trade in developing countries. In particular, there is scant evidence on the effects of trade liberalisation on unemployment within industries in developing countries due to the lack of data in household or labour force surveys on the previous industry of employment (Goldberg and Pavcnik (2007), Hasan et al. (2012)). Additionally, this is one of the first papers to study the impact of trade liberalisation on labour markets in Sub-Saharan Africa, with Erten et al. (2019) being the notable exception. Acknowledgements We thank Colleen Spencer and Inigo Verduzco-Gallo for their research assistance and Anna Majalentle and her staff at the Central Statistical Office of Botswana for assistance with understanding the data. We thank the World Bank for financial assistance. We would like to thank Jeff Chan, Bilge Erten, Azim Essaji, Keith Jefferis, Bob Rijkers, and Happy Siphambe for helpful conversations and Sara Hartley for proofreading. All errors are our own. 6 Appendix Table A1 Summary Statistics for Workers Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Note: The sample is workers aged 15 to 60 outside of agriculture. The reported information is for the individual’s primary job with the exception of income, which is the sum of paid employment in their primary or secondary job. Open in new tab Table A1 Summary Statistics for Workers Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Note: The sample is workers aged 15 to 60 outside of agriculture. The reported information is for the individual’s primary job with the exception of income, which is the sum of paid employment in their primary or secondary job. Open in new tab Table A2 Summary Statistics for Unemployed Individuals Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Note: The sample is unemployed individuals aged 15 to 60 outside of agriculture. Includes individuals who did not work despite being available to work, regardless of whether they sought work. Open in new tab Table A2 Summary Statistics for Unemployed Individuals Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Note: The sample is unemployed individuals aged 15 to 60 outside of agriculture. Includes individuals who did not work despite being available to work, regardless of whether they sought work. Open in new tab Table A3 Industry Summary Statistics Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Note: This table reports summary statistics for each of the subsectors of employment, excluding agriculture. The sample is individuals aged 15 to 60. Open in new tab Table A3 Industry Summary Statistics Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Note: This table reports summary statistics for each of the subsectors of employment, excluding agriculture. The sample is individuals aged 15 to 60. Open in new tab Table A4 SACU Tariffs and Narrow Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Note: The dependent variable is an indicator for being available to work and seeking work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. Unemployed is defined as available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A4 SACU Tariffs and Narrow Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Note: The dependent variable is an indicator for being available to work and seeking work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. Unemployed is defined as available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A5 SACU Tariffs and Broad Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. Unemployed is defined as available to work regardless of whether they sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A5 SACU Tariffs and Broad Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. Unemployed is defined as available to work regardless of whether they sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A6 SACU Tariffs and Work in Informal Firms by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Note: The dependent variable is an indicator for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. Clothing workers are omitted. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A6 SACU Tariffs and Work in Informal Firms by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Note: The dependent variable is an indicator for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. Clothing workers are omitted. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A7 SACU Tariffs and Self-Employment With Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Note: The dependent variable is an indicator variable for being self-employed with employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A7 SACU Tariffs and Self-Employment With Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Note: The dependent variable is an indicator variable for being self-employed with employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A8 SACU Tariffs and Self-Employment Without Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Note: The dependent variable is an indicator variable for being self-employed without employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A8 SACU Tariffs and Self-Employment Without Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Note: The dependent variable is an indicator variable for being self-employed without employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A9 SACU Tariffs and Unpaid Family Help Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Note: The dependent variable is an indicator variable for being self-employed as an unpaid family helper. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A9 SACU Tariffs and Unpaid Family Help Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Note: The dependent variable is an indicator variable for being self-employed as an unpaid family helper. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Footnotes 1 See Revenga (1997) and Feliciano (2001) for Mexico, Attanasio et al. (2004) for Colombia, Currie and Harrison (1997) for Morocco and Topalova (2010) for India. 2 See Menezes-Filho and Muendler (2011) and Dix-Carneiro and Kovak (2019) for responses to Brazil’s domestic trade liberalisation in the late 1980s and early 1990s and McCaig and Pavcnik (2018) for within industry adjustment to new export opportunities in Vietnam. 3 A related literature uses the region or local labour market as the unit of analysis and examines labour market adjustment to changes in trade policy in developing countries. See Topalova (2010), McCaig (2011), Kovak (2013), Dix-Carneiro and Kovak (2019) and Erten et al. (2019). 4 Artuc et al. (2015) estimate that labour mobility costs are higher in Sub-Saharan African countries than in other developing regions. 5 Based on output-side real GDP in PPP and population estimates from Penn World Tables 8.0. 6 Of course, cheaper imports also benefit consumers but because we are using labour force survey data, a complete welfare analysis is beyond the scope of this paper. 7 See Section 3.1 for a description of our definition of working in an informal firm in Botswana. 8 See, for example, Attanasio et al. (2004), Currie and Harrison (1997), Dix-Carneiro (2014), Feliciano (2001), Revenga (1997), Topalova (2010) and surveys by Goldberg and Pavcnik (2007), Harrison et al. (2011). 9 Gaddis and Pieters (2017) and Dix-Carneiro and Kovak (2019) report that Brazil’s domestic trade liberalisation was associated with a rise in non-employment (i.e., a rise in unemployment or non-participation). Erten et al. (2019) report similar evidence for South Africa. 10 For example, in work on Brazil’s trade liberalisation, Goldberg and Pavcnik (2003), Bosch et al. (2012), Menezes-Filho and Muendler (2011), Dix-Carneiro and Kovak (2019) focus on whether a worker within a formal firm has a signed work card, while Paz (2014) defines informal workers for whom payroll tax is not paid. Goldberg and Pavcnik (2003) rely on an indicator for the payment of social security taxes in Colombia. 11 The 2005/06 LFS sample included persons 7 years old and above while the 1995/96 only included persons 12 years old and above. Nevertheless, both surveys can be easily compared by controlling for age. 12 Both labour force surveys report industry of work using the Botswana Standard Industrial Classification, which is adapted from the International Standard Industrial Classification, revision 3. 13 See questions 24 through 30 in the 1995/96 LFS and questions 34 through 39 in the 2005/06 LFS. 14 Note that this definition of informality differs from many papers in the trade and informality literature that focus on workers in formal firms and define informal employment as failure to comply with labour legislation (see Goldberg and Pavcnik (2003), Paz (2014) and Dix-Carneiro and Kovak (2019) among others). 15 Although many papers do not report results for testing whether change in trade policy or trade flows are correlated with initial conditions, it is very common in the literature to include initial conditions as controls in subsequent regression analysis due to concerns about correlations with initial conditions. Examples of papers that include initial conditions as controls include Topalova (2010), McCaig (2011), Autor et al. (2013) and Costa et al. (2016). 16 See Frazer and Van Biesebroeck (2010) and Rotuno et al. (2013) for analysis of the causal effects of AGOA on trade flows. 17 Authors’ calculation based on data from UNComtrade. 18 We get very similar results if we use the share of individuals affiliated with an industry, which includes both workers and unemployed individuals, as the dependent variable. These are available from the authors on request. 19 In Tables A4 and A5, we provide disaggregated results for our two definitions of unemployment based on individual characteristics. We estimate the same model as in Panel C of Tables 7 and 8 respectively. In most of these comparisons, cautious interpretation should be given to the differences among workers due to the large standard errors. 20 In Table A6, we estimate equation (3) for various subsamples of workers. We report specifications that include initial industry characteristics interacted with time fixed effects and we exclude clothing. 21 Goldberg and Pavcnik (2003) find an increase in informality within industries in Colombia due to trade liberalisation, but only prior to a major labour market reform. However, in Brazil, they find no effect on informality within industries. This is in contrast to Bosch et al. (2012) that do find an increase in informality within industries in Brazil. Dix-Carneiro and Kovak (2019) construct regional measures of informality and exposure to trade liberalisation in Brazil. Paz (2014) also studies Brazil’s trade liberalisation, but defines informality based on the payment of payroll taxes as compared to having a signed work card as in the other Brazil studies. References Artuc E. , Lederman D. , Porto G. ( 2015 ) ‘ A mapping of mobility costs in the developing world ’, Journal of International Economics , 95 : 28 – 41 . Google Scholar Crossref Search ADS WorldCat Attanasio O. , Goldberg P. K. , Pavcnik N. ( 2004 ) ‘ Trade reforms and wage inequality in Colombia ’, Journal of Development Economics , 74 : 331 – 66 . Google Scholar Crossref Search ADS WorldCat Autor D. H. , Dorn D. , Hanson G. ( 2013 ) ‘ The China syndrome: local labor market effects of import competition in the United States ’, American Economic Review , 103 : 2121 – 68 . Google Scholar Crossref Search ADS WorldCat Bosch M. , Goñi-Pacchioni E. , Maloney W. ( 2012 ) ‘ Trade liberalization, labor reforms and formal-informal employment dynamics ’, Labour Economics , 19 : 633 – 67 . Google Scholar Crossref Search ADS WorldCat Cassim R. , Onyango D. , Van Seventer D. E. ( 2004 ) ‘ The state of trade policy in South Africa ’, Trade and Industrial Policy Strategies (TIPS) . WorldCat Center for Global Development ( 2014 ) ‘ Delivering on the Data Revolution in Sub-Saharan Africa: Final Report of the Data for African Development Working Group ’. Washington, DC : Center for Global Development . Google Preview WorldCat COPAC Central Statistics Office ( 2009 ) ‘ 2007 Informal Sector Survey Report, Central Statistics Office of Botswana ’. Botswana : Gaborone . Google Preview WorldCat COPAC Costa F. , Garred J. , Pessoa J. P. ( 2016 ) ‘ Winners and losers from a commodities-for-manufactures trade boom ’, Journal of International Economics , 102 : 50 – 69 . Google Scholar Crossref Search ADS WorldCat Currie J. , Harrison A. ( 1997 ) ‘ Sharing the costs: the impact of trade reform on capital and labor in Morocco ’, Journal of Labour Economics , 15 : S44 – 71 . Google Scholar Crossref Search ADS WorldCat Dix-Carneiro R. ( 2014 ) ‘ Trade liberalization and labor market dynamics ’, Econometrica , 82 : 825 – 85 . Google Scholar Crossref Search ADS WorldCat Dix-Carneiro R. , Kovak B. K. ( 2019 ) ‘ Margins of labor market adjustment to trade ’, Journal of International Economics , 117 : 125 – 42 . Google Scholar Crossref Search ADS WorldCat Edwards L. ( 2005 ) ‘ Has South Africa liberalised its trade? ’ South African Journal of Economics , 73 : 754 – 75 . Google Scholar Crossref Search ADS WorldCat Erten B. , Leight J. , Tregenna F. ( 2019 ) ‘ Trade liberalization and local labor market adjustment in South Africa ’, Journal of International Economics , 118 : 448 – 67 . Google Scholar Crossref Search ADS WorldCat Feliciano Z. M. ( 2001 ) ‘ Workers and trade liberalization: the impact of trade reforms in Mexico on wages and employment ’, Industrial and Labor Relations Review , 55 : 95 – 115 . Google Scholar Crossref Search ADS WorldCat Frazer G. , Van Biesebroeck J. ( 2010 ) ‘ Trade growth under the African growth and opportunities act ’, The Review of Economics and Statistics , 92 : 128 – 490 . Google Scholar Crossref Search ADS WorldCat Gaddis I. , Pieters J. ( 2017 ) ‘ The gendered labor market impacts of trade liberalization ’, The Journal of Human Resources , 52 : 457 – 90 . Google Scholar Crossref Search ADS WorldCat Goldberg P. K. , Pavcnik N. ( 2003 ) ‘ The response of the informal sector to trade liberalization ’, Journal of Development Economics , 72 : 463 – 96 . Google Scholar Crossref Search ADS WorldCat Goldberg P. K. , Pavcnik N. ( 2005 ) ‘ Trade, wages, and the political economy of trade protection: evidence from the Colombian trade reforms ’, Journal of International Economics , 66 : 75 – 105 . Google Scholar Crossref Search ADS WorldCat Goldberg P. K. , Pavcnik N. ( 2007 ) ‘ Distributional effects of globalization in developing countries ’, Journal of Economic Literature , 45 : 39 – 82 . Google Scholar Crossref Search ADS WorldCat Goldberg , P. K. and Pavcnik , N. ( 2016 ) ‘ The effects of trade policy ’, in Handbook of Commercial Policy . Amsterdam : Elsevier , vol. 1 , pp. 161 – 206 . Google Preview WorldCat COPAC Harrison A. , McLaren J. , McMillan M. ( 2011 ) ‘ Recent perspectives on trade and inequality ’, Annual Review of Economics , 3 : 261 – 89 . Google Scholar Crossref Search ADS WorldCat Hasan R. , Mitra D. , Ranjan P. , Ahsan R. N. ( 2012 ) ‘ Trade liberalization and unemployment: Theory and evidence from India ’, Journal of Development Economics , 97 : 269 – 80 . Google Scholar Crossref Search ADS WorldCat Kovak B. ( 2013 ) ‘ Regional effects of trade reform: what is the correct measure of liberalization? ’ American Economic Review , 103 : 1960 – 76 . Google Scholar Crossref Search ADS WorldCat La Porta R. , Shleifer A. ( 2008 ) ‘ The unofficial economy and economic development ’, Brookings Papers on Economic Activity , 2008 : 275 – 363 . Google Scholar Crossref Search ADS WorldCat La Porta R. , Shleifer A. ( 2014 ) ‘ Informality and development ’, Journal of Economic Perspectives , 28 : 109 – 26 . Google Scholar Crossref Search ADS WorldCat Levinsohn , J. and McCrary , J. ( 2010 ) The prevalence of HIV in Botswana . Maloney W. F. ( 2004 ) ‘ Informality revisited ’, World Development , 32 : 1159 – 78 . Google Scholar Crossref Search ADS WorldCat McCaig B. ( 2011 ) ‘ Exporting out of poverty: provincial poverty in Vietnam and U.S. market access ’, Journal of International Economics , 85 : 102 – 13 . Google Scholar Crossref Search ADS WorldCat McCaig B. , McMillan M. , Verduzco-Gallo I. , Jefferis K. ( 2017 ) in M. McMillan, D. Rodrik and C.P. Sepulveda (eds) , Stuck in the Middle? Structural Change and Productivity Growth in Botswana, in Structural Change, Fundamentals, and Growth: A Framework and Case Studies . Washington, D.C. : International Food Policy Research Institute . Google Preview WorldCat COPAC McCaig B. , Pavcnik N. ( 2018 ) ‘ Export markets and labor allocation in a low-income country ’, American Economic Review , 108 : 1899 – 941 . Google Scholar Crossref Search ADS WorldCat Menezes-Filho N. A. , Muendler M.-A. ( 2011 ) ‘ Labor reallocation in response to trade reform ’, NBER Working Paper , 17372 . WorldCat Nataraj S. ( 2011 ) ‘ The impact of trade liberalization on productivity: evidence from India’s formal and informal manufacturing sectors ’, Journal of International Economics , 85 : 292 – 301 . Google Scholar Crossref Search ADS WorldCat Paz L. S. ( 2014 ) ‘ The impacts of trade liberalization on informal labor markets: a theoretical and empirical evaluation of the Brazilian case ’, Journal of International Economics , 92 : 330 – 48 . Google Scholar Crossref Search ADS WorldCat Revenga A. ( 1997 ) ‘ Employment and wages effects of trade liberalization: the case of Mexican manufacturing ’, Journal of Labor Economics , 15 : S20 – 43 . Google Scholar Crossref Search ADS WorldCat Rotunno L. , Vézina P.-L. , Wang Z. ( 2013 ) ‘ The rise and fall of (Chinese) African apparel exports ’, Journal of Development Economics , 105 : 152 – 63 . Google Scholar Crossref Search ADS WorldCat Siphambe H. K. ( 2004 ) ‘ Botswana’s economy and labour market: are there any lessons for SADC regional integration? ’ Development Southern Africa , 21 : 353 – 64 . Google Scholar Crossref Search ADS WorldCat Southern Africa Customs Union ( 2017 ) About SACU: History of SACU , http://www.sacu.int/show.php?id=394. Topalova P. ( 2010 ) ‘ Factor immobility and regional impacts of trade liberalization: evidence on poverty from India ’, American Economic Journal: Applied Economics , 2 : 1 – 41 . Google Scholar Crossref Search ADS WorldCat Topalova P. , Khandelwal A. ( 2011 ) ‘ Trade liberalization and firm productivity: the case of India ’, Review of Economics and Statistics , 93 : 995 – 1009 . Google Scholar Crossref Search ADS WorldCat Ulyssea G. ( 2018 ) ‘ Firms, informality, and development: theory and evidence from Brazil ’, American Economic Review , 108 : 2015 – 47 . Google Scholar Crossref Search ADS WorldCat World Bank ( 2015 ) Botswana: Systematic Country Diagnostic . Washington, D.C. : World Bank Group . Google Preview WorldCat COPAC © The Author(s) 2019. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of African Economies Oxford University Press

Trade Liberalisation and Labour Market Adjustment in Botswana

Journal of African Economies , Volume Advance Article – May 1, 2020

Loading next page...
