Do firms’ exports affect analysts’ forecast errors?

Fu Xin; Shangkun Liang; Jiemin Dai; Xiaorong Du

doi:10.1080/21697213.2016.1252087

Do firms’ exports affect analysts’ forecast errors?

Xin, Fu; Liang, Shangkun; Dai, Jiemin; Du, Xiaorong 2017-01-02 00:00:00 China Journal of aCC ounting StudieS , 2017 Vol . 5, no . 1, 123–150 http://dx.doi.org/10.1080/21697213.2016.1252087 a b c a,d Fu Xin , Shangkun Liang , Jiemin Dai and Xiaorong Du a b School of Business, hohai university, nanjing, China; School of a ccountancy, Central university of f inance and economics, Beijing, China; School of international a udit, nanjing a udit university, nanjing, China; School of Business, hohai university, Jiangsu Provincial Collaborative i nnovation Center of World Water Valley and Water ecological Civilization, nanjing, China ABSTRACT KEYWORDS analysts forecast errors; This paper aims to provide insights into the forces and constraints that China; exporting; financial shape the forecast errors of local analysts when a firm enjoys both crisis; foreign analysts; domestic and foreign earnings. By applying a unique hand-collect propensity score matching; export dataset of Chinese listed firms, the key finding shows that the selection model forecast earnings of exporting firms issued by local analysts deviate more from the actual earnings than those for non-exporting firms. On the contrary, foreign analysts exhibit an informational advantage to local analysts when forecasting the earnings of exporting firms. A two-stage selection model and propensity score matching procedure are applied for correcting the endogeneity problem in a corporate exporting decision. A detailed investigation shows that the forecast errors of local analysts increased significantly during the financial crisis of 2008. Also, the forecast errors of foreign analysts were smaller after the IFRS were been adopted in China than before. Our findings indicate that the accuracy of local analysts may suffer from the macro- economic environment in an export-driven nation. 1. Introduction This paper aims to provide insights into the forces and constraints that shape the forecast errors of local analysts when a firm enjoys both domestic and foreign earnings. First of all, China’s economy has manifested a significantly obvious export-oriented characteristic for a long time and the import and export volume has accounted for over 40% of GDP since 2000. Even if the world at large has suffered the impact of a financial crisis, China’s foreign trade volume in 2014 still reached RMB 26.43 trillion, with year-on-year growth of 2.3%, which was higher than the average level of global trade growth. In particular, the Chinese government has fully supported the “Going global” strategy of Chinese firms from several directions, including the financial system and administrative approval in recent years. In May 2015, the State Council issued Advices on Improving the Cooperation of International Capacity and Equipment Manufacturing, forcefully pushing China’s manufacturing to export to the global market. When observing the sample firms used in this study, almost 50% among the listed CONTACT Jiemin dai daijiemin@163.com *Paper accepted by heng Yue. © 2017 a ccounting Society of China 124 F . XIN ET AL. firms are found to be engaged in the export trade. Therefore, the local analysts are increas- ingly confronted with the impact imposed by the fluctuations of overseas markets on the profit of listed companies, which macroeconomic facts are of vital importance for us to reexamine the behaviour of the forecast errors of local analysts. Besides, from a macroeco- nomic view, the persistence of the components of accounting earnings from different sources vary significantly (Callen, Hope, & Segal, 2009; Sloan, 1996). The export proceeds grow with the increasing growth and complexity of a firm’s export business and international transac - tions. The relatively serious asymmetry of information, universally existing in aspects such as geographical distance, cultural customs and systems environment for local analysts, leads to considerable barriers in dealing with, and processing, the market information of the firm’s exported products. All these three factors consequently cause a systematic deviation for local analysts in the firm’s future earnings forecast. In addition, a questionnaire concerning Chinese local analysts indicated that what most concerned the analysts in relation to the firm’s annual report was the firm’s primary products and market situation (Hu et al., 2003), which conformed to the basic fact of the export-oriented characteristics of China’s enter- prises. In light of this discussion, this study attempts to explore the following two issues. The first is to ask whether the firm’s export business exerts influence on the deviation of the analysts’ earnings forecast. In what circumstances have the forecast errors of local analysts been reduced? This paper has hand-collected the export data of Chinese listed firms from 2002–2012 and has constructed a firm’s export business variables, measured by both a dummy variable and a continuous variable. The study examines systematically the influence that the com- pany’s export business imposes on the forecast errors of local analysts. We find that the forecast earnings of firms with an export business, issued by Chinese local analysts, deviate significantly more than forecasts for those without an export business. Subsequently, we introduce the earnings forecast data issued by foreign analysts and find that the forecast earnings of firms with an export business, issued by foreign analysts, deviate much less. We also explore the influence that the financial crisis exerts on the earnings forecast of analysts in relation to exporting firms. We find that the deviation of the earnings forecasts of Chinese local analysts towards export enterprises has increased during the financial crisis, while the foreign analysts are less vulnerable to suffer from the financial crisis, relying on advantages of information and capability. We also study the influence on the earnings forecast for firms with export business made by foreign analysts who are confronted with the differences between domestic and overseas accounting standards. It is found that the forecast errors of foreign analysts reduce significantly after the new accounting standard has been adopted in China. This phenomenon suggests that the foreign analysts have an informational advan- tage over local analysts because a newly-adopted accounting standard system is harmonised with the International Accounting Standards. among the current research into influential factors concerning the analysts’ forecast behaviour, most researchers regard the company’s products and market distribution as control variables in dealing with the complexity of the analysts’ forecast. f or example, variables including the number of the company’s cross-industries and the number of the products’ cross-re- gions are applied to control for business complexity, which fails to go deep into the company’s products and the concrete structure of market in order to analyse the factors influencing the analysts’ forecast behaviour. t he top two are management information and shareholding change during the reporting period respectively and according to the classification based on nature; the management information and the company’s shareholding change information are regarded as belonging to the company’s internal information, with reference to form 4 of hu et al. (2003). CHINA JOURNAL OF ACCOUNTING STUDIES 125 Our study contributes to the existing literature in three ways. First, a driving factor accounting for the rapid growth in China’s economy is export trade. In the past, existing literature has paid attention only to the influence of export business on China’s macro-econ - omy. In recent years there have been studies of the relationships between export behaviour and the firm’s microeconomic behaviour including financing constraints (Luo & Li, 2014). This study aims to explore the influence of a firm’s export business on the analysts’ deci- sion-making from the perspective of the individual analyst and tries to provide insights in the relationship between the macroeconomic environment and microeconomic firm behav - iour with new evidence. Second, referring to the framework of Duru and Reeb (2002) and based on domestic studies for forecast errors of local analysts (Yuan, Zhang, & Yue, 2014; Wei & Xue, 2015), this study attempts to further interpret the reason for forecast errors of analysts from the mac- roeconomic background of international export trade. Third, the differences of forecast behaviour between domestic analysts and foreign ana- lysts are gradually receiving growing interest. For example, Wang, Chen, and Hou (2010) studied the difference of advantages between domestic and foreign analysts from the per - spective of the consistency of accounting standards. This study conducts analysis of these topics from the view of exporting and seeks to identify the concrete sources of such advan- tages, which supplements the existing literature. The remaining structure of this study is as follows: the second part deals with the literature review and research hypothesis; the third part explains the research design; the fourth part sets out the basic empirical results; the fifth part provides further exploration of the influ- encing mechanism and the last part is the conclusion. 2. Literature review and hypothesis development 2.1. Exports and the analysts’ forecasts Previous studies show that a r fi m’s disclosure of segmental reporting information was helpful to the analysts’ forecasting ability (Balakrishnan et al., 1990; Baldwin, 1984; Collins, 1976; Kinney, 1971; Nichols et al., 1995; Roberts, 1989). However, Duru and Reeb (2002) firstly discovered the impact of the firm’s export business on analysts’ forecasts. They found that the forecast accuracy of analysts in respect of the earnings in those firms with export business would decline and simultaneously there would be more tendency to overestimate. Khurana, Pereira, and Raman (2003) made a relatively systematic study of the relationships between the analysts’ forecast errors on foreign earnings and market efficiency. They found that the analysts could discern the differences of persistency between foreign and domestic earnings, which mainly embodied underestimation of the persistency of foreign earnings. Herrmann, Hope, and Thomas (2008) explored the influence that the US Regulation Fair Disclosure exerted on the analysts when they dealt with the earnings forecast in respect of those firms with overseas business. They found that upon the implementation of Regulation Fair Disclosure, the earnings forecast of analysts in respect of those firms with export business tended to be less positive and significantly more accurate. Subsequently, studies involving the analysts’ forecast would generally take into consideration whether the firm has export business or not as an influential factor (such as Henderson & Marks, 2013). 126 F. XIN ET AL. The domestic impact of the earnings components on the analysts’ forecast behaviour mainly focuses on the analysis of components including accrued profit and cash flow pro - jection. Ji and Tong (2012) studied the analysts’ responses to the persistency of different earnings components. The analysts were found in the studies to be able to differentiate the low persistency of accrued profit and high persistency of cash flow, which to some extent recognised the analytical ability of domestic analysts. However, they also found that analysts failed to differentiate the discretionary and non-discretionary accruals. In addition, Yuan, Zhang, and Yue (2014) found that the analysts’ forecast of the firm’s cash flow was helpful to better understand the earnings structure. As to the firm with an obvious motivation for earnings management, the cash flow forecast released by analysts plays a more prominent role in improving the accuracy of the earnings forecast (Yuan, Zhang, & Yue, 2014). Cai and Zeng (2010) reported the correlation existing between the diversification level of listed firms and the analysts’ attention. Their study suggests that the higher the corporate diversification level, the lower the analysts’ attention, and the higher the corporate diversi- fication with related business, the higher the analysts’ attention. However, Cai and Zeng (2010) did not explore the relationship between the regional distribution of the corporate business and the analysts’ forecast behaviour, and did not further analyse issues concerning the corporate export business. Meanwhile, they also did not take the endogeneity of diver- sification into consideration and confined the research samples to one year’s data of 2008. Huang and Huang (2013) explored the analysts’ forecast behaviour towards accounting-based performance measures (which they label as tangible information), and items such as R&D expenditure and goodwill (which they label intangible information), finding that the earnings forecast was more vulnerable to suffer from the influence of tangible information while the analysts’ recommendation tended to be easily influenced by intangible information. To sum up, in relation to research subject into the firm’s export business and the analyst’s forecast behaviour, for countries outside China the research literature has accumulated certain research findings and shaped related research conclusions. But the aforesaid research issues have not received enough attention in China so far. As China is a major exporter and an emerging nation highly reliant on an export-oriented economy, the above-mentioned issues not only mat- ter but also will generate some more meaningful and valuable topics for researchers. 2.2. Hypothesis development The impact of overseas business on the firm’s earnings falls into two aspects. On the one hand, the firms can reduce the volatility and risk of earnings based on combining businesses in dier ff ent regions or cross-industries, which is one of the reasons for many firms conducting diversification and overseas expansion. On the other hand, a growing literature indicates that the overseas revenue generated from overseas expansion tends to bring higher earnings volatility and therefore increases the firm’s operational risk. For instance, both Goldberg and Heflin (1995) and Reeb et al. (1998) found that cross-regional operation of the firm’s business would be more subject to the influence of political risk, exchange rate risk between the two countries concerned, supervision environment, economic fluctuation and other factors t he conclusion at this point is different from the subsequent discovery concerning whether analysts can restrain the enter - prise earnings management with reference to d egeorge et al. (2013), Yu (2008), Yu et al. (2011). Such as the research into the inu fl ence of the macroeconomic environment on the microcosmic company’s behaviour which have received attention recently , e.g. Chen et al. (2013), Jiang and r ao (2011), Su and Zeng (2009). CHINA JOURNAL OF ACCOUNTING STUDIES 127 where the business was located, and therefore intensified the volatility of the firm’s earnings. Li et al. (2004) also pointed out the lessons of a firm’s failed operation due to ignorance of risk management during the period of China’s listed companies conducting cross-industrial transformation. It can be observed from the descriptive statistics in this paper (VolROA, refer to Table 1) that the earnings volatility in our sample was signica fi ntly higher in the exporting firms than in other firms. Therefore, according to this logic, it is reasonable to predict that the firm’s export business will increase the difficulty of analysts’ earnings forecast, and then enlarge the forecast deviation of analysts. From the viewpoint of the analysts’ ability, the firm’s export business always requires the analysts to equip themselves with knowledge about related exporting countries or regions apart from their own country. With regard to the local analysts, most of them lack the back- ground and information of the firm’s operations concerning related exporting countries or regions. Differences in factors such as customs, systems, rivals, geographic features, super - vision in of different countries or regions impose higher requirements on the analysts’ ability than those local firms without export business and place demands on them to be more professional in interpreting, analysing and evaluating the earnings information of export firms. All these factors will increase the difficulty of the analysts’ forecast. Consequently, from the perspective of analysts, we expect that the firm’s overseas business will increase the difficulty of the analysts’ forecast and further add to the forecast errors of analysts. Hence, research hypothesis 1 is proposed. H1: The forecast earnings of analysts in respect of firms with export business deviate more than those without export business. Following the view of the analysts’ ability, the geographic location of the analysts will produce prominent influence on their forecast errors, and studies in this aspect vary. Malloy (2005) collected the data concerning the geographic location of American analysts and noticed that analysts who located closer to the listed firms tended to boast more accurate forecast. At the same time, the revision information of earnings forecast made by these local or adjacent ana- lysts exerted more significant influence on the stock price. Later, Bae, Stulz, and Tan (2008), based on the data about analysts from 32 countries, found that the local analysts held a fairly obvious forecast advantage which did not exist in companies with overseas assets. Taking the research reports of domestic analysts from 2005 to 2007 as samples, Li, Li, and Zhang (2010) discovered that analysts who located in the same province as the listed firms would forecast earnings more accurately. However, Bacmann and Bolliger (2001) got completely opposite conclusion that overseas analyst forecasted more accurately than local analysts did based on the data of capitalism market in seven Latin American countries. According to the data about analysts from seven European countries, Orpurt (2004) found that the informational advantage of local analysts only existed in Germany while the local advantages cannot be found by the standard of the location of the securities dealers to which the analysts belonged. On the con- trary, Bolliger (2004) found that the local securities dealers from European counties possessed informational advantage. This phenomenon was verified to various extents in Japan (Conroy, Fukuda, & Harris, 1997) and Taiwan. Since domestic analysts lack sensitivity of responses towards export business and overseas market, the foreign analysts will forecast earnings towards companies with export business more accurately by means of informational and capa- bility advantages. Hence, research hypothesis 2 is proposed. H2: The forecast earnings issued by foreign analysts in respect of firms with export business deviate significantly less than those issued by local analysts. 128 F. XIN ET AL. Table 1. Variable definition. Variable names Definitions Source Analysts characteristics Mean ERROR t he absolute difference between the earnings data is from CSM ar. t he construction of forecasts and actual earnings. t he firm i’s the variable is according to hong and forecast error in the year t amounts to the Kacperczyk (2010) absolute value of all the current forecast mean value (median) of analysts tracking the firm i and the current ePS divided by the closing share price of the firm i in the year t − 1. t he analysts’ forecast takes the most recent one among the ePS forecasts made before 180 days Median ERROR We take the median of earnings forecasts data is from CSM ar. t he construction of using the same data as described Mean the variable is according to hong and error Kacperczyk (2010) Number of analysts t he number of analysts following the firm i in data is from CSM ar year t Coverage t he natural logarithm of the number of data is from CSM ar analysts following the firm i in year t International analyst one when the earnings forecast is from data is from iBeS foreign analysts, zero otherwise Export sales Export Revenue (Mil. RMB) operating revenue involved outside the data is hand-collected from CSM ar mainland regions by hand based on the segmental reports (regional) disclosed by the listed firms Export/Sales t he proportion of export revenue in total data is hand-collected from CSM ar operation revenue Export/TA t he proportion of export revenue in total data is hand-collected from CSM ar assets Export/Employee (Mil RMB) t he ratio of export revenue to employees data is hand-collected from CSM ar Firm characteristics Market Value (Mil. RMB) t he firm’s total market capitalisation at the data is from CSM ar end of year LnSize t he natural logarithm of the firm’s total data is from CSM ar market capitalisation Sigma t he standard deviation of the current year’s t he construction of the variable is daily stock return according to hong and Kacperczyk (2010) Surprise one if the ePS in the current year exceeds that t he construction of the variable is in the previous year, zero otherwise according to Chen and Martin (2011) VolROA t he volatility of roa in the past three years data is from CSM ar Export determinants LnSales t he natural logarithm of sales revenue in year t he construction of the variable is −1 t − 1 according to liu and Zhang (2009) Workforce t he proportion of production workers in the t he construction of the variable is −1 total number of employees in year t − 1 according to liu and Zhang (2009) Edu t he proportion of employees possessing a t he construction of the variable is −1 university academic degree, or higher according to liu and Zhang (2009) qualification, in the total number of employees in year t − 1 PPE t he proportion of fixed assets and other t he construction of the variable is −1 long-term assets in the total assets in year according to liu and Zhang (2009) t − 1 Capd t he amount of fixed assets per capita in year t he construction of the variable is −1 t − 1 according to liu and Zhang (2009) Salesg t he growth rate of sales revenue in the last t he construction of the variable is −1 three years in year t − 1 according to liu and Zhang (2009) East one if a firm locates in the eastern regions, t he construction of the variable is −1 zero otherwise according to liu and Zhang (2009) CHINA JOURNAL OF ACCOUNTING STUDIES 129 3. Research design 3.1. Data sources We adopt the domestic A-share listed firms from 2002 to 2012 as samples and exclude the financial and insurance industries. The financial data applied in the study came from the CSMAR research database. We manually collected and arranged the basic data about export business in listed firms from 2002 to 2012 (the export regions included Hong Kong, Macau and Taiwan), including the operating revenue, operating cost and operating profits data generated from export business, export regions, and regions with export business ranking top five . The basic data about forecast earnings per share from 2002 to 2012 were acquired from the CSMAR data and data concerning the firm’s actual earnings per share came from the annual report at each year. The financial data along with data about the monthly stock return were taken from the database of CSMAR listed firms. Eventually, 7897 sample obser - vations were formed and used to test the research hypothesis 1 (see Appendix A-1 and A-2). Data in respect of the foreign analysts tracking domestic listed firms were acquired from IBES. For each sample firm, we calculate the local and foreign earnings forecast deviation respectively. Correspondingly, we construct the local and foreign analyst followings for each sample firm. Finally, we obtain a total of 11737 observations including local and foreign analyst forecasts which is applied to test the research hypothesis 2. All the continuous var- iables were winsorized by 1%. 3.2. The firm’s export business variables As to the measurement of export business, we collected the data of operating revenue, operating cost and operating profit involved outside the mainland regions by hand based on the segmental reports (regional) disclosed by the listed firms. Since the annual report did not disclose data about the overseas assets, the export revenue was consequently adopted as the basic index. On this basis, the proportion of export revenue in total operation revenue was used to measure the firm’s export business, measured as the ratio Export/Sales; in addition, Export was defined as a 0/1 dummy variable. When a firm exported to other countries and to the regions of Hong Kong, Macau and Taiwan, the value of Export was 1, otherwise 0. In this way, we obtained two variables to measure a firm’s export business: Export and Export/Sales. The definitions of the main variables are listed in Table 1. Figure 1 presents the time trend of the firm’s export variables from 2002 to 2012, based on data from our sample firms. Two features can be clearly observed: firstly, Export/Sales in Panel B demonstrates a trend of steady-state growth before 2008, and the average proportion of export revenue in that year’s total revenue was between 20 and 25%. Meanwhile, we calculated Export/TA (Total Assets) and Export/EMP (Employee) and adjusted the export revenue of the pre-IPO firms by total assets and number of employees. The growth trend was also evidenced in Export/ TA and Export/EMP. To verify the reliability of the data, we calculated the proportion of the exports in that year’s GDP (Export/GDP) based on the export data from the National Statistics Yearbook and listed the result in Panel A. The trend on the whole was basically consistent with the export trend in the study. Secondly, we can observe that the global financial crisis breaking during 2008 to 2009 had huge impact on China’s export business. Figure 1 suggests that the substantial influence exerted by the 2008 financial crisis on China emerged in 2009. 