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

Learn More →

Fertility rates and social media usage in sub‐Saharan Africa

Fertility rates and social media usage in sub‐Saharan Africa INTRODUCTIONFertility levels in sub‐Saharan Africa (SSA) are known to be influenced by exposure to traditional media, such as television and radio (Inkeles & Smith, 1974; La Ferrara et al., 2012; Lerner, 1958; Westoff & Bankole, 1997). Since 2000, rapid expansion of information and communication technology (ICT)—often referred to as the digital revolution—has caused a major shift in the availability and role of media. In 2020, 93% of Africans had access to a mobile phone, whereas this was only 53% in 2011; 30% of Africans were connected to the Internet in 2020 (World Bank, 2021a). A key characteristic of this shift is digital interactivity between media consumers, which is predominantly provided by social media. This raises the important question as to what extent social media usage is associated with fertility levels.Earlier studies have shown that women's education is the main determinant of SSA's fertility levels, acting through several mechanisms such as support for women's empowerment, facilitating their participation in the labour force and reducing child mortality (Atake & Gnakou Ali, 2019; Shapiro & Tenikue, 2017). However, another determinant is the diffusion of information, ideas and technology regarding, for example, contraceptive usage, fertility preferences and family size between individuals within communities (Bongaarts & Watkins, 1996; Coale & Watkins, 1986; Lesthaeghe, 1977). Individuals receive and interpret this information differently. The information may therefore gain new meaning in different social contexts and ultimately either promote or constrain fertility behaviour and outcomes. Schools are an important location for this diffusion (e.g., Atake & Gnakou Ali, 2019; Bongaarts, 2020; Shapiro & Tenikue, 2017), but it also takes place via traditional media (La Ferrara et al., 2012; Westoff & Koffman, 2011).Compared to traditional media, social media offers further opportunities for the diffusion of ideas. Social media transcends country borders more easily and thus promotes exposure to less context‐specific and more globalised cultural ideas, including those regarding fertility behaviour. Social media provides access to the ‘life of others’ without the need for physical proximity. Moreover, it enables its users not only to consume, but to contribute content and become members of an online community (Kietzmann et al., 2011). As such, social media provides individuals with the possibility of obtaining new information and interacting with role models beyond their own community.Earlier studies have suggested that the impact of digital media, such as social media, is likely to be greater in developing regions where other types of infrastructure are poor. Social media allows individuals living in remote areas contact with new ideas and individuals beyond their community borders, which a lack of infrastructure had previously made impossible (Aker & Mbiti, 2010; Rotondi et al., 2020). Studies have also shown that within these remote areas, the impact of the digital revolution is found to be greater for more marginalised groups, such as females who face greater barriers in accessing digital media (Rotondi et al., 2020; Varriale et al., 2022).Due to the difficulty of measuring social media access, previous research has predominantly relied on national estimates of official ICT statistics. These estimates potentially mask underlying heterogeneity within countries. The strength of the current study is that it uses new digital trace data for social media, in particular Facebook, at the subnational level to uncover these underlying differences. Digital trace data has its own advantages, such as the high speed of large‐scale data collection and the low costs involved (Cesare et al., 2018). To resolve the potential lack of accessibility, transparency and representativeness of this type of data, we enrich it with traditional survey data. This approach has already yielded promising results in previous studies with regard to data reliability (Eker et al., 2021; Kashyap et al., 2020; Rampazzo et al., 2021; Ribeiro et al., 2020).This study will use subnational digital trace data from Facebook to examine to what extent social media usage and gender gaps in usage explain subnational regional variations in fertility levels. For our analyses, we have combined Facebook data with household survey data for 311 subnational regions in 29 SSA countries. This enables us to study the relationship between fertility and social media usage in more detail than has been possible before.In the next section, we provide an overview of the relevant literature concerning the relationship between fertility behaviour and (social) media. The result section describes the spatial pattern in subnational variation in birth rates, social media usage and gender gaps in social media usage. Subsequently, the association between birth rates and social media usage will be tested using regression models.BACKGROUNDA history of media exposure and fertility behaviourThere is a long history of studies that have examined the relationship between media exposure and fertility behaviour. One of the first was by Lerner (1958), who reported that greater exposure to media (radio and film) made individuals more likely to express ‘modern’ and ‘pro‐development’ attitudes of various sorts in six countries in the Middle East. Some decades later, similar effects were found in studies focusing on SSA. Inkeles and Smith (1974) for example, found that watching television, listening to the radio and reading newspapers were good predictors for ‘modern’ norms and values, and this included the will to limit family size. Westoff and Bankole (1997) found a strong positive association between media use and fertility behaviour in six SSA countries (Burkina Faso, Ghana, Kenya, Madagascar, Namibia and Zambia). Included fertility behaviours were knowledge of modern contraceptive methods, intention to use contraceptives, current use of contraceptives and the desire to limit one's family size. An additional analysis by the same authors showed that women who were ‘heavy media consumers’ were more likely to use contraceptives and to limit their family size than women who were not. However, caution should be exercised when interpreting the causality of this relationship. According to the authors, there was a possibility that heavy media consumers were already on a different reproductive trajectory before the study started, which could have been the reason for their media consumption.Another study carried out in Brazil reported a causal link between media and fertility; here, Ferrara et al. (2012) examined the effect of exposure to soap operas on fertility rates. The main characters in most of the soap operas were living in urban settings with lifestyles that were in stark contrast to those of their rural audiences; they had small families, married later than average and had a high level of education. The study reported that the availability of a television signal, and thus exposure to soap operas, decreased the probability of giving birth to a child. This effect was strongest for women who had low socioeconomic status and whose lives were thus most in contrast with those of the soap opera protagonists.Media exposure, however, does not always result in a positive relationship with globalised norms and values. For example, media exposure at least once a week was associated with lower acceptance of domestic violence in only 14 out of 26 low‐ and middle‐income countries (Pierotti, 2013). Another study found a positive association between media access and the odds of tolerating domestic violence in Nigeria by women, but a negative association for men (Okenwa‐Emegwa et al., 2016). A similar finding was reported by Banerjee et al. (2019). They found that exposure to music television series was only associated with lower acceptance of domestic violence by men, but not women, in Nigeria. As such, media exposure can also be negatively associated with liberal and globalised fertility behaviours.From traditional to social mediaSince 2000, media has evolved from ‘old’ or ‘traditional’ media to ‘new’ media; this is often referred to as the ‘media revolution’ or ‘digital revolution’. Whereas traditional media predominantly involves listening to the radio, watching television and reading newspapers, new media is characterised by digitalisation of content. New media continuously evolves, and consequently its content changes frequently. Early on, new media was often considered as complementary to traditional media with individuals simply consuming additional online content; in other words, passively reading or watching (DiMaggio et al., 2001; Kietzmann et al., 2011).One study that focused on the relationship between new media and fertility research showed that women's internet exposure was positively related to the use of modern contraception in eight sub‐Saharan countries. The increase in use of modern contraception was greatest among poorly educated women (Toffolutti et al., 2020). Another study examined how access to mobile phones influenced fertility behaviour in Balaka, Malawi (Billari et al., 2020). This study found that the possession of a mobile phone was associated with a preference for smaller family size. Moreover, the authors concluded that phone ownership may accelerate the fertility transition in SSA.Social media has unique characteristics relative to digital media in general, such as interactivity among its users. To take a narrow view, social media involves person‐to‐person relationships on social networking services such as Twitter and Facebook. However, a broader interpretation of social media is generally accepted, in which social media is described as interactive platforms through which individuals and groups can share, cocreate, discuss and modify self‐created content (Kietzmann et al., 2011). As social interaction is known to be an important mechanism through which media influence fertility behaviour (Bongaarts & Watkins, 1996; Coale & Watkins, 1986; Ferrara et al., 2012; Lesthaeghe, 1977; Westoff & Koffman, 2011), the interactivity of social media is likely to influence ideas about fertility and fertility preferences. Online communities enable individuals to discuss ideas and information with others without the need for physical proximity, resulting in contact with persons living in areas where other fertility patterns prevail.One qualitative study examined the role of WhatsApp, a form of social media, among females living in a community in Nigeria (Abubakar & Dasuki, 2018). This study found that these women experienced greater autonomy and more opportunities to voice their personal opinions thanks to contact with other females in WhatsApp group chats. They were able to push the barriers of the social norms that restricted them to gendered roles within their community. As such, this study showed that individuals who have access to online communities may be less dependent on the societal norms and values of their immediate surroundings and more likely to have contact with globalised norms and values. Hence, even though online communities are often subject to self‐selection and homophily, they also provide users with the possibility of transcending the boundaries of their physical environment. As globalised norms tend to favour smaller family size than the average family size in SSA (Bongaarts, 2017), our first hypothesis is that social media usage is linked to lower birth rates.Inequalities in social media usageAnother aspect of new and social media is that it is not universally accessible. There are significant inequalities in access to ICT, which are often referred to as the ‘digital divide’ (DiMaggio et al., 2001; Norris, 2001). Known dimensions of the digital divide are level of development and gender. Access to new media in developing countries has been found to reflect broader structural inequalities, such as those in education, employment and income (Kashyap et al., 2020; Robinson et al., 2015). Moreover, studies have reported that women are underrepresented in online populations of Google and Facebook users in South Asia and in SSA (Fatehkia et al., 2018; Kashyap et al., 2020).The digital gender divide is also associated with cultural factors (Scheerder et al., 2017). Cultural norms in patriarchal contexts may hinder females' access to new media, especially when this access is mediated through males (Abu‐Shanab & Al‐Jamal, 2017; Gurumurthy & Chami, 2014). Other studies have shown that females' adoption of new and social media is related to their empowerment and societal well‐being, and to globalised and liberal values (Abubakar & Dasuki, 2018; Varriale et al., 2022). Hence, our second hypothesis states that when gender gaps in access to social media are smaller, fertility levels will be lower.DATAThis study examines to what extent social media usage and gender differences in social media usage are related to fertility levels in subnational regions of SSA countries. We combine data from Facebook's marketing application with development indicators computed on the basis of household surveys. The household surveys are derived from three sources: Demographic and Health Surveys (DHS, www.dhsprogram.com), UNICEF Multiple Indicator Cluster Surveys (MICS, mics.unicef.org) and Afrobarometer Surveys (www.afrobarometer.org). These three sources are publicly and freely accessible for academic purposes.Subnational regions are the unit of analysis in our study. Their selection is based on the geographic coding available in the household survey datasets and the regions used by Facebook's marketing application. The available regions in the household surveys were taken as the starting point for selecting regions used by Facebook's marketing application. We were able to make connections for 311 regions in 29 countries. An overview of the included countries, subnational regions and data sources is presented in Appendix A.Key fertility measureThe dependent variable is the crude birth rate (CBR) of the subnational regions in the year 2020. The CBR represents the number of yearly childbirths per 1000 individuals in a region. The considerable advantage of this over other fertility measures, such as the total fertility rate, is that it is a very simple measure that can be reliably computed over a short period of time. As social media use in the SSA region is expanding rapidly (Nanfuka, 2022), it is essential that the measurement of our dependent variable was close in time to our measurement of social media usage. Because the CBR might be influenced by the age structure of the female population, we included the percentage of