 
/lp/oxford-university-press/trade-liberalisation-and-labour-market-adjustment-in-botswana-c3DGcQN8SD

References (45)

Publisher
Oxford University Press
Copyright
© The Author(s) 2019. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com.
ISSN
0963-8024
eISSN
1464-3723
DOI
10.1093/jafeco/ejz027
Publisher site
See Article on Publisher Site

Abstract

Abstract We study the effects of domestic trade liberalisation on labour markets in Botswana. South Africa is the dominant member of the Southern Africa Customs Union. As such, when South Africa liberalised trade in the 1990s, this induced large and plausibly exogenous tariff reductions for the other customs union members, including Botswana. Using labour force surveys from Botswana spanning a decade, we find that trade liberalisation did not affect the relative size of industries in terms of employment. However, trade liberalisation had effects within industries. We find an increase in the prevalence of working in an informal firm and self-employment, but mixed evidence of effects on unemployment. Hours worked decreased in response to trade liberalisation, partially driven by the movement of workers to informal firms. Despite large increases in aggregate income, trade liberalisation is associated with a reduction in monthly income, but the results are imprecise. Our results also suggest that a positive export demand shock, the 2000 African Growth and Opportunities Act, is associated with a reduction in employment in informal firms in the clothing industry. 1 Introduction To fully realise the gains from trade, workers must reallocate across industries and across firms within industries. However, reallocation is costly for workers (Artuc et al., 2015, Dix-Carneiro, 2014). Evidence from various developing countries suggests that reallocation across sectors in response to trade liberalisation is uncommon.1 More recently, evidence from developing countries has emerged about additional margins of labour market adjustments, such as between formal and informal firms, into non-traded sectors and out of the workforce.2 These studies suggest that earlier attempts to look for the reallocation of workers across industries in response to trade liberalisation in developing countries may have been missing additional margins of adjustment.3 Very little evidence, however, exists on labour market adjustments to changes in trade policy in Sub-Saharan African countries, with the exception of South Africa (Erten et al., 2019). This is largely due to the lack of high quality data. Credible causal estimates of labour market adjustments to trade policy require representative and comparable data before and after a significant change in trade policy along with a viable identification strategy. These ingredients are generally missing within Sub-Saharan Africa.4 In this paper, we overcome these problems by using repeated nationally representative labour force data from Botswana that spans a period of large trade liberalisation. Botswana is a member of the Southern Africa Customs Union (SACU). SACU’s external trade policy has historically been determined by South Africa (Southern Africa Customs Union, 2017). Thus, Botswana experienced a large and externally imposed trade liberalisation when South Africa embarked upon dramatic reductions in tariffs beginning in the middle of the 1990s. The tariffs faced by Botswana have been plausibly exogenous to events within the country, making it an unusual case in which to study the causal effects of trade liberalisation. Botswana has been politically stable for many decades. It has experienced rapid growth and structural change since independence (McCaig et al., 2017). We focus on the period of 1995/96 through 2005/06 due to the availability of nationally representative labour force survey data at the beginning and end of this period. There are no other labour force surveys. During this period, real PPP adjusted GDP per capita grew by 3.4% per annum.5 This period of rapid growth was accompanied by a large increase in the size of the working age population, from 743,000 to 923,000 individuals, stubbornly high levels of unemployment and a significant increase in the rate of employment in informal firms within the manufacturing sector. However, there have been no studies exploring whether the large SACU trade liberalisation influenced labour markets in Botswana during this period. Our research exploits both the exogeneity of the tariff reductions as well as the large variation in tariff cuts across industries to estimate the effects of trade policy on the labour market in Botswana. From a theoretical standpoint, the impact of the tariff reductions on industry employment, informality and unemployment is ambiguous. Tariff reductions by South Africa are expected to lower the price of imports and increase the demand for imports in both South Africa and Botswana. From Botswana’s perspective, the increase in imports from third parties has benefits and costs.6 Cheaper intermediate inputs are likely to raise labour productivity, providing a boon to both the traded and non-traded goods sectors. In addition, to the extent that businesses in the exportables sector were competing for scarce resources with the highly protected importables sector, the reduction in import tariffs could boost the production of exportables. At the same time, cheaper imports may also displace workers and shut down businesses, potentially affecting productivity, employment and the diversity of goods produced in Botswana. Finally, because South Africa has been, and still is, the primary destination for most of Botswana’s non-mineral exports, opening up to trade is likely to erode Botswana’s market share in South Africa, possibly reducing the number of products Botswana exports to South Africa. The net effect of these competing forces will determine the impact of trade liberalisation on industry employment, unemployment, informality, hours worked and income in Botswana. We use nationally representative labour force data from the 1995/96 and 2005/06 labour force surveys. These data cover urban and rural areas, the formal and informal sectors and all industries. In this period, Botswana experienced an increase in the prevalence of informal work in aggregate and within manufacturing, the most exposed sector to the SACU trade liberalisation.7 Informality stood at 27% of the manufacturing sector in 1995/96 and rose to 31% by 2005/06. Additionally, the labour force surveys contain questions regarding the previous industry of employment for the unemployed, allowing us to examine the impacts of tariff cuts on unemployment. Unemployment is very high in Botswana, around 20% during this period. We match workers to SACU tariff cuts by industry to study the reallocation of workers across and within industries. Between 1995/96 and 2005/06, SACU tariffs fell considerably. The average tariff within manufacturing fell from 25.0 to 11.7 percentage points. Importantly, the size of the tariff cuts varied across industries, a feature we exploit for our identification strategy. Clothing experienced the largest tariff reduction, almost 60 percentage points, whereas industries such as office, accounting and computing machinery and printing and publishing experienced negligible tariff cuts. The variation in tariff cuts across industries allows us to compare differential changes in labour market outcomes across industries in relation to trade liberalisation. We find that SACU trade liberalisation was associated with little to no net movement of workers across industries. This is consistent with a large number of studies examining the impacts of domestic trade liberalisation on industry employment in developing countries.8 These studies generally find negligible changes in the allocation of industry employment within the formal sector or overall. We find mixed evidence of impacts on unemployment. We use two definitions of unemployment. Narrow unemployment includes individuals that are available to work and sought work. Broad unemployment further includes individuals that were available to work but did not seek work. The labour force surveys ask unemployed individuals about the industry they were most recently affiliated with, allowing us to match the previous industry of employment with the industry tariff cuts. Our results suggest a decrease in unemployment within manufacturing and an increase in unemployment once mining industries are added to the analysis. Although the implied magnitudes are large, the estimates are not statistically different from 0. Among workers, we find that SACU liberalisation was associated with an increase in the prevalence of working in an informal firm and in self-employment. Furthermore, workers experienced a reduction in hours worked, partially caused by the movement of workers into informal firms, and a decrease in income, although it is imprecisely estimated. Within developing countries, labour force or household survey data that tracks the previous industry of employment is rare. Thus, most of the literature focusing on trade impacts on unemployment (or non-employment, which also includes not participating in the labour market) focuses on regional exposure to trade liberalisation as unemployment cannot be directly linked to a specific industry.9 Studying India’s trade liberalisation in the 1990s, Hasan et al. (2012) are able to link the previous industry of employment to industry trade exposure. They report no increase in unemployment in Indian industries in response to domestic trade liberalisation in the 1990s. The difference in industry-level unemployment responses between India and Botswana may be due to differences in overall labour market policies and institutions. In general, unemployment is much higher in Botswana, around 20%, than in India, suggesting that this is a more common labour market outcome and thus possibly a more likely labour market adjustment mechanism to trade liberalisation. Our definition of informal employment, which is based on the registration status and size of the business, differs from that used in most of the trade and informality literature. That literature has predominantly focused on workers in formally registered firms and defined informal based on failure to comply with labour legislation.10 This approach has the drawback of implicitly ignoring workers in informal firms. Our definition more closely resembles that of McCaig and Pavcnik (2018), which defines informal employment based on the registration status of a business in Vietnam. They find a decrease in informal employment within industries in response to changes in trade policy, but unlike our study, the trade policy changes are a reduction in tariffs in a foreign market (i.e., reduced exporting costs). Hence, our study complements the existing trade and informality literature by considering a definition of informality that is based on the status of the firm (i.e., does not focus only on workers in formally registered firms) in the context of domestic trade liberalisation. This definition of informal employment is appealing due to the direct link to the high prevalence of informal firms in developing countries such as Botswana. Growing evidence suggests that there are large differences in productivity between informal and formal firms (La Porta and Shleifer (2008,2014), Nataraj (2011), McCaig and Pavcnik (2018), Ulyssea (2018)). Hence, the SACU-induced movement of workers away from formal firms to informal firms in Botswana may have lowered aggregate labour productivity, particularly within manufacturing as it was subject to the largest tariff reductions. Our results are robust to focusing on self-employment, which is strongly positively correlated with working in an informal firm. In some ways Botswana’s labour market resembles that of South Africa. Understanding these characteristics is useful for anticipating the margins of adjustment to trade liberalisation in Botswana (Erten et al., 2019). Unemployment is stubbornly high at around 20% and has been for decades. When we include discouraged workers, the unemployment rate rises to almost 30%. The reasons for high unemployment are varied but a combination of relatively high reservation wages, an underdeveloped private sector and generous social safety nets certainly contribute (World Bank, 2015). These relatively high levels of unemployment have stimulated entry into the informal sector (Siphambe, 2004). In 2005 roughly 18% of the non-agricultural labour force was employed in the informal sector. The rate of informality in Botswana is somewhat higher than it is in South Africa but significantly lower than it is in middle income Latin American countries where urban rates of informality reach as high as 70% (Maloney, 2004). By contrast, unemployment in Botswana is considerably higher than it is in middle income countries in Latin America but somewhat lower than in South Africa. Thus, apart from sectoral reallocations, we expect adjustment to trade liberalisation in Botswana to potentially impact both unemployment and informality. The remainder of this paper is organised as follows. In Section 2, we describe SACU trade liberalisation. In Section 3, we introduce the labour force surveys and present descriptive evidence on Botswana’s labour markets. In Section 4, we explore the causal effects of the trade liberalisation on industry employment size, unemployment, informality, hours worked and monthly income. Section 5 concludes. 