130 F. XIN ET AL. Figure 1. China’s export activity during the sample period of 2002–2012 (calculated by the authors from the sample data). We can see from the index of export accounting for GDP (Export/GDP) that this index had been presenting a growth trend since 2002, began to decline in 2007, then demonstrated a trend of sharp slope in 2009, decreasing rapidly from 31.97% in 2008 down to 24.06% in 2009, which indicated that the financial crisis in 2009 inflicted substantial influence on China’s export business. Similarly, the other three indexes are export indexes acquired through calculation based on the sample companies, namely, Export/Sales, Export/TA and Export/Employee, which are listed by median and mean value. The general trend of the influ - ence imposed by financial crisis on export business is basically consistent. In particular, changes in Export/TA are most obvious. Differences among indexes are mainly that the influ - ence of the financial crisis on the firm’s total assets is less than that on the revenue of com- panies relying on the export business. The financial crisis led to substantial shrinkage in the firm’s sales revenue, and factors relating to suspending and reducing production caused a dramatic decrease in the number of employees, which therefore dampened the influence of Export/Sales and Export/Employee. In general, the financial crisis led to a substantial reduc - tion in the export business in the sample companies during 2008 to 2009. 3.3. Other variables 3.3.1. The analysts’ forecast errors According to Hong and Kacperczyk (2010), the analysts’ forecast error is defined as follows: Forecast Error − Actual EPS t t FERROR = it (1) t−1 CHINA JOURNAL OF ACCOUNTING STUDIES 131 Namely, the firm i ’s forecast error in the year t amounts to the absolute value of all the current forecast mean value (median) of analysts tracking the firm i and the current EPS divided by the closing share price of the firm i in the year t − 1. The analysts’ forecast takes the most recent one among the EPS forecast made before 180 days. Based on calculating the difference between mean value and median adopted in the analysts’ forecast, we can get the Mean FERROR and Median FERROR. 3.3.2. Forecast error made by foreign analysts Acquiring the earnings forecast data of foreign analysts from IBES database, we can calculate the earnings forecast of foreign analysts for China’s listed firms in a similar way according to formula (1) calculating the analysts’ forecast error. The foreign analyst is defined as IntAnalyst, and when the earnings forecast is from foreign analysts, IntAnalyst is represented by a dummy variable 1, otherwise 0. 3.3.3. Control variables Based on previous studies, control variables include the number of analysts following (Coverage) using the natural logarithm; the volatility of the current year’s stock return is represented by the standard deviation of the current year’s daily stock return; according to the information content (Surprise) added into the current year’s EPS by Chen and Martin (2011). We represent it with a dummy variable, namely, one if the EPS in the current year exceeds that in the previous year, zero otherwise; the size of firm (LnSize) can be represented by the natural logarithm of the firm’s total market capitalization; the volatility of performance (VolROA) is represented by the volatility of ROA in the past three years. We also impose a control on the industry and time effects. 3.4. Model specifications In order to test the research hypothesis 1, we regard the analysts’ FERROR as the dependent variable and export variable as the independent variable, and simultaneously control for the variables that prior research has shown might exert influence on the analysts’ forecast errors (Hong & Kacperczyk, 2010), which can be tested by model (2). FERROR = + Export + Coverage + Sigma + Surprise i,t 0 1 i,t 2 i,t 3 i,t 4 i,t (2) + LnSize + VolROA + IND + Year + 5 i,t 6 i,t 7i i 8j j i,t In model (2), I stands for the firm and t for year. Due to the influence of the firm’s individual factors, clustering might appear in the error term. To guarantee the robustness of results, the standard deviation should be clustered by firm. Based on the differences of computing methods, FERROR corresponds to mean FERROR and median FERROR respectively. Export in the model is measured by two methods. The first one is a dummy 0/1 variable, that is, one is adopted if the firm has an export business, zero otherwise; the second one is to obtain continuous variables Export/Sales according to the proportion of export sales revenue in the firm’s total sales revenue. Based on research hypothesis 1 in the study, the coefficient β of Export is expected to be positive. In order to control for the issues of self-selection that might exist in the export firms themselves, we adopt 2SLS to construct the Probit model controlling for endogenous 132 F. XIN ET AL. problems based on prior research on the inu fl ence of the determinants concerning the local firms’ export business. The model introduces the natural logarithm of sales revenue (LnSales), workforce level (that refers to the proportion of production workers in the total number of employees, Workforce), education level (that refers to the proportion of employees possess- ing the academic degree of university or above in the total number of employees, Edu), scale of fixed assets (that refers to the proportion of fixed assets and other long-term assets in the total assets, PPE), capital intensity (that refers to the amount of fixed assets per capita, Capd ), the growth rate of sales revenue in the last three years (Salesg), and whether it locates in the eastern regions or not (East). All the variables in the Probit model belong to the period t − 1 and are used to estimate the probability of the firm’s export in the period of t. Meanwhile, Propensity Score Matching is adopted and on the basis of Probit model involving export determinants, we apply a neighbouring matching principle to get the matched sample firms. To examine the hypothesis 2, the variable of foreign analysts and its interaction terms with the export business are added on the basis of the basic model (2) to obtain the coeffi- cient direction and significance of the cross terms. FERROR = + Export + Export IntAnayst + IntAnayst + Coverage i,t 0 1 i,t 2 i,t i,t 3 i,t 4 i,t + Sigma + Surprise + LnSize + VolROA + (3) 5 i,t 6 i,t 7 i,t 8 i,t i,t According to Malloy (2005), it was found that compared with domestic analysts, the foreign analysts held more informational advantages towards firms with export business. Consequently, the coefficient β of cross term is expected to be negative in the model (3). 4. Empirical results and analysis 4.1. Descriptive statistics Table 2 compares the differences between the main variables in companies with export business and companies without export business. Above all, the forecast errors of export business companies (Mean FERROR and Median FERROR) are all significantly greater than those of companies without export business from the perspective of analysts’ characteristics. Analyst following of companies in the export business also significantly surpasses that in companies without an export business. At the same time, the foreign analyst following companies with export business significantly exceeds that of companies without export business. The average export revenue reaches almost RMB 1.4 Billion Yuan, accounting for 23% of the total revenue (Export/Sales) (the median is 14.7%), accounting for 17.9% of the total assets (the median is 9.7%), and the export revenue per capita achieves RMB 356,000 Yuan (the median is RMB 106,000 Yuan). In terms of the firm’s characteristics, market capi- talisation of companies with and without export business present no significant differences while the former mean value (LnSize) is significantly larger than that of the latter, and the stock volatility also higher than that of the latter, which suggests that the more complicated the market environment the firms face, the higher the relative risk they will be confronted with. But the unexpected changes of earnings in export companies are significantly lower than those in non-export companies. CHINA JOURNAL OF ACCOUNTING STUDIES 133 Table 2. d escriptive statistics of variables. Non-export firm Export firm Variables N Mean Median N Mean Median Analysts characteristics *** *** Mean ERROR 4,265 0.03 0.02 3,632 0.04 0.02 *** *** Median ERROR 4,265 0.03 0.02 3,632 0.04 0.02 *** *** Number of analysts 4,265 10.58 5.00 3,632 12.14 7.00 *** *** Coverage 4,265 1.67 1.61 3,632 1.87 1.95 ** ** International analyst 6,239 0.32 0.00 5,498 0.34 0.00 Export sales *** *** Export Revenue(Mil. RMB) 4,265 0.00 0.00 3,632 1,397.73 329.95 *** *** Export/Sales 4,265 0.00 0.00 3,632 0.23 0.15 *** *** Export/TA 4,265 0.00 0.00 3,632 0.18 0.10 *** *** Export/Employee (Mil RMB) 4,256 0.00 0.00 3,630 0.36 0.11 Firm characteristics Market Value(Mil. RMB) 4,265 12,499.51 4,784.56 3,632 13,585.49 4780.64 *** LnSize 4,265 22.39 22.29 3,632 22.46 22.29 * *** Sigma 4,265 0.03 0.03 3,632 0.03 0.03 *** *** Surprise 4,265 0.53 1.00 3,632 0.49 0.00 ** VolROA 4,265 0.03 0.02 3,632 0.03 0.02 Exprot determinants *** *** LnSales 3,314 21.28 21.22 2,966 21.62 21.41 −1 *** Workforce 3,314 0.66 0.59 2,966 0.63 0.66 −1 *** *** Edu 3,314 0.21 0.15 2,966 0.18 0.14 −1 *** PPE 3,314 0.34 0.30 2,966 0.34 0.31 −1 *** *** Capd 3,314 1.01 0.32 2,966 0.60 0.29 −1 Salesg 3,314 0.13 0.11 2,966 0.12 0.11 −1 *** *** East 3,314 0.57 1.00 2,966 0.67 1.00 −1 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 4.2. Hypothesis 1: export and analysts’ forecast errors 4.2.1. Regression result Columns (1) to (8) in Table 3 demonstrate the OLS regression result based on model (2). The dependent variable in columns (1) to (4) is mean FERROR, and the dependent variable in columns (5) to (8) is median FERROR. Among all the regression results, the coefficients of Export and Export/Sales are significantly positive, supporting the research hypothesis 1, that is, the earnings forecast errors made by analysts in respect of companies with export business are significantly more than those in respect of companies without export business. Among other control variables, the coefficient of Coverage is significantly negative, which is consist - ent with the results of Hong and Kacperczyk (2010). In addition, when comparing the coef- ficients of Export and Export/Sales, we can see that the coefficient of continuous variables is far larger than for the 0/1 variable (such as 0.0037 < 0.0089, 0.0019 < 0.0055). Meanwhile, when the year and industry are controlled, the value of t also gradually declines, suggesting that the factors of industry and time have relatively great influence. 4.2.2. Discussion of the endogeneity problems Even if other factors including industry and year are controlled in model (2), important variables still might be omitted. Thus, we further adopt a treatment effect model and pro - pensity score matching procedure with attempt to reduce the influence of the endogeneity of export factors on research issues. 134 F. XIN ET AL. Table 3. regressions on export and analysts’ forecast errors. Mean ERROR Median ERROR Dep. Var. (1) (2) (3) (4) (5) (6) (7) (8) *** * *** * Export 0.004 0.002 0.004 0.002 (3.66) (1.82) (3.80) (1.66) *** ** *** * Export/Sales 0.009 0.006 0.009 0.005 (3.38) (2.02) (3.33) (1.90) ** *** *** *** Coverage −0.000 −0.001 −0.000 −0.001 −0.001 −0.002 −0.001 −0.002 (−0.10) (−2.56) (−0.07) (−2.64) (−1.33) (−3.69) (−1.31) (−3.77) *** *** *** *** *** *** *** *** Sigma −0.112 0.334 −0.114 0.332 −0.119 0.312 −0.120 0.310 (−2.62) (2.85) (−2.67) (2.84) (−2.82) (2.79) (−2.87) (2.77) *** *** *** *** *** *** *** *** Surprise −0.023 −0.022 −0.023 −0.022 −0.023 −0.022 −0.023 −0.022 (−27.15) (−27.16) (−27.16) (−27.18) (−28.05) (−27.83) (−28.04) (−27.83) *** *** *** *** *** *** *** *** LnSize 0.002 0.003 0.003 0.003 0.002 0.002 0.002 0.003 (4.34) (5.12) (4.61) (5.39) (3.24) (4.40) (3.54) (4.65) *** *** *** *** *** *** *** *** VolROA 0.276 0.273 0.275 0.272 0.276 0.274 0.275 0.274 (12.06) (12.05) (11.99) (12.01) (12.26) (12.35) (12.19) (12.31) *** *** *** *** Constant −0.013 −0.066 −0.016 −0.069 0.002 −0.052 −0.001 −0.055 (−1.12) (−4.89) (−1.36) (−5.13) (0.18) (−4.09) (−0.07) (−4.32) Year fe no YeS no YeS no YeS no YeS industry fe no YeS no YeS no YeS no YeS Cluster within firms YeS YeS YeS YeS YeS YeS YeS YeS observations 7,897 7,897 7,897 7,897 7,897 7,897 7,897 7,897 a dj. R-squared 0.14 0.28 0.14 0.28 0.14 0.28 0.14 0.28 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. CHINA JOURNAL OF ACCOUNTING STUDIES 135 (1) Treatment effect model Since the dependent variables of analysts’ forecast errors can both be observed in the sample companies with or without export, a treatment effect model is adopted to deal with such kind of endogenous problems (Maddala, 1983). The first step, a binominal probit model is teased out based on the determinants of China’s export firms studied by Liu and Zhang (2009). Export = + LnSales + Workforce + Edu + PPE i,t 0 1 i,t−1 2 i,t−1 3 i,t−1 4 i,t−1 (4) + Capd + Salesg + East + 5 i,t−1 6 i,t−1 6 i,t−1 i,t The dependent variable in model (4) is 0/1 variable, and the independent variable is defined as previously. They both adopt the period of t − 1 to estimate the probability of export or not of the firm i in the period of t and simultaneously get the value of hazard of each obser - vation. In the second step, we substitute the value of hazard estimated by model (4) into model (1) to carry out regression and examine that whether the coefficient of Export confirms to the expectation. The results are displayed in Table 4. In the regression model of Probit at the first stage, the results suggests that larger firms with the more intensity of fixed assets, located in the Table 4. regressions on export and analysts’ forecast errors: t reatment effect. (1) (2) (3) Dep. Var. First Stage Mean ERROR Median ERROR *** *** Export 0.024 0.023 (5.36) (5.12) *** *** Coverage −0.003 −0.003 (−4.85) (−6.17) *** *** Sigma 0.348 0.315 (4.30) (3.98) *** *** Surprise −0.023 −0.023 (−24.49) (−24.58) *** *** LnSize 0.003 0.002 (4.05) (3.57) *** *** VolROA 0.333 0.337 (20.42) (21.12) *** Hazard 0.158 (11.69) LnSales −0.010 −1 (−1.04) *** Workforce −0.652 −1 (−5.29) ** Edu 0.208 −1 (2.21) *** PPE −0.112 −1 (−9.14) *** Capd −0.236 −1 (−3.41) *** Salesg 0.310 −1 (8.72) *** East 0.158 −1 (11.69) *** *** *** Constant −3.489 −0.051 −0.041 (−12.02) (−3.61) (−2.99) Year fixed effects no YeS YeS industry fixed effects no YeS YeS observations 5,706 5,706 5,706 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 136 F. XIN ET AL. Table 5. regressions on export and analysts’ forecast errors: PSM. (1) (2) (3) (4) Dep. Var. Mean ERROR Median ERROR Mean ERROR Median ERROR *** *** Export 0.004 0.004 (3.20) (3.27) *** *** Export/Sales 0.008 0.008 (2.80) (2.76) Coverage −0.000 −0.001 −0.000 −0.001 (−0.04) (−1.05) (−0.11) (−1.12) *** *** *** *** Sigma −0.127 −0.139 −0.130 −0.142 (−2.62) (−2.82) (−2.70) (−2.90) *** *** *** *** Surprise −0.024 −0.023 −0.024 −0.023 (−25.40) (−25.88) (−25.40) (−25.85) *** ** *** *** LnSize 0.002 0.001 0.002 0.002 (3.25) (2.26) (3.56) (2.60) *** *** *** *** VolROA 0.309 0.306 0.308 0.305 (11.11) (11.25) (11.04) (11.18) Constant −0.003 0.011 −0.007 0.008 (−0.25) (0.91) (−0.51) (0.63) Year fe no no no no industry fe no no no no Cluster within firms YeS YeS YeS YeS a dj. R-squared 0.15 0.15 0.15 0.15 observations 6,346 6,346 6,346 6,346 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. eastern regions, are more likely to export because the coefficients of LnSize, PPE, East are positively significant. However, Edu and Capd are negatively correlated to Export, which indicates that the export business of China’s listed firms is mainly characterised by low val- ue-added and labour-intensive firms. These characteristics basically conform to the fact of export business of China’s firms summarised by Liu and Zhang (2009) in their papers. In the regression model of OLS at the second stage, we control for the fixed effect of industry and year and substitute the value of hazard estimated at the first stage into OLS model. The results in columns (2)–(3) in Table 3 indicate that all the coefficients of Export are significantly positive (β = 0.0246, t = 5.39), which consequently supports the research hypothesis 1. (2) Propensity score matching (PSM) We further adopt PSM to deal with the heteroscedasticity of sample selection (Rosenbaum & Rubin, 1983). Specifically, as to each export sample firm, we identify a pairing sample (control group) according to the principle of nearest-neighbourhood from the corresponding sample firm without an export business. The pairing is conducted based on the variables in the Probit model and the variables include LnSales, Workforce, Edu, PPE, Capd, Salesg and East. Then, we put the companies sample with export business and those without export business together and again carry out regression based on model (1), obtaining columns (1)–(4) in Table 5 in which both the coefficients of Export and Export/Sales are significantly positive (coefficient of Export is 0.0032, t = 2.92, coefficient of Export/Sales is 0.0073, t = 2.65) and therefore support the research hypothesis 1. (3) Event study In Figure 2, we identify the companies that develop an export business from a non-export business and take the export business in that beginning year as the event to count the change trend of analysts’ foreign errors. We are able to clearly observe that the mean value CHINA JOURNAL OF ACCOUNTING STUDIES 137 Figure 2. analysts’ forecast errors before and after the firm’s entry to the export market. Table 6. analysts’ forecast errors before and after the firm’s export. (1) PRE (2) (3) POST Diff. [−3, −1] T = 0 [1, 3] (2)−(1) (3)−(2) Panel A: Mean Error *** Mean 0.028 0.038 0.039 0.012 −0.004 *** Median 0.016 0.028 0.025 0.010 −0.003 Panel B: Median Error *** Mean 0.028 0.035 0.038 0.011 −0.004 *** Median 0.014 0.025 0.024 0.008 −0.003 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. of mean FERROR gradually increases to 0.0378 in the year of export from 0.0161 in the pre- vious 5 years before export, and subsequently decreases to 0.0307 in 4 years after export but with a leap in the fifth year after export. The trend of median FERROR is similar. Furthermore, in Panel A and B of Table 6, after conducting statistical tests of analysts’ forecast errors before and after export and in the year of export, we can observe that forecast errors in the year of export are significantly larger than those before export (the difference of mean FERROR is 0.0121 while that of mean FERROR is 0.0108, both of which are significant at the 1% level). However, the forecast errors after export do not demonstrate any difference (the difference of mean FERROR is −0.0039 while that of median FERROR is −0.004, both of which are insignificant), and test result of median is consistent with that of mean value. The aforesaid event study did not take other features of the firm into account as controls, but the prominent influence of export on analysts’ forecast errors can be observed from this trend. This method has also fairly controlled the endogenity problem existing in the research issue itself. 4.3. Hypothesis 2: do foreign analysts possess informational advantage? 4.3.1. The basic regression results Whether foreign analysts possess informational advantage compared with domestic analysts is an important topic (Malloy, 2005; Wang, Chen, & Hou, 2010). This part carries out relatively in-depth discussion and analysis of the topic. We obtain the regression result by adding foreign analysts based on model (3), which can be seen in Table 7. In accordance with expec- tation, the regression results of columns (1) to (4) show that all coefficients of cross terms 138 F. XIN ET AL. Table 7. regression results on foreign analysts, export and analysts’ forecast errors. (1) (2) (3) (4) Dep. Var. Mean ERROR Mean ERROR Median ERROR Median ERROR ** ** Export 0.003 0.002 (2.49) (2.35) *** *** Export*IntAna −0.004 −0.004 (−3.35) (−3.42) ** ** Export/Sales 0.007 0.007 (2.52) (2.46) *** *** Export/Sales*IntAna −0.008 −0.009 (−3.04) (−3.17) *** *** *** *** IntAna −0.021 −0.022 −0.020 −0.020 (−21.22) (−24.53) (−20.34) (−23.52) *** *** *** *** Coverage −0.002 −0.002 −0.002 −0.002 (−3.27) (−3.25) (−4.51) (−4.47) *** *** *** *** Sigma 0.275 0.274 0.258 0.256 (3.10) (3.09) (3.00) (2.99) *** *** *** *** Surprise −0.018 −0.017 −0.017 −0.017 (−26.25) (−26.14) (−26.54) (−26.41) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (5.98) (6.10) (5.64) (5.73) *** *** *** *** VolROA 0.245 0.244 0.247 0.246 (13.13) (13.08) (13.36) (13.32) *** *** *** *** Constant −0.045 −0.046 −0.038 −0.039 (−3.81) (−3.86) (−3.45) (−3.49) Year fixed effect YeS YeS YeS YeS industry fixed effect YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 11,737 11,737 11,737 11,737 a djusted R-squared 0.28 0.28 0.27 0.27 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. are significantly negative at the 1% level, which suggests that foreign analysts possess rel- atively clear informational advantage in earnings forecast towards domestic listed firms with an export business. 4.3.2. The self-selection of foreign analysts Based on the discoveries of Lin, Ouyang, and Yue (2007), foreign analysts tend to analyse firms that possess a perfect governance structure, non-state ownership and desirable net profit. Therefore, would the self-selection existing in foreign analysts themselves influence the conclusion of the hypothesis 2? The section discusses this question further. In hypothesis 2, the theme of the study is whether there are significant differences between domestic and foreign analysts when making forecasts in respect of firms with export business. To reduce the effect of influence from self-selection by foreign analysts, we adopt the following methods. We choose the firms tracked simultaneously by local and foreign analysts as a subsample to explore the influence of export business on analysts’ forecast errors. In this subsample, since the local analysts choose to track the same firm as the foreign analysts do, we believe that factors influencing the corporate’s governance, own - ership structure and operation performance of the firm tracked by foreign analysts can be fairly controlled and are therefore capable of relieving the self-selection problems of foreign t hanks to the anonymous reviewers for their constructive opinions here. CHINA JOURNAL OF ACCOUNTING STUDIES 139 Table 8. Comparison between pairing sample firms. Firms followed by local analysts Firms followed by foreign analysts Diff. (0−1) Variables N Mean Median N Mean Median Mean Median *** *** Mean ERROR 3,648 0.032 0.020 3,648 0.013 0.007 0.019 0.013 *** *** Median ERROR 3,640 0.030 0.018 3,648 0.013 0.007 0.017 0.011 SOE 3,587 0.636 1.000 3,587 0.635 1.000 0.001 0.000 Shareholder 3,622 39.705 39.335 3,622 39.702 39.300 0.003 0.035 Indirector 3,628 0.364 0.333 3,628 0.364 0.333 0.000 0.000 ChairmanCEO 3,648 0.154 0.000 3,648 0.154 0.000 0.000 0.000 ROA 3,648 0.083 0.072 3,648 0.083 0.072 0.000 0.000 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. analysts. First, in Table 8, we compare the earnings forecast errors between local and foreign analysts in this pairing subsample and check whether there are significant differences in corporate governance, ownership structure and operation performance of the firm that the foreign and local analysts choose to track. Because we choose the firm sample tracked simul - taneously by local and foreign analysts, the sample firms have the same characteristics across the two analyst groups. However, when comparing the forecast errors of local analysts with those of foreign analysts, we can observe that forecast errors made by foreign analysts are prominently significantly lower than those made by local analysts. Based on the pairing samples established, according to the findings of Lin et al. (2007), we add the control variables including whether state-owned, share proportion of the largest shareholder, proportion of independent directors, dual role of the board chairman and return on assets to carry out regression on the basis of model (3) in the study, and the regression results are shown in Table 9. This result further verifies hypothesis, suggesting that the results of hypothesis 2 are robust. 4.3.3. Sensitivity test on export revenue We further study the sensitivity of domestic and foreign analysts’ forecast errors towards the export revenue in listed companies. The approach is as follows. In Table 10, we divide non-vanishing samples of Export/Sales into five groups from high to low and explore respec - tively the change trend of mean ERROR and median ERROR of forecast errors made by domes- tic and foreign analysts. Taking mean ERROR as an example, we can reduce it to discoveries in two aspects: (1) with the increase in the proportion of export revenue, forecast errors of foreign analysts are increasing as the mean value (median) changes from 0.0296 (0.0389) to 0.0627 (0.0507), with change surpassing 200% and significantly different at the 1% level, while the difference in mean value of foreign analysts’ forecast errors is only 0.0098 although it is also significant at the 1% level but its numerical value is far less than that of domestic analysts (0.0331); (2) with the increase in the proportion of export revenue, the difference between domestic and foreign analysts in forecast errors is increasingly large as its mean value expands from −0.0168 to −0.0401, with change surpassing 200% and there are signif- icant differences at the 1% level. We can infer from the above analysis that the proportion of export revenue has an influence on the value of forecast errors of domestic and foreign analysts. On the whole, export factors tend to exert a larger influence on earnings forecast errors of domestic analysts than those of foreign analysts. 