fertile women in a region's total population, and the mean age of the fertile women in a region as extra control factors in our models.The subnational CBRs for 2020 were estimated by applying the subnational variation in CBRs derived from the most recent DHS and MICS surveys to the national CBRs for 2020 derived from the World Development Indicators (WDI, World Bank, 2021b). To do this, we followed a procedure derived from Kummu et al. (2018) and Smits and Permanyer (2019) in which subnational variation derived from household surveys was normalised around national values obtained. First, the national CBR from the World Bank in 2020 was divided by the national CBR of the most recent DHS or MICS survey to create a differentiation factor. Second, the subnational CBRs from the DHS/MICS surveys were multiplied with this factor to estimate the subnational CBRs for the year 2020. By doing this, we assume that the subnational CBRs follow the same pattern of change as the national values between the survey year and 2020.Because the size of the estimation error is expected to depend on the number of years over which the estimation was made, we used only countries for which the last survey was held after 2010, and included the survey year as control variable. Moreover, we added robustness analysis in which we restricted the sample to countries in which the survey was held after 2015 instead of 2010 (Appendix A). The results are similar in terms of direction and significance.Facebook indicatorsTo create subnational social media indicators, we used Facebook's advertisement platform. For marketing purposes, Facebook allows users with a Facebook account to search the number of active Facebook users in the last month disaggregated by various geographic and demographic characteristics (detailed information can be found at: https://developers.facebook.com/docs/marketing-apis). In this way, Facebook enables potential advertisers to use advanced targeting options by creating detailed user profiles, which include demographic characteristics such as age, gender and geographic location. For example, if you want to reach females aged 20‐30 in Benin, Facebook provides you with the number of active accounts that have met these criteria in the last month. Earlier studies have already used Facebook's advertisement platform for research purposes (Alexander et al., 2019; Stewart et al., 2019; Zagheni et al., 2017).We used this facility to create active user estimates for individuals aged 15−39 within subnational regions in SSA countries. To obtain the percentage of Facebook users we compared the number of users with the number of individuals aged 15−39. The population sizes for these individuals were based on data from the most recent DHS and MICS surveys and data from the WDI, using the same estimation procedure as for the subnational CBR values (discussed above). First, we computed a differentiation factor for individuals aged 15−39 between the national population sizes in the DHS/MICS surveys and those in the 2020 WDI data. Thereafter, this differentiation factor was applied to the subnational population data in the DHS/MICS surveys to obtain the 2020 subnational population estimates.In 19 subnational regions, Facebook usage exceeded 100%. As Facebook is not transparent in how it calculates audience size, we decided to set the maximum percentage of Facebook usage at 100%. For robustness, we also ran the analysis without these subnational regions, with a maximum of 75% instead of 100 and with subnational Facebook deciles instead of percentages. As the results did not significantly differ in terms of direction and significance, we only present those that included the regions with a Facebook percentage of up to 100%.Besides the percentage of Facebook usage, we also computed the gender gap in Facebook usage in these subnational regions. First, we calculated Facebook usage for males and females separately using the technique described above. Second, we subtracted subnational female Facebook usage from male Facebook usage. As such, if the gender gap in Facebook usage is zero, this means an equal number of male and female users. If the gender gap is positive, males are overrepresented; if it is negative, females are overrepresented.Control factorsControl factors in our study were the level of development, gender inequality, phone ownership, the percentage of females aged 15−49 and the age structure of fertile women in the regions. In more developed and wealthier regions, digital access is better and fertility rates are known to be lower (Pezzulo et al., 2021). As such, regions' development should be controlled for. Lower birth rates were also expected for regions where women hold a stronger position, where there are higher levels of mobile phone ownership, where a lower percentage of females are in their fertile years (ages 15−49), and where the fertile women are on average older (Casterline, 2017; Rotondi et al., 2020).The level of development was measured by the subnational Human Development Index (sHDI) (Smits & Permanyer, 2019). The sHDI shows within‐country variation in human development across the globe and, like the Human Development Index, is an indicator of a combination of three pillars: education, health and standard of living. The sHDI is measured with a value between 0 and 1; a higher score denotes higher development. Gender inequality was measured by the subnational gender development index (sGDI), which is defined as the ratio between female and male levels of sHDI in the region (Smits & Permanyer, 2020). A sGDI of 1 indicates perfect equality between females and males; a sGDI below 1 indicates that males have a higher level of human development and a sGDI above 1 indicates that females have a higher level of human development. Phone ownership was measured by the percentage of households in the region in which at least one member owns a phone. The data was derived from the DHS, MICS and Afrobarometer surveys. Afrobarometer data on phone ownership for 2020−2021 (round 8) was used if it was more recent than the DHS or MICS data. This was the case for Benin, Cameroon, Cote d'Ivoire, Estwatini, Ethiopia, Gabon, Gambia, Ghana, Guinea, Kenya, Lesotho, Liberia, Mali, Mozambique (round 7, 2017), Namibia, Niger, Nigeria, Togo. The percentage of females aged 15−49, which is considered females' fertile period, was calculated as the percentage of females aged 15−49 of the total population of the region. The population structure of fertile females was included by calculating the mean age of females aged 15−49.METHODSExploratory spatial data analysisWe started the analyses with an exploratory spatial data analysis of the subnational CBRs, Facebook usage and gender gaps in Facebook usage. We examined whether spatial clustering exists among those three variables and to what extent there is overlap between clusters of birth rates and those of Facebook usage. Clusters refer to agglomerations of bordering subnational regions that share similar characteristics—for example, adjacent regions that both show high (or low) CBRs or high (or low) Facebook usage.First, we assessed the degree of global spatial autocorrelation with the Global Moran's I statistic for birth rates and Facebook usage (Gittleman & Kot, 1990). Moran's I measures whether a spatial pattern exists. Global Moran's I values closer to 0 indicate a random spatial pattern, positive values indicate clustering and negative values dispersion. Second, we calculated the Gi* statistic based on local Getis‐Ord Gi* hot spot analyses to examine whether and where clusters in birth rates and Facebook usage were situated (Ord & Getis, 1995). The Gi* statistic identified areas where relatively high or low values of certain characteristics, in this case birth rates and Facebook usage, tend to cluster together based on the neighbouring values. We compared the overlap in clusters of birth rates and Facebook usage and examined the characteristics of low and high birth rate clusters.Regression analysisTo test whether the association between birth rates and Facebook usage is still significant after the inclusion of subnational characteristics, we used regression analysis. As the subnational regions are nested within countries, our data has a hierarchical structure. To control for this, two types of regression models were estimated: a mixed effects regression model that addresses the hierarchical structure by including a random intercept at the national level (Snijders & Bosker, 2011) and a fixed effects regression model in which the hierarchical structure is addressed by including fixed effects dummies at the country level (Allison, 2009). The fixed effects approach uses much more degrees of freedom than the multilevel approach, but has the advantage that all direct effects of (known and unknown) confounders at the national level are controlled for (Bell et al., 2019).RESULTSGeographic patterns of fertility behaviour and social media usageOur analysis starts with exploring the geographic pattern of subnational CBRs in 2020 (Figure 1). The mean subnational CBR was 34 births per 1000 individuals in 2020, with a minimum of 17 births in North West, South Africa and a maximum of 55 births in Lunda Sul, Angola. Mean Facebook usage was 25%, with a minimum of 1% in Other Central, Malawi and a maximum of 100% in 19 regions (darkest colour in Figure 1b). The gender gap in Facebook usage was on average 8%, meaning that females are on average less connected than males. Females were relatively best represented on Facebook in Eastern Cape, South Africa (8% more than males), and males were relatively the dominant users in Maritime, Togo (59% more than females). Figure 1 shows substantial variation across subnational regions in the distribution of CBRs, Facebook usage and the gender gap in Facebook usage. The descriptive statistics can be found in Table 1.1FigureGeographic distributions of (a) crude birth rates 2020, (b) Facebook usage 2021, and (c) the gender gap in Facebook usage between males and females 20211TableDescriptive statistics: Means and percentages of variables included in this studyVariablesMean, %SDMin.Max.Subnational Crude Birth Rates 2020 (dep. var.)34.007.0016.7755.32Subnational Characteristics% Facebook24.5628.420.55100Gender gap Facebook8.189.88−7.8458.55sHDI0.530.090.270.75sGDI0.890.080.661.07% Phone77.7418.8522.5099.02% Females 15−4923.375.1516.5850.57Mean age females 15−4928.620.9825.5331.39Year of survey20162.4020112020Note: N regions = 311, N countries = 29.Abbreviations: SD, standard deviation; sGDI, subnational gender development index; sHDI, subnational Human Development Index.We tested whether the variability in these distributions showed a spatial pattern and found a Moran's I of 0.09 (standard deviation [SD] = 0.005) for birth rates, of 0.07 (SD = 0.007) for Facebook usage, and of 0.11 (SD = 0.007) for gender gaps in Facebook usage. This indicates that the variability in birth rate and Facebook usage is not random and that clustering exists, in that neighbouring subnational regions tend to have similar characteristics and form spatial clusters. Figure 2 presents the Gi* statistics, which show where clusters in CBRs, Facebook usage and gender gaps in Facebook usage are situated. Clusters with Gi* statistics below −1.96 are considered significantly low (or small) clusters, and those with Gi* statistics above 1.96 are considered significantly high (or large) clusters. The mean CBR in low birth rate clusters was 25 births per 1000 individuals (SD = 4.8), and in high birth rate clusters was 44 births per 1000 individuals (SD = 4.5). The average Facebook usage of low usage clusters was 2.7% (SD = 4.2) and of high usage clusters 48.4 per cent (SD = 20.3); the average gender gap in small gender gap clusters was 0.2% (SD = 4.2) and in large gender gap clusters 21.0% (SD = 11.5) (Appendix A).2FigureSpatial clusters of (a) crude birth rates 2020, (b) Facebook usage 2021, and (c) the gender gap in Facebook usage 2021Visual inspection of Figure 2 shows that clusters with low birth rates tend to overlap with clusters of high Facebook usage, but that this is less so for small gender gap clusters. Low birth rate clusters overlap with clusters characterised by high Facebook usage in Southern SSA (Namibia, South Africa and Swaziland). High birth rate clusters overlap with some clusters with large gender gaps, but not so much with clusters with high Facebook usage. Here, we see that high birth rate clusters and large gender gaps in Facebook usage predominantly overlap in some parts of Western SSA (Mali and Nigeria).Further examination of the relationship between the clustering of subnational CBRs and Facebook indicators reveals consistent patterns. Figure 3a shows that subnational Facebook usage is on average higher in low birth rate clusters. However, there is no clear difference between Facebook usage in the nonsignificant and high birth rate clusters. These observations seem to confirm the visual overlap in Figure 2, which showed that high Facebook usage is associated with low CBRs, but low Facebook usage is not necessarily associated with high CBRs. Figure 3b shows that the gender gap in Facebook usage is smaller in low birth rate clusters. The median of the gender gap is slightly above zero in low birth rate clusters, meaning that Facebook usage of males and females is (almost) equal in these clusters. On the other hand, the median of the gender gap in high birth rate clusters is 8, which shows that around 8% fewer females than males use Facebook in high birth rate clusters. As such, the gender gap in Facebook usage seems to be associated with high birth rate clusters.3FigureBoxplots showing the minimum (lower end line), maximum (upper end line), median (middle line of box), and fist and third quartiles (lower and upper ends of box) of low, nonsignificant and high birth rate clusters of (a) Facebook usage 2021, (b) the gender gap in Facebook usage 2021, (c) the subnational Human Development Index 2020, (d) the subnational Gender Development Index 2020, (e) phone ownership, (f) the percentage of females aged 15−49, and (g) the mean age of females aged 15−49 for crude birth rates in low, nonsignificant and high birth rate clusters.Similar patterns are observed between the CBR clusters and the control factors: sHDI, sGDI, phone ownership, the percentage of females aged 15−49 and the mean age of females aged 15−49. The sHDI, sGDI, phone ownership, percentage of females aged 15−49 and the mean age of females aged 15‐49 are higher in low birth rate clusters. This shows that low birth rate clusters have on average higher levels of development, levels of gender development, percentages of phone ownership and percentages of females aged 15−49.Regression resultsTable 2 shows the results of the regression models that test the association between birth rates and Facebook usage. The mixed effects model and the fixed effects model show comparable results in terms of coefficients, standard errors and significance.2TableRegression estimates of the determinants of subnational Crude Birth Rates (2020) for the multilevel linear regression (MLR) and Fixed Effects (FE) regression modelsMLR IFE I% Facebook−0.030* (0.014)−0.035* (0.015)Gender gap Facebook0.062* (0.031)0.067* (0.032)sHDI−21.852*** (6.357)−24.025** (7.711)sGDI−31.128*** (7.509)−33.947** (10.611)% Phone−0.064* (0.025)−0.073* (0.03)% Females 15−490.173 (0.11)0.364* (0.162)Mean age females 15−49−0.934** (0.35)−0.67 (0.379)Year0.071 (0.265)Country dummies includedYesConstant33.529*** (0.636)41.317*** (1.396)Country‐level variance9.62 (3.10)Adjusted R²0.657Note: Independent variables are centred around their mean. Standard errors are shown in parentheses. Analyses are run for 311 subnational regions in 29 countries.Abbreviations: sGDI, subnational gender development index; sHDI, subnational Human Development Index.