2 Trade liberalisation in the SACU Botswana has been a member of the SACU since 1910 and SACU’s external trade policy has been dominated by South Africa (Southern Africa Customs Union, 2017). Prior to the end of apartheid in 1994, South Africa pursued a vigorous policy of import substitution (Edwards, 2005). In 1994, the process of trade liberalisation gained momentum due to South Africa’s commitment to the GATT Uruguay Round. Between 1995 and 2006, import tariffs on all traded goods fell significantly. Levels of initial tariffs varied widely across industries, reaching over 90% in clothing. Subsequent liberalisation saw the largest tariff cuts in the industries that had the highest initial tariffs. Tariffs on trade with non-SACU members have been typically set by South Africa, with little or no input from Botswana. Thus, we use South Africa’s tariff structure to determine the level of trade protection for Botswana. We measure trade protection using tariffs (including ad valorem equivalents) plus surcharges for South Africa. Our data on trade protection, provided by Lawrence Edwards, spans the period 1990 to 2008 and is described in detail in Edwards (2005). This dataset comprises tariff rates (including ad valorem equivalents) and surcharges at the 8-digit HS level. To construct the industry tariffs that correspond to the industry classification in the labour force surveys, we construct a weighted average tariff within industries based on concordances available through the World Integrated Trade Solution. The tariff declines experienced by Botswana are advantageous for causal analysis for a number of reasons. First, the tariff reductions were large as the mean tariff within manufacturing fell from 25.0% in 1995 to 11.7% in 2005. Second, the tariff reductions varied significantly across industries. Figure 1 shows the change in industry tariff between 1995 and 2005 versus the level in 1995 across manufacturing and mining. Tariff reductions ranged from 0 to almost 60 percentage points and there is a strong negative relationship between the initial tariff level and the change. Third, the tariff changes are plausibly exogenous to Botswana. Botswana has been a member of the SACU since 1910. However, according to SACU’s website, Botswana did not have a role in setting external tariffs as ‘South Africa retained the sole decision-making power over customs and excise policies’ and set the common external tariffs to benefit South African manufacturers following an import substitution strategy. Botswana continued to lack formal influence in setting common external tariffs until the 2002 SACU Agreement, which established an independent Secretariat and headquarters in Namibia. Hence, the level of tariffs prior to the end of Apartheid and the liberalisation ushered in by South Africa are not likely to have been influenced by economic conditions within Botswana. In Section 4.1 we use regression analysis to explore whether the tariff reductions are related to initial industry conditions in Botswana. Figure 1 Open in new tabDownload slide Change in Industry Tariff Versus Initial Tariff Figure 1 Open in new tabDownload slide Change in Industry Tariff Versus Initial Tariff 3 Labour force surveys and aggregate trends 3.1 Labour force survey data We rely on two labour force surveys conducted in 1995/96 and 2005/06. The labour force surveys were conducted by the Central Statistics Office of Botswana. These surveys aim to be a source of nationally representative information on the size, structure, and main characteristics of the labour force. They include information on employment in both formal and informal firms, as well as detailed categories of work status for both workers and those not working. Data for these surveys was collected throughout the 12 months of the duration of the survey. Both the 1995/96 and 2005/06 surveys asked virtually the same questions, with the 2005/06 survey adding some questions on child employment, so data from both surveys is comparable.11 There are two main definitions of employment in the labour force surveys, each with its own time frame. The surveys ask about the main type of work the person has been doing in the past 12 months (‘usual employment’) and the type of work the person did in the past 7 days (‘current employment’). We focus on current employment during the past 7 days due to the more detailed questions asked about the current job as opposed to the usual job. Furthermore, we focus on employment outside of agriculture due to comparability issues introduced by seasonal agricultural work when using the past 7 days employment. If the respondent worked during the past 7 days or was temporarily absent, she was asked about the occupation, industry, whether the work was paid or self-employment, and a series of questions to determine if the organisation the individual worked for is formal or informal.12 We follow the conventional definition of an informal sector business in Botswana. An informal business is identified by multiple characteristics, including ‘non registration of a company with the Registrar of Companies or legal professionals; informal accounts or none; 5 or less paid employees; expenditure not easily distinguishable from household ones; enterprise often temporary or mobile or in owner’s home’ (Central Statistics Office, 2009). Under this definition, all workers in central and local government, parastatals, non-government organisations, private households and private firms with more than 10 workers are defined as working in formal organisations. Workers in private firms with 10 or fewer workers are classified by the enumerator as either being in a formal or informal business based on four questions: location (permanent; on a footpath, street or open space; market; in someone’s home; or no fixed location), how many workers are paid versus unpaid within the firm, whether the firm is a registered company or is registered with a professional organisation and whether the firm keeps a complete set of accounts. This leaves some discretion for the enumerator as a firm may not keep complete accounts, but could be a registered company or the firm could not be registered but keeps complete accounts and most of it (10 or fewer workers) are paid employees. For individuals that did not work or were not temporarily absent during the past 7 days, they were asked whether they were available for work during the past 7 days. We exclude non-working individuals that reported being unavailable to work since the questionnaires did not ask these individuals about their previous industry of work and hence we have no industry tariff to match with these individuals. Among individuals that were available to work, but did not, we create two definitions of unemployment. The first definition, narrow unemployment, is based on being available to work and trying to find work in the past 30 days. Our second definition, broad unemployment, includes individuals that were available to work, but did not report trying to find work during the past 30 days. Among individuals that were available to work, but did not seek work, the most commonly reported reason was that they thought there was no work available. We focus on unemployed individuals that had previously worked and report the industry of their most recent job. Table A1 in Appendix A provides summary statistics for employed individuals. We have 977 and 631 workers in manufacturing, 1,614 and 935 workers in mining and manufacturing and excluding workers in agriculture, we have 8,895 and 6,587 observations in 1995/96 and 2005/06 respectively. Table A2 provides summary statistics for unemployed individuals based on the industry of their most recent job. Within manufacturing, unemployed individuals are younger, on average, than employed individuals. They are also more likely to be female. We report industry summary statistics in Table A3. We have 5 industries within mining and quarrying and 27 industries in manufacturing for a total of 32 traded industries. The table reports the share of economy-wide employment within each traded industry and for non-traded industries aggregated together. Within traded industries, the largest are clothing, diamond mining and copper or nickel mining. The share of workers in traded industries fell between 1995/96 and 2005/06. The timing of the initial labour force survey is after SACU trade liberalisation had already started in 1994 (Cassim et al. (2004), Edwards (2005)). As such, any adjustment that had already taken place by the time of the 1995/96 LFS will be missed in our estimates. We thus view our estimates as a lower bound of the effect by the time of the 2005/06 LFS as we match the 1995 and 2005 tariffs to the 1995/96 and 2005/06 LFSs by industry. It would be ideal to have an additional labour force survey or population census prior to SACU liberalisation as a way to check whether pre-existing trends within Botswana’s labour markets are correlated with subsequent industry tariff reductions. Unfortunately, such data is not available. There is no labour force survey prior to the 1995/96 LFS. Additionally, the 1991 population census used more aggregated industry codes, leaving only 11 manufacturing industries, and the 1981 population census grouped all manufacturing industries together. Hence, data availability is a constraint, as is common in most sub-Saharan countries (Center for Global Development, 2014). 3.2 Changes in employment, unemployment and informality We begin by examining the reported activity of working age individuals. Table 1 shows the number and share of individuals according to whether they were working; seeking work; not available to work; or available to work, but did not look for a job during the past 30 days. The estimates are based on the individual’s activity during the past 7 days, according to the 1995/96 and 2005/06 labour force surveys, respectively. We focus on individuals between the ages of 15 and 60 inclusive. Over the sample period, the labour force participation rate increased by 8.5 percentage points as the number of working age individuals increased by 24% and the number of individuals not in the labour force remained roughly constant. Among labour force participants, the share of those actively seeking work fell from 22% to 19%. However, the share of the working age population who reported that they were available to work but did not work during the past 30 days (discouraged workers) increased from 10.9% to 13.9%. If we count both those actively seeking work and discouraged workers, the share of the working age population that is unemployed is very high in both periods at 23.3% and 26.0%, respectively. Table 1 Economic Activity of the Working Age Population Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab Table 1 Economic Activity of the Working Age Population Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Number of individuals Percentage of individuals 1995/96 2005/06 1995/96 2005/06 Total 743,403 920,646 100 100 In labour force: 415,251 592,993 55.9 64.4 Working 323,034 481,409 43.5 52.3 Actively seeking work 92,217 111,584 12.4 12.1 Not in labour force: 328,152 327,653 44.1 35.6 Not available to work 247,073 199,610 33.2 21.7 Attending school 109,821 109,904 14.8 11.9 Engaged in household duties 101,658 38,915 13.7 4.2 Too old 876 1,798 0.1 0.2 Sick 22,570 31,523 3.0 3.4 Disabled 5,826 4,101 0.8 0.4 Other 6,322 13,368 0.9 1.5 Available but did not look for work 81,079 128,043 10.9 13.9 Thought no work available 41,409 61,101 5.6 6.6 Awaiting reply for earlier inquiries 6,089 9,037 0.8 1.0 Waiting to start arranged job 2,598 3,142 0.3 0.3 Occupied with household duties 25,989 37,331 3.5 4.1 Other 4,994 17,432 0.7 1.9 Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab HIV prevalence is extremely high in Botswana. Using Botswana’s AIDS survey and correcting for selection bias, Levinsohn and McCrary (2010) estimate prevalence rates to be around 17%. According to UNAIDS, the prevalence among working age individuals is much higher, at close to 40%. Contrary to our expectations, illness does not seem to be a major determinant of labour force participation. The share of the population reporting that they were not in the labour force due to illness increased only marginally from 3% in 1995/96 to 3.4% in 2005/06. This may be a testament to the government’s aggressive campaign to treat individuals who are HIV positive. With our data, it is impossible to tell. However, given the importance of HIV in Botswana, in our econometric work we control for the individual characteristics that Levinsohn and McCrary found to be important correlates of HIV status: age, location (urban or rural) and education. Next, in Table 2 we examine the distribution of workers across sectors. The estimates are based primarily on the current job of the individual, but for agriculture we also report estimates based on the usual job over the past 12 months because of the seasonality of agricultural employment (recall that the surveys were conducted across 12 months). Additionally, for the workers usually working in agriculture, we report the number of seasonal and non-seasonal workers. An agricultural worker is defined as seasonal if the worker reported working less than 52 weeks and reported working on their own/family lands/cattlepost (hereafter referred to as own lands) for their usual job. In contrast, a non-seasonal agricultural worker reported working all 52 weeks in the sector. The distinction between the individual’s usual and current job is crucially important in agriculture. It is immediately obvious that the two definitions yield very different trends in estimates of the number of workers in agriculture. The number of current workers in agriculture grew by 182%, while the number of usual agricultural workers grew by only 19%. The most likely explanation for these differences has to do with the timing of interviews. More households were surveyed during the lean season in 1995/96 than in 2005/06. As a result, in Section 4 we focus our empirical analysis on non-agricultural workers. Table 2 Distribution of Workers Across Major Sectors Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The estimates reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The number of agricultural workers is based on the individual’s usual job and only includes individuals who reported doing some work on their own land/farm/cattlepost as usual job. The total number of workers in 1995/96 differs from Table 4 (on labour force) since there were 476 individuals with undefined industry. Non-seasonal agricultural workers are those who reported working all 52 weeks of the year and reported the sector as their usual sector of employment. Seasonal agricultural workers are those who reported not working all 52 weeks of the year and reported the sector as their usual sector of employment. Open in new tab Table 2 Distribution of Workers Across Major Sectors Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Number of workers Percentage of workers 1995/96 2005/06 1995/96 2005/06 Panel A: Non-agriculture Public administration 99,526 114,850 35.6 31.8 Foreign missions and international organisations 224 895 0.1 0.2 Mining and quarrying 15,028 14,289 5.4 4.0 Manufacturing 27,899 34,076 10.0 9.4 Electricity, gas and water supply 2,794 4,132 1.0 1.1 Construction 38,759 26,473 13.9 7.3 Wholesale and retail trade, hotels, restaurants, etc. 51,973 85,416 18.6 23.7 Transport, storage and communication 7,644 15,904 2.7 4.4 Finance, insurance, real estate and business services 11,379 30,333 4.1 8.4 Community, social, household and personal services 24,572 34,349 8.8 9.5 Total 279,798 360,718 100.0 100.0 Panel B: Agriculture Current 42,760 120,691 Usual 120,926 143,746 Non-seasonal 28,862 44,086 Seasonal 92,064 99,660 