140 F. XIN ET AL. Table 9. regression results of pairing samples. (1) (2) (3) (4) Dep. Var. Mean ERROR Mean ERROR Median ERROR Median ERROR * * Export 0.002 0.002 (1.87) (1.70) *** *** Export*IntAna −0.005 −0.005 (−4.40) (−4.58) ** ** Export/Sales 0.008 0.007 (2.35) (2.26) *** *** Export/Sales*IntAna −0.011 −0.011 (−3.61) (−3.67) *** *** *** *** IntAna −0.016 −0.017 −0.014 −0.015 (−13.59) (−14.75) (−12.49) (−13.46) Coverage 0.001 0.001 0.000 0.000 (0.91) (1.01) (0.11) (0.21) ** ** * * Sigma 0.237 0.233 0.157 0.154 (2.55) (2.53) (1.69) (1.66) *** *** *** *** Surprise −0.010 −0.010 −0.011 −0.011 (−15.68) (−15.70) (−16.72) (−16.74) *** *** *** *** LnSize 0.003 0.003 0.002 0.002 (5.43) (5.38) (5.14) (5.05) *** *** *** *** VolROA 0.232 0.232 0.230 0.230 (10.07) (10.05) (10.35) (10.31) SOE −0.001 −0.001 −0.000 0.000 (−0.82) (−0.70) (−0.09) (0.02) Shareholder −0.000 −0.000 −0.000 −0.000 (−0.25) (−0.20) (−0.07) (−0.03) * * ** ** Indirector −0.012 −0.012 −0.014 −0.014 (−1.67) (−1.67) (−2.09) (−2.09) ** ** * * ChairmanCEO −0.002 −0.002 −0.002 −0.002 (−1.97) (−2.03) (−1.89) (−1.95) *** *** *** *** ROA −0.156 −0.155 −0.146 −0.146 (−15.43) (−15.40) (−14.85) (−14.81) * * Constant −0.020 −0.019 −0.013 −0.012 (−1.78) (−1.73) (−1.19) (−1.11) Year fixed effect YeS YeS YeS YeS industry fixed effect YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 7,128 7,128 7,120 7,120 a djusted R-squared 0.378 0.378 0.365 0.364 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 5. Further analysis 5.1. External shock: the financial crisis during 2008 to 2009 The financial crisis has provided an opportunity for theoretical and practical cycles to study change of firms’ micro-decision-making behaviour when suffering an external impact. Prior research explores the influence of financial crisis on firms’ behaviour from the perspectives of investment and financing behaviour of firms, bank loan decision, corporate governance and others (Berger & Bouwman, 2013; Campello, Graham, & Harvey, 2010; Erkens, Hng, & Matos, 2012; Kahle & Stulz, 2013). Nevertheless, when searching the existing literature, we find that few involved research into the influence of the decision-making behaviour of ana- lysts in financial crisis. Only Ang and Ma (2001) studied the analysts’ behaviour in the capital market of Indonesia, South Korea, Malaysia and Thailand during the financial crisis in 1997. Our study selects the financial crisis during 2008 to 2009 in relation to the research back - ground of the Chinese capital market, Chinese economic volume and the characteristic of an export-oriented economy suffering more evident influence from financial crisis. CHINA JOURNAL OF ACCOUNTING STUDIES 141 Table 10. Sensitivity test on export revenue: local vs. foreign analysts. Export/Sales INT Lowest 2 3 4 Highest Diff. (H−L) Panel A: Mean ERROR *** l ocal analysts (0) Mean 0.030 0.034 0.032 0.027 0.063 0.033 *** Median 0.039 0.041 0.040 0.030 0.051 0.012 N 836 702 732 676 696 *** f oreign analysts (1) Mean 0.013 0.013 0.014 0.015 0.023 0.010 *** Median 0.021 0.016 0.021 0.030 0.035 0.015 N 264 397 368 423 404 *** *** *** *** *** Diff. (1−0) Mean −0.017 −0.020 −0.018 −0.012 −0.040 *** *** *** *** *** Median −0.018 −0.025 −0.019 0.001 −0.015 Panel B: Median ERROR *** l ocal analysts(0) Mean 0.030 0.032 0.030 0.025 0.060 0.032 *** Median 0.038 0.039 0.039 0.029 0.051 0.012 N 836 702 732 676 696 *** f oreign analysts(1) Mean 0.013 0.013 0.014 0.015 0.023 0.010 *** Median 0.021 0.016 0.021 0.031 0.035 0.015 N 264 397 368 423 404 *** *** *** *** *** Diff. (1−0) Mean −0.016 −0.019 −0.016 −0.010 −0.038 *** *** *** *** *** Median −0.018 −0.023 −0.018 0.002 −0.015 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. Table 11. impact of financial crisis on analysts’ forecast errors. Dep. Var. Mean ERROR Median ERROR Export 0.001 0.001 (1.06) (0.97) ** ** Export*Crisis 0.007 0.006 (2.19) (2.01) Export/Sales 0.003 0.003 (1.35) (1.32) ** * Export/Sales*Crisis 0.016 0.014 (2.02) (1.69) *** *** *** *** Crisis 0.037 0.038 0.034 0.036 (12.31) (13.28) (11.77) (12.74) *** *** *** *** Coverage −0.001 −0.001 −0.002 −0.002 (−2.59) (−2.68) (−3.71) (−3.80) *** *** *** *** Sigma 0.332 0.332 0.310 0.310 (2.84) (2.84) (2.77) (2.77) *** *** *** *** Surprise −0.022 −0.022 −0.022 −0.021 (−27.22) (−27.17) (−27.89) (−27.80) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (5.15) (5.44) (4.42) (4.69) *** *** *** *** VolROA 0.272 0.271 0.274 0.273 (12.04) (12.00) (12.33) (12.29) *** *** *** *** Constant −0.066 −0.069 −0.052 −0.056 (−4.90) (−5.17) (−4.10) (−4.35) Year fixed effect YeS YeS YeS YeS industry fixed effect YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 7,897 7,897 7,897 7,897 a djusted R-squared 0.28 0.28 0.28 0.28 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 142 F. XIN ET AL. Table 12. impact of financial crisis on foreign analysts’ forecast errors. Mean ERROR Median ERROR Dep. Var. Local analysts Foreign analysts Local analysts Foreign analysts Panel A: Export Export 0.001 0.000 0.001 0.000 (1.23) (0.29) (1.11) (0.26) *** ** Export*Crisis 0.007 0.002 0.006 0.002 (2.83) (0.86) (2.57) (0.76) *** *** Crisis 0.037 0.007 0.034 0.007 (10.03) (0.47) (9.56) (0.46) *** *** *** *** Coverage −0.001 −0.003 −0.002 −0.003 (−3.24) (−6.01) (−4.60) (−6.00) *** *** Sigma 0.332 0.059 0.310 0.057 (4.69) (0.73) (4.47) (0.71) *** *** *** *** Surprise −0.022 −0.009 −0.022 −0.008 (−27.38) (−11.38) (−27.57) (−11.23) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (6.69) (8.80) (5.58) (8.94) *** *** *** *** VolROA 0.272 0.179 0.274 0.179 (22.80) (14.14) (23.42) (14.18) *** *** *** *** Constant −0.077 −0.066 −0.063 −0.067 (−6.94) (−3.70) (−5.80) (−3.76) observations 7,907 3,830 7,907 3,830 2 * 2 * l ocal analysts = f oreign analysts Chi (1) = 2.02 Chi (1) = 1.71 t est variable: Export*Crisis p = 0.155 p = 0.190 Panel B: Export/Sales Export/Sales 0.003 0.001 0.003 0.001 (1.54) (0.29) (1.54) (0.29) *** ** Export/Sales*Crisis 0.016 0.005 0.014 0.004 (2.76) (0.96) (2.35) (0.81) *** *** Crisis 0.038 0.008 0.036 0.008 (10.65) (0.49) (10.15) (0.48) *** *** *** *** Coverage −0.001 −0.003 −0.002 −0.003 (−3.36) (−6.02) (−4.70) (−6.01) *** *** Sigma 0.3322 0.060 0.310 0.058 (4.70) (0.75) (4.47) (0.72) *** *** *** *** Surprise −0.022 −0.009 −0.021 −0.008 (−27.23) (−11.29) (−27.42) (−11.15) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (7.06) (8.97) (5.91) (9.10) *** *** *** *** VolROA 0.272 0.179 0.274 0.179 (22.72) (14.15) (23.35) (14.18) *** *** *** *** Constant −0.080 −0.067 −0.066 −0.068 (−7.29) (−3.76) (−6.12) (−3.81) observations 7,907 3,830 7,907 3,830 2 * 2 l ocal analysts = f oreign analysts Chi (1) = 1.66 Chi (1) = 1.19 t est variable: Export*Crisis p = 0.198 p = 0.276 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. The influence of the financial crisis on the analysts’ forecast mainly comes from the impact on firms’ exports, which subsequently affects analysts’ forecast errors. Panel A in Chart 1 shows that the global financial crisis during 2008 to 2009 caused a great impact on China’s exports. During 2008 to 2009, the proportion of China’s export volume in GDP declined rapidly from 31.97 to 24.06%, with a decrease of almost a quarter ((24.06% − 31.97%)/31.9 7% = −24.74%). So which influence would this impact impose on the relationships between the firm’s export business and analysts’ forecast is the question that this study seeks to further explore. CHINA JOURNAL OF ACCOUNTING STUDIES 143 Table 13A. impact of international convergence of accounting standard on foreign analysts’ forecast errors: Mean error. Mean ERROR Before new stand- After new standards Before new stand- After new standards Dep. Var. ards Year < 2007 Year >= 2007 ards Year < 2007 Year >= 2007 *** Export 0.000 0.003 (0.06) (2.67) * *** Export*IntAna −0.005 −0.004 (−1.68) (−3.23) *** Export/Sales 0.002 0.008 (0.46) (2.59) *** Export/Sales*IntAna −0.008 −0.009 (−1.04) (−2.98) *** *** *** *** IntAna −0.018 −0.021 −0.019 −0.022 (−8.20) (−20.38) (−9.65) (−23.16) *** *** *** *** Coverage −0.005 −0.001 −0.005 −0.001 (−4.06) (−2.68) (−4.01) (−2.68) * *** * *** Sigma 0.337 0.254 0.337 0.253 (1.88) (2.60) (1.87) (2.59) *** *** *** *** Surprise −0.018 −0.017 −0.018 −0.017 (−12.17) (−23.45) (−12.16) (−23.36) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (2.79) (5.83) (2.67) (5.98) *** *** *** *** VolROA 0.351 0.226 0.350 0.225 (4.64) (12.17) (4.62) (12.13) *** *** Constant −0.030 −0.049 −0.028 −0.051 (−1.28) (−4.01) (−1.18) (−4.14) Year fe YeS YeS YeS YeS industry fe YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 2,055 9,682 2,055 9,682 a dj. R-squared 0.24 0.29 0.24 0.29 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. Similar to the previous method, we add the variable of financial crisis when verifying the influence of the firm’s export business on analysts’ forecast errors to explore whether there are significant differences in the influence of the firm’s export business on analysts’ forecast errors under the financial crisis. Based on the impact of the financial crisis on China’s exports in chart 1, we select the year 2009 in which exports suffered the largest impact as the year of financial crisis (namely, Crisis=1) according to the changing trend of the proportion of exports in GDP. Table 11 lists our results according to the mean value and median of analysts’ forecast errors respectively. In the regression model adopting Export and Export/Sales to measure the export behaviour, the signs of the cross-terms in four groups of regression models are all significantly positive. The results indicate that the financial crisis intensifies the influence of export behaviour on analysts’ forecast errors and aggravates the operating environment involving companies with export business, enlarging the risk for this kind of company and accordingly increasing analysts’ forecast errors. We further add the samples of foreign analysts to conduct subsample verification and the results are showed in Table 12. It shows that domestic analysts are more easily subjected to the influence of financial crisis and accordingly cause forecast errors while foreign analysts are less vulnerable to suffering from the influence of financial crisis, which can be observed upon comparison in the cross-terms of Export*Crisis. Significant differences exist in the coef - ficients of the relevant cross-terms between domestic and foreign analysts (the value of 144 F. XIN ET AL. Table 13B. impact of international convergence of accounting standard on foreign analysts’ forecast errors: Median error. Median ERROR Before new stand- After new standards Before new stand- After new standards Dep. Var. ards Year < 2007 Year >= 2007 ards Year < 2007 Year >= 2007 ** Export 0.001 0.003 (0.23) (2.47) * *** Export*IntAna −0.005 −0.004 (−1.81) (−3.25) ** Export/Sales 0.003 0.007 (0.57) (2.52) *** Export/Sales*IntAna −0.010 −0.009 (−1.24) (−3.06) *** *** *** *** IntAna −0.018 −0.020 −0.019 −0.021 (−7.78) (−19.55) (−9.16) (−22.20) *** *** *** *** Coverage −0.005 −0.002 −0.005 −0.002 (−4.47) (−3.79) (−4.43) (−3.78) ** ** ** ** Sigma 0.361 0.227 0.360 0.227 (2.01) (2.46) (2.00) (2.46) *** *** *** *** Surprise −0.018 −0.017 −0.018 −0.017 (−12.01) (−23.77) (−12.02) (−23.66) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (2.72) (5.46) (2.61) (5.57) *** *** *** *** VolROA 0.351 0.228 0.349 0.228 (4.62) (12.45) (4.60) (12.42) *** *** Constant −0.028 −0.040 −0.026 −0.042 (−1.17) (−3.44) (−1.09) (−3.56) Year fe YeS YeS YeS YeS industry fe YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 2,055 9,682 2,055 9,682 a dj. R-squared 0.24 0.28 0.24 0.28 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. Chi-squared suggests that the single tail is statistically significant at the 10% level). This result indicates that foreign analysts possess more informational and capability advantages in dealing with the influence of financial crisis than domestic analysts do. 5.2. Effect of international convergence of accounting standards6Thanks to the anonymous reviewers for their constructive opinions here Wang, Chen, and Hou (2010) probed the differences in advantages between domestic and foreign analysts from the point of the differences of accounting standards. They found that after the implementation of a new accounting standard in China in 2007, the informational advantage of local analysts was reduced. Discoveries by Wang et al. (2010) provided a research direction that can be used for reference for this study to explain the cause of forecast advantages of foreign analysts. The sample term of the study is from 2002 to 2012, which spans the historic event of the integration of China’s accounting standards with the inter- national standards in 2007. Consequently, we take the year of 2007 as the time node of the actual convergence of accounting standards, and further verify the hypothesis 2 in the study. Specifically, we regard the year of 2007 as the watershed, divide the samples into two groups according to the time before and after the implementation of the international accounting standards respectively, and perform verification based on separate samples. The empirical CHINA JOURNAL OF ACCOUNTING STUDIES 145 results are listed as Tables 13A and 13B. Before the implementation of new standards, the interaction terms of Export*IntAna are significant at the 10% level while those of Export/ Sales*IntAna are not significant. After the implementation, all the interaction terms are sig- nificant at the 1% level, suggesting that the introduction of new standards is one of the reasons for the informational advantages of foreign analysts. 6. Conclusion This study finds that forecast earnings in respect of firms with export business, issued by domestic analysts, deviate more than the forecasts in respect of firms without export busi- ness. The results remain robust as verie fi d by treatment ee ff ct and propensity score matching after considering the firm’s export determinants. Meanwhile, we find that foreign analysts possess informational advantage in terms of earnings forecast in respect of companies with export business, showing that, forecast earnings issued by foreign analysts deviate less. The results remain robust after considering the self-selection of foreign analysts. Subsequently, the study further explores the mechanism leading to these differences. It is found that a financial crisis will enlarge the errors towards earnings forecast of export firms made by foreign analysts, but foreign analysts are less vulnerable to suffer from the influence of finan - cial crisis because of informational and capability advantages; after the implementation of new accounting standards, foreign analysts demonstrate more significant informational advantage. Evidence in this study indicates that the export business in Chinese firms is an important influential factor affecting the analysts’ forecasting. We find that study from the macroeconomic background of China as the large exporter provides favourable inspiration for interpreting the decision-making behaviour and results of Chinese analysts. Acknowledgement We thank the editor, two anonymous referees, Dr. Heng Yue and Dr. Cong Wang. This study has been supported by National Natural Science Foundation of China (71402198), Social Science Foundation of Beijing, China (15JGC176), the Beijing Municipal Commission of Education ‘Joint Construction Project’, the Beijing Municipal Commission of Education ‘Pilot Reform of Accounting Discipline Clustering’, Innovation Research of the Central University of Finance and Economics, the Social Science Foundation of Jiangsu Province, China (15GLB007), the Key Project of Philosophy and Social Science Research in Colleges and Universities in Jiangsu Province, China (2015ZDIXM011), and the Fundamental Research Funds for the Central University (2015B03414). References Ang, J.S., & Ma, Y. (2001). The behavior of financial analysts during the Asian financial crisis in Indonesia, Korea, Malaysia, and Thailand. Pacific-Basin Finance Journal, 9, 233–263. Bacmann, J.F., & Bolliger, G. (2001). Who are the best? Local versus foreign analysts on the Latin American stock markets. Unpublished working paper. Bae, K., Stulz, R.M., & Tan, H. (2008). Do local analysts know more? A cross-country study of the performance of local analysts and foreign analysts. Journal of Financial Economics, 88, 581–606. Balakrishnan, R., Harris, T., & Sen, P. (1990). The predictive ability of geographic segment disclosures. Journal of Accounting Research, 28, 305–325. Baldwin, B. (1984). Segment earnings disclosure and ability of security analysts to forecast earnings per share. The Accounting Review, 59, 376–389. 146 F. XIN ET AL. Berger, A.N., & Bouwman, C.H. (2013). How does capital affect bank performance during financial crisis? Journal of Financial Economics, 109, 146–176. Bolliger, G. (2004). The characteristics of individual analysts’ forecasts in Europe. Journal of Banking and Finance, 28, 2283–2309. Cai, W., & Zeng, C. (2010). The influence of corporate diversification on the attention of security analysts—Based on the empirical analysis from the perspective of the decision-making behavior of security analysts. Nankai Business Review, 4, 125–133 (In Chinese). Callen, J.L., Hope, O., & Segal, D. (2009). Domestic and foreign earnings, stock return variability, and the impact of investor sophistication. Journal of Accounting Research, 43, 377–412. Campello, M., Graham, J.R., & Harvey, C.R. (2010). The real ee ff cts of financial constraints: Evidence from a financial crisis. Journal of Financial Economics, 97, 470–487. Chen, D.H., Li, O.Z., & Xin, F. (2013). Five-year plan, China finance and their consequences ( Working Paper). Chen, T., & Martin, X. (2011). Do bank-affiliated analysts benefit from lending relationships? Journal of Accounting Research, 49, 633–675. Collins, D. (1976). Predicting earnings with subentity data: Some further evidence. Journal of Accounting Research, 14, 163–177. Conroy, R.M., Fukuda, Y., & Harris, R.S. (1997). Securities houses and earnings forecasts in Japan: What makes for an accurate prediction? Financial Analysts Journal, 53, 29–40. Degeorge, F., Ding, Y., Jeanjean, T., & Stolowy, H. (2013). Analyst coverage, earnings management and financial development: An international study. Journal of Accounting and Public Policy, 32, 1–25. Duru, A., & Reeb, D.M. (2002). International diversification and analysts’ forecast accuracy and bias. The Accounting Review, 77, 415–433. Erkens, D.H., Hung, M., & Matos, P. (2012). Corporate governance in the 2007–2008 financial crisis: Evidence from financial institutions worldwide. Journal of Corporate Finance, 18, 389–411. Goldberg, S., & Heflin, F. (1995). The association between the level of international diversification and risk. Journal of International Financial Management and Accounting, 6, 1–25. Henderson, B., & Marks, J.M. (2013). Predicting forecast errors through joint observation of earnings and revenue forecasts. Journal of Banking and Finance, 37, 4265–4277. Herrmann, D.R., Hope, O., & Thomas, W.B. (2008). International diversification and forecast optimism: The effects of regulation FD. Accounting Horizons, 22, 179–197. Hong, H., & Kacperczyk, M. (2010). Competition and Bias. Quarterly Journal of Economics, 1683–1725. Hu, Y., Lin, W., & Wang, W. (2003). The information sources, focus area and analysis tool of security analysts. Journal of Financial Research, (12), 52–63 (In Chinese). Huang, L., & Huang, Y. (2013). Tangible information, intangible information and analysts’ behavior. China Accounting Review, (3), 301–316 (In Chinese). Ji, K., & Tong, Z. (2012). Earnings components, continuous difference and earnings forecast of financial analysts—Based on the empirical analysis of China’s A-share listed firms. Journal of Shanxi University of Finance and Economics, (2), 106–114 (In Chinese). Jiang, G., & Rao, P. (2011). Macroeconomic policy and micro-firm’s behavior—Expanding new fields of accounting and financing researches. Accounting Research, (3), 9–18 (In Chinese). Kahle, K.M., & Stulz, R.M. (2013). Access to capital, investment, and the financial crisis. Journal of Financial Economics, 110, 280–299. Khurana, I.K., Pereira, R., & Raman, K.K. (2003). Does analyst behavior explain market mispricing of foreign earnings for U.S. multinational firms? Journal of Accounting, Auditing and Finance, 453–477. Kinney, W. (1971). Predicting earnings: Entity vs. sub-entity data. Journal of Accounting Research, 9, 127–142. Li, D., Li, X., & Zhang, B. (2010). Earnings forecast errors of analysts and local advantages. Finance & Economics, (3), 26–33 (In Chinese). Li, L., Wu, L., Xu, Y., & Yu, M. (2004). Cross-industrial transformation of the firm: Features analysis and risk control—Taking the typical case of cross-industrial transformation of China’s listed firms as the example. Management World, (1), 118–129 (In Chinese). Lin, X., Ouyang, J., & Yue, H. (2007). Who are the attentiongetting finns for foreign securities analysts? Journal of Financial Research (in Chinese), 2007, 84–98. CHINA JOURNAL OF ACCOUNTING STUDIES 147 Liu, Z., & Zhang, J. (2009). The empirical analysis of the export determinants of China’s local manufacturing firms. Economic Research Journal, (8), 99–112 (In Chinese). Luo, C., & Li, S. (2014). Whether export can be helpful to relieve the financing restraints?—Empirical research based on research data about China’s firms of the world bank. Journal of Financial Research, (9), 1–17 (In Chinese). Maddala, G.S. (1983). Limited-dependent and qualitative variables in econometrics. Cambridge: Cambridge University Press. Malloy, C. (2005). The geography of equity analysis. Journal of Finance, 60, 719–755. Nichols, D., Tunnel, L., & Seipel, C. (1995). Earnings forecast accuracy and geographic segment disclosures. Journal of International Accounting, Auditing and Taxation, 8, 113–126. Orpurt, S. (2004). Local analyst earnings forecast advantages in Europe. University of Chicago. Unpublished working study Reeb, D., Kwok, C., & Baek, Y. (1998). Systematic risk in the multinational corporation. Journal of International Business Studies, 29, 263–279. Roberts, C. (1989). Forecasting Earnings Using Geographic Segment Data: Some U.K. Evidence. Journal of International Financial Management and Accounting I, 130–151. Rosenbaum, P.R., & Rubin, D.B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41–55. Sloan, R.G. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71, 289–315. Su, D., & Zeng, H. (2009). Macroeconomic factors and changes in corporate capital structure. Economic Research Journal, (12), 52–65 (In Chinese). Wang, Y., Chen, X., & Hou, Y. (2010). The informational advantage of domestic security analysts: Geographical proximity or difference in accounting standard. Accounting Research, (12), 34–40 (In Chinese). Wei, Z., & Xue, J. (2015). Fair value accounting of financial assets and analyst forecasts. China Journal of Accounting Studies, 4, 294–319. Yu, Z., Tian, G., Qi, B., & Zhang, H. (2011). The corporate governance mechanism concerned by the media—Exploration based on the view of earnings management. Management World, (9), 127–140 (In Chinese). Yu, F. (2008). Analyst coverage and earnings management. Journal of Financial Economics, 88, 245–271. Yuan, Z., Zhang, L., & Yue, H. (2014). Does the cash flow forecast of analysts improve the accuracy of earnings forecast—Empirical evidence from China’s A-share market. Journal of Financial Research, (5), 162–177 (In Chinese). 148 F. XIN ET AL. Appendix A. The distribution of the sample (industry and year) Table A1-1 sets out the process of sample selection. Samples in the study come from all A-share listed firms from 2002 to 2012 totalling 11 years, and provides 19,306 observations after excluding small and medium firms board listed companies, We exclude companies issuing B-share simultaneously (consid - ering that the B-share might affect analysts’ forecast influenced by foreign investors) and companies that have been public for less than 2 years. We eventually have 7,897 firms’ annual observations as samples, due to missing data concerning the regional distribution of operating revenue disclosed by the firm, data forecast and tracked by the analysts and missing control variables. Tables A1-2 is the annual distribution of the samples. From the point of the proportion of analysts’ tracking, the sample as a whole conforms to the development trend of China’s analysts’ market. Table A1-1. t he process of sample selection. Firm/annual sample number The process of sample selection a-share non-financial listed firms/annual observation number: 2002–2012 23,357 Subtract: SMeS board listed companies 2,011 Companies issuing B-share simultaneously 1,187 Companies going public for less than 2 years 853 Subtotal 19,306 Subtract: observation of the missing of export data 2,999 observation of analysts’ data missing 10,788 observation of control variables data missing 621 f inal samples 7,897 CHINA JOURNAL OF ACCOUNTING STUDIES 149 Table A1-2. t he annual distribution of the samples. Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Samples in the study 119 202 284 433 570 621 829 978 1,154 1,274 1,433 number of a-share listed firms 1,224 1,287 1,377 1,381 1,434 1,550 1,625 1,718 2,063 2,342 2,494 Coverage of analysts’ forecast data 9.72% 15.70% 20.62% 31.35% 39.75% 40.06% 51.02% 56.93% 55.94% 54.40% 57.46% 150 F. XIN ET AL. Table A2. d escriptive statistics of export business. Panel A: Divided by the year Non-export Firm Export Firm Total Year N Percent N Percent N Percent 2002 85 1.99 34 0.94 119 1.51 2003 125 2.93 77 2.12 202 2.56 2004 161 3.77 123 3.39 284 3.60 2005 249 5.84 184 5.07 433 5.48 2006 327 7.67 243 6.69 570 7.22 2007 343 8.04 278 7.65 621 7.86 2008 451 10.57 378 10.41 829 10.50 2009 536 12.57 442 12.17 978 12.38 2010 611 14.33 543 14.95 1,154 14.61 2011 655 15.36 619 17.04 1,274 16.13 2012 722 16.93 711 19.58 1,433 18.15 t otal 4,265 100.00 3,632 100.00 7,897 100.00 Panel B: Divided by the industry Non-export Firm Export Firm Total Industry N Percent N Percent N Percent a 68 1.59 74 2.04 142 1.80 B 117 2.74 104 2.86 221 2.80 C 1,866 43.75 2,852 78.52 4,718 59.74 d 353 8.28 19 0.52 372 4.71 e 95 2.23 73 2.01 168 2.13 f 254 5.96 71 1.95 325 4.12 g 262 6.14 213 5.86 475 6.01 h 347 8.14 107 2.95 454 5.75 J 442 10.36 20 0.55 462 5.85 K 228 5.35 30 0.83 258 3.27 l 77 1.81 2 0.06 79 1.00 M 156 3.66 67 1.84 223 2.82 t otal 4,265 100.00 3,632 100.00 7,897 100.00 Panels A and B in Table A2 display the data distribution of the study. Panel A is the distribution by year. Samples in the study come from all A-share listed firms from 2002 to 2012 totalling 11 years. 7897 observations finally get reserved based on the data concerning the regional distribution of operation revenue disclosed by the firm and data forecast and tracked by the analysts. The industry of domestic analysts develops gradually after 2011 with a trend of year-by-year growth. Export firms accounts for 46% of the total sample changing from 28.5 to 49.6% during 2002 to 2012. Panel B is divided according to industry, among which a relatively evident trend is: the export ratio of agriculture (A) and Manufacture (C) are 52.1 and 60.4% respectively, surpassing 50%. Three industries with lowest export ratio are public utility (D), real estate (J) and (L) with the ratio of 5.1, 4.3 and 2.5% respectively. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png China Journal of Accounting Studies Taylor & Francis http://www.deepdyve.com/lp/taylor-francis/do-firms-exports-affect-analysts-forecast-errors-RBnRSby0zv