*p < 0.05;**p < 0.01;***p < 0.001.Our analysis reveals significant associations of Facebook usage and gender gaps in Facebook usage with subnational birth rates. Facebook usage is negatively associated with birth rates, which means that fertility is lower in areas with higher Facebook usage. This finding is in line with our first hypothesis.Regarding the gender gap in social media usage, we find a positive association between the gender gap in Facebook usage and birth rates. Hence in areas where females are relatively underrepresented on social media, fertility levels are higher. This finding is in line with our second hypothesis. We expected that females who have relatively less access to Facebook to be less empowered and consequently, less likely to have contact with globalised ideas and to change their behaviour accordingly.All associations are controlled for the sHDI, sGDI, subnational phone ownership, percentage of females aged 15−49, mean age of females aged 15−49 and survey year. The association between birth rates and the sHDI is significantly negative, meaning that when subnational regions have a higher level of development, CBRs are lower. Birth rates are significantly higher in regions with higher gender inequality in favour of males. Subnational phone ownership shows a significant negative association with subnational birth rates, which means that higher phone ownership is related to lower birth rates. The percentage of females aged 15−49 is positively associated with birth rates and the mean age of women in this age group negatively. Hence fertility is higher if fertile females comprise a larger share of the population and if their mean age is lower. The year of the survey is not significantly related to birth rates. All these findings for the control variables are in line with our expectations and with earlier findings (Atake & Gnakou Ali, 2019; Billari et al., 2020; Casterline, 2017; Pezzulo et al., 2021; Rotondi et al., 2020).DISCUSSIONWhere previous studies have firmly established a connection between fertility behaviour and traditional media such as television and radio (Inkeles & Smith, 1974; La Ferrara et al., 2012; Lerner, 1958; Westoff & Bankole, 1997), this study examines the link between fertility behaviour and social media. Based on data for 311 subnational regions in 29 countries in SSA, we tested two hypotheses regarding the relationship between social media and fertility outcomes: the idea that access to social media is associated with lower fertility rates (hypothesis 1) and that gender equality in social media use is associated with lower fertility rates (hypothesis 2). Our findings are in line with these hypotheses. The descriptive analysis identified spatial overlap between low birth rate clusters and high social media usage clusters, and between high birth rate clusters and clusters with a large gender gap in social media use. The regression analyses revealed that higher social media usage and smaller gender gaps in social media usage are significantly linked to lower birth rates. We therefore conclude that on the basis of our findings, the hypotheses cannot be rejected and that our study provides new empirical evidence regarding the association between social media usage and fertility decline.This study is important, because it is the first that shows the existence of a substantial relationship between Facebook usage and fertility at the level of subnational regions in SSA. Although some studies have already reported the effects of mobile phone ownership and internet access on fertility behaviour (Billari et al., 2020; Toffolutti et al., 2020), we have gone a step further by identifying similar effects of social media exposure and in particular, access to Facebook. One possible explanation for this finding is that social media enables its users to form (online) communities and to share ideas within them. Interactions between individuals on social media rely less on physical proximity, and as such individuals are likely to encounter more globalised ideas and norms, and to change their behaviour accordingly. As the discussion of new ideas and norms is a central concept within diffusion theories (Coale & Watkins, 1986; Westoff & Koffman, 2011), the findings of this study could well be interpreted within this line of theory.Our results regarding the gender gap in social media usage reconfirm the finding of earlier studies: that women are underrepresented in the online population (Fatehkia et al., 2018; Kashyap et al., 2020; Robinson et al., 2015). That better representation is associated with lower fertility outcomes is also in line with earlier studies, which have found that females adopting new and social media tend to have more modern and liberal values (Abubakar & Dasuki, 2018; Varriale et al., 2022). As such, our study also points towards negative consequences of large gender inequalities in media access.A possible limitation of the present study is that there are potential reliability issues with the social media data. Apparent Facebook usage of above 100 per cent in 19 subnational regions is an example of this. It might be that individuals in those regions have more than one Facebook account or that there are a substantial number of fake accounts. According to Meta (2019), on average about 5% of all Facebook accounts are fake. However, this could also be the result of errors in Facebook's target audience generator. The consequences could be twofold. If overestimation is systematic, it will not substantially influence the direction and significance of the regression estimates. If overestimation is random, this suggests measurement error in our data, which probably implies an underestimation of the real effects (Hutcheon et al., 2010). A second potential limitation is the possibility of selection bias in our data. It is possible that the regions with higher social media usage are also those with more globalised values and lower fertility rates. If so, this could lead to an overestimation of the association between social media access and fertility. More research is required to examine this possibility. A third limitation is that we have estimated the subnational CBRs by applying subnational variation derived from earlier surveys to national data derived from the World Bank. This approach is based on the assumption that subnational variation is relatively stable over time. To be sure, we added the year of the survey as an additional control factor and performed a robustness test by repeating our analysis with a sample restricted to countries with surveys from 2015. Findings turned out to be substantially the same as in the main analysis.In sum, this study uses an innovative approach to provide new evidence that social media usage and gender gaps in usage are significantly related to SSA fertility outcomes. This might mean that the social media revolution currently spreading across the subcontinent might profoundly affect its fertility transition. Compared with earlier studies, our research constitutes a major improvement because it uses—for the first time—social media usage data at subnational level as a potential ingredient of fertility decline in the SSA region.ACKNOWLEDGEMENTThe research team would like to thank Siem Smeets for his data work as student assistant. We would also like to thank the reviewers for their time and valuable feedback to improve this article.DATA AVAILABILITY STATEMENTThe data that support the findings of this study are available from DHS, Mics, Afrobarometer. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the author(s) with the permission of DHS, Mics, Afrobarometer.AAPPENDIXOverview of included regions, data sources (Demographic and Health survey [DHS] or Multiple Indicator Cluster [MICS] survey), year of data sources and data of accessing Facebook's target audience estimates of this study.CountryRegionData sourceYearDate of FacebookAngolaCabindaDHS20165/24/2021AngolaZairaDHS20165/24/2021AngolaUigeDHS20165/24/2021AngolaLuandaDHS20165/24/2021AngolaKuanza NorteDHS20165/24/2021AngolaKuanza SulDHS20165/24/2021AngolaMalangeDHS20165/24/2021AngolaLanda NorteDHS20165/24/2021AngolaBenguelaDHS20165/24/2021AngolaHuamboDHS20165/24/2021AngolaBieDHS20165/24/2021AngolaMoxicoDHS20165/24/2021AngolaKuanda KubangoDHS20165/24/2021AngolaNamibeDHS20165/24/2021AngolaHuilaDHS20165/24/2021AngolaCunene Lunda SulDHS20165/24/2021AngolaBengoDHS20165/24/2021BeninAtacora + DongaDHS20183/7/2021BeninAtlantique + LittorialDHS20183/7/2021BeninBorgou + AliboriDHS20183/7/2021BeninMono + KouffoDHS20183/7/2021BeninOueme + PlateauDHS20183/7/2021BeninZou + CollineDHS20183/7/2021BurundiNorth (Kayanza, Kirundo, Muyinga, Ngozi)DHS20175/2/2021BurundiSouth (Bururi, Makamba)DHS20175/2/2021BurundiEast (Cankuzo, Rutana, Ruyigi)DHS20175/2/2021BurundiWest (Bubanza, Buja Rural, Cibitoke, Mairie de Bujumbura)DHS20175/2/2021BurundiMiddle (Gitega, Karuzi, Maramvya, Mwaro)DHS20175/2/2021CameroonAdamaouaDHS20183/8/2021CameroonCentre + YaoundeDHS20183/8/2021CameroonEstDHS20183/8/2021CameroonExtreme NordDHS20183/8/2021CameroonLittoral + DoualaDHS20183/8/2021CameroonNordDHS20183/8/2021CameroonNord OuestDHS20183/8/2021CameroonOuestDHS20183/8/2021CameroonSudDHS20183/8/2021CameroonSud OuestDHS20183/8/2021ChadZone 1 (N'Djamena)MICS20195/2/2021ChadZone 2 (Borkou, Ennedi, Tibesti, Kanem, Lac)MICS20195/2/2021ChadZone 3 (Guera, Batha Est & Ouest, Salamat)MICS20195/2/2021ChadZone 4 (Ouaddai, Bilitine)MICS20195/2/2021ChadZone 5 (Chari‐Baguimi)MICS20195/2/2021ChadZone 6 (Mayo‐Kebbi)MICS20195/2/2021ChadZone 7 (Logone Occidental, Logone Oriental, Tandjilie Est & Ouest)MICS20195/2/2021ChadZone 8 (Moyen‐Chari)MICS20195/2/2021Congo BrazzavilleKouilouMICS20155/24/2021Congo BrazzavilleNiariMICS20155/24/2021Congo BrazzavilleLekoumouMICS20155/24/2021Congo BrazzavilleBouenzaMICS20155/24/2021Congo BrazzavilleBrazzaville + PoolMICS20155/24/2021Congo BrazzavillePlateauxMICS20155/24/2021Congo BrazzavilleCuvetteMICS20155/24/2021Congo BrazzavilleCuvette OuestMICS20155/24/2021Congo BrazzavilleSanghaMICS20155/24/2021Congo BrazzavilleLikoualaMICS20155/24/2021Congo BrazzavillePointe‐NoireMICS20155/24/2021Cote d'IvoireCentre (Lacs district + Yamoussoukro)MICS20165/3/2021Cote d'IvoireCentre Nord (Valée du Bandama District)MICS20165/3/2021Cote d'IvoireNord (Savanes district)MICS20165/3/2021Cote d'IvoireNord Est (Zanzan)MICS20165/3/2021Cote d'IvoireNord Ouest (Denguélé + Woroba district)MICS20165/3/2021Cote d'IvoireOuest (Montagne district)MICS20165/3/2021Cote d'IvoireSud Ouest (Bas‐Sassandra)MICS20165/3/2021Cote d'IvoireSud, Abidjan, Centre‐Ouest, Centre EstMICS20165/8/2021SwazilandHhohhoMICS20146/28/2021SwazilandManziniMICS20146/28/2021SwazilandShiselweniMICS20146/28/2021SwazilandLubomboMICS20146/28/2021EthiopiaTigrayDHS20164/26/2021EthiopiaAffarDHS20164/26/2021EthiopiaAmharaDHS20164/26/2021EthiopiaOromiaDHS20164/26/2021EthiopiaSomaliDHS20164/26/2021EthiopiaBen‐GumzDHS20164/26/2021EthiopiaYeDebubDHS20164/26/2021EthiopiaGambelaDHS20164/26/2021EthiopiaHarariDHS20164/26/2021EthiopiaAddisDHS20164/26/2021EthiopiaDire DawaDHS20164/26/2021GabonEstuaire + LibrevilleDHS20125/25/2021GabonHaut‐OgooueDHS20125/25/2021GabonMoyen‐OgooueDHS20125/25/2021GabonNgounieDHS20125/25/2021GabonNyangaDHS20125/25/2021GabonOgooue IvindoDHS20125/25/2021GabonOgooue LoloDHS20125/25/2021GabonOgooue MaritimeDHS20125/25/2021GabonWoleu NtemDHS20125/25/2021Gambia, TheBanjul + KanifingMICS20185/3/2021Gambia, TheBrikamaMICS20185/3/2021Gambia, TheMansakonkoMICS20185/3/2021Gambia, TheKerewanMICS20185/3/2021Gambia, TheKuntaurMICS20185/3/2021Gambia, TheJanjanburehMICS20185/3/2021Gambia, TheBasseMICS20185/3/2021GhanaWesternMICS20175/27/2021GhanaCentralMICS20175/27/2021GhanaGreater AccraMICS20175/27/2021GhanaVoltaMICS20175/27/2021GhanaEasternMICS20175/27/2021GhanaAshantiMICS20175/27/2021GhanaBrong AhafoMICS20175/27/2021GhanaNorthernMICS20175/27/2021GhanaUpper WestMICS20175/27/2021GhanaUpper EastMICS20175/27/2021GuineaBokeDHS20184/21/2021GuineaConakryDHS20184/21/2021GuineaFaranahDHS20184/21/2021GuineaKankanDHS20184/21/2021GuineaKindiaDHS20184/21/2021GuineaLabeDHS20184/21/2021GuineaMamouDHS20184/21/2021GuineaNzerekoreDHS20184/21/2021KenyaNairobiDHS20146/3/2021KenyaCentralDHS20146/3/2021KenyaCoastDHS20146/3/2021KenyaEasternDHS20146/3/2021KenyaNyanzaDHS20146/3/2021KenyaRift VelleyDHS20146/3/2021KenyaWesternDHS20146/3/2021KenyaNorth‐EasthernDHS20146/3/2021LesothoButha ButheMICS20184/30/2021LesothoLeribeMICS20184/30/2021LesothoBereaMICS20184/30/2021LesothoMeseruMICS20184/30/2021LesothoMafetengMICS20184/30/2021LesothoMohale's HoekMICS20184/30/2021LesothoQuthingMICS20184/30/2021LesothoQacha's NekMICS20184/30/2021LesothoMokhotlongMICS20184/30/2021LesothoThaba‐TsekaMICS20184/30/2021LiberiaBomiDHS20204/11/2021LiberiaBongDHS20204/11/2021LiberiaGbarpoluDHS20204/11/2021LiberiaGrand BassaDHS20204/11/2021LiberiaGrand Cape MountDHS20204/11/2021LiberiaGrand GedehDHS20204/11/2021LiberiaGrand KruDHS20204/11/2021LiberiaLofaDHS20204/11/2021LiberiaMargibiDHS20204/11/2021LiberiaMarylandDHS20204/11/2021LiberiaMontserradoDHS20204/11/2021LiberiaNimbaDHS20204/11/2021LiberiaRivercessDHS20204/11/2021LiberiaRiver GeeDHS20204/11/2021LiberiaSinoeDHS20204/11/2021MadagascarAnalamangaMICS20184/6/2021MadagascarVakinankaratraMICS20184/6/2021MadagascarItasyMICS20184/6/2021MadagascarBongolavaMICS20184/6/2021MadagascarHaute MatsiatraMICS20184/6/2021MadagascarAmron'i ManiaMICS20184/6/2021MadagascarVatovavy‐FitovinanyMICS20184/6/2021MadagascarIhorombeMICS20184/6/2021MadagascarAtsimo‐AtsinananaMICS20184/6/2021MadagascarAtsinananaMICS20184/6/2021MadagascarAnalanjirofoMICS20184/6/2021MadagascarAlaotra