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The estimates reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The number of agricultural workers is based on the individual’s usual job and only includes individuals who reported doing some work on their own land/farm/cattlepost as usual job. The total number of workers in 1995/96 differs from Table 4 (on labour force) since there were 476 individuals with undefined industry. Non-seasonal agricultural workers are those who reported working all 52 weeks of the year and reported the sector as their usual sector of employment. Seasonal agricultural workers are those who reported not working all 52 weeks of the year and reported the sector as their usual sector of employment. Open in new tab Outside of agriculture, as reported in Panel A, there were some important changes in the distribution of workers across sectors. The share of workers in public administration and construction fell appreciably, while the number of workers in wholesale and retail trade, hotels, restaurants, etc.; and finance, insurance, real estate and business services increased. The remaining sectors experienced only marginal changes in their shares of the non-agricultural workforce. In the context of the SACU tariff cuts, it is interesting to note that the share of workers in manufacturing fell, but only by 0.6 percentage points, and the share of workers in mining fell by 1.4 percentage points or 27%. In Table 3, we return to the issue of unemployment and labour force participation by various segments of the working age population. We report these rates for males and females, urban and rural workers, by age (30 or younger and older than 30) and by education levels. We find significant differences across groups. For example, males are less likely to be unemployed and more likely to be in the labour force than females in either year, although both genders significantly increased their labour force participation rates between 1995/96 and 2005/06. Urban individuals have a higher rate of narrow unemployment than rural individuals and the gap has grown to 5.4 percentage points by 2005/06. Urban individuals are also more likely to be participating in the labour force, although the gap has shrunk considerably over time. There is not much difference in either unemployment or labour force participation rates across individuals with primary or secondary education. However, having some education, as compared to no formal education, is an important determinant of employment status. Lastly, older workers have a higher rate of labour force participation and lower rates of both narrow and broad unemployment than younger workers. Table 3 Unemployment and Labour Force Participation Rates Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The narrow unemployment rate is defined as the ratio of the number of individuals seeking work relative to the sum of the number of workers and individuals seeking work. The broad unemployment rate is defined as the ratio of the number of individuals available to work relative to the sum of the number of workers and individuals available to work. Open in new tab Table 3 Unemployment and Labour Force Participation Rates Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Narrow unemployment rate (%) Broad unemployment rate (%) Labour force participation rate (%) 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 All 22.2 18.8 29.4 28.8 55.9 64.4 Males 20.1 16.6 25.9 23.1 62.9 71.9 Females 24.5 21.0 33.0 33.7 49.7 58.3 Urban 22.8 21.0 24.9 28.0 64.5 65.5 Rural 21.4 15.6 34.8 30.0 47.3 62.9 Less than primary 20.4 15.2 31.4 29.4 53.8 63.2 Finished primary 29.0 19.0 34.0 29.4 50.7 58.5 Finished junior secondary 11.6 22.3 15.7 32.6 62.8 61.3 Finished senior secondary 17.8 19.4 20.0 24.3 77.6 77.3 Older than 30 15.0 11.3 23.5 21.7 68.0 75.6 30 years old or younger 29.9 27.5 34.8 35.5 47.0 55.1 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: The sample is restricted to individuals aged 15 to 60. The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. The narrow unemployment rate is defined as the ratio of the number of individuals seeking work relative to the sum of the number of workers and individuals seeking work. The broad unemployment rate is defined as the ratio of the number of individuals available to work relative to the sum of the number of workers and individuals available to work. Open in new tab Table 4 Percentage of Workers in Informal Firms by Major Sector Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: Sample is persons aged 15 to 60. Informal workers are all those workers in the private sector, employed at firms with less than |$10$| employees and defined as informal in the survey by question 30 in the 1995/96 LFS and question 39 in the 2005/06 LFS. Sample excludes agricultural workers.The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab Table 4 Percentage of Workers in Informal Firms by Major Sector Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Percentage of informal workers 1995/96 2005/06 Change Public administration 0.1 0.7 0.5 Foreign missions and international organisations 0.0 0.0 0.0 Mining and quarrying 0.0 0.0 0.0 Manufacturing 27.0 31.3 4.3 Electricity, gas and water supply 0.0 0.0 0.0 Construction 17.5 26.5 9.1 Wholesale and retail trade, hotels, restaurants, etc. 32.4 41.6 9.3 Transport, storage and communication 27.0 36.0 9.0 Finance, insurance, real estate and business services 3.7 6.4 2.7 Community, social, household and personal services 12.7 9.2 -3.5 Total 13.2 18.0 4.8 Source: Authors’ calculations using 1995/96 and 2005/06 LFS data. Note: Sample is persons aged 15 to 60. Informal workers are all those workers in the private sector, employed at firms with less than |$10$| employees and defined as informal in the survey by question 30 in the 1995/96 LFS and question 39 in the 2005/06 LFS. Sample excludes agricultural workers.The numbers reported are population estimates based on using sampling weights. The estimates are based on the individual’s activity during the past 7 days. Open in new tab An additional concern about the SACU tariff cuts, aside from displacement and unemployment, is the movement of workers into informal firms. The labour force surveys define informality according to a series of questions related to the ownership sector (e.g., government, parastatal, NGO or private), the number of workers in the business, the location of the business, whether the business is registered and whether the business keeps a complete set of accounts. The enumerator evaluated whether the worker’s place of work was formal or informal based on these questions.13 Note that the questions about location, business registration and accounts were only asked for workers in businesses with 10 or fewer workers in the private sector. Consequently, for all workers that were not asked the detailed questions related to formality, we classify them as formally employed as they either worked in large private firms or in sectors more likely to be formal (e.g., government, parastatal or NGOs). This definition of informality is thus based on the status of the business, not on whether the business confirms to existing labour legislation. Furthermore, the definition of an informal business, based on a combination of ownership type, employment size, registration status and account keeping, is consistent with those from other countries. For example, La Porta and Shleifer (2008, 2014) define a business as informal based on its registration status. In Vietnam, an informal firm is one that is not registered with the central government as an enterprise under Vietnam’s Enterprise Law. Vietnamese domestic private businesses are not required to register with the government if they have less than 10 workers and do not operate in more than one location (McCaig and Pavcnik (2018)). And in India, manufacturing firms are legally required to register with the government if they have 10 or more workers and use electricity or if they have 20 or more workers, but don’t use electricity (Nataraj (2011)).14 We use this employer-based definition of informality to report on the incidence of informal work by major sector in Table 4. Overall, the percentage of work in informal firms has increased by 4.8 percentage points or 36%, a very significant increase. Importantly, the incidence of informality in the manufacturing sector increased by 4.3 percentage points, which is slightly below the overall increase, but may be partially related to the SACU tariff cuts. We explore this possibility in Section 4.3. Many sectors experienced an increase in the share of workers in informal firms. In fact, the only sector to experience a decrease was community, social, household and personal services. In summary, the share of individuals working increased between 1995/96 and 2005/06; however, the share of discouraged workers also increased. At an aggregate level, employment in manufacturing fell slightly, by 0.6 percentage points, but the share of informal workers in manufacturing increased significantly, by 4.3 percentage points from 27% in 1995/96. In Section 4, we explore whether these changes are related to SACU tariffs reductions. 4 SACU trade liberalisation and labour market adjustment in Botswana 4.1 Endogeneity of SACU tariff cuts Before estimating the impacts of SACU trade liberalisation on labour markets in Botswana, we address possible concerns about the endogeneity of the tariff reductions. As shown in Figure 1, the size of the tariff reductions was closely related to the initial tariff level. Furthermore, as previously discussed, Botswana had no ability to influence the size of the tariff reductions due to South Africa dominating trade policy within SACU. Nonetheless, it is still possible that the size of the tariff reductions may vary with conditions across industries in Botswana. To explore this possibility, we estimate the following equation: $$ \begin{equation} \Delta ln(1+\textrm{tariff}_j)= \alpha+\beta X_j + u_j \end{equation} $$ (1) where |$\Delta ln(1+\textrm{tariff}_j)$| is the change in ln tariff in industry j and |$X_j$| is a vector of initial conditions in industry j. We calculate the following initial conditions from the 1995/96 labour force survey: the share of workers within the industry that work in informal firms, the share of young workers (30 years and less), the share of urban workers, the share of highly educated workers (completed at least junior secondary) and the industry’s share of total employment. We report the results in Table 5 for manufacturing and traded (mining and manufacturing together), and for both sets of industries, but omitting clothing. We find that some of the initial conditions have a statistically significant relationship with the tariff reductions. Specifically, industries with a higher share of workers in informal firms and industries with a greater share of individuals attached to it (either currently working or most recently employed in the industry) experienced greater tariff reductions. Although the coefficient is not statistically different from 0, industries with a greater share of attached individuals that were unemployed also experienced larger tariff decreases. The employment share is a particularly strong predictor of the size of the tariff cut within manufacturing. The finding that initial conditions are correlated with subsequent tariff cuts is common in the literature. For example, Goldberg and Pavcnik (2005) find that the initial share of skilled workers is correlated with subsequent tariff reductions in Columbia. Topalova and Khandelwal (2011) find that the initial share of nonproduction workers is correlated with the size of industry tariff reductions in India.15 Table 5 SACU Tariff Cuts and Initial Industry Conditions Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Note: The dependent variable is the change in ln(1 + tariff). Heteroskedasticity robust standard errors are reported in parentheses. *** p <0.01, ** p <0.05, * p <0.1. Open in new tab Table 5 SACU Tariff Cuts and Initial Industry Conditions Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Manuf. (1) Traded (2) Manuf., no clothing (3) Traded, no clothing (4) Share informal -0.103* -0.176** -0.146* -0.211*** (0.058) (0.072) (0.074) (0.071) Share male 0.068 0.045 0.039 -0.015 (0.067) (0.074) (0.080) (0.080) Share 30 or younger -0.057 -0.072 -0.029 -0.034 (0.074) (0.063) (0.081) (0.064) Share urban -0.049 -0.112 -0.071 -0.136 (0.092) (0.095) (0.096) (0.097) Share highly educated -0.110 -0.080 -0.078 -0.046 (0.195) (0.189) (0.196) (0.192) Share attached to industry -0.735*** -0.537 -0.222 0.097 (0.214) (0.477) (0.279) (0.288) Share (broad) unemployed -0.136 -0.126 -0.143 -0.121 (0.112) (0.088) (0.106) (0.082) Observations 27 32 26 31 R|$^{2}$| 0.515 0.375 0.299 0.273 F 9.652 4.084 6.242 3.927 Prob F 0.000 0.004 0.001 0.006 Note: The dependent variable is the change in ln(1 + tariff). Heteroskedasticity robust standard errors are reported in parentheses. *** p <0.01, ** p <0.05, * p <0.1. Open in new tab In our context, the results are significantly weaker when clothing is excluded from the analysis, as shown in columns 3 and 4. Clothing is a very important industry in this analysis for two reasons. First, it is a very large industry in terms of employment, representing 25% of manufacturing employment and 15% of mining and manufacturing employment in 1995/96. Second, it is the industry that experienced the largest tariff reduction (Figure 1). These two features make it a very influential observation. Furthermore, clothing is an important industry for a second reason. The 2000 US African Growth and Opportunities Act (AGOA) led to duty free US imports from numerous Sub-Saharan African countries, including Botswana.16 Botswana’s clothing exports to the US rose by almost 700% between 2000 and 2005, from 8.9 to 61.7 million USD.17 Hence, in our labour market analysis we will check the robustness of our results controlling for the influence of AGOA on the clothing industry. 4.2 SACU trade liberalisation and industry employment Traditional trade theory predicts that workers should reallocate across industries in response to changes in tariffs. In particular, the industries that received the largest tariff cuts should contract. We begin our formal analysis of the labour market consequences of SACU’s trade liberalisation by looking for evidence of worker reallocation across industries in response to tariff changes. We estimate the following regression model: $$ \begin{equation} s_{jt}=\beta ln(1+\textrm{tariff}_{jt}) + \theta D_{2005} X_{j1995} +\lambda_j+\eta_t+u_{jt} \end{equation} $$ (2) where |$j$| indexes industries, |$t$| indexes time (i.e., 1995/96 and 2005/06), |$s_{jt}$| is the share of the workforce working in industry |$j$| at time |$t$|⁠, |$\textrm{tariff}_{jt}$| is the SACU tariff in industry |$j$| at time |$t$|⁠, |$D_{2005} X_{j1995}$| is a dummy for 2005 interacted with a vector of initial industry conditions, |$\eta _t$| is a time fixed effect and |$\lambda _j$| is an industry fixed effect. The vector of initial industry conditions is the same as in Table 5 and includes the share of workers within the industry in informal firms, the share of individuals affiliated with the industry that are male, 30 and younger, urban, have finished junior secondary, are (broad) unemployed, and the share of individuals affiliated with the industry. We are interested in the coefficient on tariff. Our identifying variation is changes in tariffs over time within an industry. Thus, we are identifying whether industries that experienced larger tariff cuts either grew or shrunk in comparison to industries that received smaller tariff cuts, conditional on initial industry conditions. Conventional trade theory predicts that the industries that experienced the largest tariff cuts should experience a reduction in the share of workers and thus that the coefficient on tariff, |$\beta $|⁠, should be positive. We estimate equation (2) for manufacturing, traded, and for all industries (excluding agriculture). For non-traded industries, we assign a tariff of 0 in both 1995/96 and 2005/06. With the inclusion of industry fixed effects, this implies that identification for non-traded industries is based on no change in the tariff. Table 6, Panel A, presents the results from estimating equation (2). For manufacturing, the coefficient is negative, suggesting an increase in employment shares in response to SACU liberalisation, but the implied magnitude is very small and the coefficient is not statistically different from 0. In column 2, the coefficient is also negative, suggesting an increase in employment share in response to the SACU tariff reductions, but again the implied magnitude is very small and the coefficient is not statistically different from 0. Lastly, in column 3 for all industries (excluding agriculture), the coefficient is positive and statistically different from 0, suggesting that employment in traded industries shrunk relative to non-traded industries in response to SACU liberalisation. Table 6 SACU Tariffs and Industry Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Note: The dependent variable is the share of workers within an industry in a given year. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. The employment shares are calculated relative to total employment within each year in the indicated industries in each column. All regressions have industry and year fixed effects and initial industry conditions (see Table 5) interacted with a 2005 indicator. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 6 SACU Tariffs and Industry Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.015 -0.024 0.021** (0.059) (0.026) (0.009) Observations 54 64 192 R|$^{2}$| 0.964 0.976 0.969 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.001 -0.002 0.021** (0.058) (0.031) (0.009) AGOA indicator 0.086* 0.034* 0.000 (0.049) (0.017) (0.003) Observations 54 64 192 R|$^{2}$| 0.970 0.978 0.969 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.001 -0.002 0.021** (0.057) (0.031) (0.009) Observations 52 62 190 R|$^{2}$| 0.925 0.968 0.968 Note: The dependent variable is the share of workers within an industry in a given year. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. The employment shares are calculated relative to total employment within each year in the indicated industries in each column. All regressions have industry and year fixed effects and initial industry conditions (see Table 5) interacted with a 2005 indicator. The industry tariff is ln(1 + tariff + surcharges). Open in new tab In Panels B and C of Table 6 we explore how sensitive our results are to the clothing industry. In Panel B we add an AGOA indicator variable as an additional control variable. This variable is 1 for the clothing industry in 2005/06 and 0 for all other observations. In Panel C we remove the clothing industry from the analysis. Recall that the clothing industry is the largest manufacturing industry in terms of employment, accounting for 25% of manufacturing workers in 1995/96. Additionally, as previously discussed, the clothing industry was subject to other large trade policy changes during this time period, specifically the AGOA. Given the size of the clothing industry and the clothing-specific effect of AGOA, it makes sense to explore how sensitive our results are to the inclusion of clothing. In both Panels B and C, the coefficients are smaller in magnitude than in Panel A for manufacturing and mining. For all industries, the estimates in Panels B and C are very similar to that in Panel A, again suggesting a relative expansion of the non-tradable sector in response to SACU tariff cuts. Lastly, note that the AGOA indicator is positive in columns 1 and 2, suggesting an expansion in the relative size of the clothing industry, as one would expect from AGOA. In summary, there is little evidence of a systematic reallocation of workers across traded industries in response to the SACU tariff cuts, but some evidence of relocation to non-traded industries.18 These results are consistent with a lack of reallocation across industries in response to domestic trade liberalisation in developing countries surveyed by Goldberg and Pavcnik (2007) and more recently found by Topalova (2010) in the case of India’s trade liberalisation in the early 1990s, Dix-Carneiro (2014) for Brazil’s trade liberalisation in the late 1980s and early 1990s and McCaig and Pavcnik (2018) in Vietnam in the early 2000s in response to reductions in a trading partner’s tariffs (i.e., export market liberalisation). Table 7 SACU Tariffs and Narrow Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Note: The dependent variable is an indicator for being available to work and seeking as work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 7 SACU Tariffs and Narrow Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.069 -0.148 -0.099** (0.385) (0.173) (0.047) Observations 1917 2934 18167 R|$^{2}$| 0.046 0.086 0.088 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.050 -0.267 -0.094 (0.353) (0.243) (0.150) AGOA indicator -0.144 -0.077 0.003 (0.169) (0.111) (0.055) Observations 1917 2934 18167 R|$^{2}$| 0.047 0.086 0.088 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.072 -0.267 -0.094 (0.358) (0.244) (0.150) Observations 1417 2434 17667 R|$^{2}$| 0.050 0.096 0.090 Note: The dependent variable is an indicator for being available to work and seeking as work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab 4.3 SACU trade liberalisation and workers Although the tariff changes do not appear to have substantially affected the distribution of workers across industries, it may be the case that there have been important within-industry changes. For example, workers could have become unemployed, moved to informal firms, changed hours worked or experienced changes in pay. We explore whether there is evidence of changes in worker outcomes in response to the SACU tariff cuts using the following regression model: $$ \begin{equation} Y_{ijt}=\beta ln(1+\textrm{tariff}_{jt}) +\delta Z_{ijt} + \theta D_{2005} X_{j1995} + \lambda_j +\eta_t +u_{ijt} \end{equation} $$ (3) where |$j$| indexes industries, |$t$| indexes time (i.e., 1995/96 and 2005/06), |$Y_{ijt}$| is an outcome variable for worker or individual |$i$|⁠, |$tariff_{jt}$| is the SACU tariff in industry |$j$| at time |$t$|⁠, |$Z_{ijt}$| is a vector of individual characteristics (age, male indicator, urban indicator, finished primary, finished junior secondary, finished senior secondary), |$D_{2005} X_{j1995}$| is a dummy for 2005 interacted with a vector of initial industry conditions (the same as used previously in Tables 5 and 6), |$\lambda _j$| is an industry fixed effect and |$\eta _t$| is a time fixed effect. We are interested in the coefficient on tariffs. Our identifying variation is changes in tariffs over time within an industry, conditional on controlling for observable individual characteristics and initial industry conditions. Controlling for initial industry conditions is important in this context due to their correlation with SACU tariff reductions. Failing to directly control for them would then imply that the SACU tariff cuts are partially picking up any effects related to initial differences in industry size and composition. Unemployment We begin by examining two definitions of unemployment. The first is based on workers that were available to work and sought work. The second is broader and includes individuals that were available for work but did not seek employment. In both cases, our sample includes individuals that are actively working and individuals that are defined as unemployed according to the respective definition. For individuals that are currently working, their industry is that in which they are currently working, but for unemployed individuals, their industry is that in which they most recently worked. Table 8 SACU Tariffs and Broad Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 8 SACU Tariffs and Broad Unemployment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.241 -0.171 -0.211*** (0.463) (0.227) (0.076) Observations 2105 3185 20336 R|$^{2}$| 0.058 0.112 0.123 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.228 -0.296 -0.147 (0.431) (0.306) (0.244) AGOA indicator -0.162 -0.082 0.030 (0.169) (0.134) (0.089) Observations 2105 3185 20336 R|$^{2}$| 0.059 0.112 0.123 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.221 -0.304 -0.148 (0.419) (0.303) (0.244) Observations 1546 2626 19777 R|$^{2}$| 0.061 0.123 0.125 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab In Table 7, we report regression results from estimating (3) for manufacturing, traded (mining and manufacturing) and all industries (excluding agriculture). As before, we explore the sensitivity of our results to the clothing industry. Within manufacturing, column 1, we consistently find a small, statistically insignificant effect on unemployment. Adding mining to the sample (column 2) causes the estimate to become negative and to increase in magnitude, but it remains statistically insignificant in all three panels. Using the estimate from Panel A, the magnitude of the coefficient for the manufacturing and mining sample is relatively large and suggests that the average reduction in manufacturing tariffs, |$-0.107$|⁠, is associated with a 1.6 percentage point increase in the likelihood of narrow unemployment, but the coefficient is estimated imprecisely, limiting our ability to draw firm conclusions. For all industries (column 3), the coefficient is also negative, but is smaller in magnitude. In Table 8, we report results for estimating equation (3) with an expanded definition of unemployment. We include unemployed individuals that sought work as well as unemployed individuals that did not seek work (i.e., discouraged). Across all specifications, the effects are noticeably larger than in Table 7, but are typically not statistically different from 0, especially once we control for the effects of AGOA (Panel B) or remove clothing (Panel C). As in Table 7, the results within manufacturing suggest a reduction in unemployment in association with SACU tariff reductions. Using the estimate from Panel B, the magnitude of the coefficient for manufacturing suggests a 2.4 percentage point decrease in unemployment, broadly defined, due to the mean tariff cut (⁠|$-0.107$|⁠), although the estimate is imprecisely estimated and not statistically different from 0. The results in columns 2 and 3 are qualitatively similar to those for narrow unemployment.19 In summary, our results examining the effects of SACU liberalisation on unemployment are suggestive of a decrease in unemployment within manufacturing and an increase in unemployment in traded industries. However, in both cases the results are imprecisely estimated and not generally statistically different from 0. Our results of a possible increase in unemployment are consistent with those reported for Brazil. Menezes-Filho and Muendler (2011), Gaddis and Pieters (2017) and Dix-Carneiro and Kovak (2019) report that Brazil’s domestic trade liberalisation was associated with a rise in non-formal employment (i.e., a rise in informal employment, unemployment or non-participation). In South Africa, Erten et al. (2019) find evidence of an increase in unemployment in local labour markets in response to SACU trade liberalisation. However, our results differ from Hasan et al. (2012) who report no increase in unemployment in India in response to domestic trade liberalisation in the 1990s. The difference between contexts likely has to do with overall labour market policies and institutions. Botswana, South Africa and Brazil have much higher rates of unemployment than India. For example, Dix-Carneiro and Kovak (2019) report mean non-employment (unemployed plus not participating in the labour force) rates of 36% to 40% across Brazilian regions and Erten et al. (2019) report broad unemployment rates of 23% for South Africa, while Hasan et al. (2012) report unemployment rates between 4.1% and 5.6% across Indian states. Botswana’s unemployment and non-employment rates are more similar in magnitude to those in Brazil and South Africa, suggesting that unemployment is, more generally, a common outcome and thus more likely to be an adjustment margin in response to domestic trade liberalisation. Work in informal firms and self-employment We next explore margins of adjustment among workers. Recent work has highlighted the role of trade in the reallocation of workers between informal and formal firms (McCaig and Pavcnik (2018) and Dix-Carneiro and Kovak (2019)). In Table 9, Panel A, we report the estimation of equation (3) using an indicator for working in an informal firm as the dependent variable. We find evidence of an increase in the prevalence of work in informal firms within manufacturing in response to the tariff reductions. The implied magnitude is large, a 7.8 percentage point increase in the prevalence of working in an informal firm in response to the average tariff reduction within manufacturing. However, the result is imprecisely estimated and not statistically different from 0. In columns 2 and 3 of Panel A, the estimate is very close to 0. Table 9 SACU Tariffs and Work in