Loading next page...

References (50)

D. Collins (1976)
Predicting Earnings With Sub-Entity Data - Some Further Evidence
Journal of Accounting Research, 14
(2010)
The informational advantage of domestic security analysts: Geographical proximity or difference in accounting standard
G. Bolliger (2001)
The Characteristics of Individual Analysts' Forecasts in Europe
FEN: Behavioral Finance (Topic)
(2004)
Cross-industrial transformation of the firm: Features analysis and risk control-Taking the typical case of cross-industrial transformation of China's listed firms as the example
Zhe Wei, Jian Xue (2015)
Fair value accounting of financial assets and analyst forecasts
China Journal of Accounting Studies, 3
Marcin Kacperczyk, Harrison Hong (2008)
Competition and Bias
FA1: Fund Management
Kee‐Hong Bae, Hongping Tan, René Stulz (2005)
Do Local Analysts Know More? A Cross-Country Study of the Performance of Local Analysts and Foreign Analysts
Financial Accounting
François Degeorge, Yuan Ding, Thomas Jeanjean, Hervé Stolowy (2012)
Analyst Coverage, Earnings Management and Financial Development: An International Study
Corporate Governance & Finance eJournal
(2014)
Whether export can be helpful to relieve the financing restraints?—Empirical research based on research data about China’s firms of the world bank
F. Magiera (1997)
Do Stock Prices Fully Reflect Information in Accruals and Cash Flows about Future Earnings
Cfa Digest, 27
(2013)
Five-year plan, China finance and their consequences (Working Paper)
Doris Weichselbaumer, R. Winter‐Ebmer, Martina Zweimüller (2007)
An International Study
(2013)
Tangible information, intangible information and analysts
David Erkens, Mingyi Hung, Pedro Matos (2012)
Corporate Governance in the 2007-2008 Financial Crisis: Evidence from Financial Institutions Worldwide
CGN: Effects on Corporate Governance in Financial & Economic Crises (Topic)
B. Henderson, Joseph Marks (2013)
Predicting forecast errors through joint observation of earnings and revenue forecasts
Journal of Banking and Finance, 37
Ting Chen, Xiumin Martin (2011)
Do Bank-affiliated Analysts Benefit from Lending Relationships?
IO: Empirical Studies of Firms & Markets eJournal
James Ang, Yulong Ma (2001)
The behavior of financial analysts during the Asian financial crisis in Indonesia, Korea, Malaysia, and Thailand
Pacific-basin Finance Journal, 9
(2012)
Earnings components, continuous difference and earnings forecast of financial
(2011)
Macroeconomic policy and micro-firm's behavior-Expanding new fields of accounting and financing researches
Jean-François Bacmann, G. Bolliger (2001)
Who are the Best? Local Versus Foreign Analysts on Latin American Stock Markets
Capital Markets eJournal
(2010)
The effects of regulation FD
(2009)
Macroeconomic factors and changes in corporate capital structure
Inder Khurana, Raynolde Pereira, K. Raman (2003)
Does Analyst Behavior Explain Market Mispricing of Foreign Earnings for U.S. Multinational Firms?
Journal of Accounting, Auditing & Finance, 18
Augustine Duru, D. Reeb (2002)
International Diversification and Analysts' Forecast Accuracy and Bias
Accounting review: A quarterly journal of the American Accounting Association, 77
(1984)
Segment earnings disclosure and ability of security analysts to forecast earnings per share
Jeffrey Callen, Ole‐Kristian Hope, D. Segal (2004)
Domestic and Foreign Earnings, Stock Return Variability, and the Impact of Investor Sophistication
SPGMI: Compustat Fundamentals (Topic)
P. Rosenbaum, Donald Rubin (1983)
The central role of the propensity score in observational studies for causal effects
Biometrika, 70
William Kinney (1971)
Predicting Earnings - Entity Versus Subentity Data
Journal of Accounting Research, 9
D. Reeb, C. Kwok, H. Baek (1998)
Systematic Risk of the Multinational Corporation
Journal of International Business Studies, 29
(2009)
The empirical analysis of the export determinants of China’s local manufacturing firms
G. Maddala (1983)
Limited-dependent and qualitative variables in econometrics: Discriminant analysis
Christopher Malloy (2003)
The Geography of Equity Analysis
S&P Global Market Intelligence Research Paper Series
Ramji Balakrishnan, T. Harris, P. Sen (1990)
The Predictive Ability Of Geographic Segment Disclosures
Journal of Accounting Research, 28
C. Roberts (1989)
Forecasting Earnings Using Geographical Segment Data: Some UK Evidence
Journal of International Financial Management and Accounting, 1
(2014)
Does the cash flow forecast of analysts improve the accuracy of earnings forecast—Empirical evidence from China’s A-share market
(2010)
The influence of corporate diversification on the attention of security
Allen Berger, Christa Bouwman (2012)
How Does Capital Affect Bank Performance During Financial Crises?
IO: Empirical Studies of Firms & Markets
Frank Yu (2008)
Analyst Coverage and Earnings Management
IO: Firm Structure
(2007)
Who are the attentiongetting finns for foreign securities analysts
Ole‐Kristian Hope, Don Herrmann, W. Thomas (2007)
International Diversification and Forecast Optimism: The Effects of Reg FD
Auditing
(2011)
The corporate governance mechanism concerned by the media—Exploration based on the view of earnings management
(2002)
Table A1-2. the annual distribution of the samples
Kathleen Kahle, René Stulz (2012)
Access to Capital, Investment, and the Financial Crisis
Corporate Finance: Valuation
(2003)
The information sources, focus area and analysis tool of security analysts
S. Orpurt (2004)
Local Analyst Earnings Forecast Advantages in Europe
S&P Global Market Intelligence Research Paper Series
M. ConroyRobert, FukudaYujiro, S. HarrisRobert (1997)
Securities Houses and Earnings Forecasts in Japan: What Makes for an Accurate Prediction?
Financial Analysts Journal, 53
Richard Sloan (1998)
Do Stock Prices Fully Reflect Information in Accruals and Cash Flows About Future Earnings
The Accounting Review, 71
Dave Nichols, Larry Tunnell, Cindy Seipel (1995)
Earnings forecast accuracy and geographic segment disclosures
Journal of International Accounting, Auditing and Taxation, 4
(2010)
Earnings forecast errors of analysts and local advantages
Stephen Goldberg, Frank Heflin (1995)
The Association Between the Level of International Diversification and Risk
Journal of International Financial Management and Accounting, 6