MangoroMICS20184/6/2021MadagascarBoenyMICS20184/6/2021MadagascarSofieMICS20184/6/2021MadagascarBetsiebokaMICS20184/6/2021MadagascarMelakyMICS20184/6/2021MadagascarAtsimo‐AndrefanaMICS20184/6/2021MadagascarAndroyMICS20184/6/2021MadagascarAnosyMICS20184/6/2021MadagascarMenableMICS20184/6/2021MadagascarDianaMICS20184/6/2021MadagascarSavaMICS20184/6/2021MalawiBlantyreMICS20204/29/2021MalawiKasunguMICS20204/29/2021MalawiMachingaMICS20204/29/2021MalawiMangochiMICS20204/29/2021MalawiMzaimbaMICS20204/29/2021MalawiSalimaMICS20204/29/2021MalawiThyoloMICS20204/29/2021MalawiZombaMICS20204/29/2021MalawiLilongweMICS20204/29/2021MalawiMulanjeMICS20204/29/2021MalawiOther NorthernMICS20204/29/2021MalawiOther CentralMICS20204/29/2021MalawiOther SouthernMICS20204/29/2021MaliKayesDHS20186/4/2021MaliKoulikoroDHS20186/4/2021MaliSikassoDHS20186/4/2021MaliSegouDHS20186/4/2021MaliMoptiDHS20186/4/2021MaliTombouctoeDHS20186/4/2021MaliGao and KidalDHS20186/4/2021MaliBamakoDHS20186/4/2021MauritaniaHodh CharghiMICS20154/21/2021MauritaniaAssabaMICS20154/21/2021MauritaniaGorgolMICS20154/21/2021MauritaniaBraknaMICS20154/21/2021MauritaniaTrarza incl. NouakchottMICS20154/21/2021MauritaniaAdrarMICS20154/21/2021MauritaniaNouadhibouMICS20154/21/2021MauritaniaTagantMICS20154/21/2021MauritaniaGuidimaghaMICS20154/21/2021MauritaniaTiris ZemmourMICS20154/21/2021MauritaniaInchiriMICS20154/21/2021MozambiqueNiassaDHS20114/13/2021MozambiqueCabo DelgadoDHS20114/13/2021MozambiqueNampulaDHS20114/13/2021MozambiqueZembeziaDHS20114/13/2021MozambiqueTeteDHS20114/13/2021MozambiqueManicaDHS20114/13/2021MozambiqueSofalaDHS20114/13/2021MozambiqueInhambaneDHS20114/13/2021MozambiqueGazaDHS20114/13/2021MozambiqueMaputoDHS20114/13/2021MozambiqueMaputo CidadeDHS20114/13/2021NamibiaCapriviDHS20136/14/2021NamibiaErongoDHS20136/14/2021NamibiaHardapDHS20136/14/2021NamibiaKarasDHS20136/14/2021NamibiaKhomasDHS20136/14/2021NamibiaKuneneDHS20136/14/2021NamibiaOhangwenaDHS20136/14/2021NamibiaKavangoDHS20136/14/2021NamibiaOmahekeDHS20136/14/2021NamibiaOmusatiDHS20136/14/2021NamibiaOshanaDHS20136/14/2021NamibiaOshikotoDHS20136/14/2021NamibiaOtjozondjupaDHS20136/14/2021NigerAgadezDHS20125/2/2021NigerDiffaDHS20125/2/2021NigerDossoDHS20125/2/2021NigerMaradiDHS20125/2/2021NigerTillaberi (incl. Niamey)DHS20125/2/2021NigerTahouaDHS20125/2/2021NigerZinderDHS20125/2/2021NigeriaAkwa IbomDHS20186/6/2021NigeriaAnambraDHS20186/6/2021NigeriaBauchiDHS20186/6/2021NigeriaEdoDHS20186/6/2021NigeriaBenueDHS20186/6/2021NigeriaBomoDHS20186/6/2021NigeriaCross RiverDHS20186/6/2021NigeriaAdamawaDHS20186/6/2021NigeriaImoDHS20186/6/2021NigeriaKadunaDHS20186/6/2021NigeriaKanoDHS20186/6/2021NigeriaKatsinaDHS20186/6/2021NigeriaKwaraDHS20186/6/2021NigeriaLagosDHS20186/6/2021NigeriaNigerDHS20186/6/2021NigeriaOgunDHS20186/6/2021NigeriaOndoDHS20186/6/2021NigeriaOyoDHS20186/6/2021NigeriaPlateauDHS20186/6/2021NigeriaRiversDHS20186/6/2021NigeriaSokotoDHS20186/6/2021NigeriaAbiaDHS20186/6/2021NigeriaDeltaDHS20186/6/2021NigeriaEnuguDHS20186/6/2021NigeriaJigawaDHS20186/6/2021NigeriaKebbiKogiDHS20186/6/2021NigeriaOsunDHS20186/6/2021NigeriaTarabaDHS20186/6/2021NigeriaYobeDHS20186/6/2021NigeriaBayelsaDHS20186/6/2021NigeriaEbonyiDHS20186/6/2021NigeriaEkitiDHS20186/6/2021NigeriaGombeDHS20186/6/2021NigeriaNassarawaDHS20186/6/2021NigeriaZamfaraDHS20186/6/2021NigeriaAbuja FCTDHS20186/6/2021South AfricaWestern CapeDHS20166/28/2021South AfricaEastern CapeDHS20166/28/2021South AfricaNorthern CapeDHS20166/28/2021South AfricaFree StateDHS20166/28/2021South AfricaKwaZulu NatalDHS20166/28/2021South AfricaNorth WestDHS20166/28/2021South AfricaGautengDHS20166/28/2021South AfricaMpumalangaDHS20166/28/2021South AfricaNorthern ProvinceDHS20166/28/2021TogoMaritime (inclusive Lome)MICS20176/28/2021TogoPlateauxMICS20176/28/2021TogoCentraleMICS20176/28/2021TogoKaraMICS20176/28/2021TogoSavanesMICS20176/28/2021UgandaCentral SouthDHS20165/4/2021UgandaCentral NorthDHS20165/4/2021UgandaKampalaDHS20165/4/2021UgandaEast CentralDHS20165/4/2021UgandaEasternDHS20165/4/2021UgandaNorthDHS20165/4/2021UgandaWest NileDHS20165/4/2021UgandaWesternDHS20165/4/2021UgandaSouthwestDHS20165/4/2021ZambiaCentralDHS20184/3/2021ZambiaCopperbeltDHS20184/3/2021ZambiaEastern + Northern + MuchingaDHS20184/3/2021ZambiaLuapulaDHS20184/3/2021ZambiaLusakaDHS20184/3/2021ZambiaNorth‐WesternDHS20184/3/2021ZambiaSouthernDHS20184/3/2021ZambiaWesternDHS20184/3/2021ZimbabweManicalandMICS20195/2/2021ZimbabweMashonaland CentralMICS20195/2/2021ZimbabweMashonaland EastMICS20195/2/2021ZimbabweMashonaland WestMICS20195/2/2021ZimbabweMashonaland NorthMICS20195/2/2021ZimbabweMashonaland SouthMICS20195/2/2021ZimbabweMidlandsMICS20195/2/2021ZimbabweMasvingoMICS20195/2/2021ZimbabweHarareMICS20195/2/2021ZimbabweBulawayoMICS20195/2/2021REFERENCESAbubakar, N. H., & Dasuki, S. I. (2018). Empowerment in their hands: Use of WhatsApp by women in Nigeria. Gender, Technology and Development, 22(2), 164–183. https://doi.org/10.1080/09718524.2018.1509490Abu‐Shanab, E., & Al‐Jamal, N. (2017). Exploring the gender digital divide in Jordan. Gender, Technology and Development, 19(1), 91–113. https://doi.org/10.1177/0971852414563201Aker, J. C., & Mbiti, I. M. (2010). Mobile phones and economic development in Africa. Journal of Economic Perspectives, 24(3), 207–232. https://doi.org/10.1257/jep.24.3.207Alexander, M., Polimis, K., & Zagheni, E. (2019). The impact of hurricane maria on out‐migration from Puerto Rico: Evidence from facebook data. Population and Development Review, 45(3), 617–630. https://doi.org/10.1111/padr.12289Allison, P. D. (2009). Fixed Effects Regression Models. SAGE Publications, Inc.Atake, E. H., & Gnakou Ali, P. (2019). Women's empowerment and fertility preferences in high fertility countries in Sub‐Saharan Africa. BMC Women's Health, 19(1), 54. https://doi.org/10.1186/s12905-019-0747-9Banerjee, A., La Ferrara, E., & Orozco, V. (2019). Entertainment, education, and attitudes toward domestic violence. AEA Papers and Proceedings, 109, 133–137. https://doi.org/10.1257/PANDP.20191073Bell, A., Fairbrother, M., & Jones, K. (2019). Fixed and random effects models: Making an informed choice. Quality & Quantity, 53(2), 1051–1074. https://doi.org/10.1007/S11135-018-0802-X/FIGURES/2Billari, F., Rotondi, V., & Trinitapoli, J. (2020). Mobile phones, digital inequality, and fertility: Longitudinal evidence from Malawi. Demographic Research, 42(37), 1057–1096. https://doi.org/10.4054/DEMRES.2020.42.37Bongaarts, J. (2017). Africa's unique fertility transition: Africa's unique fertility transition. Population and Development Review, 43, 39–58. https://doi.org/10.1111/j.1728-4457.2016.00164.xBongaarts, J. (2020). Trends in fertility and fertility preferences in sub‐Saharan Africa: The roles of education and family planning programs. Genus, 76(1), 32. https://doi.org/10.1186/s41118-020-00098-zBongaarts, J., & Watkins, S. C. (1996). Social interactions and contemporary fertility transitions. Population and Development Review, 22(4), 639. https://doi.org/10.2307/2137804Casterline, J. B. (2017). Prospects for fertility decline in Africa. Population and Development Review, 43, 3–18. https://doi.org/10.1111/padr.12055Cesare, N., Lee, H., McCormick, T., Spiro, E., & Zagheni, E. (2018). Promises and pitfalls of using digital traces for demographic research. Demography, 55(5), 1979–1999. https://doi.org/10.1007/S13524-018-0715-2/TABLES/2Coale, A. J., & Watkins, S. C. (1986). The Decline of Fertility in Europe, 484.DiMaggio, P., Hargittai, E., Neuman, W. R., & Robinson, J. P. (2001). Social implications of the Internet. Annual Review of Sociology, 27, 307–336. https://doi.org/10.1146/annurev.soc.27.1.307Eker, S., Garcia, D., Valin, H., & van Ruijven, B. (2021). Using social media audience data to analyse the drivers of low‐carbon diets. Environmental Research Letters, 16(7), 074001. https://doi.org/10.1088/1748-9326/abf770Fatehkia, M., Kashyap, R., & Weber, I. (2018). Using Facebook ad data to track the global digital gender gap. World Development, 107, 189–209. https://doi.org/10.1016/j.worlddev.2018.03.007Ferrara, E. L., Chong, A., & Duryea, S. (2012). Soap operas and fertility: Evidence from Brazil. American Economic Journal: Applied Economics, 4(4), 1–31. https://doi.org/10.1257/app.4.4.1Gittleman, J. L., & Kot, M. (1990). Adaptation: Statistics and a null model for estimating phylogenetic effects. Systematic Zoology, 39, 227–241.Gurumurthy, A., & Chami, N. (2014). Gender equality in the information society‐a review of current literature and recommendations for policy and practice.Hutcheon, J. A., Chiolero, A., & Hanley, J. A. (2010). Random measurement error and regression dilution bias. BMJ, 340, c2289. https://doi.org/10.1136/bmj.c2289Inkeles, A., & Smith, D. H. (1974). Becoming Modern. Harvard University Press.Kashyap, R., Fatehkia, M., Al Tamime, R., & Weber, I. (2020). Monitoring global digital gender inequality using the online populations of Facebook and google. Demographic research, 43, 779–816. https://doi.org/10.4054/DEMRES.2020.43.27Kietzmann, J. H., Hermkens, K., McCarthy, I. P., & Silvestre, B. S. (2011). Social media? Get serious! Understanding the functional building blocks of social media. Business Horizons, 54(3), 241–251. https://doi.org/10.1016/j.bushor.2011.01.005Kummu, M., Taka, M., & Guillaume, J. H. A. (2018). Gridded global datasets for gross domestic product and human development index over 1990–2015. Scientific Data, 5(1), 180004. https://doi.org/10.1038/sdata.2018.4Lerner, D. (1958). The Passing of Traditional Society. Free Press.Lesthaeghe, R. (1977). The decline of Belgian fertility, 1800–1970. Princeton University Press.Meta. (2019). An updata on how we are doing at enforcing our community standards. https://about.fb.com/news/2019/05/enforcing-our-community-standards-3/Nanfuka, J. (2022). Digital Access and Economic Transformation in Africa. Africa Paper Series.Norris, P. (2001). Digital Divide. Civic engagement, information poverty, and the internet worldwide. Cambridge University Press. https://doi.org/10.1017/CBO9781139164887Okenwa‐Emegwa, L., Lawoko, S., & Jansson, B. (2016). Attitudes toward physical intimate partner violence against women in Nigeria. SAGE Open, 6(4), 215824401666799. https://doi.org/10.1177/2158244016667993Ord, J. K., & Getis, A. (1995). Local spatial autocorrelation statistics: Distributional issues and an application. Geographical Analysis, 27(4), 286–306. https://doi.org/10.1111/J.1538-4632.1995.TB00912.XPezzulo, C., Nilsen, K., Carioli, A., Tejedor‐Garavito, N., Hanspal, S. E., Hilber, T., James, W. H. M., Ruktanonchai, C. W., Alegana, V., Sorichetta, A., Wigley, A. S., Hornby, G. M., Matthews, Z., & Tatem, A. J. (2021). Geographical distribution of fertility rates in 70 low‐income, lower‐middle‐income, and upper‐middle‐income countries, 2010–16: A subnational analysis of cross‐sectional surveys. The Lancet Global Health, 9(6), e802–e812. https://doi.org/10.1016/S2214-109X(21)00082-6Pierotti, R. S. (2013). Increasing rejection of intimate partner violence: Evidence of global cultural diffusion. American Sociological Review, 78(2), 240–265. https://doi.org/10.1177/0003122413480363Rampazzo, F., Bijak, J., Vitali, A., Weber, I., & Zagheni, E. (2021). A framework for estimating migrant stocks using digital traces and survey data: An application in the United Kingdom. Demography, 58, 2193–2218. https://doi.org/10.1215/00703370-9578562Ribeiro, F. N., Benevenuto, F., & Zagheni, E. (2020). How biased is the population of facebook users? comparing the demographics of facebook users with census data to generate correction factors. WebSci 2020—Proceedings of the 12th ACM Conference on Web Science, 325–334. https://doi.org/10.1145/3394231.3397923Robinson, L., Cotten, S. R., Ono, H., Quan‐Haase, A., Mesch, G., Chen, W., Schulz, J., Hale, T. M., & Stern, M. J. (2015). Digital inequalities and why they matter. Information, Communication & Society, 18(5), 569–582. https://doi.org/10.1080/1369118X.2015.1012532Rotondi, V., Kashyap, R., Pesando, L. M., Spinelli, S., & Billari, F. C. (2020). Leveraging mobile phones to attain sustainable development. Proceedings of the National Academy of Sciences, 117(24), 13413–13420. https://doi.org/10.1073/pnas.1909326117Scheerder, A., van Deursen, A., & van Dijk, J. (2017). Determinants of Internet skills, uses and outcomes. A systematic review of the second‐ and third‐level digital divide. Telematics and Informatics, 34(8), 1607–1624. https://doi.org/10.1016/J.TELE.2017.07.007Shapiro, D., & Tenikue, M. (2017). Women's education, infant and child mortality, and fertility decline in rural and urban sub‐Saharan Africa. Demographic research, 37(1), 669–708. https://doi.org/10.4054/DemRes.2017.37.21Smits, J., & Permanyer, I. (2019). The subnational human development database. Scientific Data, 6(1), 190038. https://doi.org/10.1038/sdata.2019.38Smits, J., & Permanyer, I. (2020). The Subnational Gender Development Index: Within‐country variation in gender (in)equality revealed. UNDP HDialogue Blog. https://globaldatalab.org/blog-SGDISnijders, T. A. B., & Bosker, R. J. (2011). Multilevel Analysis: An introduction to basic and advanced multilevel modeling. SAGE Publications, Inc.Stewart, I., Flores, R. D., Riffe, T., Weber, I., & Zagheni, E. (2019). Rock, rap, or reggaeton?: Assessing Mexican immigrants' cultural assimilation using Facebook data. The Web Conference 2019—Proceedings of the World Wide Web Conference, WWW 2019, 3258–3264. https://doi.org/10.1145/3308558.3313409Toffolutti, V., Ma, H., Menichelli, G., Berlot, E., Mencarini, L., & Aassve, A. (2020). How the Internet increases modern contraception uptake: Evidence from eight sub‐saharan African countries. BMJ Global Health, 5(11), e002616. https://doi.org/10.1136/bmjgh-2020-002616Varriale, C., Pesando, L. M., Kashyap, R., & Rotondi, V. (2022). Mobile phones and attitudes toward women's participation in politics. Sociology of Development, 8(1), 1–37. https://doi.org/10.1525/SOD.2020.0039Westoff, C. F., & Bankole, A. (1997). Mass Media and Reproductive Behavior in Africa. Demographic and Health Surveys Analytical Reports No 2, April.Westoff, C. F., & Koffman, D. A. (2011). The association of television and radio with reproductive behavior. Population and Development Review, 37(4), 749–759. https://doi.org/10.1111/j.1728-4457.2011.00455.xWorld Bank. (2021a). Birth rate, crude (per 1,000 people). https://data.worldbank.org/indicator/SP.DYN.CBRT.INWorld Bank. (2021b). Mobile cellular subscriptions (per 1000 people). https://data.worldbank.org/indicator/IT.CEL.SETS.P2Zagheni, E., Weber, I., & Gummadi, K. (2017). Leveraging facebook's advertising platform to monitor stocks of migrants: Leveraging facebook's advertising platform. Population and Development Review, 43(4), 721–734. https://doi.org/10.1111/padr.12102 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png "Population, Space and Place" Wiley

Fertility rates and social media usage in sub‐Saharan Africa

Loading next page...
 
/lp/wiley/fertility-rates-and-social-media-usage-in-sub-saharan-africa-FHJop3Rzzz
Publisher
Wiley
Copyright
© 2022 John Wiley & Sons Ltd.