Informal Firms Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Note: The dependent variable is an indicator variable for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 9 SACU Tariffs and Work in Informal Firms Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.726 0.020 0.053 (0.444) (0.182) (0.177) Observations 1,608 2,549 15,417 R|$^{2}$| 0.330 0.389 0.476 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.743* -0.530** -0.506** (0.394) (0.244) (0.223) AGOA indicator -0.257* -0.357*** -0.269*** (0.140) (0.096) (0.080) Observations 1,608 2,549 15,417 R|$^{2}$| 0.331 0.391 0.477 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.686* -0.540** -0.511** (0.397) (0.250) (0.222) Observations 1,210 2,151 15,019 R|$^{2}$| 0.388 0.455 0.493 Note: The dependent variable is an indicator variable for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab In Panel B, we add an indicator for AGOA. We find that AGOA is associated with a reduction in informality within clothing. Controlling for this influence leads to an increase in the magnitude of the coefficient on SACU tariffs relative to Panel A. The coefficient for manufacturing suggests a 8.0 percentage point increase in the prevalence of working for an informal firm in response to the mean tariff reduction within manufacturing (⁠|$-0.107$|⁠). The inclusion of mining in column 2 leads to a smaller effect but it is still statistically different from 0. Lastly, in column 3, the effect is again smaller than within manufacturing, but is statistically different from 0. The smaller effects in mining and non-traded industries are likely due to the lack of informal businesses in mining and the lower overall prevalence of such work in some large non-traded industries, such as public administration (Table 4). We check the sensitivity of our results to removing clothing from the specification and report the results in Panel C of Table 9. The estimated coefficients are very similar to those in Panel B, both in magnitude and statistical significance.20 The estimates of an increase in informality due to SACU tariff reductions are consistent with evidence from Colombia and Brazil where domestic trade liberalisation was associated with an increase in informality (Goldberg and Pavcnik (2003), Bosch et al. (2012), Paz (2014) and Dix-Carneiro and Kovak (2019)).21 However, those studies primarily focus on a definition of informality that is based on employment conditions for workers within formal firms and focus only on workers in formal firms. Our definition of informality is based on the registration and size of the business and more closely aligns with the definition used by McCaig and Pavcnik (2018). They find a reduction in informal employment in response to new export opportunities. The difference in effects of changes in trade policy are likely due to the positive nature of the export demand shock studied by McCaig and Pavcnik (2018) as compared to domestic trade liberalisation in Botswana. This is consistent with a recent survey by Goldberg and Pavcnik (2016) that highlights how the nature of trade policy changes matters. Table 10 SACU Tariffs and Self-Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Note: The dependent variable is an indicator variable for being self-employed. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 10 SACU Tariffs and Self-Employment Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -1.176* -0.089 0.013 (0.617) (0.240) (0.163) Observations 1,608 2,549 15,417 R|$^{2}$| 0.355 0.410 0.474 Panel B: Baseline specification, plus AGOA indicator Industry tariff -1.186* -0.664** -0.485** (0.583) (0.259) (0.222) AGOA indicator -0.153 -0.372*** -0.240*** (0.128) (0.081) (0.076) Observations 1,608 2,549 15,417 R|$^{2}$| 0.356 0.412 0.475 Panel C: Baseline specification, omitting clothing industry Industry tariff -1.092* -0.665** -0.491** (0.580) (0.273) (0.221) Observations 1,210 2,151 15,019 R|$^{2}$| 0.403 0.471 0.489 Note: The dependent variable is an indicator variable for being self-employed. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Since the definition of an informal firm varies across countries, we provide a robustness check based on self-employment. In particular, we focus on the distinction between paid employment and self-employment. Information on whether a worker is self-employed or working in wage employment is more widely available in household and labour force surveys in developing countries and McCaig and Pavcnik (2018) show that employment in an informal business is highly correlated with self-employment in Vietnam. We find that working in an informal business is highly positively correlated with self-employment in Botswana as well. The correlations are 0.82, 0.84 and 0.73 in manufacturing, traded and all industries, respectively. Thus, using an indicator for self-employment is both a useful check on the sensitivity of our results to the particular definition of an informal firm in Botswana and facilitates comparisons with other low- and middle-income countries. In Table 10, we report results from estimating equation 3 using an indicator for self-employment as the dependent variable. The specifications mimic those in Table 9. Within manufacturing, we consistently find that the SACU tariff cuts are associated with an increase in the prevalence of self-employment within an industry. The estimates are larger than those reported in Table 9 in terms of magnitude. The results in Panel C suggest an increase in self-employment of 11.7 percentage points within an industry that experienced the mean tariff reduction in manufacturing as compared to no reduction. The effects are largest within manufacturing. Self-employment within mining is unobserved in our data, which is likely why the magnitude is lower once mining industries are added to the regression. Table 11 SACU Tariffs and Weekly Hours Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 11 SACU Tariffs and Weekly Hours Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.812* 0.156 -0.045 (0.427) (0.188) (0.063) Observations 1,507 2,413 14,680 R|$^{2}$| 0.039 0.061 0.107 Panel B: Baseline specification, AGOA indicator Industry tariff 0.833** 0.463** 0.063 (0.383) (0.222) (0.121) AGOA Indicator 0.208** 0.199** 0.053 (0.080) (0.078) (0.038) Observations 1507 2413 14680 R|$^{2}$| 0.039 0.062 0.107 Panel C: Baseline specification, omitting clothing Industry tariff 0.822** 0.457** 0.065 (0.389) (0.221) (0.121) Observations 1,142 2,048 14,315 R|$^{2}$| 0.045 0.068 0.109 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab We can further disaggregate self-employment into self-employment with employees, without employees and as an unpaid family helper. These results are reported in Tables A7 through Table A9 in Appendix A. We find that more than half of the increase in self-employment due to SACU trade liberalisation occurred within unpaid family help. The combination of self-employment with employees and unpaid family help accounts for over three quarters of the SACU-induced increase in self-employment. These results are remarkably consistent with the informality results, suggesting that the particular definition of an informal business in Botswana is not driving the results. Hours worked The analysis thus far has focused on extensive margins of adjustment: changes in unemployment and changes in the type of job (working for an informal business and self-employment). In this subsection we explore whether hours worked were affected by SACU trade liberalisation. In Table 11 we report estimates of equation (3) with ln of actual hours worked during the past week as the dependent variable. The results for manufacturing consistently suggest a decrease in weekly hours worked. The estimate in Panel A suggests a decease in weekly hours of 8.7% in response to the mean tariff reduction within manufacturing. The same pattern emerges once mining is included, particularly once clothing is controlled for. Once non-traded industries are added to the analysis (column 3), the estimate is much smaller. The reduction in weekly hours worked is partially driven by the shift of workers into informal firms as workers in informal firms work fewer hours on average. For example, within manufacturing in 1995/96, the mean number of weekly hours was 35.9 and 43.6 hours in informal and formal businesses, respectively. This is confirmed in Table 12, which repeats the previous specifications, adding an indicator for working in an informal firm as a control variable. The coefficient on informal is consistently negative, suggesting a decrease in hours of between 17% and 21% across the various specifications. Moreover, the coefficient on tariffs falls by about 0.2 ln points when the informal control is added relative to Table 11. This suggests the shift of workers into informal firms is partly responsible for the decrease in weekly hours in response to the SACU tariff cuts. Table 12 SACU Tariffs and Weekly Hours Controlling for Informal Work Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table 12 SACU Tariffs and Weekly Hours Controlling for Informal Work Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.638* 0.155 -0.046 (0.370) (0.160) (0.060) Informal -0.214*** -0.219*** -0.211*** (0.067) (0.064) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.068 0.089 0.120 Panel B: Baseline specification, AGOA indicator Industry tariff 0.653* 0.340* -0.052 (0.343) (0.194) (0.134) AGOA Indicator 0.150** 0.120 -0.003 (0.066) (0.072) (0.044) Informal -0.214*** -0.218*** -0.211*** (0.067) (0.065) (0.047) Observations 1,507 2,413 14,680 R|$^{2}$| 0.069 0.089 0.120 Panel C: Baseline specification, omitting clothing Industry tariff 0.689* 0.359* -0.046 (0.361) (0.199) (0.133) Informal -0.169** -0.173** -0.203*** (0.076) (0.073) (0.050) Observations 1,142 2,048 14,315 R|$^{2}$| 0.062 0.084 0.121 Note: The dependent variable is ln of actual hours worked during the past 7 days. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Monthly income Lastly, we report results using ln monthly gross income from paid employment as the dependent variable. This includes both cash and the value of in kind payments. We present the results in Table 13. The specifications follow the same pattern as in previous tables. The results for manufacturing suggest a decrease in monthly income, although the coefficient is not very precisely estimated and is not statistically different from 0. Nonetheless, the suggested magnitude of the effect remains large. For example, the coefficient on tariffs in panel A for manufacturing suggests a reduction in monthly income of 9% in response to the mean tariff reduction. However, when we add workers in mining to the sample, the coefficient switches sign. Recall that mining industries, on average, experienced very small tariff reductions (Table A3, Figure 1). Hence, the change in the coefficient from column 1 to 2 implies that workers in mining experienced slower wage growth than workers in manufacturing industries that experienced small tariff reductions. Table 13 SACU Tariffs and ln Monthly Income Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Note: The dependent variable is the ln of monthly income from paid employment, including both cash and in kind payments. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60 in paid employment. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Income is the sum of paid employment in their primary or secondary job. Open in new tab Table 13 SACU Tariffs and ln Monthly Income Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff 0.861 -0.232 -0.031 (1.239) (0.586) (0.246) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel B: Baseline specification, plus AGOA indicator Industry tariff 0.828 -0.655 -0.720 (1.305) (0.800) (0.599) AGOA indicator -0.418 -0.260 -0.320 (0.335) (0.298) (0.208) Observations 1,128 2,056 12,889 R|$^{2}$| 0.581 0.693 0.685 Panel C: Baseline specification, omitting clothing industry Industry tariff 0.833 -0.659 -0.722 (1.317) (0.809) (0.601) Observations 842 1,770 12,603 R|$^{2}$| 0.585 0.679 0.685 Note: The dependent variable is the ln of monthly income from paid employment, including both cash and in kind payments. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60 in paid employment. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Income is the sum of paid employment in their primary or secondary job. Open in new tab 5 Conclusion We study the impact of trade liberalisation on workers in Botswana. A unique feature of Botswana’s trade liberalisation is that it was driven by South Africa’s control of the external trade policy of the SACU. This provides an excellent context for exploring causal effects since it decreases the likelihood that tariff reductions were influenced by factors within Botswana. The evidence presented in this paper indicates that SACU’s trade liberalisation had little impact on the allocation of employment across industries in Botswana. This is consistent with existing trade and labour market adjustment studies that generally find little or no changes in industry size in developing countries in response to domestic trade liberalisation. We find mixed evidence that the SACU tariff reductions impacted the prevalence of unemployment within industries. The results suggest a decrease in unemployment within manufacturing, but an increase once mining industries are added to the analysis. However, although the implied increase in unemployment is large, the results are imprecisely estimated and not statistically significant. The prevalence of working in an informal business increased within manufacturing due to SACU liberalisation, as did the prevalence of self-employment. We further find that weekly hours decreased and that this is driven in part by the increase in work in informal businesses, as these workers report working fewer hours on average. Lastly, we find evidence of a decrease in monthly income in manufacturing due to SACU liberalisation, but despite large coefficients, the results are imprecise and not statistically different from 0. The lack of statistically significant results is commonly the outcome of large standard errors, not of estimates that are quantitatively close to 0. This highlights the need for either larger labour force surveys in terms of sample sizes or more frequent labour force surveys to help derive more definitive results. This echoes concerns about the lack of data for a large number of Sub-Saharan African countries. The reallocation of workers out of formal firms and into informal firms and unemployment may have important consequences for labour productivity, particularly within manufacturing, the most impacted sector. A growing literature documents that informal firms are typically much less productive than formal firms. Thus, the movement of workers away from formal firms to informal firms is likely associated with a decrease in aggregate labour productivity. The results on unemployment, working in an informal business and self-employment add to the more recent literature on labour market adjustment to trade in developing countries. In particular, there is scant evidence on the effects of trade liberalisation on unemployment within industries in developing countries due to the lack of data in household or labour force surveys on the previous industry of employment (Goldberg and Pavcnik (2007), Hasan et al. (2012)). Additionally, this is one of the first papers to study the impact of trade liberalisation on labour markets in Sub-Saharan Africa, with Erten et al. (2019) being the notable exception. Acknowledgements We thank Colleen Spencer and Inigo Verduzco-Gallo for their research assistance and Anna Majalentle and her staff at the Central Statistical Office of Botswana for assistance with understanding the data. We thank the World Bank for financial assistance. We would like to thank Jeff Chan, Bilge Erten, Azim Essaji, Keith Jefferis, Bob Rijkers, and Happy Siphambe for helpful conversations and Sara Hartley for proofreading. All errors are our own. 6 Appendix Table A1 Summary Statistics for Workers Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Note: The sample is workers aged 15 to 60 outside of agriculture. The reported information is for the individual’s primary job with the exception of income, which is the sum of paid employment in their primary or secondary job. Open in new tab Table A1 Summary Statistics for Workers Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Manufacturing Traded All 1995/95 2005/06 1995/96 2005/06 1995/96 2005/06 Age 33.9 33.4 35.1 35.3 34.1 34.6 Male 0.419 0.460 0.591 0.597 0.498 0.487 Did not complete primary 0.340 0.197 0.325 0.210 0.333 0.176 Completed primary 0.472 0.358 0.405 0.335 0.361 0.292 Completed junior secondary 0.107 0.231 0.129 0.202 0.146 0.203 Completed senior secondary 0.081 0.214 0.141 0.253 0.161 0.328 Urban 0.643 0.734 0.762 0.793 0.541 0.755 Weekly hours worked 41.6 42.9 42.9 44.0 41.9 46.0 Informal 0.261 0.292 0.158 0.197 0.119 0.174 Paid employment 0.744 0.680 0.845 0.784 0.889 0.811 Self-employment with employees 0.040 0.063 0.024 0.043 0.025 0.047 Self-employment without employees 0.201 0.200 0.121 0.135 0.077 0.108 Unpaid family helper 0.015 0.057 0.009 0.039 0.008 0.033 Number of observations 977 631 1,614 935 8,895 6,542 Note: The sample is workers aged 15 to 60 outside of agriculture. The reported information is for the individual’s primary job with the exception of income, which is the sum of paid employment in their primary or secondary job. Open in new tab Table A2 Summary Statistics for Unemployed Individuals Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Note: The sample is unemployed individuals aged 15 to 60 outside of agriculture. Includes individuals who did not work despite being available to work, regardless of whether they sought work. Open in new tab Table A2 Summary Statistics for Unemployed Individuals Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Manufacturing Traded All 1995/96 2005/06 1995/96 2005/06 1995/96 2005/06 Age 30.0 31.5 33.1 33.8 29.3 29.7 Male 0.317 0.378 0.465 0.469 0.385 0.361 Did not complete primary 0.270 0.172 0.392 0.241 0.369 0.239 Completed primary 0.579 0.475 0.482 0.425 0.479 0.301 Completed junior secondary 0.112 0.176 0.094 0.160 0.060 0.238 Completed senior secondary 0.039 0.176 0.032 0.173 0.092 0.222 Urban 0.595 0.718 0.541 0.697 0.456 0.618 Did not seek work 0.328 0.433 0.351 0.446 0.434 0.514 Number of observations 259 238 342 294 3,748 3,708 Note: The sample is unemployed individuals aged 15 to 60 outside of agriculture. Includes individuals who did not work despite being available to work, regardless of whether they sought work. Open in new tab Table A3 Industry Summary Statistics Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Note: This table reports summary statistics for each of the subsectors of employment, excluding agriculture. The sample is individuals aged 15 to 60. Open in new tab Table A3 Industry Summary Statistics Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Tariff percentage point Employment share Share (narrow) unemployment Share (broad) unemployment Share informal Share self-employed 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 1995 2005 Coal Mining 0.0 0.0 0.1 0.1 0.0 19.1 0.0 27.9 0.0 0.0 0.0 0.0 Copper or Nickel Mining 8.0 0.0 2.1 1.6 4.2 7.6 5.5 12.9 0.0 0.0 0.0 0.0 Diamond Mining 0.0 0.0 2.5 1.6 5.3 3.6 10.0 9.9 0.0 0.0 0.0 0.0 Soda Ash and Salt Mining 26.0 2.7 0.2 0.1 14.0 6.3 14.0 6.3 0.0 0.0 0.0 0.0 Gold Mining 1.8 0.9 0.5 0.5 44.5 30.1 56.0 45.0 0.0 0.0 0.0 0.0 Meat and Meat Products 15.3 8.3 1.3 0.5 7.6 15.2 14.0 19.2 2.1 0.0 9.1 0.0 Dairy Products 45.8 20.0 0.2 0.1 6.8 6.5 15.0 6.5 0.0 0.0 6.1 8.5 Grain Mill Products 9.8 7.5 0.6 0.5 3.7 15.1 14.1 20.9 0.0 0.0 0.0 0.0 Bakery Products 40.7 16.3 0.6 0.4 19.8 30.1 27.2 45.9 19.8 23.1 12.6 32.3 Food Products Not Elsewhere Classified 33.1 16.8 0.3 0.0 8.8 34.1 18.8 43.1 37.8 48.0 37.2 48.0 Beverages 27.7 7.3 1.3 1.5 10.4 5.6 16.5 10.3 69.9 76.4 69.8 81.3 Textiles 44.0 20.1 0.5 0.6 18.4 9.1 24.8 26.5 22.3 47.4 20.5 50.5 Clothing 95.9 37.2 2.5 2.2 20.4 22.2 26.2 36.1 33.3 20.4 32.0 23.7 Tanning and Leather Products 32.0 23.0 0.2 0.1 18.8 0.0 20.3 19.7 16.7 45.1 16.7 45.1 Footwear 55.3 29.7 0.0 0.0 43.0 82.9 43.0 85.7 36.6 100.0 36.6 100.0 Wood and Wood Products 13.9 8.2 0.1 0.3 13.4 11.6 20.4 16.9 39.3 36.9 36.6 58.2 Paper and Paper Products 15.9 8.6 0.1 0.1 0.0 38.0 31.9 48.8 0.0 0.0 0.0 0.0 Printing and Publishing 4.1 3.3 0.1 0.2 24.6 21.4 24.6 21.4 18.2 5.8 23.4 6.0 Chemicals and Chemical Products 14.2 5.9 0.1 0.1 17.5 14.0 19.6 36.3 24.1 10.3 24.1 0.0 Rubber and Plastic Products 19.8 13.5 0.1 0.2 22.9 23.0 22.9 28.2 0.0 0.0 0.0 0.0 Non-Metallic Mineral Products 7.9 3.8 0.3 0.7 22.9 18.1 28.5 33.7 21.3 27.1 11.1 27.8 Basic Metals 4.4 1.2 0.2 0.1 11.8 11.0 19.9 11.0 5.0 0.0 27.6 0.0 Fabricated Metal Products 15.2 6.2 0.2 0.7 32.0 16.3 51.2 25.1 17.1 40.9 13.5 40.6 Machinery and Equipment 3.0 1.2 0.1 0.2 7.5 5.0 10.7 5.0 5.6 13.6 5.6 0.0 Office, Accounting and Computing Machinery 0.0 0.0 0.0 0.1 6.6 0.0 6.6 0.0 0.0 23.4 45.5 0.0 Electrical Machinery and Apparatus 12.1 6.5 0.1 0.2 3.6 18.2 15.3 22.3 0.0 25.9 0.0 42.9 Radio, Television and Communication Equipment 18.5 2.3 0.1 0.1 0.0 18.2 0.0 35.3 61.6 35.2 61.6 34.2 Medical, Precision and Optical Instruments 2.6 0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Motor Vehicles, Trailers and Semi Trailers 63.5 24.6 0.0 0.3 0.0 20.3 26.7 20.3 0.0 47.0 0.0 45.4 Other Transport Equipment 1.6 0.2 0.1 0.0 15.6 45.2 15.6 45.2 0.0 0.0 0.0 0.0 Furniture 28.7 17.9 0.4 0.3 22.9 18.7 23.3 30.1 26.1 30.4 22.0 33.0 Manufacturing of Products Not Elsewhere Classified 26.3 10.4 0.5 0.1 18.2 14.6 22.5 14.6 32.7 0.0 30.3 0.0 Non-traded 0.0 0.0 84.7 86.6 21.2 22.2 31.4 37.4 12.4 17.3 11.1 18.7 Note: This table reports summary statistics for each of the subsectors of employment, excluding agriculture. The sample is individuals aged 15 to 60. Open in new tab Table A4 SACU Tariffs and Narrow Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Note: The dependent variable is an indicator for being available to work and seeking work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. Unemployed is defined as available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A4 SACU Tariffs and Narrow Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff 0.049 -0.264 -0.239 (0.360) (0.317) (0.199) Observations 753 1,634 8,790 R|$^{2}$| 0.061 0.137 0.108 Females Industry tariff 0.105 -0.704 0.007 (0.668) (0.447) (0.173) Observations 663 799 8,873 R|$^{2}$| 0.117 0.119 0.091 Rural Industry tariff 2.145*** 2.127*** 0.377 (0.698) (0.504) (0.230) Observations 493 591 6,607 R|$^{2}$| 0.157 0.184 0.120 Urban Industry tariff -0.430 -0.442* -0.185 (0.378) (0.241) (0.161) Observations 923 1,841 11,050 R|$^{2}$| 0.083 0.129 0.096 Less than complete junior secondary Industry tariff 0.700 -0.215 -0.030 (0.733) (0.412) (0.221) Observations 1,003 1,611 10,787 R|$^{2}$| 0.073 0.116 0.092 Completed junior secondary Industry tariff -0.870 0.010 0.411 (1.181) (0.533) (0.257) Observations 204 355 2,912 R|$^{2}$| 0.201 0.209 0.137 30 years or younger Industry tariff -0.273 -0.760* -0.225 (0.500) (0.380) (0.229) Observations 649 921 7,842 R|$^{2}$| 0.057 0.082 0.071 Older than 30 Industry tariff 1.031 0.439 0.122 (0.634) (0.395) (0.169) Observations 767 1,512 9,821 R|$^{2}$| 0.083 0.160 0.113 Note: The dependent variable is an indicator for being available to work and seeking work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work and sought work. Unemployed is defined as available to work and sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A5 SACU Tariffs and Broad Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. Unemployed is defined as available to work regardless of whether they sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A5 SACU Tariffs and Broad Unemployment by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.034 -0.307 -0.342 (0.438) (0.445) (0.256) Observations 804 1,742 9,491 R|$^{2}$| 0.084 0.186 0.137 Females Industry tariff 0.510 -0.517 -0.007 (0.768) (0.598) (0.360) Observations 741 883 10,282 R|$^{2}$| 0.109 0.118 0.126 Rural Industry tariff 2.115** 1.448** 0.279 (0.895) (0.569) (0.365) Observations 563 697 7,911 R|$^{2}$| 0.113 0.158 0.163 Urban Industry tariff -0.101 -0.285 -0.253 (0.517) (0.288) (0.189) Observations 983 1,928 11,857 R|$^{2}$| 0.093 0.149 0.110 Less than complete junior secondary Industry tariff -0.126 -0.575 -0.181 (0.675) (0.434) (0.355) Observations 1,108 1,773 12,466 R|$^{2}$| 0.082 0.138 0.125 Completed junior secondary Industry tariff 0.521 0.526 0.403 (1.420) (0.509) (0.424) Observations 221 375 3,160 R|$^{2}$| 0.187 0.211 0.167 30 years or younger Industry tariff -0.039 -0.670 -0.183 (0.540) (0.401) (0.247) Observations 714 991 8,851 R|$^{2}$| 0.049 0.090 0.094 Older than 30 Industry tariff 0.833 0.146 -0.050 (0.577) (0.500) (0.341) Observations 831 1,634 10,922 R|$^{2}$| 0.115 0.201 0.168 Note: The dependent variable is an indicator for being available to work as opposed to working. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to individuals aged 15 to 60 that were either working or available to work. Unemployed is defined as available to work regardless of whether they sought work. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A6 SACU Tariffs and Work in Informal Firms by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Note: The dependent variable is an indicator for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. Clothing workers are omitted. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A6 SACU Tariffs and Work in Informal Firms by Worker Characteristics Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Manufacturing (1) Mining and manufacturing (2) All (3) Males Industry tariff -0.792 -0.467 -0.492* (0.527) (0.288) (0.284) Observations 649 1,462 7,548 R|$^{2}$| 0.187 0.256 0.365 Females Industry tariff -0.546 -0.041 -0.454** (0.395) (0.215) (0.228) Observations 559 687 7,466 R|$^{2}$| 0.574 0.615 0.625 Rural Industry tariff 1.470 0.338 -0.636 (1.677) (0.851) (0.575) Observations 431 492 5,597 R|$^{2}$| 0.517 0.560 0.594 Urban Industry tariff -0.574 -0.087 -0.367 (0.389) (0.265) (0.329) Observations 776 1,656 9,412 R|$^{2}$| 0.231 0.286 0.452 Less than complete junior secondary Industry tariff -0.018 -0.513* -0.801*** (0.635) (0.266) (0.304) Observations 864 1,408 8,944 R|$^{2}$| 0.448 0.520 0.508 Completed junior secondary Industry tariff 2.307*** 1.298** -0.216 (0.643) (0.533) (0.562) Observations 173 319 2,537 R|$^{2}$| 0.334 0.388 0.551 30 years or younger Industry tariff -1.101** -0.946** -0.685** (0.511) (0.357) (0.321) Observations 524 780 6,274 R|$^{2}$| 0.327 0.369 0.456 Older than 30 Industry tariff -0.288 -0.201 -0.380* (0.623) (0.309) (0.225) Observations 684 1,369 8,742 R|$^{2}$| 0.474 0.549 0.546 Note: The dependent variable is an indicator for working in an informal firm. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, **p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. Clothing workers are omitted. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A7 SACU Tariffs and Self-Employment With Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Note: The dependent variable is an indicator variable for being self-employed with employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A7 SACU Tariffs and Self-Employment With Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.351 0.029 -0.001 (0.422) (0.135) (0.045) Observations 1,608 2,549 15,417 R|$^{2}$| 0.096 0.095 0.091 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.339 0.042 -0.003 (0.420) (0.213) (0.136) AGOA indicator 0.168* 0.008 -0.001 (0.097) (0.085) (0.048) Observations 1,608 2,549 15,417 R|$^{2}$| 0.097 0.095 0.091 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.325 0.030 -0.004 (0.418) (0.216) (0.136) Observations 1,210 2,151 15,019 R|$^{2}$| 0.098 0.096 0.092 Note: The dependent variable is an indicator variable for being self-employed with employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A8 SACU Tariffs and Self-Employment Without Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Note: The dependent variable is an indicator variable for being self-employed without employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A8 SACU Tariffs and Self-Employment Without Employees Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.116 0.049 -0.027 (0.228) (0.103) (0.074) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.391 0.418 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.127 -0.269* -0.174 (0.211) (0.156) (0.154) AGOA indicator -0.163* -0.206*** -0.071 (0.083) (0.058) (0.052) Observations 1,608 2,549 15,417 R|$^{2}$| 0.364 0.392 0.418 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.069 -0.257 -0.178 (0.212) (0.171) (0.154) Observations 1,210 2,151 15,019 R|$^{2}$| 0.431 0.463 0.433 Note: The dependent variable is an indicator variable for being self-employed without employees. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A9 SACU Tariffs and Unpaid Family Help Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Note: The dependent variable is an indicator variable for being self-employed as an unpaid family helper. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Table A9 SACU Tariffs and Unpaid Family Help Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Manufacturing (1) Mining and manufacturing (2) All (3) Panel A: Baseline specification Industry tariff -0.709** -0.167 0.041 (0.276) (0.128) (0.103) Observations 1,608 2,549 15,417 R|$^{2}$| 0.111 0.105 0.120 Panel B: Baseline specification, plus AGOA indicator Industry tariff -0.720*** -0.437** -0.308*** (0.223) (0.165) (0.115) AGOA indicator -0.158** -0.175*** -0.168*** (0.062) (0.062) (0.041) Observations 1,608 2,549 15,417 R|$^{2}$| 0.113 0.109 0.122 Panel C: Baseline specification, omitting clothing industry Industry tariff -0.698*** -0.438** -0.309*** (0.221) (0.166) (0.115) Observations 1,210 2,151 15,019 R|$^{2}$| 0.120 0.116 0.123 Note: The dependent variable is an indicator variable for being self-employed as an unpaid family helper. Standard errors are reported in parentheses and clustered by industry. *** p <0.01, ** p <0.05, * p <0.1. The sample is restricted to workers aged 15 to 60. All regressions include individual characteristics (age, gender, urban indicator and indicators for finishing primary, junior secondary and senior secondary), initial industry characteristics (see Table 5) interacted with a 2005 indicator and industry and year fixed effects. The industry tariff is ln(1 + tariff + surcharges). Open in new tab Footnotes 1 See Revenga (1997) and Feliciano (2001) for Mexico, Attanasio et al. (2004) for Colombia, Currie and Harrison (1997) for Morocco and Topalova (2010) for India. 2 See Menezes-Filho and Muendler (2011) and Dix-Carneiro and Kovak (2019) for responses to Brazil’s domestic trade liberalisation in the late 1980s and early 1990s and McCaig and Pavcnik (2018) for within industry adjustment to new export opportunities in Vietnam. 3 A related literature uses the region or local labour market as the unit of analysis and examines labour market adjustment to changes in trade policy in developing countries. See Topalova (2010), McCaig (2011), Kovak (2013), Dix-Carneiro and Kovak (2019) and Erten et al. (2019). 4 Artuc et al. (2015) estimate that labour mobility costs are higher in Sub-Saharan African countries than in other developing regions. 5 Based on output-side real GDP in PPP and population estimates from Penn World Tables 8.0. 6 Of course, cheaper imports also benefit consumers but because we are using labour force survey data, a complete welfare analysis is beyond the scope of this paper. 7 See Section 3.1 for a description of our definition of working in an informal firm in Botswana. 8 See, for example, Attanasio et al. (2004), Currie and Harrison (1997), Dix-Carneiro (2014), Feliciano (2001), Revenga (1997), Topalova (2010) and surveys by Goldberg and Pavcnik (2007), Harrison et al. (2011). 9 Gaddis and Pieters (2017) and Dix-Carneiro and Kovak (2019) report that Brazil’s domestic trade liberalisation was associated with a rise in non-employment (i.e., a rise in unemployment or non-participation). Erten et al. (2019) report similar evidence for South Africa. 10 For example, in work on Brazil’s trade liberalisation, Goldberg and Pavcnik (2003), Bosch et al. (2012), Menezes-Filho and Muendler (2011), Dix-Carneiro and Kovak (2019) focus on whether a worker within a formal firm has a signed work card, while Paz (2014) defines informal workers for whom payroll tax is not paid. Goldberg and Pavcnik (2003) rely on an indicator for the payment of social security taxes in Colombia. 11 The 2005/06 LFS sample included persons 7 years old and above while the 1995/96 only included persons 12 years old and above. Nevertheless, both surveys can be easily compared by controlling for age. 12 Both labour force surveys report industry of work using the Botswana Standard Industrial Classification, which is adapted from the International Standard Industrial Classification, revision 3. 13 See questions 24 through 30 in the 1995/96 LFS and questions 34 through 39 in the 2005/06 LFS. 14 Note that this definition of informality differs from many papers in the trade and informality literature that focus on workers in formal firms and define informal employment as failure to comply with labour legislation (see Goldberg and Pavcnik (2003), Paz (2014) and Dix-Carneiro and Kovak (2019) among others). 15 Although many papers do not report results for testing whether change in trade policy or trade flows are correlated with initial conditions, it is very common in the literature to include initial conditions as controls in subsequent regression analysis due to concerns about correlations with initial conditions. Examples of papers that include initial conditions as controls include Topalova (2010), McCaig (2011), Autor et al. (2013) and Costa et al. (2016). 16 See Frazer and Van Biesebroeck (2010) and Rotuno et al. (2013) for analysis of the causal effects of AGOA on trade flows. 17 Authors’ calculation based on data from UNComtrade. 18 We get very similar results if we use the share of individuals affiliated with an industry, which includes both workers and unemployed individuals, as the dependent variable. These are available from the authors on request. 19 In Tables A4 and A5, we provide disaggregated results for our two definitions of unemployment based on individual characteristics. We estimate the same model as in Panel C of Tables 7 and 8 respectively. In most of these comparisons, cautious interpretation should be given to the differences among workers due to the large standard errors. 20 In Table A6, we estimate equation (3) for various subsamples of workers. We report specifications that include initial industry characteristics interacted with time fixed effects and we exclude clothing. 21 Goldberg and Pavcnik (2003) find an increase in informality within industries in Colombia due to trade liberalisation, but only prior to a major labour market reform. However, in Brazil, they find no effect on informality within industries. This is in contrast to Bosch et al. (2012) that do find an increase in informality within industries in Brazil. Dix-Carneiro and Kovak (2019) construct regional measures of informality and exposure to trade liberalisation in Brazil. Paz (2014) also studies Brazil’s trade liberalisation, but defines informality based on the payment of payroll taxes as compared to having a signed work card as in the other Brazil studies. References Artuc E. , Lederman D. , Porto G. ( 2015 ) ‘ A mapping of mobility costs in the developing world ’, Journal of International Economics , 95 : 28 – 41 . Google Scholar Crossref Search ADS WorldCat Attanasio O. , Goldberg P. K. , Pavcnik N. ( 2004 ) ‘ Trade reforms and wage inequality in Colombia ’, Journal of Development Economics , 74 : 331 – 66 . Google Scholar Crossref Search ADS WorldCat Autor D. H. , Dorn D. , Hanson G. ( 2013 ) ‘ The China syndrome: local labor market effects of import competition in the United States ’, American Economic Review , 103 : 2121 – 68 . Google Scholar Crossref Search ADS WorldCat Bosch M. , Goñi-Pacchioni E. , Maloney W. ( 2012 ) ‘ Trade liberalization, labor reforms and formal-informal employment dynamics ’, Labour Economics , 19 : 633 – 67 . Google Scholar Crossref Search ADS WorldCat Cassim R. , Onyango D. , Van Seventer D. E. ( 2004 ) ‘ The state of trade policy in South Africa ’, Trade and Industrial Policy Strategies (TIPS) . WorldCat Center for Global Development ( 2014 ) ‘ Delivering on the Data Revolution in Sub-Saharan Africa: Final Report of the Data for African Development Working Group ’. Washington, DC : Center for Global Development . Google Preview WorldCat COPAC Central Statistics Office ( 2009 ) ‘ 2007 Informal Sector Survey Report, Central Statistics Office of Botswana ’. Botswana : Gaborone . Google Preview WorldCat COPAC Costa F. , Garred J. , Pessoa J. P. ( 2016 ) ‘ Winners and losers from a commodities-for-manufactures trade boom ’, Journal of International Economics , 102 : 50 – 69 . Google Scholar Crossref Search ADS WorldCat Currie J. , Harrison A. ( 1997 ) ‘ Sharing the costs: the impact of trade reform on capital and labor in Morocco ’, Journal of Labour Economics , 15 : S44 – 71 . Google Scholar Crossref Search ADS WorldCat Dix-Carneiro R. ( 2014 ) ‘ Trade liberalization and labor market dynamics ’, Econometrica , 82 : 825 – 85 . Google Scholar Crossref Search ADS WorldCat Dix-Carneiro R. , Kovak B. K. ( 2019 ) ‘ Margins of labor market adjustment to trade ’, Journal of International Economics , 117 : 125 – 42 . Google Scholar Crossref Search ADS WorldCat Edwards L. ( 2005 ) ‘ Has South Africa liberalised its trade? ’ South African Journal of Economics , 73 : 754 – 75 . Google Scholar Crossref Search ADS WorldCat Erten B. , Leight J. , Tregenna F. ( 2019 ) ‘ Trade liberalization and local labor market adjustment in South Africa ’, Journal of International Economics , 118 : 448 – 67 . Google Scholar Crossref Search ADS WorldCat Feliciano Z. M. ( 2001 ) ‘ Workers and trade liberalization: the impact of trade reforms in Mexico on wages and employment ’, Industrial and Labor Relations Review , 55 : 95 – 115 . Google Scholar Crossref Search ADS WorldCat Frazer G. , Van Biesebroeck J. ( 2010 ) ‘ Trade growth under the African growth and opportunities act ’, The Review of Economics and Statistics , 92 : 128 – 490 . Google Scholar Crossref Search ADS WorldCat Gaddis I. , Pieters J. ( 2017 ) ‘ The gendered labor market impacts of trade liberalization ’, The Journal of Human Resources , 52 : 457 – 90 . Google Scholar Crossref Search ADS WorldCat Goldberg P. K. , Pavcnik N. ( 2003 ) ‘ The response of the informal sector to trade liberalization ’, Journal of Development Economics , 72 : 463 – 96 . Google Scholar Crossref Search ADS WorldCat Goldberg P. K. , Pavcnik N. ( 2005 ) ‘ Trade, wages, and the political economy of trade protection: evidence from the Colombian trade reforms ’, Journal of International Economics , 66 : 75 – 105 . Google Scholar Crossref Search ADS WorldCat Goldberg P. K. , Pavcnik N. ( 2007 ) ‘ Distributional effects of globalization in developing countries ’, Journal of Economic Literature , 45 : 39 – 82 . Google Scholar Crossref Search ADS WorldCat Goldberg , P. K. and Pavcnik , N. ( 2016 ) ‘ The effects of trade policy ’, in Handbook of Commercial Policy . Amsterdam : Elsevier , vol. 1 , pp. 161 – 206 . Google Preview WorldCat COPAC Harrison A. , McLaren J. , McMillan M. ( 2011 ) ‘ Recent perspectives on trade and inequality ’, Annual Review of Economics , 3 : 261 – 89 . Google Scholar Crossref Search ADS WorldCat Hasan R. , Mitra D. , Ranjan P. , Ahsan R. N. ( 2012 ) ‘ Trade liberalization and unemployment: Theory and evidence from India ’, Journal of Development Economics , 97 : 269 – 80 . Google Scholar Crossref Search ADS WorldCat Kovak B. ( 2013 ) ‘ Regional effects of trade reform: what is the correct measure of liberalization? ’ American Economic Review , 103 : 1960 – 76 . Google Scholar Crossref Search ADS WorldCat La Porta R. , Shleifer A. ( 2008 ) ‘ The unofficial economy and economic development ’, Brookings Papers on Economic Activity , 2008 : 275 – 363 . Google Scholar Crossref Search ADS WorldCat La Porta R. , Shleifer A. ( 2014 ) ‘ Informality and development ’, Journal of Economic Perspectives , 28 : 109 – 26 . Google Scholar Crossref Search ADS WorldCat Levinsohn , J. and McCrary , J. ( 2010 ) The prevalence of HIV in Botswana . Maloney W. F. ( 2004 ) ‘ Informality revisited ’, World Development , 32 : 1159 – 78 . Google Scholar Crossref Search ADS WorldCat McCaig B. ( 2011 ) ‘ Exporting out of poverty: provincial poverty in Vietnam and U.S. market access ’, Journal of International Economics , 85 : 102 – 13 . Google Scholar Crossref Search ADS WorldCat McCaig B. , McMillan M. , Verduzco-Gallo I. , Jefferis K. ( 2017 ) in M. McMillan, D. Rodrik and C.P. Sepulveda (eds) , Stuck in the Middle? Structural Change and Productivity Growth in Botswana, in Structural Change, Fundamentals, and Growth: A Framework and Case Studies . Washington, D.C. : International Food Policy Research Institute . Google Preview WorldCat COPAC McCaig B. , Pavcnik N. ( 2018 ) ‘ Export markets and labor allocation in a low-income country ’, American Economic Review , 108 : 1899 – 941 . Google Scholar Crossref Search ADS WorldCat Menezes-Filho N. A. , Muendler M.-A. ( 2011 ) ‘ Labor reallocation in response to trade reform ’, NBER Working Paper , 17372 . WorldCat Nataraj S. ( 2011 ) ‘ The impact of trade liberalization on productivity: evidence from India’s formal and informal manufacturing sectors ’, Journal of International Economics , 85 : 292 – 301 . Google Scholar Crossref Search ADS WorldCat Paz L. S. ( 2014 ) ‘ The impacts of trade liberalization on informal labor markets: a theoretical and empirical evaluation of the Brazilian case ’, Journal of International Economics , 92 : 330 – 48 . Google Scholar Crossref Search ADS WorldCat Revenga A. ( 1997 ) ‘ Employment and wages effects of trade liberalization: the case of Mexican manufacturing ’, Journal of Labor Economics , 15 : S20 – 43 . Google Scholar Crossref Search ADS WorldCat Rotunno L. , Vézina P.-L. , Wang Z. ( 2013 ) ‘ The rise and fall of (Chinese) African apparel exports ’, Journal of Development Economics , 105 : 152 – 63 . Google Scholar Crossref Search ADS WorldCat Siphambe H. K. ( 2004 ) ‘ Botswana’s economy and labour market: are there any lessons for SADC regional integration? ’ Development Southern Africa , 21 : 353 – 64 . Google Scholar Crossref Search ADS WorldCat Southern Africa Customs Union ( 2017 ) About SACU: History of SACU , http://www.sacu.int/show.php?id=394. Topalova P. ( 2010 ) ‘ Factor immobility and regional impacts of trade liberalization: evidence on poverty from India ’, American Economic Journal: Applied Economics , 2 : 1 – 41 . Google Scholar Crossref Search ADS WorldCat Topalova P. , Khandelwal A. ( 2011 ) ‘ Trade liberalization and firm productivity: the case of India ’, Review of Economics and Statistics , 93 : 995 – 1009 . Google Scholar Crossref Search ADS WorldCat Ulyssea G. ( 2018 ) ‘ Firms, informality, and development: theory and evidence from Brazil ’, American Economic Review , 108 : 2015 – 47 . Google Scholar Crossref Search ADS WorldCat World Bank ( 2015 ) Botswana: Systematic Country Diagnostic . Washington, D.C. : World Bank Group . Google Preview WorldCat COPAC © The Author(s) 2019. Published by Oxford University Press on behalf of the Centre for the Study of African Economies, all rights reserved. For Permissions, please email: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)

Journal

Journal of African EconomiesOxford University Press

Published: May 1, 2020

There are no references for this article.