Publisher: Taylor & Francis
Copyright: © 2017 Accounting Society of China
ISSN: 2169-7221
eISSN: 2169-7213
DOI: 10.1080/21697213.2016.1252087
Publisher site: See Article on Publisher Site

Abstract

China Journal of aCC ounting StudieS , 2017 Vol . 5, no . 1, 123–150 http://dx.doi.org/10.1080/21697213.2016.1252087 a b c a,d Fu Xin , Shangkun Liang , Jiemin Dai and Xiaorong Du a b School of Business, hohai university, nanjing, China; School of a ccountancy, Central university of f inance and economics, Beijing, China; School of international a udit, nanjing a udit university, nanjing, China; School of Business, hohai university, Jiangsu Provincial Collaborative i nnovation Center of World Water Valley and Water ecological Civilization, nanjing, China ABSTRACT KEYWORDS analysts forecast errors; This paper aims to provide insights into the forces and constraints that China; exporting; financial shape the forecast errors of local analysts when a firm enjoys both crisis; foreign analysts; domestic and foreign earnings. By applying a unique hand-collect propensity score matching; export dataset of Chinese listed firms, the key finding shows that the selection model forecast earnings of exporting firms issued by local analysts deviate more from the actual earnings than those for non-exporting firms. On the contrary, foreign analysts exhibit an informational advantage to local analysts when forecasting the earnings of exporting firms. A two-stage selection model and propensity score matching procedure are applied for correcting the endogeneity problem in a corporate exporting decision. A detailed investigation shows that the forecast errors of local analysts increased significantly during the financial crisis of 2008. Also, the forecast errors of foreign analysts were smaller after the IFRS were been adopted in China than before. Our findings indicate that the accuracy of local analysts may suffer from the macro- economic environment in an export-driven nation. 1. Introduction This paper aims to provide insights into the forces and constraints that shape the forecast errors of local analysts when a firm enjoys both domestic and foreign earnings. First of all, China’s economy has manifested a significantly obvious export-oriented characteristic for a long time and the import and export volume has accounted for over 40% of GDP since 2000. Even if the world at large has suffered the impact of a financial crisis, China’s foreign trade volume in 2014 still reached RMB 26.43 trillion, with year-on-year growth of 2.3%, which was higher than the average level of global trade growth. In particular, the Chinese government has fully supported the “Going global” strategy of Chinese firms from several directions, including the financial system and administrative approval in recent years. In May 2015, the State Council issued Advices on Improving the Cooperation of International Capacity and Equipment Manufacturing, forcefully pushing China’s manufacturing to export to the global market. When observing the sample firms used in this study, almost 50% among the listed CONTACT Jiemin dai daijiemin@163.com *Paper accepted by heng Yue. © 2017 a ccounting Society of China 124 F . XIN ET AL. firms are found to be engaged in the export trade. Therefore, the local analysts are increas- ingly confronted with the impact imposed by the fluctuations of overseas markets on the profit of listed companies, which macroeconomic facts are of vital importance for us to reexamine the behaviour of the forecast errors of local analysts. Besides, from a macroeco- nomic view, the persistence of the components of accounting earnings from different sources vary significantly (Callen, Hope, & Segal, 2009; Sloan, 1996). The export proceeds grow with the increasing growth and complexity of a firm’s export business and international transac - tions. The relatively serious asymmetry of information, universally existing in aspects such as geographical distance, cultural customs and systems environment for local analysts, leads to considerable barriers in dealing with, and processing, the market information of the firm’s exported products. All these three factors consequently cause a systematic deviation for local analysts in the firm’s future earnings forecast. In addition, a questionnaire concerning Chinese local analysts indicated that what most concerned the analysts in relation to the firm’s annual report was the firm’s primary products and market situation (Hu et al., 2003), which conformed to the basic fact of the export-oriented characteristics of China’s enter- prises. In light of this discussion, this study attempts to explore the following two issues. The first is to ask whether the firm’s export business exerts influence on the deviation of the analysts’ earnings forecast. In what circumstances have the forecast errors of local analysts been reduced? This paper has hand-collected the export data of Chinese listed firms from 2002–2012 and has constructed a firm’s export business variables, measured by both a dummy variable and a continuous variable. The study examines systematically the influence that the com- pany’s export business imposes on the forecast errors of local analysts. We find that the forecast earnings of firms with an export business, issued by Chinese local analysts, deviate significantly more than forecasts for those without an export business. Subsequently, we introduce the earnings forecast data issued by foreign analysts and find that the forecast earnings of firms with an export business, issued by foreign analysts, deviate much less. We also explore the influence that the financial crisis exerts on the earnings forecast of analysts in relation to exporting firms. We find that the deviation of the earnings forecasts of Chinese local analysts towards export enterprises has increased during the financial crisis, while the foreign analysts are less vulnerable to suffer from the financial crisis, relying on advantages of information and capability. We also study the influence on the earnings forecast for firms with export business made by foreign analysts who are confronted with the differences between domestic and overseas accounting standards. It is found that the forecast errors of foreign analysts reduce significantly after the new accounting standard has been adopted in China. This phenomenon suggests that the foreign analysts have an informational advan- tage over local analysts because a newly-adopted accounting standard system is harmonised with the International Accounting Standards. among the current research into influential factors concerning the analysts’ forecast behaviour, most researchers regard the company’s products and market distribution as control variables in dealing with the complexity of the analysts’ forecast. f or example, variables including the number of the company’s cross-industries and the number of the products’ cross-re- gions are applied to control for business complexity, which fails to go deep into the company’s products and the concrete structure of market in order to analyse the factors influencing the analysts’ forecast behaviour. t he top two are management information and shareholding change during the reporting period respectively and according to the classification based on nature; the management information and the company’s shareholding change information are regarded as belonging to the company’s internal information, with reference to form 4 of hu et al. (2003). CHINA JOURNAL OF ACCOUNTING STUDIES 125 Our study contributes to the existing literature in three ways. First, a driving factor accounting for the rapid growth in China’s economy is export trade. In the past, existing literature has paid attention only to the influence of export business on China’s macro-econ - omy. In recent years there have been studies of the relationships between export behaviour and the firm’s microeconomic behaviour including financing constraints (Luo & Li, 2014). This study aims to explore the influence of a firm’s export business on the analysts’ deci- sion-making from the perspective of the individual analyst and tries to provide insights in the relationship between the macroeconomic environment and microeconomic firm behav - iour with new evidence. Second, referring to the framework of Duru and Reeb (2002) and based on domestic studies for forecast errors of local analysts (Yuan, Zhang, & Yue, 2014; Wei & Xue, 2015), this study attempts to further interpret the reason for forecast errors of analysts from the mac- roeconomic background of international export trade. Third, the differences of forecast behaviour between domestic analysts and foreign ana- lysts are gradually receiving growing interest. For example, Wang, Chen, and Hou (2010) studied the difference of advantages between domestic and foreign analysts from the per - spective of the consistency of accounting standards. This study conducts analysis of these topics from the view of exporting and seeks to identify the concrete sources of such advan- tages, which supplements the existing literature. The remaining structure of this study is as follows: the second part deals with the literature review and research hypothesis; the third part explains the research design; the fourth part sets out the basic empirical results; the fifth part provides further exploration of the influ- encing mechanism and the last part is the conclusion. 2. Literature review and hypothesis development 2.1. Exports and the analysts’ forecasts Previous studies show that a r fi m’s disclosure of segmental reporting information was helpful to the analysts’ forecasting ability (Balakrishnan et al., 1990; Baldwin, 1984; Collins, 1976; Kinney, 1971; Nichols et al., 1995; Roberts, 1989). However, Duru and Reeb (2002) firstly discovered the impact of the firm’s export business on analysts’ forecasts. They found that the forecast accuracy of analysts in respect of the earnings in those firms with export business would decline and simultaneously there would be more tendency to overestimate. Khurana, Pereira, and Raman (2003) made a relatively systematic study of the relationships between the analysts’ forecast errors on foreign earnings and market efficiency. They found that the analysts could discern the differences of persistency between foreign and domestic earnings, which mainly embodied underestimation of the persistency of foreign earnings. Herrmann, Hope, and Thomas (2008) explored the influence that the US Regulation Fair Disclosure exerted on the analysts when they dealt with the earnings forecast in respect of those firms with overseas business. They found that upon the implementation of Regulation Fair Disclosure, the earnings forecast of analysts in respect of those firms with export business tended to be less positive and significantly more accurate. Subsequently, studies involving the analysts’ forecast would generally take into consideration whether the firm has export business or not as an influential factor (such as Henderson & Marks, 2013). 126 F. XIN ET AL. The domestic impact of the earnings components on the analysts’ forecast behaviour mainly focuses on the analysis of components including accrued profit and cash flow pro - jection. Ji and Tong (2012) studied the analysts’ responses to the persistency of different earnings components. The analysts were found in the studies to be able to differentiate the low persistency of accrued profit and high persistency of cash flow, which to some extent recognised the analytical ability of domestic analysts. However, they also found that analysts failed to differentiate the discretionary and non-discretionary accruals. In addition, Yuan, Zhang, and Yue (2014) found that the analysts’ forecast of the firm’s cash flow was helpful to better understand the earnings structure. As to the firm with an obvious motivation for earnings management, the cash flow forecast released by analysts plays a more prominent role in improving the accuracy of the earnings forecast (Yuan, Zhang, & Yue, 2014). Cai and Zeng (2010) reported the correlation existing between the diversification level of listed firms and the analysts’ attention. Their study suggests that the higher the corporate diversification level, the lower the analysts’ attention, and the higher the corporate diversi- fication with related business, the higher the analysts’ attention. However, Cai and Zeng (2010) did not explore the relationship between the regional distribution of the corporate business and the analysts’ forecast behaviour, and did not further analyse issues concerning the corporate export business. Meanwhile, they also did not take the endogeneity of diver- sification into consideration and confined the research samples to one year’s data of 2008. Huang and Huang (2013) explored the analysts’ forecast behaviour towards accounting-based performance measures (which they label as tangible information), and items such as R&D expenditure and goodwill (which they label intangible information), finding that the earnings forecast was more vulnerable to suffer from the influence of tangible information while the analysts’ recommendation tended to be easily influenced by intangible information. To sum up, in relation to research subject into the firm’s export business and the analyst’s forecast behaviour, for countries outside China the research literature has accumulated certain research findings and shaped related research conclusions. But the aforesaid research issues have not received enough attention in China so far. As China is a major exporter and an emerging nation highly reliant on an export-oriented economy, the above-mentioned issues not only mat- ter but also will generate some more meaningful and valuable topics for researchers. 2.2. Hypothesis development The impact of overseas business on the firm’s earnings falls into two aspects. On the one hand, the firms can reduce the volatility and risk of earnings based on combining businesses in dier ff ent regions or cross-industries, which is one of the reasons for many firms conducting diversification and overseas expansion. On the other hand, a growing literature indicates that the overseas revenue generated from overseas expansion tends to bring higher earnings volatility and therefore increases the firm’s operational risk. For instance, both Goldberg and Heflin (1995) and Reeb et al. (1998) found that cross-regional operation of the firm’s business would be more subject to the influence of political risk, exchange rate risk between the two countries concerned, supervision environment, economic fluctuation and other factors t he conclusion at this point is different from the subsequent discovery concerning whether analysts can restrain the enter - prise earnings management with reference to d egeorge et al. (2013), Yu (2008), Yu et al. (2011). Such as the research into the inu fl ence of the macroeconomic environment on the microcosmic company’s behaviour which have received attention recently , e.g. Chen et al. (2013), Jiang and r ao (2011), Su and Zeng (2009). CHINA JOURNAL OF ACCOUNTING STUDIES 127 where the business was located, and therefore intensified the volatility of the firm’s earnings. Li et al. (2004) also pointed out the lessons of a firm’s failed operation due to ignorance of risk management during the period of China’s listed companies conducting cross-industrial transformation. It can be observed from the descriptive statistics in this paper (VolROA, refer to Table 1) that the earnings volatility in our sample was signica fi ntly higher in the exporting firms than in other firms. Therefore, according to this logic, it is reasonable to predict that the firm’s export business will increase the difficulty of analysts’ earnings forecast, and then enlarge the forecast deviation of analysts. From the viewpoint of the analysts’ ability, the firm’s export business always requires the analysts to equip themselves with knowledge about related exporting countries or regions apart from their own country. With regard to the local analysts, most of them lack the back- ground and information of the firm’s operations concerning related exporting countries or regions. Differences in factors such as customs, systems, rivals, geographic features, super - vision in of different countries or regions impose higher requirements on the analysts’ ability than those local firms without export business and place demands on them to be more professional in interpreting, analysing and evaluating the earnings information of export firms. All these factors will increase the difficulty of the analysts’ forecast. Consequently, from the perspective of analysts, we expect that the firm’s overseas business will increase the difficulty of the analysts’ forecast and further add to the forecast errors of analysts. Hence, research hypothesis 1 is proposed. H1: The forecast earnings of analysts in respect of firms with export business deviate more than those without export business. Following the view of the analysts’ ability, the geographic location of the analysts will produce prominent influence on their forecast errors, and studies in this aspect vary. Malloy (2005) collected the data concerning the geographic location of American analysts and noticed that analysts who located closer to the listed firms tended to boast more accurate forecast. At the same time, the revision information of earnings forecast made by these local or adjacent ana- lysts exerted more significant influence on the stock price. Later, Bae, Stulz, and Tan (2008), based on the data about analysts from 32 countries, found that the local analysts held a fairly obvious forecast advantage which did not exist in companies with overseas assets. Taking the research reports of domestic analysts from 2005 to 2007 as samples, Li, Li, and Zhang (2010) discovered that analysts who located in the same province as the listed firms would forecast earnings more accurately. However, Bacmann and Bolliger (2001) got completely opposite conclusion that overseas analyst forecasted more accurately than local analysts did based on the data of capitalism market in seven Latin American countries. According to the data about analysts from seven European countries, Orpurt (2004) found that the informational advantage of local analysts only existed in Germany while the local advantages cannot be found by the standard of the location of the securities dealers to which the analysts belonged. On the con- trary, Bolliger (2004) found that the local securities dealers from European counties possessed informational advantage. This phenomenon was verified to various extents in Japan (Conroy, Fukuda, & Harris, 1997) and Taiwan. Since domestic analysts lack sensitivity of responses towards export business and overseas market, the foreign analysts will forecast earnings towards companies with export business more accurately by means of informational and capa- bility advantages. Hence, research hypothesis 2 is proposed. H2: The forecast earnings issued by foreign analysts in respect of firms with export business deviate significantly less than those issued by local analysts. 