ISSN
1544-8444
eISSN
1544-8452
DOI
10.1002/psp.2635
Publisher site
See Article on Publisher Site

Abstract

INTRODUCTIONFertility levels in sub‐Saharan Africa (SSA) are known to be influenced by exposure to traditional media, such as television and radio (Inkeles & Smith, 1974; La Ferrara et al., 2012; Lerner, 1958; Westoff & Bankole, 1997). Since 2000, rapid expansion of information and communication technology (ICT)—often referred to as the digital revolution—has caused a major shift in the availability and role of media. In 2020, 93% of Africans had access to a mobile phone, whereas this was only 53% in 2011; 30% of Africans were connected to the Internet in 2020 (World Bank, 2021a). A key characteristic of this shift is digital interactivity between media consumers, which is predominantly provided by social media. This raises the important question as to what extent social media usage is associated with fertility levels.Earlier studies have shown that women's education is the main determinant of SSA's fertility levels, acting through several mechanisms such as support for women's empowerment, facilitating their participation in the labour force and reducing child mortality (Atake & Gnakou Ali, 2019; Shapiro & Tenikue, 2017). However, another determinant is the diffusion of information, ideas and technology regarding, for example, contraceptive usage, fertility preferences and family size between individuals within communities (Bongaarts & Watkins, 1996; Coale & Watkins, 1986; Lesthaeghe, 1977). Individuals receive and interpret this information differently. The information may therefore gain new meaning in different social contexts and ultimately either promote or constrain fertility behaviour and outcomes. Schools are an important location for this diffusion (e.g., Atake & Gnakou Ali, 2019; Bongaarts, 2020; Shapiro & Tenikue, 2017), but it also takes place via traditional media (La Ferrara et al., 2012; Westoff & Koffman, 2011).Compared to traditional media, social media offers further opportunities for the diffusion of ideas. Social media transcends country borders more easily and thus promotes exposure to less context‐specific and more globalised cultural ideas, including those regarding fertility behaviour. Social media provides access to the ‘life of others’ without the need for physical proximity. Moreover, it enables its users not only to consume, but to contribute content and become members of an online community (Kietzmann et al., 2011). As such, social media provides individuals with the possibility of obtaining new information and interacting with role models beyond their own community.Earlier studies have suggested that the impact of digital media, such as social media, is likely to be greater in developing regions where other types of infrastructure are poor. Social media allows individuals living in remote areas contact with new ideas and individuals beyond their community borders, which a lack of infrastructure had previously made impossible (Aker & Mbiti, 2010; Rotondi et al., 2020). Studies have also shown that within these remote areas, the impact of the digital revolution is found to be greater for more marginalised groups, such as females who face greater barriers in accessing digital media (Rotondi et al., 2020; Varriale et al., 2022).Due to the difficulty of measuring social media access, previous research has predominantly relied on national estimates of official ICT statistics. These estimates potentially mask underlying heterogeneity within countries. The strength of the current study is that it uses new digital trace data for social media, in particular Facebook, at the subnational level to uncover these underlying differences. Digital trace data has its own advantages, such as the high speed of large‐scale data collection and the low costs involved (Cesare et al., 2018). To resolve the potential lack of accessibility, transparency and representativeness of this type of data, we enrich it with traditional survey data. This approach has already yielded promising results in previous studies with regard to data reliability (Eker et al., 2021; Kashyap et al., 2020; Rampazzo et al., 2021; Ribeiro et al., 2020).This study will use subnational digital trace data from Facebook to examine to what extent social media usage and gender gaps in usage explain subnational regional variations in fertility levels. For our analyses, we have combined Facebook data with household survey data for 311 subnational regions in 29 SSA countries. This enables us to study the relationship between fertility and social media usage in more detail than has been possible before.In the next section, we provide an overview of the relevant literature concerning the relationship between fertility behaviour and (social) media. The result section describes the spatial pattern in subnational variation in birth rates, social media usage and gender gaps in social media usage. Subsequently, the association between birth rates and social media usage will be tested using regression models.BACKGROUNDA history of media exposure and fertility behaviourThere is a long history of studies that have examined the relationship between media exposure and fertility behaviour. One of the first was by Lerner (1958), who reported that greater exposure to media (radio and film) made individuals more likely to express ‘modern’ and ‘pro‐development’ attitudes of various sorts in six countries in the Middle East. Some decades later, similar effects were found in studies focusing on SSA. Inkeles and Smith (1974) for example, found that watching television, listening to the radio and reading newspapers were good predictors for ‘modern’ norms and values, and this included the will to limit family size. Westoff and Bankole (1997) found a strong positive association between media use and fertility behaviour in six SSA countries (Burkina Faso, Ghana, Kenya, Madagascar, Namibia and Zambia). Included fertility behaviours were knowledge of modern contraceptive methods, intention to use contraceptives, current use of contraceptives and the desire to limit one's family size. An additional analysis by the same authors showed that women who were ‘heavy media consumers’ were more likely to use contraceptives and to limit their family size than women who were not. However, caution should be exercised when interpreting the causality of this relationship. According to the authors, there was a possibility that heavy media consumers were already on a different reproductive trajectory before the study started, which could have been the reason for their media consumption.Another study carried out in Brazil reported a causal link between media and fertility; here, Ferrara et al. (2012) examined the effect of exposure to soap operas on fertility rates. The main characters in most of the soap operas were living in urban settings with lifestyles that were in stark contrast to those of their rural audiences; they had small families, married later than average and had a high level of education. The study reported that the availability of a television signal, and thus exposure to soap operas, decreased the probability of giving birth to a child. This effect was strongest for women who had low socioeconomic status and whose lives were thus most in contrast with those of the soap opera protagonists.Media exposure, however, does not always result in a positive relationship with globalised norms and values. For example, media exposure at least once a week was associated with lower acceptance of domestic violence in only 14 out of 26 low‐ and middle‐income countries (Pierotti, 2013). Another study found a positive association between media access and the odds of tolerating domestic violence in Nigeria by women, but a negative association for men (Okenwa‐Emegwa et al., 2016). A similar finding was reported by Banerjee et al. (2019). They found that exposure to music television series was only associated with lower acceptance of domestic violence by men, but not women, in Nigeria. As such, media exposure can also be negatively associated with liberal and globalised fertility behaviours.From traditional to social mediaSince 2000, media has evolved from ‘old’ or ‘traditional’ media to ‘new’ media; this is often referred to as the ‘media revolution’ or ‘digital revolution’. Whereas traditional media predominantly involves listening to the radio, watching television and reading newspapers, new media is characterised by digitalisation of content. New media continuously evolves, and consequently its content changes frequently. Early on, new media was often considered as complementary to traditional media with individuals simply consuming additional online content; in other words, passively reading or watching (DiMaggio et al., 2001; Kietzmann et al., 2011).One study that focused on the relationship between new media and fertility research showed that women's internet exposure was positively related to the use of modern contraception in eight sub‐Saharan countries. The increase in use of modern contraception was greatest among poorly educated women (Toffolutti et al., 2020). Another study examined how access to mobile phones influenced fertility behaviour in Balaka, Malawi (Billari et al., 2020). This study found that the possession of a mobile phone was associated with a preference for smaller family size. Moreover, the authors concluded that phone ownership may accelerate the fertility transition in SSA.Social media has unique characteristics relative to digital media in general, such as interactivity among its users. To take a narrow view, social media involves person‐to‐person relationships on social networking services such as Twitter and Facebook. However, a broader interpretation of social media is generally accepted, in which social media is described as interactive platforms through which individuals and groups can share, cocreate, discuss and modify self‐created content (Kietzmann et al., 2011). As social interaction is known to be an important mechanism through which media influence fertility behaviour (Bongaarts & Watkins, 1996; Coale & Watkins, 1986; Ferrara et al., 2012; Lesthaeghe, 1977; Westoff & Koffman, 2011), the interactivity of social media is likely to influence ideas about fertility and fertility preferences. Online communities enable individuals to discuss ideas and information with others without the need for physical proximity, resulting in contact with persons living in areas where other fertility patterns prevail.One qualitative study examined the role of WhatsApp, a form of social media, among females living in a community in Nigeria (Abubakar & Dasuki, 2018). This study found that these women experienced greater autonomy and more opportunities to voice their personal opinions thanks to contact with other females in WhatsApp group chats. They were able to push the barriers of the social norms that restricted them to gendered roles within their community. As such, this study showed that individuals who have access to online communities may be less dependent on the societal norms and values of their immediate surroundings and more likely to have contact with globalised norms and values. Hence, even though online communities are often subject to self‐selection and homophily, they also provide users with the possibility of transcending the boundaries of their physical environment. As globalised norms tend to favour smaller family size than the average family size in SSA (Bongaarts, 2017), our first hypothesis is that social media usage is linked to lower birth rates.Inequalities in social media usageAnother aspect of new and social media is that it is not universally accessible. There are significant inequalities in access to ICT, which are often referred to as the ‘digital divide’ (DiMaggio et al., 2001; Norris, 2001). Known dimensions of the digital divide are level of development and gender. Access to new media in developing countries has been found to reflect broader structural inequalities, such as those in education, employment and income (Kashyap et al., 2020; Robinson et al., 2015). Moreover, studies have reported that women are underrepresented in online populations of Google and Facebook users in South Asia and in SSA (Fatehkia et al., 2018; Kashyap et al., 2020).The digital gender divide is also associated with cultural factors (Scheerder et al., 2017). Cultural norms in patriarchal contexts may hinder females' access to new media, especially when this access is mediated through males (Abu‐Shanab & Al‐Jamal, 2017; Gurumurthy & Chami, 2014). Other studies have shown that females' adoption of new and social media is related to their empowerment and societal well‐being, and to globalised and liberal values (Abubakar & Dasuki, 2018; Varriale et al., 2022). Hence, our second hypothesis states that when gender gaps in access to social media are smaller, fertility levels will be lower.DATAThis study examines to what extent social media usage and gender differences in social media usage are related to fertility levels in subnational regions of SSA countries. We combine data from Facebook's marketing application with development indicators computed on the basis of household surveys. The household surveys are derived from three sources: Demographic and Health Surveys (DHS, www.dhsprogram.com), UNICEF Multiple Indicator Cluster Surveys (MICS, mics.unicef.org) and Afrobarometer Surveys (www.afrobarometer.org). These three sources are publicly and freely accessible for academic purposes.Subnational regions are the unit of analysis in our study. Their selection is based on the geographic coding available in the household survey datasets and the regions used by Facebook's marketing application. The available regions in the household surveys were taken as the starting point for selecting regions used by Facebook's marketing application. We were able to make connections for 311 regions in 29 countries. An overview of the included countries, subnational regions and data sources is presented in Appendix A.Key fertility measureThe dependent variable is the crude birth rate (CBR) of the subnational regions in the year 2020. The CBR represents the number of yearly childbirths per 1000 individuals in a region. The considerable advantage of this over other fertility measures, such as the total fertility rate, is that it is a very simple measure that can be reliably computed over a short period of time. As social media use in the SSA region is expanding rapidly (Nanfuka, 2022), it is essential that the measurement of our dependent variable was close in time to our measurement of social media usage. Because the CBR might be influenced by the age structure of the female population, we included the percentage of fertile women in a region's total population, and the mean age of the fertile women in a region as extra control factors in our models.The subnational CBRs for 2020 were estimated by applying the subnational variation in CBRs derived from the most recent DHS and MICS surveys to the national CBRs for 2020 derived from the World Development Indicators (WDI, World Bank, 2021b). To do this, we followed a procedure derived from Kummu et al. (2018) and Smits and Permanyer (2019) in which subnational variation derived from household surveys was normalised around national values obtained. First, the national CBR from the World Bank in 2020 was divided by the national CBR of the most recent DHS or MICS survey to create a differentiation factor. Second, the subnational CBRs from the DHS/MICS surveys were multiplied with this factor to estimate the subnational CBRs for the year 2020. By doing this, we assume that the subnational CBRs follow the same pattern of change as the national values between the survey year and 2020.Because the size of the estimation error is expected to depend on the number of years over which the estimation was made, we used only countries for which the last survey was held after 2010, and included the survey year as control variable. Moreover, we added robustness analysis in which we restricted the sample to countries in which the survey was held after 2015 instead of 2010 (Appendix A). The results are similar in terms of direction and significance.Facebook indicatorsTo create subnational social media indicators, we used Facebook's advertisement platform. For marketing purposes, Facebook allows users with a Facebook account to search the number of active Facebook users in the last month disaggregated by various geographic and demographic characteristics (detailed information can be found at: https://developers.facebook.com/docs/marketing-apis). In this way, Facebook enables potential advertisers to use advanced targeting options by creating detailed user profiles, which include demographic characteristics such as age, gender and geographic location. For example, if you want to reach females aged 20‐30 in Benin, Facebook provides you with the number of active accounts that have met these criteria in the last month. Earlier studies have already used Facebook's advertisement platform for research purposes (Alexander et al., 2019; Stewart et al., 2019; Zagheni et al., 2017).We used this facility to create active user estimates for individuals aged 15−39 within subnational regions in SSA countries. To obtain the percentage of Facebook users we compared the number of users with the number of individuals aged 15−39. The population sizes for these individuals were based on data from the most recent DHS and MICS surveys and data from the WDI, using the same estimation procedure as for the subnational CBR values (discussed above). First, we computed a differentiation factor for individuals aged 15−39 between the national population sizes in the DHS/MICS surveys and those in the 2020 WDI data. Thereafter, this differentiation factor was applied to the subnational population data in the DHS/MICS surveys to obtain the 2020 subnational population estimates.In 19 subnational regions, Facebook usage exceeded 100%. As Facebook is not transparent in how it calculates audience size, we decided to set the maximum percentage of Facebook usage at 100%. For robustness, we also ran the analysis without these subnational regions, with a maximum of 75% instead of 100 and with subnational Facebook deciles instead of percentages. As the results did not significantly differ in terms of direction and significance, we only present those that included the regions with a Facebook percentage of up to 100%.Besides the percentage of Facebook usage, we also computed the gender gap in Facebook usage in these subnational regions. First, we calculated Facebook usage for males and females separately using the technique described above. Second, we subtracted subnational female Facebook usage from male Facebook usage. As such, if the gender gap in Facebook usage is zero, this means an equal number of male and female users. If the gender gap is positive, males are overrepresented; if it is negative, females are overrepresented.Control factorsControl factors in our study were the level of development, gender inequality, phone ownership, the percentage of females aged 15−49 and the age structure of fertile women in the regions. In more developed and wealthier regions, digital access is better and fertility rates are known to be lower (Pezzulo et al., 2021). As such, regions' development should be controlled for. Lower birth rates were also expected for regions where women hold a stronger position, where there are higher levels of mobile phone ownership, where a lower percentage of females are in their fertile years (ages 15−49), and where the fertile women are on average older (Casterline, 2017; Rotondi et al., 2020).The level of development was measured by the subnational Human Development Index (sHDI) (Smits & Permanyer, 2019). The sHDI shows within‐country variation in human development across the globe and, like the Human Development Index, is an indicator of a combination of three pillars: education, health and standard of living. The sHDI is measured with a value between 0 and 1; a higher score denotes higher development. Gender inequality was measured by the subnational gender development index (sGDI), which is defined as the ratio between female and male levels of sHDI in the region (Smits & Permanyer, 2020). A sGDI of 1 indicates perfect equality between females and males; a sGDI below 1 indicates that males have a higher level of human development and a sGDI above 1 indicates that females have a higher level of human development. Phone ownership was measured by the percentage of households in the region in which at least one member owns a phone. The data was derived from the DHS, MICS and Afrobarometer surveys. Afrobarometer data on phone ownership for 2020−2021 (round 8) was used if it was more recent than the DHS or MICS data. This was the case for Benin, Cameroon, Cote d'Ivoire, Estwatini, Ethiopia, Gabon, Gambia, Ghana, Guinea, Kenya, Lesotho, Liberia, Mali, Mozambique (round 7, 2017), Namibia, Niger, Nigeria, Togo. The percentage of females aged 15−49, which is considered females' fertile period, was calculated as the percentage of females aged 15−49 of the total population of the region. The population structure of fertile females was included by calculating the mean age of females aged 15−49.METHODSExploratory spatial data analysisWe started the analyses with an exploratory spatial data analysis of the subnational CBRs, Facebook usage and gender gaps in Facebook usage. We examined whether spatial clustering exists among those three variables and to what extent there is overlap between clusters of birth rates and those of Facebook usage. Clusters refer to agglomerations of bordering subnational regions that share similar characteristics—for example, adjacent regions that both show high (or low) CBRs or high (or low) Facebook usage.First, we assessed the degree of global spatial autocorrelation with the Global Moran's I statistic for birth rates and Facebook usage (Gittleman & Kot, 1990). Moran's I measures whether a spatial pattern exists. Global Moran's I values closer to 0 indicate a random spatial pattern, positive values indicate clustering and negative values dispersion. Second, we calculated the Gi* statistic based on local Getis‐Ord Gi* hot spot analyses to examine whether and where clusters in birth rates and Facebook usage were situated (Ord & Getis, 1995). The Gi* statistic identified areas where relatively high or low values of certain characteristics, in this case birth rates and Facebook usage, tend to cluster together based on the neighbouring values. We compared the overlap in clusters of birth rates and Facebook usage and examined the characteristics of low and high birth rate clusters.Regression analysisTo test whether the association between birth rates and Facebook usage is still significant after the inclusion of subnational characteristics, we used regression analysis. As the subnational regions are nested within countries, our data has a hierarchical structure. To control for this, two types of regression models were estimated: a mixed effects regression model that addresses the hierarchical structure by including a random intercept at the national level (Snijders & Bosker, 2011) and a fixed effects regression model in which the hierarchical structure is addressed by including fixed effects dummies at the country level (Allison, 2009). The fixed effects approach uses much more degrees of freedom than the multilevel approach, but has the advantage that all direct effects of (known and unknown) confounders at the national level are controlled for (Bell et al., 2019).RESULTSGeographic patterns of fertility behaviour and social media usageOur analysis starts with exploring the geographic pattern of subnational CBRs in 2020 (Figure 1). The mean subnational CBR was 34 births per 1000 individuals in 2020, with a minimum of 17 births in North West, South Africa and a maximum of 55 births in Lunda Sul, Angola. Mean Facebook usage was 25%, with a minimum of 1% in Other Central, Malawi and a maximum of 100% in 19 regions (darkest colour in Figure 1b). The gender gap in Facebook usage was on average 8%, meaning that females are on average less connected than males. Females were relatively best represented on Facebook in Eastern Cape, South Africa (8% more than males), and males were relatively the dominant users in Maritime, Togo (59% more than females). Figure 1 shows substantial variation across subnational regions in the distribution of CBRs, Facebook usage and the gender gap in Facebook usage. The descriptive statistics can be found in Table 1.1FigureGeographic distributions of (a) crude birth rates 2020, (b) Facebook usage 2021, and (c) the gender gap in Facebook usage between males and females 20211TableDescriptive statistics: Means and percentages of variables included in this studyVariablesMean, %SDMin.Max.Subnational Crude Birth Rates 2020 (dep. var.)34.007.0016.7755.32Subnational Characteristics% Facebook24.5628.420.55100Gender gap Facebook8.189.88−7.8458.55sHDI0.530.090.270.75sGDI0.890.080.661.07% Phone77.7418.8522.5099.02% Females 15−4923.375.1516.5850.57Mean age females 15−4928.620.9825.5331.39Year of survey20162.4020112020Note: N regions = 311, N countries = 29.Abbreviations: SD, standard deviation; sGDI, subnational gender development index; sHDI, subnational Human Development Index.We tested whether the variability in these distributions showed a spatial pattern and found a Moran's I of 0.09 (standard deviation [SD] = 0.005) for birth rates, of 0.07 (SD = 0.007) for Facebook usage, and of 0.11 (SD = 0.007) for gender gaps in Facebook usage. This indicates that the variability in birth rate and Facebook usage is not random and that clustering exists, in that neighbouring subnational regions tend to have similar characteristics and form spatial clusters. Figure 2 presents the Gi* statistics, which show where clusters in CBRs, Facebook usage and gender gaps in Facebook usage are situated. Clusters with Gi* statistics below −1.96 are considered significantly low (or small) clusters, and those with Gi* statistics above 1.96 are considered significantly high (or large) clusters. The mean CBR in low birth rate clusters was 25 births per 1000 individuals (SD = 4.8), and in high birth rate clusters was 44 births per 1000 individuals (SD = 4.5). The average Facebook usage of low usage clusters was 2.7% (SD = 4.2) and of high usage clusters 48.4 per cent (SD = 20.3); the average gender gap in small gender gap clusters was 0.2% (SD = 4.2) and in large gender gap clusters 21.0% (SD = 11.5) (Appendix A).2FigureSpatial clusters of (a) crude birth rates 2020, (b) Facebook usage 2021, and (c) the gender gap in Facebook usage 2021Visual inspection of Figure 2 shows that clusters with low birth rates tend to overlap with clusters of high Facebook usage, but that this is less so for small gender gap clusters. Low birth rate clusters overlap with clusters characterised by high Facebook usage in Southern SSA (Namibia, South Africa and Swaziland). High birth rate clusters overlap with some clusters with large gender gaps, but not so much with clusters with high Facebook usage. Here, we see that high birth rate clusters and large gender gaps in Facebook usage predominantly overlap in some parts of Western SSA (Mali and Nigeria).Further examination of the relationship between the clustering of subnational CBRs and Facebook indicators reveals consistent patterns. Figure 3a shows that subnational Facebook usage is on average higher in low birth rate clusters. However, there is no clear difference between Facebook usage in the nonsignificant and high birth rate clusters. These observations seem to confirm the visual overlap in Figure 2, which showed that high Facebook usage is associated with low CBRs, but low Facebook usage is not necessarily associated with high CBRs. Figure 3b shows that the gender gap in Facebook usage is smaller in low birth rate clusters. The median of the gender gap is slightly above zero in low birth rate clusters, meaning that Facebook usage of males and females is (almost) equal in these clusters. On the other hand, the median of the gender gap in high birth rate clusters is 8, which shows that around 8% fewer females than males use Facebook in high birth rate clusters. As such, the gender gap in Facebook usage seems to be associated with high birth rate clusters.3FigureBoxplots showing the minimum (lower end line), maximum (upper end line), median (middle line of box), and fist and third quartiles (lower and upper ends of box) of low, nonsignificant and high birth rate clusters of (a) Facebook usage 2021, (b) the gender gap in Facebook usage 2021, (c) the subnational Human Development Index 2020, (d) the subnational Gender Development Index 2020, (e) phone ownership, (f) the percentage of females aged 15−49, and (g) the mean age of females aged 15−49 for crude birth rates in low, nonsignificant and high birth rate clusters.Similar patterns are observed between the CBR clusters and the control factors: sHDI, sGDI, phone ownership, the percentage of females aged 15−49 and the mean age of females aged 15−49. The sHDI, sGDI, phone ownership, percentage of females aged 15−49 and the mean age of females aged 15‐49 are higher in low birth rate clusters. This shows that low birth rate clusters have on average higher levels of development, levels of gender development, percentages of phone ownership and percentages of females aged 15−49.Regression resultsTable 2 shows the results of the regression models that test the association between birth rates and Facebook usage. The mixed effects model and the fixed effects model show comparable results in terms of coefficients, standard errors and significance.2TableRegression estimates of the determinants of subnational Crude Birth Rates (2020) for the multilevel linear regression (MLR) and Fixed Effects (FE) regression modelsMLR IFE I% Facebook−0.030* (0.014)−0.035* (0.015)Gender gap Facebook0.062* (0.031)0.067* (0.032)sHDI−21.852*** (6.357)−24.025** (7.711)sGDI−31.128*** (7.509)−33.947** (10.611)% Phone−0.064* (0.025)−0.073* (0.03)% Females 15−490.173 (0.11)0.364* (0.162)Mean age females 15−49−0.934** (0.35)−0.67 (0.379)Year0.071 (0.265)Country dummies includedYesConstant33.529*** (0.636)41.317*** (1.396)Country‐level variance9.62 (3.10)Adjusted R²0.657Note: Independent variables are centred around their mean. Standard errors are shown in parentheses. Analyses are run for 311 subnational regions in 29 countries.Abbreviations: sGDI, subnational gender development index; sHDI, subnational Human Development Index.*p < 0.05;**p < 0.01;***p < 0.001.Our analysis reveals significant associations of Facebook usage and gender gaps in Facebook usage with subnational birth rates. Facebook usage is negatively associated with birth rates, which means that fertility is lower in areas with higher Facebook usage. This finding is in line with our first hypothesis.Regarding the gender gap in social media usage, we find a positive association between the gender gap in Facebook usage and birth rates. Hence in areas where females are relatively underrepresented on social media, fertility levels are higher. This finding is in line with our second hypothesis. We expected that females who have relatively less access to Facebook to be less empowered and consequently, less likely to have contact with globalised ideas and to change their behaviour accordingly.All associations are controlled for the sHDI, sGDI, subnational phone ownership, percentage of females aged 15−49, mean age of females aged 15−49 and survey year. The association between birth rates and the sHDI is significantly negative, meaning that when subnational regions have a higher level of development, CBRs are lower. Birth rates are significantly higher in regions with higher gender inequality in favour of males. Subnational phone ownership shows a significant negative association with subnational birth rates, which means that higher phone ownership is related to lower birth rates. The percentage of females aged 15−49 is positively associated with birth rates and the mean age of women in this age group negatively. Hence fertility is higher if fertile females comprise a larger share of the population and if their mean age is lower. The year of the survey is not significantly related to birth rates. All these findings for the control variables are in line with our expectations and with earlier findings (Atake & Gnakou Ali, 2019; Billari et al., 2020; Casterline, 2017; Pezzulo et al., 2021; Rotondi et al., 2020).DISCUSSIONWhere previous studies have firmly established a connection between fertility behaviour and traditional media such as television and radio (Inkeles & Smith, 1974; La Ferrara et al., 2012; Lerner, 1958; Westoff & Bankole, 1997), this study examines the link between fertility behaviour and social media. Based on data for 311 subnational regions in 29 countries in SSA, we tested two hypotheses regarding the relationship between social media and fertility outcomes: the idea that access to social media is associated with lower fertility rates (hypothesis 1) and that gender equality in social media use is associated with lower fertility rates (hypothesis 2). Our findings are in line with these hypotheses. The descriptive analysis identified spatial overlap between low birth rate clusters and high social media usage clusters, and between high birth rate clusters and clusters with a large gender gap in social media use. The regression analyses revealed that higher social media usage and smaller gender gaps in social media usage are significantly linked to lower birth rates. We therefore conclude that on the basis of our findings, the hypotheses cannot be rejected and that our study provides new empirical evidence regarding the association between social media usage and fertility decline.This study is important, because it is the first that shows the existence of a substantial relationship between Facebook usage and fertility at the level of subnational regions in SSA. Although some studies have already reported the effects of mobile phone ownership and internet access on fertility behaviour (Billari et al., 2020; Toffolutti et al., 2020), we have gone a step further by identifying similar effects of social media exposure and in particular, access to Facebook. One possible explanation for this finding is that social media enables its users to form (online) communities and to share ideas within them. Interactions between individuals on social media rely less on physical proximity, and as such individuals are likely to encounter more globalised ideas and norms, and to change their behaviour accordingly. As the discussion of new ideas and norms is a central concept within diffusion theories (Coale & Watkins, 1986; Westoff & Koffman, 2011), the findings of this study could well be interpreted within this line of theory.Our results regarding the gender gap in social media usage reconfirm the finding of earlier studies: that women are underrepresented in the online population (Fatehkia et al., 2018; Kashyap et al., 2020; Robinson et al., 2015). That better representation is associated with lower fertility outcomes is also in line with earlier studies, which have found that females adopting new and social media tend to have more modern and liberal values (Abubakar & Dasuki, 2018; Varriale et al., 2022). As such, our study also points towards negative consequences of large gender inequalities in media access.A possible limitation of the present study is that there are potential reliability issues with the social media data. Apparent Facebook usage of above 100 per cent in 19 subnational regions is an example of this. It might be that individuals in those regions have more than one Facebook account or that there are a substantial number of fake accounts. According to Meta (2019), on average about 5% of all Facebook accounts are fake. However, this could also be the result of errors in Facebook's target audience generator. The consequences could be twofold. If overestimation is systematic, it will not substantially influence the direction and significance of the regression estimates. If overestimation is random, this suggests measurement error in our data, which probably implies an underestimation of the real effects (Hutcheon et al., 2010). A second potential limitation is the possibility of selection bias in our data. It is possible that the regions with higher social media usage are also those with more globalised values and lower fertility rates. If so, this could lead to an overestimation of the association between social media access and fertility. More research is required to examine this possibility. A third limitation is that we have estimated the subnational CBRs by applying subnational variation derived from earlier surveys to national data derived from the World Bank. This approach is based on the assumption that subnational variation is relatively stable over time. To be sure, we added the year of the survey as an additional control factor and performed a robustness test by repeating our analysis with a sample restricted to countries with surveys from 2015. Findings turned out to be substantially the same as in the main analysis.In sum, this study uses an innovative approach to provide new evidence that social media usage and gender gaps in usage are significantly related to SSA fertility outcomes. This might mean that the social media revolution currently spreading across the subcontinent might profoundly affect its fertility transition. Compared with earlier studies, our research constitutes a major improvement because it uses—for the first time—social media usage data at subnational level as a potential ingredient of fertility decline in the SSA region.ACKNOWLEDGEMENTThe research team would like to thank Siem Smeets for his data work as student assistant. We would also like to thank the reviewers for their time and valuable feedback to improve this article.DATA AVAILABILITY STATEMENTThe data that support the findings of this study are available from DHS, Mics, Afrobarometer. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the author(s) with the permission of DHS, Mics, Afrobarometer.AAPPENDIXOverview of included regions, data sources (Demographic and Health survey [DHS] or Multiple Indicator Cluster [MICS] survey), year of data sources and data of accessing Facebook's target audience estimates of this study.CountryRegionData sourceYearDate of FacebookAngolaCabindaDHS20165/24/2021AngolaZairaDHS20165/24/2021AngolaUigeDHS20165/24/2021AngolaLuandaDHS20165/24/2021AngolaKuanza NorteDHS20165/24/2021AngolaKuanza SulDHS20165/24/2021AngolaMalangeDHS20165/24/2021AngolaLanda NorteDHS20165/24/2021AngolaBenguelaDHS20165/24/2021AngolaHuamboDHS20165/24/2021AngolaBieDHS20165/24/2021AngolaMoxicoDHS20165/24/2021AngolaKuanda KubangoDHS20165/24/2021AngolaNamibeDHS20165/24/2021AngolaHuilaDHS20165/24/2021AngolaCunene Lunda SulDHS20165/24/2021AngolaBengoDHS20165/24/2021BeninAtacora + DongaDHS20183/7/2021BeninAtlantique + LittorialDHS20183/7/2021BeninBorgou + AliboriDHS20183/7/2021BeninMono + KouffoDHS20183/7/2021BeninOueme + PlateauDHS20183/7/2021BeninZou + CollineDHS20183/7/2021BurundiNorth (Kayanza, Kirundo, Muyinga, Ngozi)DHS20175/2/2021BurundiSouth (Bururi, Makamba)DHS20175/2/2021BurundiEast (Cankuzo, Rutana, Ruyigi)DHS20175/2/2021BurundiWest (Bubanza, Buja Rural, Cibitoke, Mairie de Bujumbura)DHS20175/2/2021BurundiMiddle (Gitega, Karuzi, Maramvya, Mwaro)DHS20175/2/2021CameroonAdamaouaDHS20183/8/2021CameroonCentre + YaoundeDHS20183/8/2021CameroonEstDHS20183/8/2021CameroonExtreme NordDHS20183/8/2021CameroonLittoral + DoualaDHS20183/8/2021CameroonNordDHS20183/8/2021CameroonNord OuestDHS20183/8/2021CameroonOuestDHS20183/8/2021CameroonSudDHS20183/8/2021CameroonSud OuestDHS20183/8/2021ChadZone 1 (N'Djamena)MICS20195/2/2021ChadZone 2 (Borkou, Ennedi, Tibesti, Kanem, Lac)MICS20195/2/2021ChadZone 3 (Guera, Batha Est & Ouest, Salamat)MICS20195/2/2021ChadZone 4 (Ouaddai, Bilitine)MICS20195/2/2021ChadZone 5 (Chari‐Baguimi)MICS20195/2/2021ChadZone 6 (Mayo‐Kebbi)MICS20195/2/2021ChadZone 7 (Logone Occidental, Logone Oriental, Tandjilie Est & Ouest)MICS20195/2/2021ChadZone 8 (Moyen‐Chari)MICS20195/2/2021Congo BrazzavilleKouilouMICS20155/24/2021Congo BrazzavilleNiariMICS20155/24/2021Congo BrazzavilleLekoumouMICS20155/24/2021Congo BrazzavilleBouenzaMICS20155/24/2021Congo BrazzavilleBrazzaville + PoolMICS20155/24/2021Congo BrazzavillePlateauxMICS20155/24/2021Congo BrazzavilleCuvetteMICS20155/24/2021Congo BrazzavilleCuvette OuestMICS20155/24/2021Congo BrazzavilleSanghaMICS20155/24/2021Congo BrazzavilleLikoualaMICS20155/24/2021Congo BrazzavillePointe‐NoireMICS20155/24/2021Cote d'IvoireCentre (Lacs district + Yamoussoukro)MICS20165/3/2021Cote d'IvoireCentre Nord (Valée du Bandama District)MICS20165/3/2021Cote d'IvoireNord (Savanes district)MICS20165/3/2021Cote d'IvoireNord Est (Zanzan)MICS20165/3/2021Cote d'IvoireNord Ouest (Denguélé + Woroba district)MICS20165/3/2021Cote d'IvoireOuest (Montagne district)MICS20165/3/2021Cote d'IvoireSud Ouest (Bas‐Sassandra)MICS20165/3/2021Cote d'IvoireSud, Abidjan, Centre‐Ouest, Centre EstMICS20165/8/2021SwazilandHhohhoMICS20146/28/2021SwazilandManziniMICS20146/28/2021SwazilandShiselweniMICS20146/28/2021SwazilandLubomboMICS20146/28/2021EthiopiaTigrayDHS20164/26/2021EthiopiaAffarDHS20164/26/2021EthiopiaAmharaDHS20164/26/2021EthiopiaOromiaDHS20164/26/2021EthiopiaSomaliDHS20164/26/2021EthiopiaBen‐GumzDHS20164/26/2021EthiopiaYeDebubDHS20164/26/2021EthiopiaGambelaDHS20164/26/2021EthiopiaHarariDHS20164/26/2021EthiopiaAddisDHS20164/26/2021EthiopiaDire DawaDHS20164/26/2021GabonEstuaire + LibrevilleDHS20125/25/2021GabonHaut‐OgooueDHS20125/25/2021GabonMoyen‐OgooueDHS20125/25/2021GabonNgounieDHS20125/25/2021GabonNyangaDHS20125/25/2021GabonOgooue IvindoDHS20125/25/2021GabonOgooue LoloDHS20125/25/2021GabonOgooue MaritimeDHS20125/25/2021GabonWoleu NtemDHS20125/25/2021Gambia, TheBanjul + KanifingMICS20185/3/2021Gambia, TheBrikamaMICS20185/3/2021Gambia, TheMansakonkoMICS20185/3/2021Gambia, TheKerewanMICS20185/3/2021Gambia, TheKuntaurMICS20185/3/2021Gambia, TheJanjanburehMICS20185/3/2021Gambia, TheBasseMICS20185/3/2021GhanaWesternMICS20175/27/2021GhanaCentralMICS20175/27/2021GhanaGreater AccraMICS20175/27/2021GhanaVoltaMICS20175/27/2021GhanaEasternMICS20175/27/2021GhanaAshantiMICS20175/27/2021GhanaBrong AhafoMICS20175/27/2021GhanaNorthernMICS20175/27/2021GhanaUpper WestMICS20175/27/2021GhanaUpper EastMICS20175/27/2021GuineaBokeDHS20184/21/2021GuineaConakryDHS20184/21/2021GuineaFaranahDHS20184/21/2021GuineaKankanDHS20184/21/2021GuineaKindiaDHS20184/21/2021GuineaLabeDHS20184/21/2021GuineaMamouDHS20184/21/2021GuineaNzerekoreDHS20184/21/2021KenyaNairobiDHS20146/3/2021KenyaCentralDHS20146/3/2021KenyaCoastDHS20146/3/2021KenyaEasternDHS20146/3/2021KenyaNyanzaDHS20146/3/2021KenyaRift VelleyDHS20146/3/2021KenyaWesternDHS20146/3/2021KenyaNorth‐EasthernDHS20146/3/2021LesothoButha ButheMICS20184/30/2021LesothoLeribeMICS20184/30/2021LesothoBereaMICS20184/30/2021LesothoMeseruMICS20184/30/2021LesothoMafetengMICS20184/30/2021LesothoMohale's HoekMICS20184/30/2021LesothoQuthingMICS20184/30/2021LesothoQacha's NekMICS20184/30/2021LesothoMokhotlongMICS20184/30/2021LesothoThaba‐TsekaMICS20184/30/2021LiberiaBomiDHS20204/11/2021LiberiaBongDHS20204/11/2021LiberiaGbarpoluDHS20204/11/2021LiberiaGrand BassaDHS20204/11/2021LiberiaGrand Cape MountDHS20204/11/2021LiberiaGrand GedehDHS20204/11/2021LiberiaGrand KruDHS20204/11/2021LiberiaLofaDHS20204/11/2021LiberiaMargibiDHS20204/11/2021LiberiaMarylandDHS20204/11/2021LiberiaMontserradoDHS20204/11/2021LiberiaNimbaDHS20204/11/2021LiberiaRivercessDHS20204/11/2021LiberiaRiver GeeDHS20204/11/2021LiberiaSinoeDHS20204/11/2021MadagascarAnalamangaMICS20184/6/2021MadagascarVakinankaratraMICS20184/6/2021MadagascarItasyMICS20184/6/2021MadagascarBongolavaMICS20184/6/2021MadagascarHaute MatsiatraMICS20184/6/2021MadagascarAmron'i ManiaMICS20184/6/2021MadagascarVatovavy‐FitovinanyMICS20184/6/2021MadagascarIhorombeMICS20184/6/2021MadagascarAtsimo‐AtsinananaMICS20184/6/2021MadagascarAtsinananaMICS20184/6/2021MadagascarAnalanjirofoMICS20184/6/2021MadagascarAlaotra MangoroMICS20184/6/2021MadagascarBoenyMICS20184/6/2021MadagascarSofieMICS20184/6/2021MadagascarBetsiebokaMICS20184/6/2021MadagascarMelakyMICS20184/6/2021MadagascarAtsimo‐AndrefanaMICS20184/6/2021MadagascarAndroyMICS20184/6/2021MadagascarAnosyMICS20184/6/2021MadagascarMenableMICS20184/6/2021MadagascarDianaMICS20184/6/2021MadagascarSavaMICS20184/6/2021MalawiBlantyreMICS20204/29/2021MalawiKasunguMICS20204/29/2021MalawiMachingaMICS20204/29/2021MalawiMangochiMICS20204/29/2021MalawiMzaimbaMICS20204/29/2021MalawiSalimaMICS20204/29/2021MalawiThyoloMICS20204/29/2021MalawiZombaMICS20204/29/2021MalawiLilongweMICS20204/29/2021MalawiMulanjeMICS20204/29/2021MalawiOther NorthernMICS20204/29/2021MalawiOther CentralMICS20204/29/2021MalawiOther SouthernMICS20204/29/2021MaliKayesDHS20186/4/2021MaliKoulikoroDHS20186/4/2021MaliSikassoDHS20186/4/2021MaliSegouDHS20186/4/2021MaliMoptiDHS20186/4/2021MaliTombouctoeDHS20186/4/2021MaliGao and KidalDHS20186/4/2021MaliBamakoDHS20186/4/2021MauritaniaHodh CharghiMICS20154/21/2021MauritaniaAssabaMICS20154/21/2021MauritaniaGorgolMICS20154/21/2021MauritaniaBraknaMICS20154/21/2021MauritaniaTrarza incl. NouakchottMICS20154/21/2021MauritaniaAdrarMICS20154/21/2021MauritaniaNouadhibouMICS20154/21/2021MauritaniaTagantMICS20154/21/2021MauritaniaGuidimaghaMICS20154/21/2021MauritaniaTiris ZemmourMICS20154/21/2021MauritaniaInchiriMICS20154/21/2021MozambiqueNiassaDHS20114/13/2021MozambiqueCabo DelgadoDHS20114/13/2021MozambiqueNampulaDHS20114/13/2021MozambiqueZembeziaDHS20114/13/2021MozambiqueTeteDHS20114/13/2021MozambiqueManicaDHS20114/13/2021MozambiqueSofalaDHS20114/13/2021MozambiqueInhambaneDHS20114/13/2021MozambiqueGazaDHS20114/13/2021MozambiqueMaputoDHS20114/13/2021MozambiqueMaputo CidadeDHS20114/13/2021NamibiaCapriviDHS20136/14/2021NamibiaErongoDHS20136/14/2021NamibiaHardapDHS20136/14/2021NamibiaKarasDHS20136/14/2021NamibiaKhomasDHS20136/14/2021NamibiaKuneneDHS20136/14/2021NamibiaOhangwenaDHS20136/14/2021NamibiaKavangoDHS20136/14/2021NamibiaOmahekeDHS20136/14/2021NamibiaOmusatiDHS20136/14/2021NamibiaOshanaDHS20136/14/2021NamibiaOshikotoDHS20136/14/2021NamibiaOtjozondjupaDHS20136/14/2021NigerAgadezDHS20125/2/2021NigerDiffaDHS20125/2/2021NigerDossoDHS20125/2/2021NigerMaradiDHS20125/2/2021NigerTillaberi (incl. Niamey)DHS20125/2/2021NigerTahouaDHS20125/2/2021NigerZinderDHS20125/2/2021NigeriaAkwa IbomDHS20186/6/2021NigeriaAnambraDHS20186/6/2021NigeriaBauchiDHS20186/6/2021NigeriaEdoDHS20186/6/2021NigeriaBenueDHS20186/6/2021NigeriaBomoDHS20186/6/2021NigeriaCross RiverDHS20186/6/2021NigeriaAdamawaDHS20186/6/2021NigeriaImoDHS20186/6/2021NigeriaKadunaDHS20186/6/2021NigeriaKanoDHS20186/6/2021NigeriaKatsinaDHS20186/6/2021NigeriaKwaraDHS20186/6/2021NigeriaLagosDHS20186/6/2021NigeriaNigerDHS20186/6/2021NigeriaOgunDHS20186/6/2021NigeriaOndoDHS20186/6/2021NigeriaOyoDHS20186/6/2021NigeriaPlateauDHS20186/6/2021NigeriaRiversDHS20186/6/2021NigeriaSokotoDHS20186/6/2021NigeriaAbiaDHS20186/6/2021NigeriaDeltaDHS20186/6/2021NigeriaEnuguDHS20186/6/2021NigeriaJigawaDHS20186/6/2021NigeriaKebbiKogiDHS20186/6/2021NigeriaOsunDHS20186/6/2021NigeriaTarabaDHS20186/6/2021NigeriaYobeDHS20186/6/2021NigeriaBayelsaDHS20186/6/2021NigeriaEbonyiDHS20186/6/2021NigeriaEkitiDHS20186/6/2021NigeriaGombeDHS20186/6/2021NigeriaNassarawaDHS20186/6/2021NigeriaZamfaraDHS20186/6/2021NigeriaAbuja FCTDHS20186/6/2021South AfricaWestern CapeDHS20166/28/2021South AfricaEastern CapeDHS20166/28/2021South AfricaNorthern CapeDHS20166/28/2021South AfricaFree StateDHS20166/28/2021South AfricaKwaZulu NatalDHS20166/28/2021South AfricaNorth WestDHS20166/28/2021South AfricaGautengDHS20166/28/2021South AfricaMpumalangaDHS20166/28/2021South AfricaNorthern ProvinceDHS20166/28/2021TogoMaritime (inclusive Lome)MICS20176/28/2021TogoPlateauxMICS20176/28/2021TogoCentraleMICS20176/28/2021TogoKaraMICS20176/28/2021TogoSavanesMICS20176/28/2021UgandaCentral SouthDHS20165/4/2021UgandaCentral NorthDHS20165/4/2021UgandaKampalaDHS20165/4/2021UgandaEast CentralDHS20165/4/2021UgandaEasternDHS20165/4/2021UgandaNorthDHS20165/4/2021UgandaWest NileDHS20165/4/2021UgandaWesternDHS20165/4/2021UgandaSouthwestDHS20165/4/2021ZambiaCentralDHS20184/3/2021ZambiaCopperbeltDHS20184/3/2021ZambiaEastern + Northern + MuchingaDHS20184/3/2021ZambiaLuapulaDHS20184/3/2021ZambiaLusakaDHS20184/3/2021ZambiaNorth‐WesternDHS20184/3/2021ZambiaSouthernDHS20184/3/2021ZambiaWesternDHS20184/3/2021ZimbabweManicalandMICS20195/2/2021ZimbabweMashonaland CentralMICS20195/2/2021ZimbabweMashonaland EastMICS20195/2/2021ZimbabweMashonaland WestMICS20195/2/2021ZimbabweMashonaland NorthMICS20195/2/2021ZimbabweMashonaland SouthMICS20195/2/2021ZimbabweMidlandsMICS20195/2/2021ZimbabweMasvingoMICS20195/2/2021ZimbabweHarareMICS20195/2/2021ZimbabweBulawayoMICS20195/2/2021REFERENCESAbubakar, N. H., & Dasuki, S. I. (2018). Empowerment in their hands: Use of WhatsApp by women in Nigeria. Gender, Technology and Development, 22(2), 164–183. https://doi.org/10.1080/09718524.2018.1509490Abu‐Shanab, E., & Al‐Jamal, N. (2017). Exploring the gender digital divide in Jordan. Gender, Technology and Development, 19(1), 91–113. https://doi.org/10.1177/0971852414563201Aker, J. C., & Mbiti, I. M. (2010). Mobile phones and economic development in Africa. Journal of Economic Perspectives, 24(3), 207–232. https://doi.org/10.1257/jep.24.3.207Alexander, M., Polimis, K., & Zagheni, E. (2019). The impact of hurricane maria on out‐migration from Puerto Rico: Evidence from facebook data. Population and Development Review, 45(3), 617–630. https://doi.org/10.1111/padr.12289Allison, P. D. (2009). Fixed Effects Regression Models. SAGE Publications, Inc.Atake, E. H., & Gnakou Ali, P. (2019). Women's empowerment and fertility preferences in high fertility countries in Sub‐Saharan Africa. BMC Women's Health, 19(1), 54. https://doi.org/10.1186/s12905-019-0747-9Banerjee, A., La Ferrara, E., & Orozco, V. (2019). Entertainment, education, and attitudes toward domestic violence. AEA Papers and Proceedings, 109, 133–137. https://doi.org/10.1257/PANDP.20191073Bell, A., Fairbrother, M., & Jones, K. (2019). Fixed and random effects models: Making an informed choice. Quality & Quantity, 53(2), 1051–1074. https://doi.org/10.1007/S11135-018-0802-X/FIGURES/2Billari, F., Rotondi, V., & Trinitapoli, J. (2020). Mobile phones, digital inequality, and fertility: Longitudinal evidence from Malawi. Demographic Research, 42(37), 1057–1096. https://doi.org/10.4054/DEMRES.2020.42.37Bongaarts, J. (2017). Africa's unique fertility transition: Africa's unique fertility transition. Population and Development Review, 43, 39–58. https://doi.org/10.1111/j.1728-4457.2016.00164.xBongaarts, J. (2020). Trends in fertility and fertility preferences in sub‐Saharan Africa: The roles of education and family planning programs. Genus, 76(1), 32. https://doi.org/10.1186/s41118-020-00098-zBongaarts, J., & Watkins, S. C. (1996). Social interactions and contemporary fertility transitions. Population and Development Review, 22(4), 639. https://doi.org/10.2307/2137804Casterline, J. B. (2017). Prospects for fertility decline in Africa. Population and Development Review, 43, 3–18. https://doi.org/10.1111/padr.12055Cesare, N., Lee, H., McCormick, T., Spiro, E., & Zagheni, E. (2018). Promises and pitfalls of using digital traces for demographic research. Demography, 55(5), 1979–1999. https://doi.org/10.1007/S13524-018-0715-2/TABLES/2Coale, A. J., & Watkins, S. C. (1986). The Decline of Fertility in Europe, 484.DiMaggio, P., Hargittai, E., Neuman, W. R., & Robinson, J. P. (2001). Social implications of the Internet. Annual Review of Sociology, 27, 307–336. https://doi.org/10.1146/annurev.soc.27.1.307Eker, S., Garcia, D., Valin, H., & van Ruijven, B. (2021). Using social media audience data to analyse the drivers of low‐carbon diets. Environmental Research Letters, 16(7), 074001. https://doi.org/10.1088/1748-9326/abf770Fatehkia, M., Kashyap, R., & Weber, I. (2018). Using Facebook ad data to track the global digital gender gap. World Development, 107, 189–209. https://doi.org/10.1016/j.worlddev.2018.03.007Ferrara, E. L., Chong, A., & Duryea, S. (2012). Soap operas and fertility: Evidence from Brazil. American Economic Journal: Applied Economics, 4(4), 1–31. https://doi.org/10.1257/app.4.4.1Gittleman, J. L., & Kot, M. (1990). Adaptation: Statistics and a null model for estimating phylogenetic effects. Systematic Zoology, 39, 227–241.Gurumurthy, A., & Chami, N. (2014). Gender equality in the information society‐a review of current literature and recommendations for policy and practice.Hutcheon, J. A., Chiolero, A., & Hanley, J. A. (2010). Random measurement error and regression dilution bias. BMJ, 340, c2289. https://doi.org/10.1136/bmj.c2289Inkeles, A., & Smith, D. H. (1974). Becoming Modern. Harvard University Press.Kashyap, R., Fatehkia, M., Al Tamime, R., & Weber, I. (2020). Monitoring global digital gender inequality using the online populations of Facebook and google. Demographic research, 43, 779–816. https://doi.org/10.4054/DEMRES.2020.43.27Kietzmann, J. H., Hermkens, K., McCarthy, I. P., & Silvestre, B. S. (2011). Social media? Get serious! Understanding the functional building blocks of social media. Business Horizons, 54(3), 241–251. https://doi.org/10.1016/j.bushor.2011.01.005Kummu, M., Taka, M., & Guillaume, J. H. A. (2018). Gridded global datasets for gross domestic product and human development index over 1990–2015. Scientific Data, 5(1), 180004. https://doi.org/10.1038/sdata.2018.4Lerner, D. (1958). The Passing of Traditional Society. Free Press.Lesthaeghe, R. (1977). The decline of Belgian fertility, 1800–1970. Princeton University Press.Meta. (2019). An updata on how we are doing at enforcing our community standards. https://about.fb.com/news/2019/05/enforcing-our-community-standards-3/Nanfuka, J. (2022). Digital Access and Economic Transformation in Africa. Africa Paper Series.Norris, P. (2001). Digital Divide. Civic engagement, information poverty, and the internet worldwide. Cambridge University Press. https://doi.org/10.1017/CBO9781139164887Okenwa‐Emegwa, L., Lawoko, S., & Jansson, B. (2016). Attitudes toward physical intimate partner violence against women in Nigeria. SAGE Open, 6(4), 215824401666799. https://doi.org/10.1177/2158244016667993Ord, J. K., & Getis, A. (1995). Local spatial autocorrelation statistics: Distributional issues and an application. Geographical Analysis, 27(4), 286–306. https://doi.org/10.1111/J.1538-4632.1995.TB00912.XPezzulo, C., Nilsen, K., Carioli, A., Tejedor‐Garavito, N., Hanspal, S. E., Hilber, T., James, W. H. M., Ruktanonchai, C. W., Alegana, V., Sorichetta, A., Wigley, A. S., Hornby, G. M., Matthews, Z., & Tatem, A. J. (2021). Geographical distribution of fertility rates in 70 low‐income, lower‐middle‐income, and upper‐middle‐income countries, 2010–16: A subnational analysis of cross‐sectional surveys. The Lancet Global Health, 9(6), e802–e812. https://doi.org/10.1016/S2214-109X(21)00082-6Pierotti, R. S. (2013). Increasing rejection of intimate partner violence: Evidence of global cultural diffusion. American Sociological Review, 78(2), 240–265. https://doi.org/10.1177/0003122413480363Rampazzo, F., Bijak, J., Vitali, A., Weber, I., & Zagheni, E. (2021). A framework for estimating migrant stocks using digital traces and survey data: An application in the United Kingdom. Demography, 58, 2193–2218. https://doi.org/10.1215/00703370-9578562Ribeiro, F. N., Benevenuto, F., & Zagheni, E. (2020). How biased is the population of facebook users? comparing the demographics of facebook users with census data to generate correction factors. WebSci 2020—Proceedings of the 12th ACM Conference on Web Science, 325–334. https://doi.org/10.1145/3394231.3397923Robinson, L., Cotten, S. R., Ono, H., Quan‐Haase, A., Mesch, G., Chen, W., Schulz, J., Hale, T. M., & Stern, M. J. (2015). Digital inequalities and why they matter. Information, Communication & Society, 18(5), 569–582. https://doi.org/10.1080/1369118X.2015.1012532Rotondi, V., Kashyap, R., Pesando, L. M., Spinelli, S., & Billari, F. C. (2020). Leveraging mobile phones to attain sustainable development. Proceedings of the National Academy of Sciences, 117(24), 13413–13420. https://doi.org/10.1073/pnas.1909326117Scheerder, A., van Deursen, A., & van Dijk, J. (2017). Determinants of Internet skills, uses and outcomes. A systematic review of the second‐ and third‐level digital divide. Telematics and Informatics, 34(8), 1607–1624. https://doi.org/10.1016/J.TELE.2017.07.007Shapiro, D., & Tenikue, M. (2017). Women's education, infant and child mortality, and fertility decline in rural and urban sub‐Saharan Africa. Demographic research, 37(1), 669–708. https://doi.org/10.4054/DemRes.2017.37.21Smits, J., & Permanyer, I. (2019). The subnational human development database. Scientific Data, 6(1), 190038. https://doi.org/10.1038/sdata.2019.38Smits, J., & Permanyer, I. (2020). The Subnational Gender Development Index: Within‐country variation in gender (in)equality revealed. UNDP HDialogue Blog. https://globaldatalab.org/blog-SGDISnijders, T. A. B., & Bosker, R. J. (2011). Multilevel Analysis: An introduction to basic and advanced multilevel modeling. SAGE Publications, Inc.Stewart, I., Flores, R. D., Riffe, T., Weber, I., & Zagheni, E. (2019). Rock, rap, or reggaeton?: Assessing Mexican immigrants' cultural assimilation using Facebook data. The Web Conference 2019—Proceedings of the World Wide Web Conference, WWW 2019, 3258–3264. https://doi.org/10.1145/3308558.3313409Toffolutti, V., Ma, H., Menichelli, G., Berlot, E., Mencarini, L., & Aassve, A. (2020). How the Internet increases modern contraception uptake: Evidence from eight sub‐saharan African countries. BMJ Global Health, 5(11), e002616. https://doi.org/10.1136/bmjgh-2020-002616Varriale, C., Pesando, L. M., Kashyap, R., & Rotondi, V. (2022). Mobile phones and attitudes toward women's participation in politics. Sociology of Development, 8(1), 1–37. https://doi.org/10.1525/SOD.2020.0039Westoff, C. F., & Bankole, A. (1997). Mass Media and Reproductive Behavior in Africa. Demographic and Health Surveys Analytical Reports No 2, April.Westoff, C. F., & Koffman, D. A. (2011). The association of television and radio with reproductive behavior. Population and Development Review, 37(4), 749–759. https://doi.org/10.1111/j.1728-4457.2011.00455.xWorld Bank. (2021a). Birth rate, crude (per 1,000 people). https://data.worldbank.org/indicator/SP.DYN.CBRT.INWorld Bank. (2021b). Mobile cellular subscriptions (per 1000 people). https://data.worldbank.org/indicator/IT.CEL.SETS.P2Zagheni, E., Weber, I., & Gummadi, K. (2017). Leveraging facebook's advertising platform to monitor stocks of migrants: Leveraging facebook's advertising platform. Population and Development Review, 43(4), 721–734. https://doi.org/10.1111/padr.12102

Journal

"Population, Space and Place"Wiley

Published: Nov 24, 2022

Keywords: fertility; social media; sub‐Saharan Africa

References