128 F. XIN ET AL. Table 1. Variable definition. Variable names Definitions Source Analysts characteristics Mean ERROR t he absolute difference between the earnings data is from CSM ar. t he construction of forecasts and actual earnings. t he firm i’s the variable is according to hong and forecast error in the year t amounts to the Kacperczyk (2010) absolute value of all the current forecast mean value (median) of analysts tracking the firm i and the current ePS divided by the closing share price of the firm i in the year t − 1. t he analysts’ forecast takes the most recent one among the ePS forecasts made before 180 days Median ERROR We take the median of earnings forecasts data is from CSM ar. t he construction of using the same data as described Mean the variable is according to hong and error Kacperczyk (2010) Number of analysts t he number of analysts following the firm i in data is from CSM ar year t Coverage t he natural logarithm of the number of data is from CSM ar analysts following the firm i in year t International analyst one when the earnings forecast is from data is from iBeS foreign analysts, zero otherwise Export sales Export Revenue (Mil. RMB) operating revenue involved outside the data is hand-collected from CSM ar mainland regions by hand based on the segmental reports (regional) disclosed by the listed firms Export/Sales t he proportion of export revenue in total data is hand-collected from CSM ar operation revenue Export/TA t he proportion of export revenue in total data is hand-collected from CSM ar assets Export/Employee (Mil RMB) t he ratio of export revenue to employees data is hand-collected from CSM ar Firm characteristics Market Value (Mil. RMB) t he firm’s total market capitalisation at the data is from CSM ar end of year LnSize t he natural logarithm of the firm’s total data is from CSM ar market capitalisation Sigma t he standard deviation of the current year’s t he construction of the variable is daily stock return according to hong and Kacperczyk (2010) Surprise one if the ePS in the current year exceeds that t he construction of the variable is in the previous year, zero otherwise according to Chen and Martin (2011) VolROA t he volatility of roa in the past three years data is from CSM ar Export determinants LnSales t he natural logarithm of sales revenue in year t he construction of the variable is −1 t − 1 according to liu and Zhang (2009) Workforce t he proportion of production workers in the t he construction of the variable is −1 total number of employees in year t − 1 according to liu and Zhang (2009) Edu t he proportion of employees possessing a t he construction of the variable is −1 university academic degree, or higher according to liu and Zhang (2009) qualification, in the total number of employees in year t − 1 PPE t he proportion of fixed assets and other t he construction of the variable is −1 long-term assets in the total assets in year according to liu and Zhang (2009) t − 1 Capd t he amount of fixed assets per capita in year t he construction of the variable is −1 t − 1 according to liu and Zhang (2009) Salesg t he growth rate of sales revenue in the last t he construction of the variable is −1 three years in year t − 1 according to liu and Zhang (2009) East one if a firm locates in the eastern regions, t he construction of the variable is −1 zero otherwise according to liu and Zhang (2009) CHINA JOURNAL OF ACCOUNTING STUDIES 129 3. Research design 3.1. Data sources We adopt the domestic A-share listed firms from 2002 to 2012 as samples and exclude the financial and insurance industries. The financial data applied in the study came from the CSMAR research database. We manually collected and arranged the basic data about export business in listed firms from 2002 to 2012 (the export regions included Hong Kong, Macau and Taiwan), including the operating revenue, operating cost and operating profits data generated from export business, export regions, and regions with export business ranking top five . The basic data about forecast earnings per share from 2002 to 2012 were acquired from the CSMAR data and data concerning the firm’s actual earnings per share came from the annual report at each year. The financial data along with data about the monthly stock return were taken from the database of CSMAR listed firms. Eventually, 7897 sample obser - vations were formed and used to test the research hypothesis 1 (see Appendix A-1 and A-2). Data in respect of the foreign analysts tracking domestic listed firms were acquired from IBES. For each sample firm, we calculate the local and foreign earnings forecast deviation respectively. Correspondingly, we construct the local and foreign analyst followings for each sample firm. Finally, we obtain a total of 11737 observations including local and foreign analyst forecasts which is applied to test the research hypothesis 2. All the continuous var- iables were winsorized by 1%. 3.2. The firm’s export business variables As to the measurement of export business, we collected the data of operating revenue, operating cost and operating profit involved outside the mainland regions by hand based on the segmental reports (regional) disclosed by the listed firms. Since the annual report did not disclose data about the overseas assets, the export revenue was consequently adopted as the basic index. On this basis, the proportion of export revenue in total operation revenue was used to measure the firm’s export business, measured as the ratio Export/Sales; in addition, Export was defined as a 0/1 dummy variable. When a firm exported to other countries and to the regions of Hong Kong, Macau and Taiwan, the value of Export was 1, otherwise 0. In this way, we obtained two variables to measure a firm’s export business: Export and Export/Sales. The definitions of the main variables are listed in Table 1. Figure 1 presents the time trend of the firm’s export variables from 2002 to 2012, based on data from our sample firms. Two features can be clearly observed: firstly, Export/Sales in Panel B demonstrates a trend of steady-state growth before 2008, and the average proportion of export revenue in that year’s total revenue was between 20 and 25%. Meanwhile, we calculated Export/TA (Total Assets) and Export/EMP (Employee) and adjusted the export revenue of the pre-IPO firms by total assets and number of employees. The growth trend was also evidenced in Export/ TA and Export/EMP. To verify the reliability of the data, we calculated the proportion of the exports in that year’s GDP (Export/GDP) based on the export data from the National Statistics Yearbook and listed the result in Panel A. The trend on the whole was basically consistent with the export trend in the study. Secondly, we can observe that the global financial crisis breaking during 2008 to 2009 had huge impact on China’s export business. Figure 1 suggests that the substantial influence exerted by the 2008 financial crisis on China emerged in 2009. 130 F. XIN ET AL. Figure 1. China’s export activity during the sample period of 2002–2012 (calculated by the authors from the sample data). We can see from the index of export accounting for GDP (Export/GDP) that this index had been presenting a growth trend since 2002, began to decline in 2007, then demonstrated a trend of sharp slope in 2009, decreasing rapidly from 31.97% in 2008 down to 24.06% in 2009, which indicated that the financial crisis in 2009 inflicted substantial influence on China’s export business. Similarly, the other three indexes are export indexes acquired through calculation based on the sample companies, namely, Export/Sales, Export/TA and Export/Employee, which are listed by median and mean value. The general trend of the influ - ence imposed by financial crisis on export business is basically consistent. In particular, changes in Export/TA are most obvious. Differences among indexes are mainly that the influ - ence of the financial crisis on the firm’s total assets is less than that on the revenue of com- panies relying on the export business. The financial crisis led to substantial shrinkage in the firm’s sales revenue, and factors relating to suspending and reducing production caused a dramatic decrease in the number of employees, which therefore dampened the influence of Export/Sales and Export/Employee. In general, the financial crisis led to a substantial reduc - tion in the export business in the sample companies during 2008 to 2009. 3.3. Other variables 3.3.1. The analysts’ forecast errors According to Hong and Kacperczyk (2010), the analysts’ forecast error is defined as follows: Forecast Error − Actual EPS t t FERROR = it (1) t−1 CHINA JOURNAL OF ACCOUNTING STUDIES 131 Namely, the firm i ’s forecast error in the year t amounts to the absolute value of all the current forecast mean value (median) of analysts tracking the firm i and the current EPS divided by the closing share price of the firm i in the year t − 1. The analysts’ forecast takes the most recent one among the EPS forecast made before 180 days. Based on calculating the difference between mean value and median adopted in the analysts’ forecast, we can get the Mean FERROR and Median FERROR. 3.3.2. Forecast error made by foreign analysts Acquiring the earnings forecast data of foreign analysts from IBES database, we can calculate the earnings forecast of foreign analysts for China’s listed firms in a similar way according to formula (1) calculating the analysts’ forecast error. The foreign analyst is defined as IntAnalyst, and when the earnings forecast is from foreign analysts, IntAnalyst is represented by a dummy variable 1, otherwise 0. 3.3.3. Control variables Based on previous studies, control variables include the number of analysts following (Coverage) using the natural logarithm; the volatility of the current year’s stock return is represented by the standard deviation of the current year’s daily stock return; according to the information content (Surprise) added into the current year’s EPS by Chen and Martin (2011). We represent it with a dummy variable, namely, one if the EPS in the current year exceeds that in the previous year, zero otherwise; the size of firm (LnSize) can be represented by the natural logarithm of the firm’s total market capitalization; the volatility of performance (VolROA) is represented by the volatility of ROA in the past three years. We also impose a control on the industry and time effects. 3.4. Model specifications In order to test the research hypothesis 1, we regard the analysts’ FERROR as the dependent variable and export variable as the independent variable, and simultaneously control for the variables that prior research has shown might exert influence on the analysts’ forecast errors (Hong & Kacperczyk, 2010), which can be tested by model (2). FERROR = + Export + Coverage + Sigma + Surprise i,t 0 1 i,t 2 i,t 3 i,t 4 i,t (2) + LnSize + VolROA + IND + Year + 5 i,t 6 i,t 7i i 8j j i,t In model (2), I stands for the firm and t for year. Due to the influence of the firm’s individual factors, clustering might appear in the error term. To guarantee the robustness of results, the standard deviation should be clustered by firm. Based on the differences of computing methods, FERROR corresponds to mean FERROR and median FERROR respectively. Export in the model is measured by two methods. The first one is a dummy 0/1 variable, that is, one is adopted if the firm has an export business, zero otherwise; the second one is to obtain continuous variables Export/Sales according to the proportion of export sales revenue in the firm’s total sales revenue. Based on research hypothesis 1 in the study, the coefficient β of Export is expected to be positive. In order to control for the issues of self-selection that might exist in the export firms themselves, we adopt 2SLS to construct the Probit model controlling for endogenous 132 F. XIN ET AL. problems based on prior research on the inu fl ence of the determinants concerning the local firms’ export business. The model introduces the natural logarithm of sales revenue (LnSales), workforce level (that refers to the proportion of production workers in the total number of employees, Workforce), education level (that refers to the proportion of employees possess- ing the academic degree of university or above in the total number of employees, Edu), scale of fixed assets (that refers to the proportion of fixed assets and other long-term assets in the total assets, PPE), capital intensity (that refers to the amount of fixed assets per capita, Capd ), the growth rate of sales revenue in the last three years (Salesg), and whether it locates in the eastern regions or not (East). All the variables in the Probit model belong to the period t − 1 and are used to estimate the probability of the firm’s export in the period of t. Meanwhile, Propensity Score Matching is adopted and on the basis of Probit model involving export determinants, we apply a neighbouring matching principle to get the matched sample firms. To examine the hypothesis 2, the variable of foreign analysts and its interaction terms with the export business are added on the basis of the basic model (2) to obtain the coeffi- cient direction and significance of the cross terms. FERROR = + Export + Export IntAnayst + IntAnayst + Coverage i,t 0 1 i,t 2 i,t i,t 3 i,t 4 i,t + Sigma + Surprise + LnSize + VolROA + (3) 5 i,t 6 i,t 7 i,t 8 i,t i,t According to Malloy (2005), it was found that compared with domestic analysts, the foreign analysts held more informational advantages towards firms with export business. Consequently, the coefficient β of cross term is expected to be negative in the model (3). 4. Empirical results and analysis 4.1. Descriptive statistics Table 2 compares the differences between the main variables in companies with export business and companies without export business. Above all, the forecast errors of export business companies (Mean FERROR and Median FERROR) are all significantly greater than those of companies without export business from the perspective of analysts’ characteristics. Analyst following of companies in the export business also significantly surpasses that in companies without an export business. At the same time, the foreign analyst following companies with export business significantly exceeds that of companies without export business. The average export revenue reaches almost RMB 1.4 Billion Yuan, accounting for 23% of the total revenue (Export/Sales) (the median is 14.7%), accounting for 17.9% of the total assets (the median is 9.7%), and the export revenue per capita achieves RMB 356,000 Yuan (the median is RMB 106,000 Yuan). In terms of the firm’s characteristics, market capi- talisation of companies with and without export business present no significant differences while the former mean value (LnSize) is significantly larger than that of the latter, and the stock volatility also higher than that of the latter, which suggests that the more complicated the market environment the firms face, the higher the relative risk they will be confronted with. But the unexpected changes of earnings in export companies are significantly lower than those in non-export companies. CHINA JOURNAL OF ACCOUNTING STUDIES 133 Table 2. d escriptive statistics of variables. Non-export firm Export firm Variables N Mean Median N Mean Median Analysts characteristics *** *** Mean ERROR 4,265 0.03 0.02 3,632 0.04 0.02 *** *** Median ERROR 4,265 0.03 0.02 3,632 0.04 0.02 *** *** Number of analysts 4,265 10.58 5.00 3,632 12.14 7.00 *** *** Coverage 4,265 1.67 1.61 3,632 1.87 1.95 ** ** International analyst 6,239 0.32 0.00 5,498 0.34 0.00 Export sales *** *** Export Revenue(Mil. RMB) 4,265 0.00 0.00 3,632 1,397.73 329.95 *** *** Export/Sales 4,265 0.00 0.00 3,632 0.23 0.15 *** *** Export/TA 4,265 0.00 0.00 3,632 0.18 0.10 *** *** Export/Employee (Mil RMB) 4,256 0.00 0.00 3,630 0.36 0.11 Firm characteristics Market Value(Mil. RMB) 4,265 12,499.51 4,784.56 3,632 13,585.49 4780.64 *** LnSize 4,265 22.39 22.29 3,632 22.46 22.29 * *** Sigma 4,265 0.03 0.03 3,632 0.03 0.03 *** *** Surprise 4,265 0.53 1.00 3,632 0.49 0.00 ** VolROA 4,265 0.03 0.02 3,632 0.03 0.02 Exprot determinants *** *** LnSales 3,314 21.28 21.22 2,966 21.62 21.41 −1 *** Workforce 3,314 0.66 0.59 2,966 0.63 0.66 −1 *** *** Edu 3,314 0.21 0.15 2,966 0.18 0.14 −1 *** PPE 3,314 0.34 0.30 2,966 0.34 0.31 −1 *** *** Capd 3,314 1.01 0.32 2,966 0.60 0.29 −1 Salesg 3,314 0.13 0.11 2,966 0.12 0.11 −1 *** *** East 3,314 0.57 1.00 2,966 0.67 1.00 −1 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 4.2. Hypothesis 1: export and analysts’ forecast errors 4.2.1. Regression result Columns (1) to (8) in Table 3 demonstrate the OLS regression result based on model (2). The dependent variable in columns (1) to (4) is mean FERROR, and the dependent variable in columns (5) to (8) is median FERROR. Among all the regression results, the coefficients of Export and Export/Sales are significantly positive, supporting the research hypothesis 1, that is, the earnings forecast errors made by analysts in respect of companies with export business are significantly more than those in respect of companies without export business. Among other control variables, the coefficient of Coverage is significantly negative, which is consist - ent with the results of Hong and Kacperczyk (2010). In addition, when comparing the coef- ficients of Export and Export/Sales, we can see that the coefficient of continuous variables is far larger than for the 0/1 variable (such as 0.0037 < 0.0089, 0.0019 < 0.0055). Meanwhile, when the year and industry are controlled, the value of t also gradually declines, suggesting that the factors of industry and time have relatively great influence. 4.2.2. Discussion of the endogeneity problems Even if other factors including industry and year are controlled in model (2), important variables still might be omitted. Thus, we further adopt a treatment effect model and pro - pensity score matching procedure with attempt to reduce the influence of the endogeneity of export factors on research issues. 134 F. XIN ET AL. Table 3. regressions on export and analysts’ forecast errors. Mean ERROR Median ERROR Dep. Var. (1) (2) (3) (4) (5) (6) (7) (8) *** * *** * Export 0.004 0.002 0.004 0.002 (3.66) (1.82) (3.80) (1.66) *** ** *** * Export/Sales 0.009 0.006 0.009 0.005 (3.38) (2.02) (3.33) (1.90) ** *** *** *** Coverage −0.000 −0.001 −0.000 −0.001 −0.001 −0.002 −0.001 −0.002 (−0.10) (−2.56) (−0.07) (−2.64) (−1.33) (−3.69) (−1.31) (−3.77) *** *** *** *** *** *** *** *** Sigma −0.112 0.334 −0.114 0.332 −0.119 0.312 −0.120 0.310 (−2.62) (2.85) (−2.67) (2.84) (−2.82) (2.79) (−2.87) (2.77) *** *** *** *** *** *** *** *** Surprise −0.023 −0.022 −0.023 −0.022 −0.023 −0.022 −0.023 −0.022 (−27.15) (−27.16) (−27.16) (−27.18) (−28.05) (−27.83) (−28.04) (−27.83) *** *** *** *** *** *** *** *** LnSize 0.002 0.003 0.003 0.003 0.002 0.002 0.002 0.003 (4.34) (5.12) (4.61) (5.39) (3.24) (4.40) (3.54) (4.65) *** *** *** *** *** *** *** *** VolROA 0.276 0.273 0.275 0.272 0.276 0.274 0.275 0.274 (12.06) (12.05) (11.99) (12.01) (12.26) (12.35) (12.19) (12.31) *** *** *** *** Constant −0.013 −0.066 −0.016 −0.069 0.002 −0.052 −0.001 −0.055 (−1.12) (−4.89) (−1.36) (−5.13) (0.18) (−4.09) (−0.07) (−4.32) Year fe no YeS no YeS no YeS no YeS industry fe no YeS no YeS no YeS no YeS Cluster within firms YeS YeS YeS YeS YeS YeS YeS YeS observations 7,897 7,897 7,897 7,897 7,897 7,897 7,897 7,897 a dj. R-squared 0.14 0.28 0.14 0.28 0.14 0.28 0.14 0.28 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. CHINA JOURNAL OF ACCOUNTING STUDIES 135 (1) Treatment effect model Since the dependent variables of analysts’ forecast errors can both be observed in the sample companies with or without export, a treatment effect model is adopted to deal with such kind of endogenous problems (Maddala, 1983). The first step, a binominal probit model is teased out based on the determinants of China’s export firms studied by Liu and Zhang (2009). Export = + LnSales + Workforce + Edu + PPE i,t 0 1 i,t−1 2 i,t−1 3 i,t−1 4 i,t−1 (4) + Capd + Salesg + East + 5 i,t−1 6 i,t−1 6 i,t−1 i,t The dependent variable in model (4) is 0/1 variable, and the independent variable is defined as previously. They both adopt the period of t − 1 to estimate the probability of export or not of the firm i in the period of t and simultaneously get the value of hazard of each obser - vation. In the second step, we substitute the value of hazard estimated by model (4) into model (1) to carry out regression and examine that whether the coefficient of Export confirms to the expectation. The results are displayed in Table 4. In the regression model of Probit at the first stage, the results suggests that larger firms with the more intensity of fixed assets, located in the Table 4. regressions on export and analysts’ forecast errors: t reatment effect. (1) (2) (3) Dep. Var. First Stage Mean ERROR Median ERROR *** *** Export 0.024 0.023 (5.36) (5.12) *** *** Coverage −0.003 −0.003 (−4.85) (−6.17) *** *** Sigma 0.348 0.315 (4.30) (3.98) *** *** Surprise −0.023 −0.023 (−24.49) (−24.58) *** *** LnSize 0.003 0.002 (4.05) (3.57) *** *** VolROA 0.333 0.337 (20.42) (21.12) *** Hazard 0.158 (11.69) LnSales −0.010 −1 (−1.04) *** Workforce −0.652 −1 (−5.29) ** Edu 0.208 −1 (2.21) *** PPE −0.112 −1 (−9.14) *** Capd −0.236 −1 (−3.41) *** Salesg 0.310 −1 (8.72) *** East 0.158 −1 (11.69) *** *** *** Constant −3.489 −0.051 −0.041 (−12.02) (−3.61) (−2.99) Year fixed effects no YeS YeS industry fixed effects no YeS YeS observations 5,706 5,706 5,706 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 136 F. XIN ET AL. Table 5. regressions on export and analysts’ forecast errors: PSM. (1) (2) (3) (4) Dep. Var. Mean ERROR Median ERROR Mean ERROR Median ERROR *** *** Export 0.004 0.004 (3.20) (3.27) *** *** Export/Sales 0.008 0.008 (2.80) (2.76) Coverage −0.000 −0.001 −0.000 −0.001 (−0.04) (−1.05) (−0.11) (−1.12) *** *** *** *** Sigma −0.127 −0.139 −0.130 −0.142 (−2.62) (−2.82) (−2.70) (−2.90) *** *** *** *** Surprise −0.024 −0.023 −0.024 −0.023 (−25.40) (−25.88) (−25.40) (−25.85) *** ** *** *** LnSize 0.002 0.001 0.002 0.002 (3.25) (2.26) (3.56) (2.60) *** *** *** *** VolROA 0.309 0.306 0.308 0.305 (11.11) (11.25) (11.04) (11.18) Constant −0.003 0.011 −0.007 0.008 (−0.25) (0.91) (−0.51) (0.63) Year fe no no no no industry fe no no no no Cluster within firms YeS YeS YeS YeS a dj. R-squared 0.15 0.15 0.15 0.15 observations 6,346 6,346 6,346 6,346 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. eastern regions, are more likely to export because the coefficients of LnSize, PPE, East are positively significant. However, Edu and Capd are negatively correlated to Export, which indicates that the export business of China’s listed firms is mainly characterised by low val- ue-added and labour-intensive firms. These characteristics basically conform to the fact of export business of China’s firms summarised by Liu and Zhang (2009) in their papers. In the regression model of OLS at the second stage, we control for the fixed effect of industry and year and substitute the value of hazard estimated at the first stage into OLS model. The results in columns (2)–(3) in Table 3 indicate that all the coefficients of Export are significantly positive (β = 0.0246, t = 5.39), which consequently supports the research hypothesis 1. (2) Propensity score matching (PSM) We further adopt PSM to deal with the heteroscedasticity of sample selection (Rosenbaum & Rubin, 1983). Specifically, as to each export sample firm, we identify a pairing sample (control group) according to the principle of nearest-neighbourhood from the corresponding sample firm without an export business. The pairing is conducted based on the variables in the Probit model and the variables include LnSales, Workforce, Edu, PPE, Capd, Salesg and East. Then, we put the companies sample with export business and those without export business together and again carry out regression based on model (1), obtaining columns (1)–(4) in Table 5 in which both the coefficients of Export and Export/Sales are significantly positive (coefficient of Export is 0.0032, t = 2.92, coefficient of Export/Sales is 0.0073, t = 2.65) and therefore support the research hypothesis 1. (3) Event study In Figure 2, we identify the companies that develop an export business from a non-export business and take the export business in that beginning year as the event to count the change trend of analysts’ foreign errors. We are able to clearly observe that the mean value CHINA JOURNAL OF ACCOUNTING STUDIES 137 Figure 2. analysts’ forecast errors before and after the firm’s entry to the export market. Table 6. analysts’ forecast errors before and after the firm’s export. (1) PRE (2) (3) POST Diff. [−3, −1] T = 0 [1, 3] (2)−(1) (3)−(2) Panel A: Mean Error *** Mean 0.028 0.038 0.039 0.012 −0.004 *** Median 0.016 0.028 0.025 0.010 −0.003 Panel B: Median Error *** Mean 0.028 0.035 0.038 0.011 −0.004 *** Median 0.014 0.025 0.024 0.008 −0.003 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. of mean FERROR gradually increases to 0.0378 in the year of export from 0.0161 in the pre- vious 5 years before export, and subsequently decreases to 0.0307 in 4 years after export but with a leap in the fifth year after export. The trend of median FERROR is similar. Furthermore, in Panel A and B of Table 6, after conducting statistical tests of analysts’ forecast errors before and after export and in the year of export, we can observe that forecast errors in the year of export are significantly larger than those before export (the difference of mean FERROR is 0.0121 while that of mean FERROR is 0.0108, both of which are significant at the 1% level). However, the forecast errors after export do not demonstrate any difference (the difference of mean FERROR is −0.0039 while that of median FERROR is −0.004, both of which are insignificant), and test result of median is consistent with that of mean value. The aforesaid event study did not take other features of the firm into account as controls, but the prominent influence of export on analysts’ forecast errors can be observed from this trend. This method has also fairly controlled the endogenity problem existing in the research issue itself. 4.3. Hypothesis 2: do foreign analysts possess informational advantage? 4.3.1. The basic regression results Whether foreign analysts possess informational advantage compared with domestic analysts is an important topic (Malloy, 2005; Wang, Chen, & Hou, 2010). This part carries out relatively in-depth discussion and analysis of the topic. We obtain the regression result by adding foreign analysts based on model (3), which can be seen in Table 7. In accordance with expec- tation, the regression results of columns (1) to (4) show that all coefficients of cross terms 138 F. XIN ET AL. Table 7. regression results on foreign analysts, export and analysts’ forecast errors. (1) (2) (3) (4) Dep. Var. Mean ERROR Mean ERROR Median ERROR Median ERROR ** ** Export 0.003 0.002 (2.49) (2.35) *** *** Export*IntAna −0.004 −0.004 (−3.35) (−3.42) ** ** Export/Sales 0.007 0.007 (2.52) (2.46) *** *** Export/Sales*IntAna −0.008 −0.009 (−3.04) (−3.17) *** *** *** *** IntAna −0.021 −0.022 −0.020 −0.020 (−21.22) (−24.53) (−20.34) (−23.52) *** *** *** *** Coverage −0.002 −0.002 −0.002 −0.002 (−3.27) (−3.25) (−4.51) (−4.47) *** *** *** *** Sigma 0.275 0.274 0.258 0.256 (3.10) (3.09) (3.00) (2.99) *** *** *** *** Surprise −0.018 −0.017 −0.017 −0.017 (−26.25) (−26.14) (−26.54) (−26.41) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (5.98) (6.10) (5.64) (5.73) *** *** *** *** VolROA 0.245 0.244 0.247 0.246 (13.13) (13.08) (13.36) (13.32) *** *** *** *** Constant −0.045 −0.046 −0.038 −0.039 (−3.81) (−3.86) (−3.45) (−3.49) Year fixed effect YeS YeS YeS YeS industry fixed effect YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 11,737 11,737 11,737 11,737 a djusted R-squared 0.28 0.28 0.27 0.27 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. are significantly negative at the 1% level, which suggests that foreign analysts possess rel- atively clear informational advantage in earnings forecast towards domestic listed firms with an export business. 4.3.2. The self-selection of foreign analysts Based on the discoveries of Lin, Ouyang, and Yue (2007), foreign analysts tend to analyse firms that possess a perfect governance structure, non-state ownership and desirable net profit. Therefore, would the self-selection existing in foreign analysts themselves influence the conclusion of the hypothesis 2? The section discusses this question further. In hypothesis 2, the theme of the study is whether there are significant differences between domestic and foreign analysts when making forecasts in respect of firms with export business. To reduce the effect of influence from self-selection by foreign analysts, we adopt the following methods. We choose the firms tracked simultaneously by local and foreign analysts as a subsample to explore the influence of export business on analysts’ forecast errors. In this subsample, since the local analysts choose to track the same firm as the foreign analysts do, we believe that factors influencing the corporate’s governance, own - ership structure and operation performance of the firm tracked by foreign analysts can be fairly controlled and are therefore capable of relieving the self-selection problems of foreign t hanks to the anonymous reviewers for their constructive opinions here. CHINA JOURNAL OF ACCOUNTING STUDIES 139 Table 8. Comparison between pairing sample firms. Firms followed by local analysts Firms followed by foreign analysts Diff. (0−1) Variables N Mean Median N Mean Median Mean Median *** *** Mean ERROR 3,648 0.032 0.020 3,648 0.013 0.007 0.019 0.013 *** *** Median ERROR 3,640 0.030 0.018 3,648 0.013 0.007 0.017 0.011 SOE 3,587 0.636 1.000 3,587 0.635 1.000 0.001 0.000 Shareholder 3,622 39.705 39.335 3,622 39.702 39.300 0.003 0.035 Indirector 3,628 0.364 0.333 3,628 0.364 0.333 0.000 0.000 ChairmanCEO 3,648 0.154 0.000 3,648 0.154 0.000 0.000 0.000 ROA 3,648 0.083 0.072 3,648 0.083 0.072 0.000 0.000 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. analysts. First, in Table 8, we compare the earnings forecast errors between local and foreign analysts in this pairing subsample and check whether there are significant differences in corporate governance, ownership structure and operation performance of the firm that the foreign and local analysts choose to track. Because we choose the firm sample tracked simul - taneously by local and foreign analysts, the sample firms have the same characteristics across the two analyst groups. However, when comparing the forecast errors of local analysts with those of foreign analysts, we can observe that forecast errors made by foreign analysts are prominently significantly lower than those made by local analysts. Based on the pairing samples established, according to the findings of Lin et al. (2007), we add the control variables including whether state-owned, share proportion of the largest shareholder, proportion of independent directors, dual role of the board chairman and return on assets to carry out regression on the basis of model (3) in the study, and the regression results are shown in Table 9. This result further verifies hypothesis, suggesting that the results of hypothesis 2 are robust. 4.3.3. Sensitivity test on export revenue We further study the sensitivity of domestic and foreign analysts’ forecast errors towards the export revenue in listed companies. The approach is as follows. In Table 10, we divide non-vanishing samples of Export/Sales into five groups from high to low and explore respec - tively the change trend of mean ERROR and median ERROR of forecast errors made by domes- tic and foreign analysts. Taking mean ERROR as an example, we can reduce it to discoveries in two aspects: (1) with the increase in the proportion of export revenue, forecast errors of foreign analysts are increasing as the mean value (median) changes from 0.0296 (0.0389) to 0.0627 (0.0507), with change surpassing 200% and significantly different at the 1% level, while the difference in mean value of foreign analysts’ forecast errors is only 0.0098 although it is also significant at the 1% level but its numerical value is far less than that of domestic analysts (0.0331); (2) with the increase in the proportion of export revenue, the difference between domestic and foreign analysts in forecast errors is increasingly large as its mean value expands from −0.0168 to −0.0401, with change surpassing 200% and there are signif- icant differences at the 1% level. We can infer from the above analysis that the proportion of export revenue has an influence on the value of forecast errors of domestic and foreign analysts. On the whole, export factors tend to exert a larger influence on earnings forecast errors of domestic analysts than those of foreign analysts. 140 F. XIN ET AL. Table 9. regression results of pairing samples. (1) (2) (3) (4) Dep. Var. Mean ERROR Mean ERROR Median ERROR Median ERROR * * Export 0.002 0.002 (1.87) (1.70) *** *** Export*IntAna −0.005 −0.005 (−4.40) (−4.58) ** ** Export/Sales 0.008 0.007 (2.35) (2.26) *** *** Export/Sales*IntAna −0.011 −0.011 (−3.61) (−3.67) *** *** *** *** IntAna −0.016 −0.017 −0.014 −0.015 (−13.59) (−14.75) (−12.49) (−13.46) Coverage 0.001 0.001 0.000 0.000 (0.91) (1.01) (0.11) (0.21) ** ** * * Sigma 0.237 0.233 0.157 0.154 (2.55) (2.53) (1.69) (1.66) *** *** *** *** Surprise −0.010 −0.010 −0.011 −0.011 (−15.68) (−15.70) (−16.72) (−16.74) *** *** *** *** LnSize 0.003 0.003 0.002 0.002 (5.43) (5.38) (5.14) (5.05) *** *** *** *** VolROA 0.232 0.232 0.230 0.230 (10.07) (10.05) (10.35) (10.31) SOE −0.001 −0.001 −0.000 0.000 (−0.82) (−0.70) (−0.09) (0.02) Shareholder −0.000 −0.000 −0.000 −0.000 (−0.25) (−0.20) (−0.07) (−0.03) * * ** ** Indirector −0.012 −0.012 −0.014 −0.014 (−1.67) (−1.67) (−2.09) (−2.09) ** ** * * ChairmanCEO −0.002 −0.002 −0.002 −0.002 (−1.97) (−2.03) (−1.89) (−1.95) *** *** *** *** ROA −0.156 −0.155 −0.146 −0.146 (−15.43) (−15.40) (−14.85) (−14.81) * * Constant −0.020 −0.019 −0.013 −0.012 (−1.78) (−1.73) (−1.19) (−1.11) Year fixed effect YeS YeS YeS YeS industry fixed effect YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 7,128 7,128 7,120 7,120 a djusted R-squared 0.378 0.378 0.365 0.364 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 5. Further analysis 5.1. External shock: the financial crisis during 2008 to 2009 The financial crisis has provided an opportunity for theoretical and practical cycles to study change of firms’ micro-decision-making behaviour when suffering an external impact. Prior research explores the influence of financial crisis on firms’ behaviour from the perspectives of investment and financing behaviour of firms, bank loan decision, corporate governance and others (Berger & Bouwman, 2013; Campello, Graham, & Harvey, 2010; Erkens, Hng, & Matos, 2012; Kahle & Stulz, 2013). Nevertheless, when searching the existing literature, we find that few involved research into the influence of the decision-making behaviour of ana- lysts in financial crisis. Only Ang and Ma (2001) studied the analysts’ behaviour in the capital market of Indonesia, South Korea, Malaysia and Thailand during the financial crisis in 1997. Our study selects the financial crisis during 2008 to 2009 in relation to the research back - ground of the Chinese capital market, Chinese economic volume and the characteristic of an export-oriented economy suffering more evident influence from financial crisis. CHINA JOURNAL OF ACCOUNTING STUDIES 141 Table 10. Sensitivity test on export revenue: local vs. foreign analysts. Export/Sales INT Lowest 2 3 4 Highest Diff. (H−L) Panel A: Mean ERROR *** l ocal analysts (0) Mean 0.030 0.034 0.032 0.027 0.063 0.033 *** Median 0.039 0.041 0.040 0.030 0.051 0.012 N 836 702 732 676 696 *** f oreign analysts (1) Mean 0.013 0.013 0.014 0.015 0.023 0.010 *** Median 0.021 0.016 0.021 0.030 0.035 0.015 N 264 397 368 423 404 *** *** *** *** *** Diff. (1−0) Mean −0.017 −0.020 −0.018 −0.012 −0.040 *** *** *** *** *** Median −0.018 −0.025 −0.019 0.001 −0.015 Panel B: Median ERROR *** l ocal analysts(0) Mean 0.030 0.032 0.030 0.025 0.060 0.032 *** Median 0.038 0.039 0.039 0.029 0.051 0.012 N 836 702 732 676 696 *** f oreign analysts(1) Mean 0.013 0.013 0.014 0.015 0.023 0.010 *** Median 0.021 0.016 0.021 0.031 0.035 0.015 N 264 397 368 423 404 *** *** *** *** *** Diff. (1−0) Mean −0.016 −0.019 −0.016 −0.010 −0.038 *** *** *** *** *** Median −0.018 −0.023 −0.018 0.002 −0.015 notes: ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. Table 11. impact of financial crisis on analysts’ forecast errors. Dep. Var. Mean ERROR Median ERROR Export 0.001 0.001 (1.06) (0.97) ** ** Export*Crisis 0.007 0.006 (2.19) (2.01) Export/Sales 0.003 0.003 (1.35) (1.32) ** * Export/Sales*Crisis 0.016 0.014 (2.02) (1.69) *** *** *** *** Crisis 0.037 0.038 0.034 0.036 (12.31) (13.28) (11.77) (12.74) *** *** *** *** Coverage −0.001 −0.001 −0.002 −0.002 (−2.59) (−2.68) (−3.71) (−3.80) *** *** *** *** Sigma 0.332 0.332 0.310 0.310 (2.84) (2.84) (2.77) (2.77) *** *** *** *** Surprise −0.022 −0.022 −0.022 −0.021 (−27.22) (−27.17) (−27.89) (−27.80) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (5.15) (5.44) (4.42) (4.69) *** *** *** *** VolROA 0.272 0.271 0.274 0.273 (12.04) (12.00) (12.33) (12.29) *** *** *** *** Constant −0.066 −0.069 −0.052 −0.056 (−4.90) (−5.17) (−4.10) (−4.35) Year fixed effect YeS YeS YeS YeS industry fixed effect YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 7,897 7,897 7,897 7,897 a djusted R-squared 0.28 0.28 0.28 0.28 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. 142 F. XIN ET AL. Table 12. impact of financial crisis on foreign analysts’ forecast errors. Mean ERROR Median ERROR Dep. Var. Local analysts Foreign analysts Local analysts Foreign analysts Panel A: Export Export 0.001 0.000 0.001 0.000 (1.23) (0.29) (1.11) (0.26) *** ** Export*Crisis 0.007 0.002 0.006 0.002 (2.83) (0.86) (2.57) (0.76) *** *** Crisis 0.037 0.007 0.034 0.007 (10.03) (0.47) (9.56) (0.46) *** *** *** *** Coverage −0.001 −0.003 −0.002 −0.003 (−3.24) (−6.01) (−4.60) (−6.00) *** *** Sigma 0.332 0.059 0.310 0.057 (4.69) (0.73) (4.47) (0.71) *** *** *** *** Surprise −0.022 −0.009 −0.022 −0.008 (−27.38) (−11.38) (−27.57) (−11.23) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (6.69) (8.80) (5.58) (8.94) *** *** *** *** VolROA 0.272 0.179 0.274 0.179 (22.80) (14.14) (23.42) (14.18) *** *** *** *** Constant −0.077 −0.066 −0.063 −0.067 (−6.94) (−3.70) (−5.80) (−3.76) observations 7,907 3,830 7,907 3,830 2 * 2 * l ocal analysts = f oreign analysts Chi (1) = 2.02 Chi (1) = 1.71 t est variable: Export*Crisis p = 0.155 p = 0.190 Panel B: Export/Sales Export/Sales 0.003 0.001 0.003 0.001 (1.54) (0.29) (1.54) (0.29) *** ** Export/Sales*Crisis 0.016 0.005 0.014 0.004 (2.76) (0.96) (2.35) (0.81) *** *** Crisis 0.038 0.008 0.036 0.008 (10.65) (0.49) (10.15) (0.48) *** *** *** *** Coverage −0.001 −0.003 −0.002 −0.003 (−3.36) (−6.02) (−4.70) (−6.01) *** *** Sigma 0.3322 0.060 0.310 0.058 (4.70) (0.75) (4.47) (0.72) *** *** *** *** Surprise −0.022 −0.009 −0.021 −0.008 (−27.23) (−11.29) (−27.42) (−11.15) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (7.06) (8.97) (5.91) (9.10) *** *** *** *** VolROA 0.272 0.179 0.274 0.179 (22.72) (14.15) (23.35) (14.18) *** *** *** *** Constant −0.080 −0.067 −0.066 −0.068 (−7.29) (−3.76) (−6.12) (−3.81) observations 7,907 3,830 7,907 3,830 2 * 2 l ocal analysts = f oreign analysts Chi (1) = 1.66 Chi (1) = 1.19 t est variable: Export*Crisis p = 0.198 p = 0.276 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. The influence of the financial crisis on the analysts’ forecast mainly comes from the impact on firms’ exports, which subsequently affects analysts’ forecast errors. Panel A in Chart 1 shows that the global financial crisis during 2008 to 2009 caused a great impact on China’s exports. During 2008 to 2009, the proportion of China’s export volume in GDP declined rapidly from 31.97 to 24.06%, with a decrease of almost a quarter ((24.06% − 31.97%)/31.9 7% = −24.74%). So which influence would this impact impose on the relationships between the firm’s export business and analysts’ forecast is the question that this study seeks to further explore. CHINA JOURNAL OF ACCOUNTING STUDIES 143 Table 13A. impact of international convergence of accounting standard on foreign analysts’ forecast errors: Mean error. Mean ERROR Before new stand- After new standards Before new stand- After new standards Dep. Var. ards Year < 2007 Year >= 2007 ards Year < 2007 Year >= 2007 *** Export 0.000 0.003 (0.06) (2.67) * *** Export*IntAna −0.005 −0.004 (−1.68) (−3.23) *** Export/Sales 0.002 0.008 (0.46) (2.59) *** Export/Sales*IntAna −0.008 −0.009 (−1.04) (−2.98) *** *** *** *** IntAna −0.018 −0.021 −0.019 −0.022 (−8.20) (−20.38) (−9.65) (−23.16) *** *** *** *** Coverage −0.005 −0.001 −0.005 −0.001 (−4.06) (−2.68) (−4.01) (−2.68) * *** * *** Sigma 0.337 0.254 0.337 0.253 (1.88) (2.60) (1.87) (2.59) *** *** *** *** Surprise −0.018 −0.017 −0.018 −0.017 (−12.17) (−23.45) (−12.16) (−23.36) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (2.79) (5.83) (2.67) (5.98) *** *** *** *** VolROA 0.351 0.226 0.350 0.225 (4.64) (12.17) (4.62) (12.13) *** *** Constant −0.030 −0.049 −0.028 −0.051 (−1.28) (−4.01) (−1.18) (−4.14) Year fe YeS YeS YeS YeS industry fe YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 2,055 9,682 2,055 9,682 a dj. R-squared 0.24 0.29 0.24 0.29 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. Similar to the previous method, we add the variable of financial crisis when verifying the influence of the firm’s export business on analysts’ forecast errors to explore whether there are significant differences in the influence of the firm’s export business on analysts’ forecast errors under the financial crisis. Based on the impact of the financial crisis on China’s exports in chart 1, we select the year 2009 in which exports suffered the largest impact as the year of financial crisis (namely, Crisis=1) according to the changing trend of the proportion of exports in GDP. Table 11 lists our results according to the mean value and median of analysts’ forecast errors respectively. In the regression model adopting Export and Export/Sales to measure the export behaviour, the signs of the cross-terms in four groups of regression models are all significantly positive. The results indicate that the financial crisis intensifies the influence of export behaviour on analysts’ forecast errors and aggravates the operating environment involving companies with export business, enlarging the risk for this kind of company and accordingly increasing analysts’ forecast errors. We further add the samples of foreign analysts to conduct subsample verification and the results are showed in Table 12. It shows that domestic analysts are more easily subjected to the influence of financial crisis and accordingly cause forecast errors while foreign analysts are less vulnerable to suffering from the influence of financial crisis, which can be observed upon comparison in the cross-terms of Export*Crisis. Significant differences exist in the coef - ficients of the relevant cross-terms between domestic and foreign analysts (the value of 144 F. XIN ET AL. Table 13B. impact of international convergence of accounting standard on foreign analysts’ forecast errors: Median error. Median ERROR Before new stand- After new standards Before new stand- After new standards Dep. Var. ards Year < 2007 Year >= 2007 ards Year < 2007 Year >= 2007 ** Export 0.001 0.003 (0.23) (2.47) * *** Export*IntAna −0.005 −0.004 (−1.81) (−3.25) ** Export/Sales 0.003 0.007 (0.57) (2.52) *** Export/Sales*IntAna −0.010 −0.009 (−1.24) (−3.06) *** *** *** *** IntAna −0.018 −0.020 −0.019 −0.021 (−7.78) (−19.55) (−9.16) (−22.20) *** *** *** *** Coverage −0.005 −0.002 −0.005 −0.002 (−4.47) (−3.79) (−4.43) (−3.78) ** ** ** ** Sigma 0.361 0.227 0.360 0.227 (2.01) (2.46) (2.00) (2.46) *** *** *** *** Surprise −0.018 −0.017 −0.018 −0.017 (−12.01) (−23.77) (−12.02) (−23.66) *** *** *** *** LnSize 0.003 0.003 0.003 0.003 (2.72) (5.46) (2.61) (5.57) *** *** *** *** VolROA 0.351 0.228 0.349 0.228 (4.62) (12.45) (4.60) (12.42) *** *** Constant −0.028 −0.040 −0.026 −0.042 (−1.17) (−3.44) (−1.09) (−3.56) Year fe YeS YeS YeS YeS industry fe YeS YeS YeS YeS Cluster within firms YeS YeS YeS YeS observations 2,055 9,682 2,055 9,682 a dj. R-squared 0.24 0.28 0.24 0.28 notes: robust T value is in parentheses. ***significant at the 1% level; **significant at the 5% level; *significant at the 10% level. Chi-squared suggests that the single tail is statistically significant at the 10% level). This result indicates that foreign analysts possess more informational and capability advantages in dealing with the influence of financial crisis than domestic analysts do. 5.2. Effect of international convergence of accounting standards6Thanks to the anonymous reviewers for their constructive opinions here Wang, Chen, and Hou (2010) probed the differences in advantages between domestic and foreign analysts from the point of the differences of accounting standards. They found that after the implementation of a new accounting standard in China in 2007, the informational advantage of local analysts was reduced. Discoveries by Wang et al. (2010) provided a research direction that can be used for reference for this study to explain the cause of forecast advantages of foreign analysts. The sample term of the study is from 2002 to 2012, which spans the historic event of the integration of China’s accounting standards with the inter- national standards in 2007. Consequently, we take the year of 2007 as the time node of the actual convergence of accounting standards, and further verify the hypothesis 2 in the study. Specifically, we regard the year of 2007 as the watershed, divide the samples into two groups according to the time before and after the implementation of the international accounting standards respectively, and perform verification based on separate samples. The empirical CHINA JOURNAL OF ACCOUNTING STUDIES 145 results are listed as Tables 13A and 13B. Before the implementation of new standards, the interaction terms of Export*IntAna are significant at the 10% level while those of Export/ Sales*IntAna are not significant. After the implementation, all the interaction terms are sig- nificant at the 1% level, suggesting that the introduction of new standards is one of the reasons for the informational advantages of foreign analysts. 6. Conclusion This study finds that forecast earnings in respect of firms with export business, issued by domestic analysts, deviate more than the forecasts in respect of firms without export busi- ness. The results remain robust as verie fi d by treatment ee ff ct and propensity score matching after considering the firm’s export determinants. Meanwhile, we find that foreign analysts possess informational advantage in terms of earnings forecast in respect of companies with export business, showing that, forecast earnings issued by foreign analysts deviate less. The results remain robust after considering the self-selection of foreign analysts. Subsequently, the study further explores the mechanism leading to these differences. It is found that a financial crisis will enlarge the errors towards earnings forecast of export firms made by foreign analysts, but foreign analysts are less vulnerable to suffer from the influence of finan - cial crisis because of informational and capability advantages; after the implementation of new accounting standards, foreign analysts demonstrate more significant informational advantage. Evidence in this study indicates that the export business in Chinese firms is an important influential factor affecting the analysts’ forecasting. We find that study from the macroeconomic background of China as the large exporter provides favourable inspiration for interpreting the decision-making behaviour and results of Chinese analysts. Acknowledgement We thank the editor, two anonymous referees, Dr. Heng Yue and Dr. Cong Wang. This study has been supported by National Natural Science Foundation of China (71402198), Social Science Foundation of Beijing, China (15JGC176), the Beijing Municipal Commission of Education ‘Joint Construction Project’, the Beijing Municipal Commission of Education ‘Pilot Reform of Accounting Discipline Clustering’, Innovation Research of the Central University of Finance and Economics, the Social Science Foundation of Jiangsu Province, China (15GLB007), the Key Project of Philosophy and Social Science Research in Colleges and Universities in Jiangsu Province, China (2015ZDIXM011), and the Fundamental Research Funds for the Central University (2015B03414). References Ang, J.S., & Ma, Y. (2001). The behavior of financial analysts during the Asian financial crisis in Indonesia, Korea, Malaysia, and Thailand. Pacific-Basin Finance Journal, 9, 233–263. Bacmann, J.F., & Bolliger, G. (2001). Who are the best? Local versus foreign analysts on the Latin American stock markets. Unpublished working paper. Bae, K., Stulz, R.M., & Tan, H. (2008). Do local analysts know more? A cross-country study of the performance of local analysts and foreign analysts. Journal of Financial Economics, 88, 581–606. Balakrishnan, R., Harris, T., & Sen, P. (1990). The predictive ability of geographic segment disclosures. Journal of Accounting Research, 28, 305–325. Baldwin, B. (1984). Segment earnings disclosure and ability of security analysts to forecast earnings per share. The Accounting Review, 59, 376–389. 146 F. XIN ET AL. Berger, A.N., & Bouwman, C.H. (2013). How does capital affect bank performance during financial crisis? Journal of Financial Economics, 109, 146–176. Bolliger, G. (2004). The characteristics of individual analysts’ forecasts in Europe. Journal of Banking and Finance, 28, 2283–2309. Cai, W., & Zeng, C. (2010). The influence of corporate diversification on the attention of security analysts—Based on the empirical analysis from the perspective of the decision-making behavior of security analysts. Nankai Business Review, 4, 125–133 (In Chinese). Callen, J.L., Hope, O., & Segal, D. (2009). Domestic and foreign earnings, stock return variability, and the impact of investor sophistication. Journal of Accounting Research, 43, 377–412. Campello, M., Graham, J.R., & Harvey, C.R. (2010). The real ee ff cts of financial constraints: Evidence from a financial crisis. Journal of Financial Economics, 97, 470–487. Chen, D.H., Li, O.Z., & Xin, F. (2013). Five-year plan, China finance and their consequences ( Working Paper). Chen, T., & Martin, X. (2011). Do bank-affiliated analysts benefit from lending relationships? Journal of Accounting Research, 49, 633–675. Collins, D. (1976). Predicting earnings with subentity data: Some further evidence. Journal of Accounting Research, 14, 163–177. Conroy, R.M., Fukuda, Y., & Harris, R.S. (1997). Securities houses and earnings forecasts in Japan: What makes for an accurate prediction? Financial Analysts Journal, 53, 29–40. Degeorge, F., Ding, Y., Jeanjean, T., & Stolowy, H. (2013). Analyst coverage, earnings management and financial development: An international study. Journal of Accounting and Public Policy, 32, 1–25. Duru, A., & Reeb, D.M. (2002). International diversification and analysts’ forecast accuracy and bias. The Accounting Review, 77, 415–433. Erkens, D.H., Hung, M., & Matos, P. (2012). Corporate governance in the 2007–2008 financial crisis: Evidence from financial institutions worldwide. Journal of Corporate Finance, 18, 389–411. Goldberg, S., & Heflin, F. (1995). The association between the level of international diversification and risk. Journal of International Financial Management and Accounting, 6, 1–25. Henderson, B., & Marks, J.M. (2013). Predicting forecast errors through joint observation of earnings and revenue forecasts. Journal of Banking and Finance, 37, 4265–4277. Herrmann, D.R., Hope, O., & Thomas, W.B. (2008). International diversification and forecast optimism: The effects of regulation FD. Accounting Horizons, 22, 179–197. Hong, H., & Kacperczyk, M. (2010). Competition and Bias. Quarterly Journal of Economics, 1683–1725. Hu, Y., Lin, W., & Wang, W. (2003). The information sources, focus area and analysis tool of security analysts. Journal of Financial Research, (12), 52–63 (In Chinese). Huang, L., & Huang, Y. (2013). Tangible information, intangible information and analysts’ behavior. China Accounting Review, (3), 301–316 (In Chinese). Ji, K., & Tong, Z. (2012). Earnings components, continuous difference and earnings forecast of financial analysts—Based on the empirical analysis of China’s A-share listed firms. Journal of Shanxi University of Finance and Economics, (2), 106–114 (In Chinese). Jiang, G., & Rao, P. (2011). Macroeconomic policy and micro-firm’s behavior—Expanding new fields of accounting and financing researches. Accounting Research, (3), 9–18 (In Chinese). Kahle, K.M., & Stulz, R.M. (2013). Access to capital, investment, and the financial crisis. Journal of Financial Economics, 110, 280–299. Khurana, I.K., Pereira, R., & Raman, K.K. (2003). Does analyst behavior explain market mispricing of foreign earnings for U.S. multinational firms? Journal of Accounting, Auditing and Finance, 453–477. Kinney, W. (1971). Predicting earnings: Entity vs. sub-entity data. Journal of Accounting Research, 9, 127–142. Li, D., Li, X., & Zhang, B. (2010). Earnings forecast errors of analysts and local advantages. Finance & Economics, (3), 26–33 (In Chinese). Li, L., Wu, L., Xu, Y., & Yu, M. (2004). Cross-industrial transformation of the firm: Features analysis and risk control—Taking the typical case of cross-industrial transformation of China’s listed firms as the example. Management World, (1), 118–129 (In Chinese). Lin, X., Ouyang, J., & Yue, H. (2007). Who are the attentiongetting finns for foreign securities analysts? Journal of Financial Research (in Chinese), 2007, 84–98. CHINA JOURNAL OF ACCOUNTING STUDIES 147 Liu, Z., & Zhang, J. (2009). The empirical analysis of the export determinants of China’s local manufacturing firms. Economic Research Journal, (8), 99–112 (In Chinese). Luo, C., & Li, S. (2014). Whether export can be helpful to relieve the financing restraints?—Empirical research based on research data about China’s firms of the world bank. Journal of Financial Research, (9), 1–17 (In Chinese). Maddala, G.S. (1983). Limited-dependent and qualitative variables in econometrics. Cambridge: Cambridge University Press. Malloy, C. (2005). The geography of equity analysis. Journal of Finance, 60, 719–755. Nichols, D., Tunnel, L., & Seipel, C. (1995). Earnings forecast accuracy and geographic segment disclosures. Journal of International Accounting, Auditing and Taxation, 8, 113–126. Orpurt, S. (2004). Local analyst earnings forecast advantages in Europe. University of Chicago. Unpublished working study Reeb, D., Kwok, C., & Baek, Y. (1998). Systematic risk in the multinational corporation. Journal of International Business Studies, 29, 263–279. Roberts, C. (1989). Forecasting Earnings Using Geographic Segment Data: Some U.K. Evidence. Journal of International Financial Management and Accounting I, 130–151. Rosenbaum, P.R., & Rubin, D.B. (1983). The central role of the propensity score in observational studies for causal effects. Biometrika, 70, 41–55. Sloan, R.G. (1996). Do stock prices fully reflect information in accruals and cash flows about future earnings? The Accounting Review, 71, 289–315. Su, D., & Zeng, H. (2009). Macroeconomic factors and changes in corporate capital structure. Economic Research Journal, (12), 52–65 (In Chinese). Wang, Y., Chen, X., & Hou, Y. (2010). The informational advantage of domestic security analysts: Geographical proximity or difference in accounting standard. Accounting Research, (12), 34–40 (In Chinese). Wei, Z., & Xue, J. (2015). Fair value accounting of financial assets and analyst forecasts. China Journal of Accounting Studies, 4, 294–319. Yu, Z., Tian, G., Qi, B., & Zhang, H. (2011). The corporate governance mechanism concerned by the media—Exploration based on the view of earnings management. Management World, (9), 127–140 (In Chinese). Yu, F. (2008). Analyst coverage and earnings management. Journal of Financial Economics, 88, 245–271. Yuan, Z., Zhang, L., & Yue, H. (2014). Does the cash flow forecast of analysts improve the accuracy of earnings forecast—Empirical evidence from China’s A-share market. Journal of Financial Research, (5), 162–177 (In Chinese). 148 F. XIN ET AL. Appendix A. The distribution of the sample (industry and year) Table A1-1 sets out the process of sample selection. Samples in the study come from all A-share listed firms from 2002 to 2012 totalling 11 years, and provides 19,306 observations after excluding small and medium firms board listed companies, We exclude companies issuing B-share simultaneously (consid - ering that the B-share might affect analysts’ forecast influenced by foreign investors) and companies that have been public for less than 2 years. We eventually have 7,897 firms’ annual observations as samples, due to missing data concerning the regional distribution of operating revenue disclosed by the firm, data forecast and tracked by the analysts and missing control variables. Tables A1-2 is the annual distribution of the samples. From the point of the proportion of analysts’ tracking, the sample as a whole conforms to the development trend of China’s analysts’ market. Table A1-1. t he process of sample selection. Firm/annual sample number The process of sample selection a-share non-financial listed firms/annual observation number: 2002–2012 23,357 Subtract: SMeS board listed companies 2,011 Companies issuing B-share simultaneously 1,187 Companies going public for less than 2 years 853 Subtotal 19,306 Subtract: observation of the missing of export data 2,999 observation of analysts’ data missing 10,788 observation of control variables data missing 621 f inal samples 7,897 CHINA JOURNAL OF ACCOUNTING STUDIES 149 Table A1-2. t he annual distribution of the samples. Year 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Samples in the study 119 202 284 433 570 621 829 978 1,154 1,274 1,433 number of a-share listed firms 1,224 1,287 1,377 1,381 1,434 1,550 1,625 1,718 2,063 2,342 2,494 Coverage of analysts’ forecast data 9.72% 15.70% 20.62% 31.35% 39.75% 40.06% 51.02% 56.93% 55.94% 54.40% 57.46% 150 F. XIN ET AL. Table A2. d escriptive statistics of export business. Panel A: Divided by the year Non-export Firm Export Firm Total Year N Percent N Percent N Percent 2002 85 1.99 34 0.94 119 1.51 2003 125 2.93 77 2.12 202 2.56 2004 161 3.77 123 3.39 284 3.60 2005 249 5.84 184 5.07 433 5.48 2006 327 7.67 243 6.69 570 7.22 2007 343 8.04 278 7.65 621 7.86 2008 451 10.57 378 10.41 829 10.50 2009 536 12.57 442 12.17 978 12.38 2010 611 14.33 543 14.95 1,154 14.61 2011 655 15.36 619 17.04 1,274 16.13 2012 722 16.93 711 19.58 1,433 18.15 t otal 4,265 100.00 3,632 100.00 7,897 100.00 Panel B: Divided by the industry Non-export Firm Export Firm Total Industry N Percent N Percent N Percent a 68 1.59 74 2.04 142 1.80 B 117 2.74 104 2.86 221 2.80 C 1,866 43.75 2,852 78.52 4,718 59.74 d 353 8.28 19 0.52 372 4.71 e 95 2.23 73 2.01 168 2.13 f 254 5.96 71 1.95 325 4.12 g 262 6.14 213 5.86 475 6.01 h 347 8.14 107 2.95 454 5.75 J 442 10.36 20 0.55 462 5.85 K 228 5.35 30 0.83 258 3.27 l 77 1.81 2 0.06 79 1.00 M 156 3.66 67 1.84 223 2.82 t otal 4,265 100.00 3,632 100.00 7,897 100.00 Panels A and B in Table A2 display the data distribution of the study. Panel A is the distribution by year. Samples in the study come from all A-share listed firms from 2002 to 2012 totalling 11 years. 7897 observations finally get reserved based on the data concerning the regional distribution of operation revenue disclosed by the firm and data forecast and tracked by the analysts. The industry of domestic analysts develops gradually after 2011 with a trend of year-by-year growth. Export firms accounts for 46% of the total sample changing from 28.5 to 49.6% during 2002 to 2012. Panel B is divided according to industry, among which a relatively evident trend is: the export ratio of agriculture (A) and Manufacture (C) are 52.1 and 60.4% respectively, surpassing 50%. Three industries with lowest export ratio are public utility (D), real estate (J) and (L) with the ratio of 5.1, 4.3 and 2.5% respectively.

Journal

China Journal of Accounting Studies – Taylor & Francis

Published: Jan 2, 2017

Keywords: Analysts forecast errors; China; exporting; financial crisis; foreign analysts; propensity score matching; selection model

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

Learn More →

Do firms’ exports affect analysts’ forecast errors?

Do firms’ exports affect analysts’ forecast errors?

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

Learn More →

Do firms’ exports affect analysts’ forecast errors?

Do firms’ exports affect analysts’ forecast errors?

References (50)

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies