Is Education Neglected in Natural Resources-Rich Countries? An Intergenerational Approach in Africa

Jean-Marc B. Atsebi; Rasmané Ouedraogo; Regina S. Séri

doi:10.5089/9798400218071.001.A001

Jump to Content Jump to Main Navigation

Is Education Neglected in Natural Resources-Rich Countries? An Intergenerational Approach in Africa

Author:

Jean-Marc B. Atsebi

Jean-Marc B. Atsebi
Search for other papers by Jean-Marc B. Atsebi in
Current site
Google Scholar
PubMed

Rasmané Ouedraogo

0000000404811396

https://isni.org/isni/0000000404811396 International Monetary Fund

Search for other papers by Rasmané Ouedraogo in
Current site
Google Scholar
PubMed

, and

Regina S. Séri

Regina S. Séri
Search for other papers by Regina S. Séri in
Current site
Google Scholar
PubMed

Publication Date:: 29 Jul 2022

eISBN:: 9798400218071

ISBN:: 9798400218071

Language:: English

Keywords:: educational IM; mineral discovery; upward mobility; mining production; tertiary IM; Non-renewable resources; Natural resources; Africa

Download PDF (5 MB)

The literature on the effects of natural resources on education is mixed and inconclusive. In this paper, we adopt an innovative approach by exploring the effects of mineral discoveries and productions on intergenerational educational mobility (IM), linking parents to the children education levels for more than 14 million individuals across 28 African countries and 2,890 districts. We find that mineral discoveries and productions positively affect educational IM for primary education in Africa for individuals exposed to the mineral sites and living in districts with discoveries. Specifically, the probability of upward primary IM increases by 2.7 percentage points (pp.) following mineral discoveries and 6.7 pp. following mineral productions. Downward primary IM decreases by 1.2 pp. following both mineral discoveries and productions. These positive effects are increasing for individuals born later after discoveries and productions, for males, and individuals living in the urban area. However, no significant effects are found for secondary and tertiary educational IM. Finally, we explore the income and returns to education channels through which mineral discoveries and productions affect educational IM.

Abstract

I. Introduction

Pre-COVID-19, Africa emerged out of decades of stagnant and unstable economic growth since mid-1990s, with significant progress made on education and human capital. According to Young (2012), Sub-Saharan living standards have, for the past two decades, been growing about 3.4 to 3.7 percent per annum, reflecting the African “growth miracle”. This “growth miracle” has been accompanied by significant improvements in education. Indeed, in Sub-Saharan Africa (SSA), gross enrollment in primary education almost doubled from 54 percent in 1970 to 99 percent in recent years. For secondary and tertiary education, it has been more than three and six times higher in recent years compared to 1970, from 13 to 43 percent and 1.4 to 9.4 percent, respectively (World Bank, 2020). These improvements have been observed for both rural and urban areas, as well as females and males. As a result, intergenerational mobility (IM) in education––which measures the levels of children education relative to their parents––has also significantly increased across African countries, above the level in Latin America. But it remains lower in the region compared to Western, Asian, and Eastern Europe countries (Hertz et al., 2008; Azomahou and Yitbarek, 2020; Henn and Robinson, 2021). As shown by Henn and Robinson (2021), actual and perceived social and educational intergenerational mobility constitute one of the three Africa’s latent assets that will drive its economic prosperity and bright future.

Taking stock of this progress in education, we analyze the potential effects of mineral discoveries and productions on educational IM across 28 African countries and 2,890 districts. Few studies have investigated the determinants of educational intergenerational mobility in Africa. They found that individuals and local characteristics (e.g., gender, race, urbanization, etc.), access to markets, quality of education, globalization, and investments in physical and human capital affect educational mobility (Alesina et al., 2021; Azomahou and Yitbarek, 2020; Baah and Eshun, 2020; Nimubona and Vencatachellum, 2007). To the best of our knowledge, only Alesina et al. (2021) discussed the effects of natural resources on educational IM. They show a weak association between oil, gas, and diamond discoveries and educational IM. They explain this weak association by opposing mechanisms as natural resources might be a curse, and they may also represent a wealth spurring human capital accumulation and structural transformation. Still, their analysis of the effects of natural resources on educational IM is very limited and discussed in a short paragraph as this is not their focus.

In this paper, we fill this gap in the existing and inconclusive literature. Mineral discoveries are generally known as being a curse or sources of difficulties and fragility for most African countries as illustrated by the resource curse literature. The growth in the mining sector does not necessary shift an economy towards better industry processing, services, i.e., structural transformation, education, health, job creation, and inclusive growth and development. In contrast, they might have some positive effects, especially on educational IM. Mining activities may create new opportunities that will increase households’ income (Becker and Tomes, 1979 Weber-Fahr et al., 2002; Loayza et al., 2013), enabling them to invest more in children education. They may also favor a structural transformation in the districts with natural resources (Cavalcanti et al. 2019), and therefore an increase of the returns to education, i.e., income or weath induce by education (Torche, 2014; Bütikofer et al., 2018). Finally, they may also support the provision of insfrastructures (in education in particular) financed by the revenues from the resources (Witter and Jakobsen, 2017). This analysis is particularly relevant as the African continent is home to an abundance of mineral resources.¹ It hosts 30 percent of the world’s mineral reserves, 40 percent of the world’s gold and up to 90 percent of some minerals like chromium and platinum.² According to Minex Consulting database (2019),³ 969 normal to super-giant mineral discoveries occurred in Africa between 1950 and 2019, and 396 of them (40 percent) since 2000. Moreover, the exploitation of mineral resources makes a significant contribution to the development of African economy. According to the IMF, the mining sector accounted for 8.8 percent of GDP and 51.2 percent of total exports in Sub-Saharan countries over the period 2009–19.

There is a vast and inconclusive literature on the effects of natural resources. At the macro level, most papers found that natural resources have been a curse than a blessing, as well illustrated by the resource curse literature (e.g., Corden and Neary, 1982; Sachs and Warner, 1995, 2001; Kretzmann and Nooruddin, 2005).⁴ Others studies found positive effects on foreign direct investments in non-resource sectors (Toews and Vezina, 2017), or ambiguous effects on macroeconomic activity and financial conditions (Arezki et al., 2017; Seri, 2021). At the local level, a positive effect of natural resources has emerged in recent analyses focusing on African countries or other developing countries. (e.g., Fisher et al., 2009; Cust and Mensah, 2020; Bhattacharyya and Mamo, 2021). Specifically, the literature of the effects of natural resources on education is also inconclusive. Some papers found that natural resources exert a decrease in education level in developing countries (Leamer et al., 1999; Gylfason, 2001 and Ahlerup et al., 2020), while others revealed that natural resources are positively associated with human capital accumulation, notably through the increase of public spending in education (Kim and Lin, 2017; Pegg, 2010; Stijns, 2006). Other studies highlighted that the effects depend on the quantity or the quality of education, the levels of education or the characteristics of individuals (e.g., Farzanegan and Thum, 2020; Gradstein and Ishak, 2020).

Our paper complements these studies by exploring the effects of mineral discoveries and productions on educational IM. We use a large dataset of more than 14 million individuals from 28 African countries, 2,890 districts. To do so, we rely on two main sources: i) the Integrated Public Use Microdata Series (IPUMS), and ii) the Minex Consulting datasets. We first start by providing a panorama of the trends, dynamics, and disparities of educational IM across countries, regions of the continent, and the characteristics of the individuals, using the conditional absolute measure of IM as in Alesina et al. (2021). Second, we empirically study the effects of mineral discoveries and productions on primary and secondary/tertiary educational IM by employing a generalized difference-in-differences method in a quasi-natural experiment. Our quasi-natural experiment relies on the plausible exogeneity of mineral discoveries that revert specific characteristics, specifically the unpredicted time of discoveries, the unpredicted geographical location, and the lag between the natural resources discoveries and beginning of production (Horn, 2011; Khan et al., 2016; Arezki et al., 2017; Cavalcanti et al., 2017). Third, we explore the channels through which mineral resources discoveries and productions may affect educational IM such as job creation, and the returns to education.⁵

We find that mineral discoveries and productions positively affect educational IM for primary education in Africa. Indeed, the probability of upward primary educational IM, i.e. the probability for a child born from uneducated parents or parents with less than primary education attainment to achieve at least primary education, increases by 2.7 pp. following mineral discoveries and 6.7 pp. following mineral productions. The probability of downward primary educational IM, i.e. the probability for a child born from parents with at least primary education attainment to be uneducated or have less than primary education attainment, decreases by 1.2 pp. following both mineral discoveries and productions. About the size of the effects, it is relatively small when compared to the increase of educational IM across the different cohorts, signaling that, other factors have also played a significant role in improving IM in Africa.

To put our findings into perspective and extrapolate them to Africa, we show that the number of individuals born up to 15 years after mineral discoveries and who have completed at least primary education while their parents have not, increases by 662 thousand in Africa over the period 1950-2000. This figure stands at 581 thousand for individuals born up to 15 years after mineral productions. Similarly, the number of individuals born up to 15 years after mineral discoveries and who have not completed at least primary education while their parents have completed it, decreases by 371 thousand. This figure stands at 124 thousand for individuals born up to 15 years after mineral productions. These figures would have been even higher to millions of individuals if we would have considered all the individuals born after the discoveries and productions, and not only those born up to 15 years after the event. However, our results show that the effects of mineral resource discoveries and productions on the probability of upward and downward secondary and tertiary educational IM are not statistically significant.

Moreover, we also explore the dynamic effects of mineral discoveries and productions on educational IM, i.e. whether the effects vary with the distance between individuals’ birth years and the year of mineral discoveries and productions, and test the assumption of parallel trend of the GDID model by estimating a leads and lags model following Angrist and Pischke (2009) and Maurer (2019). Our results show that mining activities positively affect primary educational IM for all age groups, but the effects are non-significant for secondary/tertiary educational IM, confirming our baseline findings. Interestingly, we find that the positive effects of mineral activities on educational IM are increasing for individuals born later after the discovery and beginning of mining production, while being non-significant or low for individuals born before the discovery or production, also confirming that the parallel trend assumption is verified. Our baseline results are robust to several robustness checks. We also analyze the sensitivity of our findings across African regions, size of mineral discoveries, gender, and urban-rural living area. Overall, our findings show that the effects of mining activities are different by African regions and size of the discoveries. Also, the positive effects are higher for males than females (only for primary education), and individuals living in urban than rural areas (both for primary and secondary/tertiary education).

Furthermore, we discuss two transmission channels through which the positive effects of mineral discoveries and productions on educational primary IM operate, including the income effect proxied by parents working in the mining sector, and the returns to education. First, our results show that the mining sector creates new job and income opportunities for parents, allowing them to invest more in their child’s education attainment (Becker and Tomes, 1979). Second, we uncover that the economic dynamism and creation of new jobs following the discoveries of mineral resources lead to an increase in the demand for skilled workers and thereby boosting the returns to education (Torche, 2014).

Our paper unveils the potential impact of mineral discoveries and productions on educational IM in Africa. It shows that, on average, mineral discoveries and productions have led to an improvements of primary education and intergenerational mobility at local levels. It adds to the existing knowledge of the effects of natural resources in general, and on education in particular, and identifies some potential channels through which they impact educational IM. Our paper has many policy implications. We discussed them in the concluding section. In short, this paper calls for a better management of the resources by the government and companies. Adequate policies should be put in place to extract the benefits of the resources, including policies aiming at creating jobs and facilitating entreprise development, and ensuring equitable access to the benefits of the resources by all people independently of gender and location.

The rest of the paper is organized as follows. Section II reviews the literature and places this study among the existing papers. Section III discusses the data, explains the construction of the educational IM, and presents some stylized facts on educational IM in Africa by decade, gender, at district and country level. Section IV provides stylized facts on both educational IM and mineral discoveries and productions. Section V describes the methodology. Section VI presents our main findings. Section VII explains the transmission channels. Sections VIII and IX discuss the robustness checks and the sensitivity of our findings, respectively. Section X concludes and discusses potential policy implications.

II. Review of literature

Our paper closely relates to three strands of the literature, notably the general effects of natural resources, the relationship between natural resources and education, and the literature on the determinants of intergenerational mobility.

A. General effects of natural resources

At the macroeconomic level, most papers in this literature show that natural resources have been a curse than a blessing, as well illustrated by the resource curse literature. They found that natural resources are generally associated with the deterioration of economic and institutional conditions, the occurrence of conflicts, an appreciation of real exchange rate, which induces a loss of competitiveness and de-industrialization of the economy, as well as with weak fiscal policy stance and unsustainable debt accumulation (e.g., Corden and Neary, 1982; Sachs and Warner, 1995, 2001; Kretzmann and Nooruddin, 2005; Collier and Hoeffler, 2005; Ross, 2004, 2006; Van Der Ploeg, 2011; Keen, 2012; Van Der Ploeg and Poelhekke, 2017). However, the macroeconomic effects of natural resource discoveries of oil, gas, and minerals on economic activity seem to be mixed. While some papers find negative impacts of giant discoveries on fiscal policy, debt, conflict, poverty, and inequality (Kretzmann and Nooruddin, 2005; Harding et al., 2020; Lei and Michaels, 2014; Tsui, 2011; Smith and Wills, 2018), others show positive effects on foreign direct investments in non-resource sectors (Toews and Vezina, 2017), or ambiguous effects on macroeconomic activity and financial conditions (Arezki et al., 2017; Seri, 2021).

At the local level, a positive effect of natural resources has emerged in more recent analysis focusing on African countries or other developing countries. They show that natural resources are associated with a reduction of inequality, poverty, and an increase of living standards, income, and welfare. In fact, Goderis and Malone (2011) find that resource exploitation booms reduce income inequality in resource-rich countries, while Fisher et al. (2009) show an evidence of the reduction of poverty in the mineworkers’ population in Tanzanian artisanal mines of gold and diamond. Zabsonré et al. (2018) reveal for Burkina Faso that gold exploitation led to better living standards, an increase in per capita household expenditures, and a reduction of poverty in the mining areas. Marlet (2020), using mining exploitation in Ghana, finds that mining activities tend to increase migration flows up to 200 km from the treated district by reducing migration costs through the construction of roads and infrastructures. Moreover, they also induce an increase of income and improvement of welfare by 1.3 percent. In contrast, some papers find that mining activities can create some environmental issues by increasing pollution and metal toxicity (see, e.g., Hausermann et al., 2018; von der Goltz and Barnwal, 2019).

The literature also supports the benefits and positive role of natural resources discoveries on local economic development, governance and conflicts, provisions of public goods and welfare. Cavalcanti et al. (2019) find evidence of a positive impact of oil and gas discoveries on local development and urbanization in Brazil. Cust and Mensah (2020) reveal that oil, gas, and mineral discoveries positively impact the citizen’s expectations, which is materialized by a decrease in outward migration and an increase in fertility in the short term. Bhattacharyya and Mamo (2021) show that oil and mineral discoveries reduce the likelihood of conflict in 48 African countries, which is mainly driven by an improvement of economic development and efficient political distribution patronage in districts with discoveries.

B. Natural resources and education

The literature on the effects of natural resources on education is also inconclusive, but it identifies channels through which natural resources may affect education. While some papers find that natural resources favor a decrease in education levels in developing countries, others rather point out to a positive effect. On the negative effects, Leamer et al. (1999) find that the abundance of natural resources entails a delay of industrialization, and it lowers education levels in Latin American resource-rich countries as workers do not need high skills to work in the natural resources sector. Gylfason (2001), Cokx and Francken (2016), and Ahlerup et al. (2020) show that natural resources crowd out investments in education in resource-rich countries, decrease public education expenditures relative to GDP, and reduce educational attainment, respectively. On the positive effects, some papers find that natural resources abundance is positively associated with human capital accumulation, notably through the increase of public spending in education (see, e.g., Stijns, 2006; Kim and Lin, 2017, and Pegg, 2010). A possible channel is that the mining activity may create new opportunities that will increase households’ income (Becker and Tomes, 1978, Weber-Fahr et al., 2002; Loayza et al., 2013), enabling them to invest more in children education. It may also favor a structural transformation in the district with natural resources (Cavalcanti et al. 2019), and therefore an increase of the returns to education, i.e., income or wealth induce by education (Torche, 2014; Bütikofer et al., 2018). Finally, it may also support the provision of insfrastructures (in education in particular) financed by the revenues from natural resources (Witter and Jakobsen, 2017). We test these channels for the effects of mineral discoveries and productions on educational IM.

Other studies find that the effects of natural resources on education may diverge depending on whether the focus is the quantity or the quality of education, the levels of education, and characteristics of individuals (age, gender). In fact, Farzanegan and Thum (2020) show a positive effect of oil rents on the quantity of education measured by government spending in primary and secondary education, particularly in countries with sound quality of institutions. In contrast, they find a negative effect of oil rents on the quality of education, defined as “an increase in cognitive skills obtained from an additional year of schooling”. This negative effect is driven by the low demand and supply for high-quality education. On the demand side, the phenomenon of resource curse in those countries, by leading to an increase in the size of the non-tradable sector, requires less skilled workers with lower level of human capital. On the supply side, the lower incentive to attract local qualified teachers and the lack of long-term opportunities for foreign or migrant teachers reduce the quality of education. Moreover, Gradstein and Ishak (2020), using IPUMS data on 18 African countries, find that oil price booms occuring in early childhood (ages 0-4) enhance educational attainment and other derived outcomes, but reduce them when occuring in the adolescence (ages 10-14), especially for girls.

C. Intergenerational mobility in education and its determinants

Very few papers exclusively focus on educational IM in Africa. Focusing on South Africa, Nimubona and Vencatachellum (2007) show that educational IM is higher for white than black people. They find that access to credit market and quality of schools are the main determinants of lower educational IM for black people. Baah and Eshun (2020) reveal that economic and educational IM in Ghana is one of the lowest in the world. In addition, they find that globalization enhances IM, thereby recommending policies aiming at expanding globalization. Moreover, they find that FDI and expansionary fiscal policy improve IM while unemployment has an exactly opposite effect on it. Other papers have conducted cross-country analysis based on several African countries. Alesina et al. (2021) employ measures of absolute mobility to estimate educational intergenerational mobility since independence using census data on 27 African countries. After maping IM cross-country and within country variation, they find that colonial investments in the transportation network and missionary activities were associated with higher upward mobility. Intergenerational mobility was also higher in regions close to the coast and national capitals as well as in rugged areas without malaria. Upward mobility is higher and downward mobility is lower in regions that were more developed at independence, with higher urbanization and employment in services and manufacturing. Finally, they also reveal that early exposure of children to regions with higher (lower) upward IM significantly improve (decrease) the likelihood of completing primary schooling. In addition, Azomahou and Yitbarek (2020) analyze the educational IM across 9 Sub-Saharan African countries over 50 years, using two measures of intergenerational educational persistence. They reveal that educational intergenerational persistence has reduced among the birth cohorts in all countries, particularly after the 1960s due to huge investments in human capital following independence and drastic changes in the educational systems. Even in the light of declining educational intergenerational persistence in the region, countries such as Ghana, Guinea, Nigeria, and Uganda experienced higher intergenerational mobility while Comoros and Madagascar had the lowest. Also, intergenerational persistence in education was found to be stronger from mothers to their children, and daughters’ education is more correlated with their parents’ education than that of sons.

More generally, the trends and drivers of intergenerational mobility or persistence in education have been studied in the literature (see, e.g., Corak, 2013; Chetty et al., 2014; Chetty and Hendren, 2018; Howell, 2019; Engzell and Tropf, 2019). Overall, the intergenerational mobility in education has increased over time, but some heterogeneities and disparities across regions remain. Hertz et al. (2008) analyze trends in the intergenerational persistence of education over 50 years in 42 countries, including 19 developing countries and 3 SSA countries.⁶ They find that the educational IM has improved in almost all regions of the world.⁷ The western developed countries have higher educational IM than in any region of the world, especially for the Nordic countries. They are followed by the Eastern bloc countries and Asian countries. However, educational IM is lower in Latin American countries and African countries. Interestingly, Henn and Robinson (2021), using the 2015 World Bank Intergenerational Database, find that educational IM is higher in Africa than in South Asia, MENA, and Latin American countries, with countries like Botswana, Kenya, Mauritania, and Cape Verde displaying approximately the same educational IM as high-income countries.⁸

III. Data sources and IM index

A. Data sources

Our data mainly come from two sources. We use data on education and individual characteristics from Integrated Public Use Microdata Series (IPUMS)⁹ and data on mineral discoveries and production from Minex Consulting Dataset (2019).

1) IPUMS data

The Integrated Public Use Microdata Series (IPUMS) database covers 82 national censuses surveys from 28 African countries: Benin, Botswana, Burkina Faso, Cameroon, Egypt, Ethiopia, Ghana, Guinea, Kenya, Lesotho, Liberia, Malawi, Mali, Mauritius, Morocco, Mozambique, Nigeria, Rwanda, Senegal, Sierra Leone, South Africa, South Sudan, Sudan, Tanzania, Togo, Uganda, Zambia, and Zimbabwe.¹⁰ It contains information on more than 130 million of individuals, including on their demographic characteristics, occupation, household members, the relationship between household members, and place of residence. Regarding education, IPUMS reports data on the total years of schooling and whether the individuals completed primary, secondary, tertiary education levels. For this study, we follow Alesina et al. (2021) and use the educational attainment for both parents and childrens instead of the years of schooling given their higher coverage. Our sample is based on the availability of information on district level and residency, education and individuals characteristics (gender, age) as well as whether individuals co-reside with their biological/step parents or immediate older generation. We have harmonized the boundaries of districts following Alesina et al. (2021) to deal with admininistrative boundaries changes.¹¹ We focus on the individuals aged between 16 and 50, and born between 1950 and 2000.¹² The final sample covers more than 14 million individuals across 2,890 districts. Table 20 in Appendix A describes the data for each census and country.

2) Mineral discoveries data

Our data on mineral discoveries and productions in Africa come from Minex Consulting Dataset (2019). This dataset provides geolocalized information on discoveries, their size (moderate, major, giant, super-giant), the status of the mine (closed, feasibility study, operating, underdeveloped), and the type of minerals.¹³ After merging this dataset with the IPUMS data, we identify 331 districts out of the total of 2890 in which mineral sites were discovered or entered in production. Figure 11 in Appendix C displays the evolution of the number of discoveries over time in all African countries. 969 mineral discoveries occurred in Africa between 1950 and 2019, of which 573 (60 percent) between 1950 and 2000. This study covers 406 mineral discoveries in 28 African countries over the period 1950–2000 (i.e., 71 percent of the 573 mineral discoveries). We therefore cover a large share of mineral discoveries in Africa.

Figure 1 maps the location of these discoveries across Africa by the size of mineral discoveries. We observe that mineral discoveries have been concentrated in Southern Africa (48.3 percent) and Western and Central Africa (34.2 percent). A relatively few discoveries occurred in Eastern Africa (11.9 percent) and Northern Africa (5.7 percent). Looking at the sizes of mineral discoveries, they have been mostly moderate (45.3 percent), followed by major (29.8 percent), giant (21.7 percent) and super-giant giant (3.2 percent). Moderate and major discoveries were mainly found in Western and Central Africa while giant and super-giant were located in Southern Africa. The minerals discovered were mostly gold (34 percent), bulk metals (18.4 percent), precious minerals (15.8 percent), and base metals (15 percent). Table 22 and Table 23 in Appendix C present some statistics on the mineral discoveries by regions, size, and types.

B. Construction of educational IM

In this paper, we use absolute educational IM measures as in Alesina et al. (2021).¹⁴ We define both an upward and downward educational IM for the primary and secondary/tertiary education levels. First, upward primary educational IM is defined as the probability for a child born from uneducated parents or parents with less than primary education attainment to achieve at least primary education. Downward primary educational IM is defined as the probability for a child born from parents with at least primary education attainment to be uneducated or have less than primary education attainment. Second, upward secondary/tertiary educational IM is defined as the probability for a child born from parents with at most primary education background to achieve at least secondary education. Downward secondary/tertiary educational IM is defined as the probability for a child born from parents with at least secondary education background to achieve primary education or be uneducated.¹⁵ To identify the old generation benchmark for each child, we use the average of education attainment for their biological/step parents, rounded to the nearest integer. In the robustness section, we use the minimum or maximum of the education levels of the biological/step parents as benchmark. We also consider the immediate older generation and broaden the definition of parental authority to include uncles/aunts (in law), parents-in-law, grand-parents, and grand-uncles/aunts in the reference group, to take into account fostered, abandoned, or orphan children.

Practically, first, for each individual (parents and children), we compute two educational attainment variables P_jith and ST_jith measuring the primary and secondary/tertiary educational attainment, respectively. P_jith takes that value of one if the individual j born in district i and year h, and surveyed in year t has completed at least the primary education, and zero otherwise. Similarly, ST_jith takes that value of one if the individual j born in district i and year h, and surveyed in year t has completed at least the secondary education. Second, for each child j, we compute two averaged measures of parents’ educational attainment, PP_jith and PST_jith, as the average of P_jith and ST_jith rounded to the nearest integer, respectively, for the biological/step parents if both cohabit with the child, or for only the father/step-father or mother/step-mother if the child lives with only one of its parents. Third, we compare the educational attainment for each child j cohabiting with at least one parent to the average educational attainment of its parents and obtain our absolute measures of educational as follows:¹⁶

i. Upward primary IM: $\begin{matrix} {I M U P}_{j i t h} = {\begin{cases} 1 i f P_{j i t h} = 1 a n d {P P}_{j i t h} = 0 \\ 0 i f P_{j i t h} = 0 a n d {P P}_{j i t h} = 0 \end{cases} & (1) \end{matrix}$
ii. Downward primary IM: $\begin{matrix} {I M D P}_{j i t h} = {\begin{cases} 1 i f P_{j i t h} = 0 a n d {P P}_{j i t h} = 1 \\ 0 i f P_{j i t h} = 1 a n d {P P}_{j i t h} = 1 \end{cases} & (2) \end{matrix}$
iii. Upward secondary/tertiary IM: $\begin{matrix} {I M U S T}_{j i t h} = {\begin{cases} 1 i f {S T}_{j i t h} = 1 a n d {P S T}_{j i t h} = 0 \\ 0 i f {S T}_{j i t h} = 0 a n d {P S T}_{j i t h} = 0 \end{cases} & (3) \end{matrix}$
iv. Downward secondary/tertiary IM: $\begin{matrix} {I M D S T}_{j i t h} = {\begin{cases} 1 i f {S T}_{j i t h} = 0 a n d {P S T}_{j i t h} = 1 \\ 0 i f {S T}_{j i t h} = 1 a n d {P S T}_{j i t h} = 1 \end{cases} & (4) \end{matrix}$

C. Cohabitation-selection issues

Our analysis might be subject to cohabitation-selection issues, as our sample includes both youth children and adults, and individuals’ educational attainment is also determined by their cohabitation with their parents or not.¹⁷ For adults, the cohabitation-selection issues are more severe as the intensity of self-selection is increasing with age. Moreover, it is more accentuated for adults’ women, especially for those who got married at younger age, as some African countries are patrilocal (Heckert et al., 2021). Since our focus is also on secondary education, our sample must include adults at age where secondary education is mostly completed despite the likelihood of co-residing with parents at this age being low (see Figure 2).

Interestingly, we show in Figure 2 that the cohabitation selection is independent from the discoveries of natural resources. Both districts with and without discoveries exhibit the same patterns of cohabitation by age and gender, therefore the cohabitation selection issues might have limited impact on our analysis.¹⁸ If the cohabitation rate was higher in districts with discoveries than districts without discoveries, it would have caused our estimates of the effects of discoveries on upward (downward) IM to be upward (downward) bias. Indeed, as we expect discoveries to have a positive effect on educational IM and given that individuals living with their parents are likely to have higher educational attainments, then, higher cohabitation rates in districts with discoveries than without discoveries would have resulted in higher positive estimates of the effects of discoveries on educational IM.

We further investigate the potential bias of the cohabitation selection. We do so in Table 1 by computing the differences in educational attainment for both individuals co-residing with biological/step parents (selected individuals) and those that do not, both in districts with and without discoveries, separately. We test the significance of the differences through a Khi-2 test. We show that the unconditional likelihood of not completing primary education is higher for individuals not co-residing with their biological/step parents than those who do.

^{Table 1:}

Differences in educational attainment between individuals living with biological/step parents or not.

^{Table 1:}

Differences in educational attainment between individuals living with biological/step parents or not.

	(A) With discoveries				(B) Without discoveries
	Without relatives	With relatives	Differences	(4) Without relatives	(5) With relatives	(6) Differences
Less than primary completed	47.31	40.52	-6.79	56.63	40.08	-16.55
Primary completed	36.06	42.38	6.32	27.19	37.22	10.03
Secondary completed	14.4	15.72	1.32	13	19.42	6.42
Tertiary completed	2.23	1.37	-0.86	3.18	3.28	0.1
Total	100	100		100	100
Khi-2 tests p-value		0.000			0.000

^{Table 1:}

Differences in educational attainment between individuals living with biological/step parents or not.

	(A) With discoveries				(B) Without discoveries
	Without relatives	With relatives	Differences	(4) Without relatives	(5) With relatives	(6) Differences
Less than primary completed	47.31	40.52	-6.79	56.63	40.08	-16.55
Primary completed	36.06	42.38	6.32	27.19	37.22	10.03
Secondary completed	14.4	15.72	1.32	13	19.42	6.42
Tertiary completed	2.23	1.37	-0.86	3.18	3.28	0.1
Total	100	100		100	100
Khi-2 tests p-value		0.000			0.000

This difference is more accentuated in districts without discoveries than districts with discoveries. Similarly, the likelihood of completing primary and secondary education is higher for those living with their biological/step parents than those who do not, also more pronounced in districts without discoveries. These differences are significant as indicated by the Khi-2 tests. As a result, if we were to consider individuals not co-residing with parents in our study, the level of educational attainment would have been on average lower in districts without discoveries than districts with discoveries. Thus, if the cohabitation selection creates any biases, our estimates of the effects of discoveries on upward (downward) IM would be downward (upward) bias. Therefore, as we expect discoveries to have a positive effect on IM (i.e., increase upward IM and decrease downward IM), these effects should be considered as a lower bound.

D. Stylized facts on educational IM in Africa

In this section, we briefly describe the trends of educational IM in Africa as well as their disparities across countries, gender, and residency. To do so, we calculate conditional educational IM, netting country/districts, cohort and census effects. Specifically, we regress educational IM indices on country or district fixed effects α_i, cohort fixed effects γ_t, and census-year fixed effects λ_t. The model is as follows:

IM_jit = α_i + γ_t + δ_t + ε_jit (5)

Country or district fixed effects α_i reflect the conditional likelihood of each type of educational IM at the country or district levels, netting the cohort and census effects. Cohort fixed effects γ_t reflect the conditional likelihood of each type of educational IM, netting the country/district and census effects. We do so to better compare the educational IM across individuals, cohorts, districts, and over time, especially by purging the differences between countries/districts, cohorts, and census-year specific effects. In addition, we estimate conditional educational IM by country and gender, country and individuals’ residency (urban and rural), by cohort and gender, cohort and residency, and cohort and discovery dummy by introducing country and gender fixed effects, country and residency fixed effects, cohort and gender fixed effects, cohort and residency fixed effects, and cohort and discovery fixed effects, respectively.¹⁹

1) Trends of IM by decade

Overall, we observe that primary and secondary/tertiary educational IM have significantly improved in Africa over time, independently of gender and residency (urban/rural areas). The average trends of primary and secondary/tertiary educational IM are displayed in Figure 3 for both upward and downward mobility, respectively, and for five cohorts between 1950 and 2000. We show that upward primary IM has steadily increased across cohorts, from 35.1 percent for the 1950s cohort to 57.7 percent for the 1990s cohort. Similarly, downward primary IM has steadily decreased, but at a slower pace, from 29.8 to 23 percent between the 1950s and 1990s cohorts, respectively. Moreover, secondary and tertiary educational IM have experienced similar trends. Upward secondary and tertiary educational IM has steadily increased from 10.8 to 32.9 percent, while downward secondary and tertiary educational IM has steadily decreased from 45.1 to 35.8 percent between the 1950s and 1990s cohorts, respectively. In contrast to primary education level, downward IM has always been elevated than upward IM for secondary and tertiary levels, but the gap has closed over time. Finally, upward (downward) educational IM has been higher (lower) at primary level than secondary and tertiary level.

We next look at these trends by gender in Figure 4. In general, both males and females have had an increase in the probability of upward educational IM and a decline of the probability of downward educational IM for both primary and secondary/tertiary levels. We also observe that the gender gap in favor of males has narrowed over time, for instance, with the probability of upward (downward) secondary/tertiary educational IM for females being higher (lower) than that of males in the last decades.

Finally, we explore the trends by residency as presented in Figure 5. We find that the general trends are also confirmed for both individuals living in urban and rural areas. More specifically, we show that educational IM has always been higher in urban areas than in rural areas. Although the residency gap has diminished over time, it has remained significant across cohorts.

2) Educational IM at country and district levels

a. Country-level educational IM

We display in Figure 6, the country-level educational IM. It shows that educational IM is uneven across countries, and that upward and downward IM are negatively correlated, i.e., countries with the highest upward IM tend to have the lowest downward IM, and inversely. These findings hold for both primary and secondary/tertiary education levels. Upward primary IM ranges between 13 percent in South Sudan and 98 percent in Mauritius, and downward primary IM between close to zero for Egypt, Mauritius and Botswana and 58 percent in South Sudan. Upward secondary/tertiary IM ranges about 14 percent in Sierra Leone and Sudan and more than 70 percent in Egypt and Nigeria, while downward secondary/tertiary IM is between 1 percent in Egypt and 71 percent in Togo.

b. Country-level educational IM by gender

We also present the country-level gender gap in educational IM in Figure 12 in Appendix D. First, we show that gender gap, in favor of males for the primary level, is more pronounced for individuals living in countries with lowest values of upward IM, and in countries with highest values of Downward IM. More specifically, upward primary IM is higher for males than females in Togo (14.9 percent), Liberia (11.6 percent), Sierra Leone, Zambia, Uganda (around 8–9 percent).²⁰ It is rather higher for females than males in Lesotho (25.9 percent), Botswana (15.5 percent), Nigeria (5 percent), and South Africa (4.2 percent). Similarly, downward primary IM is higher for females than males in Togo (9.8 percent), Liberia (7 percent), Guinea (6.2 percent), and Sierra Leone, Benin, and South Sudan (4–6 percent). It is rather higher for males than females in Lesotho (16.1 percent), and Botswana, Burkina Faso (5.4 percent). Second, in contrast, gender gap in favor of males, for the secondary/tertiary level, is less related to whether individuals are living in countries with highest values of upward or downward educational IM. For instance, upward secondary/tertiary IM is higher for males than females in Rwanda, Egypt, Liberia (8–9 percent), Uganda, Malawi and Zambia (6–6.5 percent). It is rather higher for females and males in South Africa, Morocco (4.5–5 percent), Sudan, Lesotho (3.9 percent), Mauritius (3 percent), Nigeria (2.5 percent), and Burkina Faso (2 percent). Moreover, downward secondary/tertiary IM is higher for females than males in Botswana (14.2 percent), Malawi (11.3 percent), Togo, Sierra Leone (around 9 percent), Ghana, and Benin (around 8 percent). It is rather higher for males than females in Morocco (7.2 percent), Mauritius (6.8 percent), South Africa (5.8 percent), Zimbabwe, Burkina Faso (around 4.5 percent), Sudan, Lesotho and Mozambique (3 percent).

c. Country-level educational IM by residency

We also report the country level residency gap in educational IM in Figure 13 in Appendix D. First, we show that upward (downward) IM are higher (lower) for individuals living in urban than rural areas, for both primary and secondary/tertiary levels, and for all countries. Second, individuals living in urban areas tend to do far better than those in rural areas in countries with the lowest values of upward IM and the highest values of downward IM. Indeed, the countries with the highest values of upward primary IM for individuals living in urban than rural areas are Ethiopia (57.6 percent), Sudan (44.8 percent), Burkina Faso (42.3 percent), and Guinea (40.2 percent). The countries with the lowest residency gap for upward primary IM (always in favor of individuals living in urban areas) are Mauritius (5 percent), South Africa (7.6 percent), Nigeria (10.5 percent), and South Sudan (12.1 percent). Similarly, the countries with the highest values of downward primary IM for individuals living in urban than rural areas are Ethiopia (50.8 percent), Burkina Faso (43.3 percent), and Sierra Leone (34.4 percent). The countries with the lowest residency gap for downward primary IM are Mauritius (0.9 percent), Nigeria, South Africa (close to 3 percent), and Egypt (4.7 percent). Third, we find that the residency gap for upward secondary/tertiary IM is higher in Ethiopia (32.9 percent), Malawi (21.8 percent), Morocco (21.6 percent), and Guinea (21.2 percent), while it is lower in South Sudan, Sierra Leone, and Botswana (5–6 percent). We also show that the residency gap for downward secondary/tertiary IM is higher in Burkina Faso (51.9 percent), Ethiopia (40.9 percent), and Morocco (34.9 percent), while it is lower in South Sudan (-2.8 percent), Nigeria (7.7 percent), and Egypt (9.1 percent).

d. Mapping of district-level educational IM across Africa

Finally, we map educational IM across 2,890 districts in Africa in Figure 7.²¹ Table 24 and Table 25 in Appendix E also report the summary statistics of district-level educational IM by country for primary and secondary/tertiary levels, respectively. They show both large within country and cross-districts variations. Overall, we find that within country disparities are larger in countries with lower educational mobility, with some exceptions. First, we show that upward primary IM is more unequal in South Sudan, Sudan, Ethiopia, and Burkina Faso (countries with lower average upward primary IM) and less unequal in Mauritius, South Africa, and Zimbabwe (most countries with higher average upward primary IM).²² Regarding the probability of downward primary IM, it is more unequal in Botswana, Mauritius, and Egypt, and less unequal in South Sudan, Rwanda, Liberia, and Lesotho. Moreover, upward primary IM varies less across regions than downward primary IM (coefficient of variation is 1.6 times higher for the latter than the former). Second, we find that upward secondary/tertiary IM varies more across districts in Ethiopia, Sudan, Malawi, and Cameroon (many countries with lower average upward secondary/tertiary IM). It varies less across districts in Botswana, Mali, Lesotho, and Senegal (most countries with higher or milder average upward secondary/tertiary IM). Moreover, downward secondary/tertiary IM varies more across districts in Botswana, Egypt, Ethiopia, and Malawi (paradoxically, most countries with lower average downward secondary/tertiary IM). It varies less across districts in Senegal, Ghana, Morocco, and Tanzania (paradoxically, most countries with higher or milder average downward secondary/tertiary IM).

IV. Stylized facts on educational IM and mineral discoveries and production

As a forestate of the effects of mineral resource discoveries and productions on intergenerational educational IM, we present in this section some stylized facts on both conditional IM in education and mineral discoveries and productions. South Africa, the country with the highest number of discoveries in our sample (108 total discoveries), illustrates well the case of countries where educational IM is higher in districts with discoveries/productions (see Figure 8). In the 60 districts where mineral sites were found, upward IM was 92 and 46 percent for primary and secondary/tertiary levels, respectively, higher by 2–3 pp. than in the 156 districts without any discoveries. Similarly, downward IM was 2 and 25 percent in districts with discoveries for primary and secondary/tertiary levels, respectively, lower by 1–2 pp. than in districts without discoveries.

We also show the mean differences of district-level conditional IM between districts with and without discoveries over the period of study (see Table 26 and Table 27).²³ We find that upward (downward) IM for primary education is, on average, higher (lower) in districts with discoveries than in districts without discoveries by around 4 pp. The opposite result holds for secondary education: upward (downward) IM is on average lower (higher) in districts with discoveries than districts without discoveries by around 4–6 pp. Second, we also present the summary statistics of IM by districts with and without discoveries for each country in Table 26 and Table 27. Upward IM is on average higher in districts with discoveries than districts without discoveries in 12 and 7 countries out of 26 countries (with both districts with and without discoveries) for primary and secondary levels, respectively. Downward IM is, on average, lower in districts with discoveries than districts without discoveries in 11 and 9 countries out of 26 for primary and secondary levels, respectively (see also Figure 16).

However, these stylized facts could hide differences in the dynamics of IM in districts with and without discoveries, and particularly progress that has happened in districts with discoveries. Figure 9 shows that while upward IM for primary and secondary/tertiary education was lower in districts with discoveries among the old cohorts (1950s and 1960s), it has significantly increased and closed the gap in these districts for more recent cohorts (1980s and 1990s) to stand above the one in districts without discoveries. Similarly, downward IM for primary and secondary/tertiary education was higher in districts with discoveries than without discoveries for old cohorts, and the gap has narrowed over time and for more recent cohorts. Therefore, IM has been significantly more dynamic in districts with discoveries. As a result, mineral discoveries and productions seem to have contributed to change the geography of the land of opportunities across African regions.

In addition, we find that improvements in IM have occurred for both females and males, with females doing better than males in districts with discoveries (see

Figure 14 in Appendix D). The gender gap of upward primary IM in favor of males has closed early in districts with discoveries (for 1970s cohort) than districts without discoveries (for 1990s cohort). The gender gap of secondary and tertiary IM in favor of males quickly turns to be in favor of females in districts with discoveries (for cohorts 1960s–1990s), which happened 20 years later in districts without discoveries (for cohorts 1980s– 1990s). Likewise, downward IM for primary and secondary/tertiary was higher for females than males in districts without discoveries contrary to districts with discoveries. Still, the gap has closed or widened for recent cohorts in districts without and with discoveries, respectively, with females performing better than males. We also find that IM improvements have occurred both in urban and rural areas in all districts (see Figure 15 in Appendix D). However, the gap between urban and rural areas has remained significant despite greater improvements in rural areas. There are, however, no significant differences in the dynamics of IM for rural and urban districts with and without discoveries.

Finally, we look at the dynamics of IM for individuals born around the first discovery and beginning of production. They are presented in Figure 10. These dynamics show that the timing of mineral discoveries and productions may constitute structural breaks in IM dynamics for individuals born before and after the discovery or production. Indeed, we show that the likelihood of upward IM for primary and secondary education has significantly increased after mineral discoveries and productions for individuals born after the discovery or production. Similarly, the likelihood of downward IM for primary and secondary education has decreased after the mineral discoveries and productions for individuals born after the discovery or production, while it has sometimes increased for individuals born in years running up to discovery or production. While we cannot plot the dynamics of the counterfactuals, we show that IM has on average significantly accelerated after mineral

V. Empirical methodology

To estimate the potential effects of mineral discoveries and productions on educational IM, we adopt an experimental approach and exploit the exogeneity of natural resource discoveries. First, it is plausible that the timing of mineral resource discoveries is exogenous due to the uncertainty related to the timing of the discovery and exploration success. While the technology used for exploration has improved over time, it is still highly improbable to predict the timing and success likelihood of finding a mineral field in a particular region (Khan et al., 2016; Arezki et al., 2017; Cavalcanti and others, 2019; Seri, 2021). Moreover, the exact location of mineral resources discoveries is purely exogenous as it depends on random geographical factors of the area. Therefore, while some regions may be endowed with mineral resources, it is improbable to find any resources in others. Second, mineral discoveries provide a significant source of revenues and represent a major economic shock that can affect the trajectory of the development in countries and districts they are found. They can also change the habit of individuals and their expectations about their own and children's future. Third, as shown by Horn (2011) and Arezki et al. (2017), there is a significant lag between the discoveries of natural resources and the beginning of their production, around five to six years. This allows us to study the effects of both mineral discoveries and productions on educational IM, separately. These features stand at the heart of our identification strategy and allow capturing a causal effect of mineral discoveries and productions on educational IM in African countries. Throughout the paper, we analyze both the effects of the first discoveries and productions on upward and downward mobility, and later in the robustness section, of multiple and successive discoveries and productions.

We employ a generalized difference-in-differences (GDID) strategy in a quasi-natural experiment and estimate treatment effects by comparing the changes in educational IM between a treatment group (people with exposure to the mineral discoveries and productions) and a control group, across pre-discovery/production and post-discovery/production. By doing so, our goal is to identify how educational IM has evolved following discovery/production for a group of people with exposition compared to a group of people born in the same district and around the discovery or production but not exposed to it, while controlling for the dynamics of educational IM in other districts without any discovery/production.

To capture the effects of the discovery/production, first, we focus in our baseline on a period spanning 30 years around it, i.e., we consider in the regressions, individuals born 15 years before and after the discovery or production. In the robustness, we expand this window and consider larger window periods of 40, 50, 60 years around the discovery or production. We define different expositions to the discovery or production as well as various control groups. First, focusing on a window of 30 years, we consider all the individuals born the year of the discovery or production to up to 15 years after it to be in the treatment group. In an alternative specification, we assume that individuals born 5 years before the discovery or production will still be exposed to it as they start their education around the date of the discovery or production; therefore, the treatment group comprises individuals born five years before the discovery or production to up to 15 years after. Second, in the first specification, we consider as control groups (i) individuals born in districts with a discovery or production and between 15 years before to one year before the discovery or production, and (ii) individuals born in districts without any discovery or production over the period of study. Similarly, the second specification includes in the control groups, (i) individuals born in districts with a discovery or production and born between 15 years to 5 years before the discovery or production, and (ii) individuals born in districts without any discovery or production.

Given the nature of our data, multiple discoveries or productions and multiple treatments and control groups within and across districts, our GDID model allows for (i) specific IM across districts by introducing district fixed-effects α_i, (ii) the common change in IM to vary across cohorts (decade in baseline, and each year of birth in the robustness) and years of census/survey by introducing cohort fixed-effects or year-birth fixed effects γ_t and census-year fixed effects δ_t, respectively, and (iii) different timing of the discovery or production for different treated groups. This allows to filter out all rigid characteristics specific to districts, cohorts or years of birth, and census year found to be critical in explaining education IM by Alesina et al. (2021). The model is estimated using a linear probability specification and obtained as follows:

IM_jith = α_i + γ_t + δ_t + βd_h + X_jitθ + ε_jit (6)

Where IM_jith is our measure of upward (and downward) mobility for primary or secondary/tertiary levels that takes the value one if the child j in district i has on average a higher (lower) education than its biological/step-parents knowing that its biological/step-parents have not completed (have completed) primary or secondary/tertiary education, respectively. d_h is a dummy that takes the value of one if the individual is in the treated group (e.g., born between the year of the discovery or production to up to 15 years later) and zero if the individual is in the control group (e.g., either born between 15 years to one year before the discovery or production, or born in districts without any discovery or production). β is the coefficient of interest. It captures the treatment effect of mineral discovery/production on upward and downward educational IM by comparing educational IM in the treated and control groups. X_it is a set of control variables including the gender of individuals and of their household head, the occupation of household head, the dummies of cohabitation with biological/step-parents (i.e, with only biological/step-father, only biological/step-mother, or both biological/step-father and mother), the size of the household, and urban/rural residency. ε_jit is the idiosyncratic term.

Our model requires the parallel trends assumption to hold, i.e., in the absence of the discovery or production, the change of educational IM would have been the same in both the treated and control groups. This assumption is violated when there are unobserved factors that are correlated with both the exposition to the discovery or production and the timing of the discovery or production. As discussed above, we have good reasons to believe that the timing of mineral discoveries is exogenous. Regarding the exposition to the discovery or production, since we focus on a relatively short period around the discovery or production in our baseline and include either cohort fixed effects or year-birth fixed effects, we limit the risks that other shocks or interventions polluted our findings. However, since we cannot test for the parallel trends’ assumption in our GDID, we apply the following strategy to test the robustness of our findings and implicitly verified whether this assumption holds. First, we analyze the dynamic effects and conduct a standard leads-and-lags test following the literature (see, e.g. Angrist & Pischke, 2009, Maurer, 2019). This allows testing whether the effects of discoveries occurred after the discovery or production and tend to intensify thereafter. Second, we cross validate our findings by using different control groups while dropping from the control group all individuals in (i) countries without mineral discoveries (Mauritius and South Sudan) and (ii) in districts without mineral discoveries. This reduces the heterogeneity and differences in characteristics between our treated and control groups.

VI. Results

In this section, we present our main findings, starting with the baseline results for the primary and secondary/tertiary levels of education, before discussing the dynamic effects of mineral discoveries and productions on educational IM.

A. Baseline results

1) Educational primary IM

Table 2 reports the baseline results for the various sample compositions and definitions of the variables. The dependent variable is the occurrence of upward (columns 1–4) or downward (columns 5–8) educational IM. As explained in the previous section, we consider a time window of 15 years before and after the mineral discoveries or productions. We start with the results in Panel (A) where control variables are not included in the estimates.

^{Table 2:}

Baseline results, primary education

Note: This table presents the baseline results for the probability of upward (columns 1–4) and downward (columns 5–8) educational IM as a function of mining discoveries and production without (Panel A) and with (Panel B) a number of other control variables. Standard errors are in parentheses. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Disc-5 is when individuals born 5 years before the discovery are included in the treatment group; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production; Prod-5 is when individuals born 5 years before mining production are included in the treatment group. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

In columns (1) and (5), the treatment group includes individuals born after the mineral discoveries or productions. The estimates then provide the differences in the likelihood of upward or downward primary educational IM for individuals in districts with mineral discoveries or productions born after (treatment group) and before (control group) the discovery within the exposure window period of 15 years. We do so by controlling for the likelihood of educational IM in districts without mineral discoveries/productions. The results show that the probability of experiencing an upward educational IM is 2.7 pp. higher for an individual born after a discovery of a mining site compared to an individual born before the discovery (column 1). In other words, individuals who are born from parents who are uneducated have better chances of achieving at least primary education if exposed and living to a district with a mining discovery. Inversely, the likelihood of experiencing a downward educational IM for individuals born after the discovery of a mining site is 1.2 pp. lower than those before the mining discovery (column 5). That said, individuals who are born from educated parents are less likely to do less than their parents if exposed and living to a district with a mining discovery. About the size of the effects, it is relatively small when compared to the increase of IM across the different cohorts, signaling that, other factors have also played a significant role in improving IM in Africa.

In columns (2) and (6), we expand the treatment group to include individuals who are born 5 years before the discovery. In fact, these individuals may have not started their education at the time of the discovery as they do not meet the minimum years of schooling in most African countries. The coefficient of interest in column (2) remains broadly the same as in column (1), suggesting that the change in the time exposition to the mining discovery does not after the likelihood of upward primary educational IM. However, the coefficient is significantly lower in column (6) than in column (5), implying that individuals born after the discovery of the mining sites would have a lower likelihood of downward mobility than those born just before (five years before) the discovery.

In columns (3) and (7), we use an alternative definition of mining activity where the binary variable is now equal to one for districts where mining production has started and zero otherwise. The estimates compare the likelihood of upward or downward educational IM between individuals born after and before the beginning of the mining production. We find that the probability of experiencing an upward educational IM is 6.7 pp. higher for an individual born after the mining production compared to those born before the beginning of the mining production. The coefficient associated with mining is higher in column (3) than in column (1), implying that mining productions tend to have, on average, higher positive effects on the likelihood of upward primary educational IM than the discoveries. These higher effects could be explained by the increase in investments required to start production, which may create more jobs and income opportunities for district residents than the proceeds associated with the exploration and discovery of mining sites. Indeed, the start of the production reveals a new information that would affect the parents own and child’s life expectation. We also observe that our coefficient of interest remains broadly unchanged in column (7) compared to column (5), suggesting that individuals born after the discovery and the beginning of the mining production have the same likelihood of experiencing downward primary educational IM. We also include the individuals born 5 years before the beginning of production in the treatment group in columns (4) and (8). We find that the coefficient associated with mining production is lower in column (4) than in column (3), while remaining the same in columns (7) and (8).

In Panel (B), we control for several covariates that can affect the likelihood of upward and downward educational IM as presented in the previous section (see Table 28 in Appendix F for the coefficients associated with each control variable). We find that the coefficients associated with mining are highly significant at 1 percent level in all columns, and their magnitude are broadly equal to those found in Panel (A). Thus, our findings remain unchanged when we control for individual characteristics.

2) Educational secondary and tertiary IM

We now turn to secondary/tertiary education level. The results are reported in Table 3. We find that the coefficient associated with mineral discoveries and productions is not statistically significant in all columns, except the slightly significance in column (5) at 10 percent level.

^{Table 3:}

Baseline results, secondary and tertiary education

The result suggests that the likelihood for individuals born after the mineral discoveries and productions to experience an upward or downward secondary/tertiary educational IM is not statistically different from that of individuals born before the start of mining activities discovery or production. The insignificance of the effects of mining activities on secondary and tertiary educational IM could be due to the presence of mixed effects, both positive and negative, that are offsetting each other. For instance, as illustrated by Gradstein and Ishak (2020), a positive oil price shock is found to have a positive effect on educational attainment for childs (ages 0–4), and a negative effect for adolescents (ages 10–14). Indeed, when youth children or adults reach secondary and tertiary education levels, they face a significant trade-off between education and employment. They are now able to do some domestic tasks or work outside of the household, and therefore be forced or decide to drop out of school. This situation could be particularly prone when there is a mining site within the district. For instance, Ahlerup et al. (2020) found that individuals who had gold mines within their district when they were adolescent have significantly lower educational attainment. This is explained by some myopic educational decisions when employment in gold mining is an alternative. However, some other papers reveal a positive effect of subsoil wealth on average years of primary, secondary, and tertiary education level (Stjins, 2001). In addition, the same reasons driving the benefits of mineral discoveries and productions on the primary education may still be playing a role, therefore rebalancing the negative effects at higher levels of education.

In Panel (B), we include control variables (see Table 29 in Appendix F for the coefficient associated with each control variables). The results remain unchanged. The coefficients associated with mineral discoveries and productions are still statistically not significant in all columns, except column (5). In the latter, the coefficient is positive and significant at 5 percent level, suggesting that the likelihood of downward secondary/tertiary educational IM of individuals born after mining discoveries tend to be higher than those born before the discoveries. However, this finding does not hold when we expand the treatment group by including individuals born 5 years before the discovery in column (6), meaning it is not robust.

B. Dynamic effects of mineral discoveries and productions on educational IM

In this subsection, we explore the dynamic effects of mining activities based on the time distance between the years of discoveries or productions and the birth years of individuals to test the parallel trends’ assumption. As explained previously, we conduct a leads-and-lags test following the literature (Angrist & Pischke, 2009, Maurer, 2019) to analyze whether the effects of mineral discoveries on IM tend to intensify the years after the shock. To do so, we estimate the likelihood of upward and downward educational IM for individuals born 0–5, 6–10 and 11–15 years after the discoveries or the beginning of mining production, and those born 5–10 and 10–6 years before the discoveries or the first year of mining production. The reference group is given by individuals born 11–15 years before the discoveries or the beginning of mining productions. The results are reported in Table 4 for upward IM (columns 1–2 and 5–6) and downward IM (columns 3–4 and 7–8).

^{Table 4:}

Dynamic effects of mining activities on educational IM

Note: This table presents the dynamic effects of mineral discoveries and productions on primary (columns 1–4) and secondary/tertiary (columns 5–8) educational IM. Standard errors are in parentheses. Disc stands for mining discovery. Prod stands for mining. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

The results are two folds. First, we find that mining activities affect the likelihood of primary educational IM for all age groups, while the effect is not statistically significant for secondary/tertiary educational IM. In fact, the results show that the coefficients associated with mineral discoveries and productions are positive and significant in columns 1–2, suggesting that mining activities tend to increase the likelihood of upward primary educational IM for all age groups of individuals. Inversely, the coefficients associated with mineral discoveries and productions are negative and significant in columns 3–4, meaning that mining activities are correlated with lower likelihood of downward primary educational IM for all age groups. On the other hand, the coefficients associated with mineral discoveries and productions are not statistically significant for all age groups of individuals in columns 5–8. This is in line with our findings in

that mining activities do not affect the likelihood of upward and downward secondary and tertiary educational IM. Second and more importantly, we find that the coefficients associated with mineral discoveries and productions are higher for individuals born later after the discovery and the beginning of mining production. For instance, the probability of upward primary educational IM is 7.6 pp. higher for individuals born 6–10 years after the beginning of mining production against only 1.1 pp. higher for those born 10–6 years before the beginning of mining production, as compared to those born 15–11 years before the beginning of mining of production (column 2). This higher probability could be explained by the fact that it takes time for mining activities to have an impact on the local communities, particularly in terms of infrastructure provision. It also supports that the parallel trend assumption might be verified, and therefore the effects we are capturing can be fully attributed to mineral discoveries and start of productions.

VII. Transmission channels

We explore the channels through which mining activities affect the likelihood of educational IM. We focus on two channels including the income effect proxied by parents working in the mining sector, and the returns to education.²⁴ First, the increase of income for parents working in the mining sector, due to novel abundant opportunities, will allow them to invest more in their children education attainment (Becker and Tomes, 1979). Second, this new economic dynamism, creation of new jobs, will lead to an increasing demand for skilled workers. Thus, the higher returns or benefits to education in terms of wealth and income will motivate individuals to increase their educational attaintment relative to their parents (Torche, 2014).

To capture the income effect, we include an interactive variable between the discovery or production variable and a dummy equals to one if individuals have one of their parents working in the mining sector, and the latter itself as additional variables. For the return to education channel, we employ a two-step strategy where, in the first step, the effects of the mining discovery or production on the transmission channel variable is analyzed, and in the second step, we check the correlation between the transmission channel variable and upward/downward educational IM. This allows us to test whether the effect of mineral discoveries and productions on the probability of upward and downward educational IM transits through our transmission channel variable.

A. Income channel: parents working in mining sector

We first test whether the income effect proxied by parents working in the mining sector could be a channel through which mining activities affect the likelihood of upward/downward educational IM. One would expect that mining activities will create jobs in the local communities, therefore generating a source of income which could allow parents to invest more in their children education. We define a binary variable taking the value of one if one of the parents of the child is working in the mining sector, and zero otherwise and interact it with the treatment variable of mineral discoveries or productions. The results are displayed in Table 5, with the estimates about primary education in columns 1–4, and secondary and tertiary education in columns 5–8. We find that the coefficients associated with mineral discoveries or productions and the interactive term in columns 1–4 are statistically significant and have the expected sign. First, the mineral discoveries and productions increase (decrease) the likelihood of upward (downward) primary IM. Second and more importantly, these benefits of mining discovery or production are accentuated for individuals those one of the parents works in the mining sector. We find that having one of the parents working in the mining sector raises the likelihood of upward primary IM by 2.2 and 6.3 pp. following mineral discoveries and productions, respectively. Moreover, it diminishes the likelihood of downward primary IM by 1.1 pp. following mineral discovery. For secondary and tertiary IM, we find no significant effects of both the mineral discovery or production and its interactive term with parents working in mining. This suggests that the insignificant effects of mineral discoveries and productions hold for all individuals, independently of whether they have a parent working in the mining sector or not.

^{Table 5:}

Income channel: parents working in the mining sector

Note: This table presents the effects of mineral discoveries and productions on primary (columns 1–4) and secondary/tertiary (columns 5–8) educational IM, conditional on having of the parents working in the mining sectors. Standard errors are in parentheses. Yes Parents work in mining is a dummy taking the value of one for individuals with one of the parents working in the mining sector. Yes Mining is the treatment dummy variable for either discovery or production (depending on the columns). Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

B. Returns to education: Wealth index and LIDO score

We further test a second channel of return to education to check whether invididuals in districts with mineral discoveries and productions increase their level of education to get a higher wealth and income. Indeed, individuals exposed to the mineral discoveries and productions have a higher likelihood to work in the manufacturing and services sectors, and a lower likelihood to work in the agriculture sector. Given that the manufacturing and services sector may require a higher educational attainment and skills, individuals would have to increase their education level to work in these sectors, and thus having a better socio-economic status. To study the effect of mineral discoveries and productions on the returns to education, we first estimate a Mincer-like equation using a GDID model. The dependent variables are a proxy of wealth and income, given the lack of data on income and consumption in our database. The explanatory variables include the educational levels (primary and secondary/tertiary), the mineral discoveries and productions dummy, and their interactive term.

To do so, we construct the wealth index using a principal component analysis on several variables at the household level reflecting the economic status of household, following closely the Demographic and Health Survey wealth index (Rutstein and Staveteig, 2014).²⁵ Second, we apply the lasso-adjusted industry, demographic, and occupational (LIDO) scores computed using actual data of labor market for the United States in 1950 by Saavedra and Twinam (2020) to our individuals in Africa. This LIDO score is an occupational income ranking score used as an alternative measure of income, socioeconomic status, and labor market outcome. It is dependent on (i) the fine categories of sectors of employment based on the industry classification (e.g., agriculture, mining and extraction, manufacturing, construction, hotels and restaurants, etc.), (ii) the occupation within employment based on the occupation classification (e.g., legislators, senior officials and managers, technicians and associate professionals, service works and market sales, elementary occupations), and (iii) individuals characteristics (e.g., age, gender). When applies to workers in Africa, the cross-individual, district and country differences at each period and over time would only come from the differences in the labor market conditions (sectors of activities, and occupation within employment) and demographics. We neutralize the effects of the demographic variables in the estimations by controlling for gender and age. We show in the Figure 17 and Figure 18 in Appendix G that the LIDO Score and Wealth index are strongly correlated with PPP GDP per capita at the country level, thereby implying that they are good proxies of income in Africa.

The results for the LIDO score and Wealth score are reported in Table 6 in columns 1–3 and 4–6, respectively. We find that an increase in the levels of education are associated with a higher LIDO score and Wealth index in all districts (with or without mineral discoveries) and all individuals (exposed to mineral discoveries/productions or not). This implies higher returns to education in Africa. More importantly, we find that the returns to education are higher in districts with mineral discoveries or productions (columns 1 and 4), and they are even greater for individuals exposed to the mineral discoveries and productions (columns 2–3 and 5–6, respectively). In a second step equation, we look how the LIDO score and Wealth index correlate with the educational IM. Table 7 shows a strong and positive (negative) association between the LIDO score and Wealth index with the upward (downward) primary educational IM. Thus, our findings suggest that in districts with mineral discoveries and productions, and specifically for individuals born after the mineral discoveries and productions, the returns to higher education has played a key role as an incentive to achieve higher educational levels and greater educational IM.

^{Table 6:}

Returns to education channel: Mincer-type equation

Note: This table presents a mincer-type equation where the effects of mineral discoveries and productions on the LIDO score (columns 1–3) and Wealth index (columns 4–6), conditional on educational attainment. Standard errors are in parentheses. Yes Mining is a dummy for district with discovery or not (columns 1 and 2) or the treatment dummy variable for either discovery (columns 2 and 5) or production (columns 3 and 6). With Disc. means with discovery. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 6:}

Returns to education channel: Mincer-type equation

	(1)	(2)	(3)	(4)	(5)	(6)
	LIDO Score			Wealth Index
	With Disc.	Disc-B/A	Prod-B/A	With Disc.	Disc-B/A	Prod-B/A
Primary completed	0.021***	0.021***	0.021***	0.041***	0.041***	0.041***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Secondary and/or tertiary completed	0.048***	0.049***	0.049***	0.097***	0.097***	0.098***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Primary completed X Yes Mining	0.001***	0.003***	0.007***	0.013***	0.027***	0.045***
	(0.000)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)
Secondary and/or tertiary completed X Yes Mining	0.009***	0.010***	0.012***	0.017***	0.042***	0.064***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)
Yes Mining		0.003***	0.001		-0.012***	-0.052***
		(0.001)	(0.001)		(0.001)	(0.002)
Observations	4615725	4665165	4665165	4254638	4254638	4254638
R-squared	0.518	0.520	0.520	0.789	0.789	0.789
# Treated	396901	136323	46519	387670	178970	66437
District FE	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 6:}

Returns to education channel: Mincer-type equation

	(1)	(2)	(3)	(4)	(5)	(6)
	LIDO Score			Wealth Index
	With Disc.	Disc-B/A	Prod-B/A	With Disc.	Disc-B/A	Prod-B/A
Primary completed	0.021***	0.021***	0.021***	0.041***	0.041***	0.041***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Secondary and/or tertiary completed	0.048***	0.049***	0.049***	0.097***	0.097***	0.098***
	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)	(0.000)
Primary completed X Yes Mining	0.001***	0.003***	0.007***	0.013***	0.027***	0.045***
	(0.000)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)
Secondary and/or tertiary completed X Yes Mining	0.009***	0.010***	0.012***	0.017***	0.042***	0.064***
	(0.001)	(0.001)	(0.001)	(0.001)	(0.001)	(0.002)
Yes Mining		0.003***	0.001		-0.012***	-0.052***
		(0.001)	(0.001)		(0.001)	(0.002)
Observations	4615725	4665165	4665165	4254638	4254638	4254638
R-squared	0.518	0.520	0.520	0.789	0.789	0.789
# Treated	396901	136323	46519	387670	178970	66437
District FE	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 7:}

Returns to education channel: primary educational IM and LIDO score, Wealth index

Note: This table presents the effects of mineral discoveries and productions as well as of LIDO score and wealth index on primary upward (columns 1– 2) and secondary/tertiary (columns 3–4) educational IM. Standard errors are in parentheses. Mining is the treatment dummy variable for either discovery or production (depending on the columns). Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 7:}

Returns to education channel: primary educational IM and LIDO score, Wealth index

	(1)	(2)	(3)	(4)
	Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	0.014**	0.118***	-0.014*	-0.003
	(0.006)	(0.009)	(0.008)	(0.018)
LIDO score, 0-1	0.592***	0.592***	-0.352***	-0.351***
	(0.007)	(0.007)	(0.009)	(0.009)
Wealth index, 0-1	0.433***	0.434***	-0.252***	-0.252***
	(0.004)	(0.004)	(0.005)	(0.005)
Observations	1032317	1032317	369737	369737
R-squared	0.227	0.228	0.177	0.177
# Treated	48751	23733	15531	4664
District FE	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes

^{Table 7:}

Returns to education channel: primary educational IM and LIDO score, Wealth index

	(1)	(2)	(3)	(4)
	Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	0.014**	0.118***	-0.014*	-0.003
	(0.006)	(0.009)	(0.008)	(0.018)
LIDO score, 0-1	0.592***	0.592***	-0.352***	-0.351***
	(0.007)	(0.007)	(0.009)	(0.009)
Wealth index, 0-1	0.433***	0.434***	-0.252***	-0.252***
	(0.004)	(0.004)	(0.005)	(0.005)
Observations	1032317	1032317	369737	369737
R-squared	0.227	0.228	0.177	0.177
# Treated	48751	23733	15531	4664
District FE	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes

To further reinforce this channel, we also whether mineral discoveries and productions affect the reallocation of individuals across the broad sectors, i.e., agriculture, manufacturing, and services. Specifically, the mineral discoveries and productions, by creating new opportunities, may change the economic propects at the local level. They can accelerate the demand for skilled workers by more capital-intensive and better paid activities in the manufacturing and services sectors, whereas individuals may be less prone to work in labor-intensive agricutlture sector. The results are reported in Table 8. First, we show that the likelihood to work in the agriculture sector decreases by 1.1 pp. for individuals exposed to the mineral discoveries or productions (see columns 1–2). Second, we also find that the likelihood to work in the manufacturing or services sector increases by around 0.8–1.4 pp., respectively (see columns 3–4 and 5–6). These findings suggest that mineral discoveries and productions lead to the shift from lower skilled to higher skilled jobs, and therefore boost the demand for better educated people in Africa.

^{Table 8:}

Returns to education channel: effects of mineral discoveries and productions on sectoral employment

Note: This table presents the effects of mineral discoveries and productions on the likelihood to work in the agricultural (columns 1–2), manufacturing (columns 3–4), and services (columns 5–6) sectors. Standard errors are in parentheses. Mining is the treatment dummy variable for either discovery or production (depending on the columns). Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 8:}

Returns to education channel: effects of mineral discoveries and productions on sectoral employment

	(1)	(2)	(3)	(4)	(5)	(6)
	Agriculture		Manufacturing		Services
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	-0.011*** (0.001)	-0.010*** (0.002)	0.008*** (0.001)	0.012*** (0.001)	0.003*** (0.001)	0.014*** (0.002)
Observations	7167154	7167154	7167154	7167154	7167154	7167154
R-squared	0.659	0.659	0.145	0.145	0.220	0.220
# Treated	335257	138082	335257	138082	335257	138082
District FE	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 8:}

Returns to education channel: effects of mineral discoveries and productions on sectoral employment

	(1)	(2)	(3)	(4)	(5)	(6)
	Agriculture		Manufacturing		Services
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	-0.011*** (0.001)	-0.010*** (0.002)	0.008*** (0.001)	0.012*** (0.001)	0.003*** (0.001)	0.014*** (0.002)
Observations	7167154	7167154	7167154	7167154	7167154	7167154
R-squared	0.659	0.659	0.145	0.145	0.220	0.220
# Treated	335257	138082	335257	138082	335257	138082
District FE	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes

VIII. Robustness checks

We undertake various robustness tests to check the validity of our results to alternative samples and specifications.

A. Use of alternative control groups (only countries/districts with discoveries)

We start by using an alternative definition of the control groups. In our baseline, we compared the likelihood of upward and downward educational IM between individuals born after the discovery or the beginning of mining production, and those born before the discovery or beginning of production or in regions and countries without discoveries, regardless of their residing countries or districts. In this robustness check, we restrict the control group to individuals living in countries or districts where mining sites are discovered or with production activities. The results are reported in Table 9, with the estimates for primary educational IM being in Panel (A) and those for secondary and tertiary educational IM in Panel (B). As in the baseline, we find that the coefficients associated with mining activities are significant for primary educational IM, and insignificant for secondary and tertiary educational IM. Therefore, mining activities affect the likelihood of upward and downward primary educational IM in countries and districts with mineral resources, while the effect on secondary and tertiary educational IM is not statistically significant. Our results remain unchanged and are robust to the change of control groups and samples.

^{Table 9:}

Robustness check: using different control groups

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM on a sample of countries (columns 1–4) and districts (columns 5–8) with mining activities. Standard errors are in parentheses. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 9:}

Robustness check: using different control groups

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Countries with mining activities				Districts with mining activities
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.027*** (0.002)	0.069*** (0.003)	-0.013*** (0.002)	-0.012*** (0.002)	0.045*** (0.003)	0.072*** (0.004)	-0.012*** (0.002)	-0.005** (0.002)
Observations	8226648	8458031	4320910	4424877	318826	550209	230075	334042
R-squared	0.266	0.267	0.130	0.130	0.342	0.326	0.121	0.116
# Treated	148633	53986	98793	36768	148633	53986	98793	36768
Panel (B) Secondary and tertiary education
Mining	-0.007	0.015	0.038**	-0.010	0.011	0.017	0.024	-0.008
	(0.013)	(0.021)	(0.019)	(0.024)	(0.008)	(0.020)	(0.020)	(0.024)
Observations	3289349	3415101	321879	329499	149215	274967	10822	18442
R-squared	0.217	0.217	0.168	0.168	0.152	0.169	0.119	0.127
# Treated	67525	38002	6197	2813	67525	38002	6197	2813
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 9:}

Robustness check: using different control groups

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Countries with mining activities				Districts with mining activities
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.027*** (0.002)	0.069*** (0.003)	-0.013*** (0.002)	-0.012*** (0.002)	0.045*** (0.003)	0.072*** (0.004)	-0.012*** (0.002)	-0.005** (0.002)
Observations	8226648	8458031	4320910	4424877	318826	550209	230075	334042
R-squared	0.266	0.267	0.130	0.130	0.342	0.326	0.121	0.116
# Treated	148633	53986	98793	36768	148633	53986	98793	36768
Panel (B) Secondary and tertiary education
Mining	-0.007	0.015	0.038**	-0.010	0.011	0.017	0.024	-0.008
	(0.013)	(0.021)	(0.019)	(0.024)	(0.008)	(0.020)	(0.020)	(0.024)
Observations	3289349	3415101	321879	329499	149215	274967	10822	18442
R-squared	0.217	0.217	0.168	0.168	0.152	0.169	0.119	0.127
# Treated	67525	38002	6197	2813	67525	38002	6197	2813
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

B. Use of alternative structures of fixed effects and time variable

We then perform a series of consistency checks based on the structure of the fixed effects. The results are in Table 10, with Panel (A) being for primary education and Panel (B) for secondary and tertiary education. In columns 1–4, we replace the cohort fixed effects by birth year fixed effects and therefore compare individuals born within the same year instead of cohort since they may have experienced different shocks in 10 years. In columns 5–8, we include both the birth year fixed effects and the common time trend to capture the evolution of IM and rule out the possibility that individuals born before and after the discoveries or the beginning of mining production were already on differential growth trajectories in their education outcomes, i.e., a change in the educational IM indices that would have happened even in the absence of the mining activities. These factors could include particularly the family background of individuals (rich, poor and others). In columns 9–12, we control for cohort fixed effects and common time trend to filter out all persistent cohort related differences that could affect the likelihood of educational IM of individuals born before and after the discoveries of mining sites: for example, the availability of school infrastructure and change in the education system. Table 10 shows that our main findings remain unchanged even after accounting for all these different structures of fixed effects and time variable.

^{Table 10:}

Robustness check: inclusion of fixed effects

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM, using different sets of fixed effects reported at the bottom of the table. Standard errors are in parentheses. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 10:}

Robustness check: inclusion of fixed effects

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	Upward mobility		Downward mobility		Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/	AProd-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.016***	0.039***	-0.009***	-0.009***	0.016***	0.039***	-0.009***	-0.009***	0.014***	0.052***	-0.009***	-0.008***
	(0.002)	(0.003)	(0.002)	(0.002)	(0.002)	(0.003)	(0.002)	(0.002)	(0.002)	(0.003)	(0.002)	(0.002)
Observations	8306024	8537407	4374423	4478390	8306024	8537407	4374423	4478390	8306024	8537407	4374423	4478390
R-squared	0.273	0.274	0.135	0.134	0.273	0.274	0.135	0.134	0.272	0.273	0.134	0.133
# Treated	148633	53986	98793	36768	148633	53986	98793	36768	148633	53986	98793	36768
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	No	No	No	No	No	No	No	No	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	No
Common Time-Trend	No	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
District X Cohort FE	No	No	No	No	No	No	No	No	No	No	No	No
Panel (B) Secondary and tertiary education
Mining	-0.017	0.011	0.050***	-0.001	-0.017	0.011	0.050***	-0.001	-0.017	0.011	0.048**	-0.001
	(0.012)	(0.020)	(0.019)	(0.024)	(0.012)	(0.020)	(0.019)	(0.024)	(0.012)	(0.020)	(0.019)	(0.025)
Observations	3335415	3461167	323998	331618	3335415	3461167	323998	331618	3335415	3461167	323998	331618
R-squared	0.221	0.221	0.171	0.171	0.221	0.221	0.171	0.171	0.221	0.221	0.170	0.170
# Treated	67525	38002	6197	2813	67525	38002	6197	2813	67525	38002	6197	2813
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	No	No	No	No	No	No	No	No	No	No	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No
Common Time-Trend	No	No	No	No	Yes	Yes	Yes	Yes	No	No	Yes	Yes
District X Cohort FE	No	No	No	No	No	No	No	No	No	No	No	No

^{Table 10:}

Robustness check: inclusion of fixed effects

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	Upward mobility		Downward mobility		Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/	AProd-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.016***	0.039***	-0.009***	-0.009***	0.016***	0.039***	-0.009***	-0.009***	0.014***	0.052***	-0.009***	-0.008***
	(0.002)	(0.003)	(0.002)	(0.002)	(0.002)	(0.003)	(0.002)	(0.002)	(0.002)	(0.003)	(0.002)	(0.002)
Observations	8306024	8537407	4374423	4478390	8306024	8537407	4374423	4478390	8306024	8537407	4374423	4478390
R-squared	0.273	0.274	0.135	0.134	0.273	0.274	0.135	0.134	0.272	0.273	0.134	0.133
# Treated	148633	53986	98793	36768	148633	53986	98793	36768	148633	53986	98793	36768
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	No	No	No	No	No	No	No	No	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No	No	No
Common Time-Trend	No	No	No	No	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
District X Cohort FE	No	No	No	No	No	No	No	No	No	No	No	No
Panel (B) Secondary and tertiary education
Mining	-0.017	0.011	0.050***	-0.001	-0.017	0.011	0.050***	-0.001	-0.017	0.011	0.048**	-0.001
	(0.012)	(0.020)	(0.019)	(0.024)	(0.012)	(0.020)	(0.019)	(0.024)	(0.012)	(0.020)	(0.019)	(0.025)
Observations	3335415	3461167	323998	331618	3335415	3461167	323998	331618	3335415	3461167	323998	331618
R-squared	0.221	0.221	0.171	0.171	0.221	0.221	0.171	0.171	0.221	0.221	0.170	0.170
# Treated	67525	38002	6197	2813	67525	38002	6197	2813	67525	38002	6197	2813
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	No	No	No	No	No	No	No	No	No	No	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Birth Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No	No
Common Time-Trend	No	No	No	No	Yes	Yes	Yes	Yes	No	No	Yes	Yes
District X Cohort FE	No	No	No	No	No	No	No	No	No	No	No	No

C. Use of alternative time window around the mineral discoveries or productions

Another important element of our analysis is the time window around the mineral discovery and production. In the baseline, we used a time window of 30 years (15 years before and after the discovery or the beginning of production). We test the robustness to alternative time windows, including 40 years (i.e., 20 years before and after), 50 years (25 years before and after) and 60 years (30 years before and after the discovery/production) to account for the individuals who take more time to complete their education and the potential long-lasting effect of mining activities. The results are reported in Table 11, with Panel (A) for primary education, and Panel (B) for secondary and tertiary education. We find that the coefficients associated with our variable of interest are significant for the primary education, while insignificant for the secondary and tertiary education, thus in line with the findings in the baseline estimates. This confirms that our baseline results are not driven by the choice of the time window around the discovery or the beginning of mining production. Moreover, the effects of mineral discoveries or productions tend to be long-lasting, also for cohorts born years after the discovery or production.

^{Table 11:}

Robustness check: using alternative time window

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM, considering different time windows around discovery or production. Standard errors are in parentheses. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 11:}

Robustness check: using alternative time window

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	Window of 40 years				Window of 50 years				Window of 60 years
	Upward mobility		Downward mobility		Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.030*** (0.002)	0.074*** (0.004)	-0.015*** (0.002)	-0.018*** (0.003)	0.032*** (0.002)	0.074*** (0.004)	-0.016*** (0.002)	-0.019*** (0.002)	0.033*** (0.002)	0.073*** (0.003)	-0.016*** (0.002)	-0.016*** (0.002)
Observations	8397851	8397851	4459701	4459701	8487171	8487171	4534408	4534408	8563949	8563949	4603751	4603751
R-squared	0.270	0.270	0.134	0.134	0.271	0.271	0.134	0.134	0.271	0.271	0.134	0.134
# Treated	182250	37372	128892	27622	211230	50369	154302	38299	242877	66867	180868	55795
Panel (B) Secondary and tertiary education
Mining	-0.003	0.026	0.039**	-0.021	-0.001	0.029	0.040**	-0.021	-0.001	0.022	0.040**	-0.011
	(0.012)	(0.016)	(0.019)	(0.023)	(0.012)	(0.020)	(0.020)	(0.023)	(0.012)	(0.019)	(0.019)	(0.022)
Observations	3388392	3388392	329884	329884	3442816	3442816	337347	337347	3495819	3495819	343205	343205
R-squared	0.217	0.217	0.168	0.168	0.218	0.218	0.167	0.167	0.217	0.217	0.166	0.166
# Treated	88176	26107	7715	1948	106151	36649	9003	2871	126826	50705	11022	4475
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 11:}

Robustness check: using alternative time window

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	Window of 40 years				Window of 50 years				Window of 60 years
	Upward mobility		Downward mobility		Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.030*** (0.002)	0.074*** (0.004)	-0.015*** (0.002)	-0.018*** (0.003)	0.032*** (0.002)	0.074*** (0.004)	-0.016*** (0.002)	-0.019*** (0.002)	0.033*** (0.002)	0.073*** (0.003)	-0.016*** (0.002)	-0.016*** (0.002)
Observations	8397851	8397851	4459701	4459701	8487171	8487171	4534408	4534408	8563949	8563949	4603751	4603751
R-squared	0.270	0.270	0.134	0.134	0.271	0.271	0.134	0.134	0.271	0.271	0.134	0.134
# Treated	182250	37372	128892	27622	211230	50369	154302	38299	242877	66867	180868	55795
Panel (B) Secondary and tertiary education
Mining	-0.003	0.026	0.039**	-0.021	-0.001	0.029	0.040**	-0.021	-0.001	0.022	0.040**	-0.011
	(0.012)	(0.016)	(0.019)	(0.023)	(0.012)	(0.020)	(0.020)	(0.023)	(0.012)	(0.019)	(0.019)	(0.022)
Observations	3388392	3388392	329884	329884	3442816	3442816	337347	337347	3495819	3495819	343205	343205
R-squared	0.217	0.217	0.168	0.168	0.218	0.218	0.167	0.167	0.217	0.217	0.166	0.166
# Treated	88176	26107	7715	1948	106151	36649	9003	2871	126826	50705	11022	4475
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

D. Use of alternative time exposition to the mineral discoveries or productions

We further explore whether our results are robust to the change in the time exposition to the mining activities. We expand the definition of the treatment group to account for individuals who may not have completed their education before the discovery or the beginning of mining production. Our baseline estimates compare the likelihood of individuals born before and after the start of mining activities. However, some individuals born before the start of mining activities may complete their education after the beginning of mining activities. In this case, those individuals will benefit from the economic and social impact from the exploitation of mining resources. We have already tried to account for these individuals by including in the treatment group the individuals born 5 years before the discovery/production in Table 2 and Table 3. In this robustness check, we include in the treatment group individuals born 10 and 15 years before the discovery or the start of mining production. To this end, we consider a longer time window of 60 years (30 years before and after the discovery/production). The results are reported in Table 12, with the primary educational IM in Panel (A), and secondary and tertiary educational IM in Panel (B). In all cases, we confirm our findings that mining activities have a statistically significant positive (negative) effect on the probability of upward (downward) primary educational IM, while the effect on the probability of secondary/tertiary educational IM is not statistically significant.

^{Table 12:}

Robustness check: use of alternative time exposition to mining activities

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM, considering different treated group expositions. Standard errors are in parentheses. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 12:}

Robustness check: use of alternative time exposition to mining activities

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	10 years				15 years
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-10	Prod-10	Disc-10	Prod-10	Disc-15	Prod-15	Disc-15	Prod-15
Panel (A) Primary education
Mining	0.022*** (0.002)	0.052*** (0.003)	-0.008*** (0.002)	-0.012*** (0.002)	0.005** (0.002)	0.048*** (0.003)	-0.003* (0.001)	-0.013*** (0.002)
Observations	8397851	8397851	4459701	4459701	8397851	8397851	4459701	4459701
R-squared	0.270	0.270	0.134	0.134	0.270	0.270	0.134	0.134
# Treated	281879	65666	194290	56478	341347	90201	248730	74213
Panel (B) Secondary and tertiary education
Mining	0.009	0.003	0.012	0.022	0.009	-0.001	-0.007	-0.002
	(0.010)	(0.015)	(0.015)	(0.032)	(0.009)	(0.016)	(0.019)	(0.026)
Observations	3388392	3388392	329884	329884	3388392	3388392	329884	329884
R-squared	0.217	0.217	0.168	0.168	0.217	0.217	0.168	0.168
# Treated	138784	38651	10596	3304	165254	51506	11988	4546
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 12:}

Robustness check: use of alternative time exposition to mining activities

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	10 years				15 years
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-10	Prod-10	Disc-10	Prod-10	Disc-15	Prod-15	Disc-15	Prod-15
Panel (A) Primary education
Mining	0.022*** (0.002)	0.052*** (0.003)	-0.008*** (0.002)	-0.012*** (0.002)	0.005** (0.002)	0.048*** (0.003)	-0.003* (0.001)	-0.013*** (0.002)
Observations	8397851	8397851	4459701	4459701	8397851	8397851	4459701	4459701
R-squared	0.270	0.270	0.134	0.134	0.270	0.270	0.134	0.134
# Treated	281879	65666	194290	56478	341347	90201	248730	74213
Panel (B) Secondary and tertiary education
Mining	0.009	0.003	0.012	0.022	0.009	-0.001	-0.007	-0.002
	(0.010)	(0.015)	(0.015)	(0.032)	(0.009)	(0.016)	(0.019)	(0.026)
Observations	3388392	3388392	329884	329884	3388392	3388392	329884	329884
R-squared	0.217	0.217	0.168	0.168	0.217	0.217	0.168	0.168
# Treated	138784	38651	10596	3304	165254	51506	11988	4546
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

E. Use of alternative IM definitions

We also check the robustness of our findings to alternative definitions of our dependent variables. In the baseline, we considered children living with their biological/step parents and the average values of parents’ education achievements to construct the intergenerational mobility indices. In this subsection, we first broaden the definition of parental authority to include all other immediate relatives from older generations such as uncles/aunts (in law), parents-in-law, grand-parents, and grand-uncles/aunts to account for abandoned or orphan children sent to relatives, and biological parents deliberately sending their children to relatives or places where education conditions are better. The results are reported in Table 13, columns 1–4. Panel (A) is for primary education and Panel (B) is for secondary and tertiary education. Second, we use the minimum and the maximum of the parents’ education attainment instead of the average education attainment to better capture potential parents’ education inequalities as women tend to have lower education attainment then men in Africa. In this case, a child will experience upward (downward) educational IM if his/her education attainment is higher (lower) than the minimum or the maximum of his/her parents’ education attainment.²⁶ The results of the estimates are displayed in Table 13, columns 5–12. We still find that mining discoveries and productions increase (reduce) the likelihood of upward (downward) primary educational IM in Panel (A), while the effects on secondary/tertiary educational IM is insignificant in Panel (B). That said, the use of alternative intergenerational mobility definitions does not alter our findings.

^{Table 13:}

Robustness check: using alternative IM definitions

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM, using alternative definitions of IM to account for fostered, abandoned, or orphan children. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 13:}

Robustness check: using alternative IM definitions

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	All immediate older generations				Minimum parents' education				Maximum parents' education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.022*** (0.002)	0.070*** (0.003)	-0.016*** (0.002)	-0.008*** (0.002)	0.028*** (0.002)	0.067*** (0.003)	-0.008*** (0.002)	-0.006*** (0.002)	0.028*** (0.002)	0.070*** (0.003)	-0.013*** (0.002)	-0.012*** (0.002)
Observations	9317365	9583522	4759270	4876916	9656297	9920198	3024150	3095599	8306024	8537407	4374423	4478390
R-squared	0.279	0.280	0.137	0.136	0.280	0.281	0.0917	0.0909	0.269	0.270	0.134	0.133
# Treated	168029	63557	112392	41220	172451	62158	74975	28596	148633	53986	98793	36768
Panel (B) Secondary and tertiary education
Mining	-0.009	0.011	0.028	-0.011	-0.009	0.017	0.031*	0.011	-0.012	-0.001	0.025*	-0.003
	(0.012)	(0.018)	(0.022)	(0.022)	(0.013)	(0.021)	(0.017)	(0.032)	(0.010)	(0.006)	(0.013)	(0.008)
Observations	4026641	4178603	381313	390657	3472544	3601346	186869	191439	3390543	3609255	330215	345236
R-squared	0.220	0.219	0.183	0.183	0.234	0.233	0.162	0.162	0.216	0.216	0.167	0.166
# Treated	80221	45703	7960	3549	69345	38973	4377	1842	86544	45422	7904	3240
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 13:}

Robustness check: using alternative IM definitions

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
	All immediate older generations				Minimum parents' education				Maximum parents' education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Primary education
Mining	0.022*** (0.002)	0.070*** (0.003)	-0.016*** (0.002)	-0.008*** (0.002)	0.028*** (0.002)	0.067*** (0.003)	-0.008*** (0.002)	-0.006*** (0.002)	0.028*** (0.002)	0.070*** (0.003)	-0.013*** (0.002)	-0.012*** (0.002)
Observations	9317365	9583522	4759270	4876916	9656297	9920198	3024150	3095599	8306024	8537407	4374423	4478390
R-squared	0.279	0.280	0.137	0.136	0.280	0.281	0.0917	0.0909	0.269	0.270	0.134	0.133
# Treated	168029	63557	112392	41220	172451	62158	74975	28596	148633	53986	98793	36768
Panel (B) Secondary and tertiary education
Mining	-0.009	0.011	0.028	-0.011	-0.009	0.017	0.031*	0.011	-0.012	-0.001	0.025*	-0.003
	(0.012)	(0.018)	(0.022)	(0.022)	(0.013)	(0.021)	(0.017)	(0.032)	(0.010)	(0.006)	(0.013)	(0.008)
Observations	4026641	4178603	381313	390657	3472544	3601346	186869	191439	3390543	3609255	330215	345236
R-squared	0.220	0.219	0.183	0.183	0.234	0.233	0.162	0.162	0.216	0.216	0.167	0.166
# Treated	80221	45703	7960	3549	69345	38973	4377	1842	86544	45422	7904	3240
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

F. Use of all mineral discoveries and productions

We test for robustness to the coverage of all mineral discoveries and productions. As explained in Section V, we focused on the first discovery and the first production to cancel out any potential anticipation and duplication effects as resource-rich districts are more likely to experience several discoveries or have many production sites. In this subsection, we use all discoveries and productions of mineral resources. The results are displayed in Table 14. We still find that mining activities affect the likelihood of primary educational IM, while the effect on secondary and tertiary educational IM is not statistically significant. Therefore, the change in coverage of mineral discoveries and productions does not alter our baseline findings.

^{Table 14:}

Robustness check: using all mining sites

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM, using all mineral discoveries and productions. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 14:}

Robustness check: using all mining sites

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	0.014***	0.028***	-0.008***	-0.007***	-0.012	-0.001	0.025*	-0.003
	(0.002)	(0.002)	(0.001)	(0.001)	(0.010)	(0.006)	(0.013)	(0.008)
Observations	8432049	8797840	4485265	4679709	3390543	3609255	330215	345236
R-squared	0.270	0.272	0.134	0.133	0.216	0.216	0.167	0.166
# Treated	185161	64163	136094	45205	86544	45422	7904	3240
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 14:}

Robustness check: using all mining sites

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	0.014***	0.028***	-0.008***	-0.007***	-0.012	-0.001	0.025*	-0.003
	(0.002)	(0.002)	(0.001)	(0.001)	(0.010)	(0.006)	(0.013)	(0.008)
Observations	8432049	8797840	4485265	4679709	3390543	3609255	330215	345236
R-squared	0.270	0.272	0.134	0.133	0.216	0.216	0.167	0.166
# Treated	185161	64163	136094	45205	86544	45422	7904	3240
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

G. Use of conflicts as additional control variables

Finally, we verify if our baseline results hold after the inclusion of conflicts as additional control variables, given their negative association with mineral discoveries, extensively found in the literature. We have not included this variable in the baseline since geolocalized data on conflicts at district level is available from 1989, thereby restraining our sample of study. We use the Georeferenced Event Dataset (GED) on conflicts from Uppsala Conflict Data Program. We create a dummy equal to one if individuals aged 0–16 years old were exposed to a conflict with more than 25 deaths at the district level and use it as explanatory of primary educational IM. For secondary and tertiary educational IM, we rather consider individuals aged 0–25 years old for exposition to conflicts. The results are reported in Table 15. We show that conflicts are negatively associated with upward primary IM, while they have no significant effects on downward primary IM as well as both upward and downward secondary and tertiary IM. However, the effect of mining activities (discoveries or productions) on educational upward (downward) IM remain positive (negative) for the primary level, and not significant for the secondary and tertiary level. Then, additional conflicts as an explanatory of educational IM does not alter our baseline findings.

^{Table 15:}

Robustness check: adding conflicts as explanatory of educational IM

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM, adding conflicts has an explanatory variable. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 15:}

Robustness check: adding conflicts as explanatory of educational IM

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	0.028***	0.020***	-0.016***	0.001	0.012	0.024	0.019	0.002
	(0.003)	(0.005)	(0.003)	(0.002)	(0.026)	(0.029)	(0.037)	(0.022)
Conflict	-0.009***	-0.009***	-0.000	-0.000	0.010	0.010	-0.012	-0.014*
	(0.001)	(0.001)	(0.001)	(0.001)	(0.010)	(0.009)	(0.008)	(0.007)
Observations	2240418	2327927	1529992	1566001	1051050	1094890	111055	115467
R-squared	0.380	0.382	0.162	0.161	0.180	0.180	0.100	0.100
# Treated	57972	11529	34160	21096	31019	12668	3016	2093
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 15:}

Robustness check: adding conflicts as explanatory of educational IM

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Mining	0.028***	0.020***	-0.016***	0.001	0.012	0.024	0.019	0.002
	(0.003)	(0.005)	(0.003)	(0.002)	(0.026)	(0.029)	(0.037)	(0.022)
Conflict	-0.009***	-0.009***	-0.000	-0.000	0.010	0.010	-0.012	-0.014*
	(0.001)	(0.001)	(0.001)	(0.001)	(0.010)	(0.009)	(0.008)	(0.007)
Observations	2240418	2327927	1529992	1566001	1051050	1094890	111055	115467
R-squared	0.380	0.382	0.162	0.161	0.180	0.180	0.100	0.100
# Treated	57972	11529	34160	21096	31019	12668	3016	2093
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

IX. Sensitivity

In this section, we undertake some sensitivity tests to explore whether our findings vary across African regions, size of the mining sites, gender, and urban-rural residency.

A. Depending on the African regions

We first explore whether the regional subdivision matters. We split the African continent into four regions: Eastern Africa, Northern Africa, Southern Africa, and Western and Central Africa. We then estimate the effects of mining activities on the probability of upward/downward educational IM for each region. The results reported in Table 16 show that there are some heterogeneities across regions. We find that the coefficients associated with mineral discoveries or productions are positive and strongly significant at the 1 percent level (columns 1– 2), suggesting that mining activities tend to increase the probability of upward primary educational IM in all African regions. However, mining activities reduce the likelihood of downward primary educational IM only in Eastern Africa and Northern Africa, as the coefficients associated with mineral discoveries or productions are negative and significant only for these two regions (columns 3–4). Regarding secondary and tertiary educational IM, the results are more divergent. We find that the coefficient associated with mining activities are negative and significant in Eastern Africa and Northern Africa, while positive and significant in Southern Africa, and Western and Central Africa (columns 5–6). In other words, mining activities increase the probability of upward secondary and tertiary educational IM in Southern Africa, and Western and Central Africa, while reducing the probability of upward secondary and tertiary educational IM in Eastern Africa and Northern Africa. On the other hand, the coefficients associated with downward secondary and tertiary educational IM are not statistically significant in all regions, except some positive associations in Eastern and Northern Africa.

^{Table 16:}

Sensitivity: African regions

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM across African regions. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 16:}

Sensitivity: African regions

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
Mining	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Eastern Africa	0.011***	0.054***	-0.024***	-0.017***	-0.013***	-0.029***	0.037*	0.036
	(0.003)	(0.005)	(0.004)	(0.003)	(0.004)	(0.008)	(0.022)	(0.036)
Northern Africa	0.039***	0.105***	-0.008***	-0.012***	-0.016***	0.015	0.045***	0.030
	(0.003)	(0.006)	(0.002)	(0.004)	(0.005)	(0.010)	(0.016)	(0.036)
Southern Africa	0.044***	0.026**	-0.005	-0.008	0.048***	0.066***	0.003	-0.056*
	(0.008)	(0.011)	(0.005)	(0.006)	(0.010)	(0.009)	(0.060)	(0.032)
Western and Central Africa	0.078***	0.062***	0.007	0.007	0.063***	0.050***	-0.015	-0.063
	(0.016)	(0.017)	(0.011)	(0.007)	(0.016)	(0.018)	(0.060)	(0.058)
Observations	8306024	8537407	4374423	4478390	3335415	3461167	323998	331618
R-squared	0.269	0.270	0.134	0.133	0.217	0.217	0.169	0.169
# Treated; Eastern Africa	71453	32587	35765	11765	27505	19397	1880	541
# Treated; Northern Africa	57616	11335	45223	10666	25016	6121	2935	476
# Treated; Southern Africa	16186	7765	12252	10160	11255	10185	888	1457
# Treated; Western and Central Africa	3378	2299	5553	4177	3749	2299	494	339
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

^{Table 16:}

Sensitivity: African regions

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
Mining	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Eastern Africa	0.011***	0.054***	-0.024***	-0.017***	-0.013***	-0.029***	0.037*	0.036
	(0.003)	(0.005)	(0.004)	(0.003)	(0.004)	(0.008)	(0.022)	(0.036)
Northern Africa	0.039***	0.105***	-0.008***	-0.012***	-0.016***	0.015	0.045***	0.030
	(0.003)	(0.006)	(0.002)	(0.004)	(0.005)	(0.010)	(0.016)	(0.036)
Southern Africa	0.044***	0.026**	-0.005	-0.008	0.048***	0.066***	0.003	-0.056*
	(0.008)	(0.011)	(0.005)	(0.006)	(0.010)	(0.009)	(0.060)	(0.032)
Western and Central Africa	0.078***	0.062***	0.007	0.007	0.063***	0.050***	-0.015	-0.063
	(0.016)	(0.017)	(0.011)	(0.007)	(0.016)	(0.018)	(0.060)	(0.058)
Observations	8306024	8537407	4374423	4478390	3335415	3461167	323998	331618
R-squared	0.269	0.270	0.134	0.133	0.217	0.217	0.169	0.169
# Treated; Eastern Africa	71453	32587	35765	11765	27505	19397	1880	541
# Treated; Northern Africa	57616	11335	45223	10666	25016	6121	2935	476
# Treated; Southern Africa	16186	7765	12252	10160	11255	10185	888	1457
# Treated; Western and Central Africa	3378	2299	5553	4177	3749	2299	494	339
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes

B. Depending on the size of mineral discoveries

We now study whether our results generalize to all sizes of mineral discoveries or if some subgroups of minerals have specific effects on education outcomes. The Minex Consulting Dataset (2019) splits mineral discoveries into four categories: moderate, major, giant and super-giant mining. We merge the last two categories as there was not sufficient observations to include each of them in the estimates. The results are in Table 17, columns 1–4 for primary education and columns 5–6 for secondary and tertiary education. We find that the coefficient associated with all sizes of mineral discoveries are positive and significant at the 1 percent in columns 1–2, suggesting that mineral discoveries or productions, regardless of its size, is positively correlated with higher likelihood of primary educational upward IM for individuals born after the discovery than those born before. However, we observe in columns 3–4 that the coefficients associated with giant and super-giant mining are not statistically significant, while those associated with moderate and major mining are significant and in line with our baseline findings. Therefore, individuals living in districts with moderate and major mining operations are less likely to experience downward primary educational IM. Furthermore, we find that the coefficients associated with major and moderate mining are higher in absolute terms than those associated with giant and super-giant mining. On the other hand, only the coefficients associated with giant and super-giant mining discoveries are statistically significant in columns 5–6, meaning that individuals exposed to the discoveries and productions of giant and super-giant mining have a higher likelihood of upward secondary and tertiary educational IM.

^{Table 17:}

Sensitivity: size of mineral discoveries

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM by sizes of discoveries. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 17:}

Sensitivity: size of mineral discoveries

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Moderate Mining	0.011***	0.054***	-0.024***	-0.017***	-0.013	-0.029**	0.037	0.036**
	(0.003)	(0.005)	(0.004)	(0.003)	(0.024)	(0.013)	(0.037)	(0.017)
Major Mining	0.039***	0.105***	-0.008***	-0.012***	-0.016	0.015	0.045**	0.030
	(0.003)	(0.006)	(0.002)	(0.004)	(0.010)	(0.047)	(0.021)	(0.032)
Giant and Super-Giant Mining	0.051***	0.036***	-0.003	-0.003	0.053***	0.062***	-0.007	-0.057***
	(0.007)	(0.009)	(0.005)	(0.005)	(0.019)	(0.015)	(0.045)	(0.014)
Observations	8306024	8537407	4374423	4478390	3335415	3461167	323998	331618
R-squared	0.269	0.270	0.134	0.133	0.217	0.217	0.169	0.169
# Treated; Moderate	71453	32587	35765	11765	27505	19397	1880	541
# Treated; Major	57616	11335	45223	10666	25016	6121	2935	476
# Treated; Giant and Super-Giant	19564	10064	17805	14337	15004	12484	1382	1796
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Moderate <> Major, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Moderate <> Giant / Super Giant, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Major <> Giant / Super Giant, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

^{Table 17:}

Sensitivity: size of mineral discoveries

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Moderate Mining	0.011***	0.054***	-0.024***	-0.017***	-0.013	-0.029**	0.037	0.036**
	(0.003)	(0.005)	(0.004)	(0.003)	(0.024)	(0.013)	(0.037)	(0.017)
Major Mining	0.039***	0.105***	-0.008***	-0.012***	-0.016	0.015	0.045**	0.030
	(0.003)	(0.006)	(0.002)	(0.004)	(0.010)	(0.047)	(0.021)	(0.032)
Giant and Super-Giant Mining	0.051***	0.036***	-0.003	-0.003	0.053***	0.062***	-0.007	-0.057***
	(0.007)	(0.009)	(0.005)	(0.005)	(0.019)	(0.015)	(0.045)	(0.014)
Observations	8306024	8537407	4374423	4478390	3335415	3461167	323998	331618
R-squared	0.269	0.270	0.134	0.133	0.217	0.217	0.169	0.169
# Treated; Moderate	71453	32587	35765	11765	27505	19397	1880	541
# Treated; Major	57616	11335	45223	10666	25016	6121	2935	476
# Treated; Giant and Super-Giant	19564	10064	17805	14337	15004	12484	1382	1796
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Moderate <> Major, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Moderate <> Giant / Super Giant, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Major <> Giant / Super Giant, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

C. Depending on the gender

We then look at whether the effect of mining activities varies based on the gender status of individuals. The models are separately estimated for males and females. The results are reported in Table 18 for both primary education (columns 1–4) and secondary and tertiary education (columns 5–6), with male gender being in Panel (A) and female gender in Panel (B). The p-values of the difference-in-means test between males and females are presented at the bottom of the table. Interestingly, the results in Table 18 show that the coefficient associated with mining in Panel (A) and (B) in column (1) are not statistically different, suggesting that mining discoveries affect by the same magnitude the probability of upward primary educational IM of males and females. However, the coefficient associated with mining in column (2) is nearly 2 times higher in Panel (A) than in Panel (B), reflecting the gender gap in benefits associated with mining production, in favor of males. Indeed, the probability for males to experience an upward primary educational IM is 8.4 percent, against 4.9 percent for females. We also find that the coefficients associated with mining in columns (3) and (4) are higher in absolute terms in Panel (A) than in Panel (B). Males are therefore less likely to experience downward primary educational IM than females. Regarding secondary and tertiary education, Table 18 shows that the coefficient associated with mining is mostly not statistically significant or inconsistently estimated both for males and females.

^{Table 18:}

Sensitivity: gender

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM by gender. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 18:}

Sensitivity: gender

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Male
Mining	0.027***	0.084***	-0.016***	-0.017***	-0.006	0.017***	0.045**	-0.024
	(0.003)	(0.005)	(0.003)	(0.003)	(0.004)	(0.006)	(0.018)	(0.026)
Observations	5101686	5241058	2361599	2417884	2089939	2165635	190636	194782
R-squared	0.257	0.258	0.126	0.125	0.232	0.232	0.172	0.172
# Treated	85358	32490	51792	19482	37088	23104	3269	1553
Panel (B) Female
Mining	0.027***	0.049***	-0.010***	-0.006*	-0.014***	0.014*	0.029	0.011
	(0.003)	(0.005)	(0.003)	(0.003)	(0.005)	(0.008)	(0.018)	(0.027)
Observations	3204338	3296349	2012824	2060506	1245476	1295532	133362	136836
R-squared	0.308	0.308	0.153	0.153	0.217	0.217	0.194	0.194
# Treated	63275	21496	47001	17286	30437	14898	2928	1260
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Mean difference, p-value	1.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

^{Table 18:}

Sensitivity: gender

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Male
Mining	0.027***	0.084***	-0.016***	-0.017***	-0.006	0.017***	0.045**	-0.024
	(0.003)	(0.005)	(0.003)	(0.003)	(0.004)	(0.006)	(0.018)	(0.026)
Observations	5101686	5241058	2361599	2417884	2089939	2165635	190636	194782
R-squared	0.257	0.258	0.126	0.125	0.232	0.232	0.172	0.172
# Treated	85358	32490	51792	19482	37088	23104	3269	1553
Panel (B) Female
Mining	0.027***	0.049***	-0.010***	-0.006*	-0.014***	0.014*	0.029	0.011
	(0.003)	(0.005)	(0.003)	(0.003)	(0.005)	(0.008)	(0.018)	(0.027)
Observations	3204338	3296349	2012824	2060506	1245476	1295532	133362	136836
R-squared	0.308	0.308	0.153	0.153	0.217	0.217	0.194	0.194
# Treated	63275	21496	47001	17286	30437	14898	2928	1260
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Mean difference, p-value	1.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

D. Depending on the urban-rural living area

We split the sample into two subsamples based on the urban-rural residency, and then run the estimates of the effect of mining activities on the likelihood of upward and downward educational IM for each subgroup, separately. The results are reported in Table 19 for both primary education (columns 1–4) and secondary and tertiary education (columns 5–6). The estimates for urban residents are in Panel (A), while those of rural residents are in Panel (B). We report at the bottom of the table the p-value of the significance of the difference in coefficients between urban and rural areas. In columns 1–4, we find that the coefficients associated with mining are broadly higher in absolute terms in urban areas than in rural areas, suggesting that the effect of mining activities on the probability of educational IM tends to be high for individuals living in urban areas. In columns 5–6, we observe that the coefficients associated with our variable of interest are not statistically significant in rural areas, while there are significant in urban areas, meaning that the place of living also matters regarding the effect of mining activities on the probability of secondary and tertiary educational IM. Indeed, individuals living in rural areas face unique barriers to economic and educational opportunities (including poverty, access to public goods, lack of teachers, weaker local economy, etc.) than those living in urban areas. This supports that it is more likely that the preconditions for a positive effect of mineral discoveries and productions on educational IM are more likely to be reunited in urban than rural areas.

^{Table 19:}

Sensitivity: residency

Note: This table presents the effects of mineral discoveries and productions on upward and downward educational IM by urban-rural residency. Disc-B/A means that the treatment group is after the discovery and the control group is before the discovery; Prod-B/A means that the treatment group is after the beginning of production and the control group is before production. *Indicates significance at 10% level, **significance at 5% level, and ***significance at 1% level.

^{Table 19:}

Sensitivity: residency

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Urban
Mining	0.055***	0.083***	-0.015***	-0.010***	-0.013**	0.020***	0.047***	-0.022
	(0.005)	(0.006)	(0.003)	(0.003)	(0.006)	(0.007)	(0.017)	(0.021)
Observations	2620209	2675545	2639013	2672691	1407813	1447625	258097	263176
R-squared	0.118	0.118	0.0772	0.0766	0.156	0.155	0.151	0.150
# Treated	31407	20106	41516	22049	23830	19390	3612	2319
Panel (B) Rural
Mining	0.018***	0.060***	-0.005**	-0.009***	-0.007	0.006	0.046	0.050*
	(0.002)	(0.004)	(0.003)	(0.003)	(0.015)	(0.021)	(0.030)	(0.029)
Observations	5685815	5861862	1735410	1805699	1927602	2013542	65901	68442
R-squared	0.290	0.291	0.152	0.152	0.230	0.229	0.208	0.208
# Treated	117226	33880	57277	14719	43695	18612	2585	494
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Mean difference, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

^{Table 19:}

Sensitivity: residency

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Primary education				Secondary and tertiary education
	Upward mobility		Downward mobility		Upward mobility		Downward mobility
	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A	Disc-B/A	Prod-B/A
Panel (A) Urban
Mining	0.055***	0.083***	-0.015***	-0.010***	-0.013**	0.020***	0.047***	-0.022
	(0.005)	(0.006)	(0.003)	(0.003)	(0.006)	(0.007)	(0.017)	(0.021)
Observations	2620209	2675545	2639013	2672691	1407813	1447625	258097	263176
R-squared	0.118	0.118	0.0772	0.0766	0.156	0.155	0.151	0.150
# Treated	31407	20106	41516	22049	23830	19390	3612	2319
Panel (B) Rural
Mining	0.018***	0.060***	-0.005**	-0.009***	-0.007	0.006	0.046	0.050*
	(0.002)	(0.004)	(0.003)	(0.003)	(0.015)	(0.021)	(0.030)	(0.029)
Observations	5685815	5861862	1735410	1805699	1927602	2013542	65901	68442
R-squared	0.290	0.291	0.152	0.152	0.230	0.229	0.208	0.208
# Treated	117226	33880	57277	14719	43695	18612	2585	494
District FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Census-Year FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Mean difference, p-value	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000

X. Conclusion and Policy implications

This paper sheds light on the effects of mineral discoveries and productions on the educational IM on more than 14 million individuals across 28 countries and 2,890 districts. Using this large and unique dataset, we compute absolute measures of intergenerational mobility and provide a panorama of stylized facts about the trend, dynamics, and disparities of educational IM across countries, and regions of the continent. We show that primary and secondary/tertiary educational IM have significantly improved in Africa over time, with a more significant increase in primary IM (higher upward IM and lower downward IM) than in secondary/tertiary IM. We uncover that the gender gap in favor of males has narrowed over time, with sometimes females doing better than males in most recent cohorts. Regarding the living area, we show that although the residency gap has slightly diminished over time, educational IM has always been better in urban than rural areas and the gap still remains significant. We also provide country-specific characteristics regarding educational IM as well as cross-country and within-country disparities. In addition, we identify where is the land of opportunities by mapping the district-level educational IM to unveil the heterogeneities across the African continent.

We then empirically study the potential role of mineral discoveries and productions on educational IM in Africa given the abundance of mineral resources across the continent. To so do, we employ a generalized difference-in-differences method in a quasi-natural experiment to identify the causal relationship between mineral discoveries/production and educational IM. Our findings suggest that mineral discoveries and productions positively affect primary educational IM in Africa for individuals exposed to the mineral sites and living in districts with mineral discoveries and productions. However, no significant effects are found for secondary/tertiary educational IM. We also unveil two transmission channels through which the positive effects of mineral discoveries and productions on educational primary IM operate, including the income effect proxied by parents working in the mining sector and the returns to education.

Our paper has many policy implications. We show that mineral discoveries and productions have helped improved the social and educational intergenerational mobility in Africa by creating job opportunities, and the returns to education. For these opportunities to be seized, several conditions need to be in place. First, governments should implement accommodative policies that will support enterprise development and the full potential for the creation of jobs to be harnessed. These policies can include labor market flexibility and the creation of business linkages between large mining companies and local small and medium enterprises (SMEs) that will strengthen the sector’s capacity to create jobs. Second, as shown in this paper, mineral discoveries and productions tend to profit males and females as well as people living in urban and rural areas differently. Targeted policies that aim at reducing the inequality of opportunities, especially following mineral discoveries, are also welcome. In addition, our results suggest that districts with discoveries could benefit more from the discoveries or productions, therefore calling for the channeling of the mining revenues in a fund and redistributing it among the districts with the objective to reduce regional disparities.

As avenue for further research, it would be interesting to investigate in details how mineral discoveries and productions induce a local structural transformation, by analyzing their impact on intergenerational mobility in occupation. It might be that mineral discoveries and productions also incur an increase of the likelihood that children do better than their parents in terms of employment at local level.

References

Ahlerup, P., T. Baskaran, and A. Bigsten (2020): “Gold mining and education: a long-run resource curse in Africa?” The Journal of Development Studies, 56, 1745–1762.
- Search Google Scholar
- Export Citation
Alesina, A., S. Hohmann, S. Michalopoulos, and E. Papaioannou (2021): “Intergenerational mobility in Africa,” Econometrica, 89, 1–35.
- Search Google Scholar
- Export Citation
Angrist, J. D. and J.-S. Pischke (2009): “Instrumental variables in action: sometimes you get what you need,” Mostly harmless econometrics: an empiricist’s companion, 113–220.
- Search Google Scholar
- Export Citation
Arezki, R., V. A. Ramey, and L. Sheng (2017): “News shocks in open economies: Evidence from giant oil discoveries,” The Quarterly Journal of Economics, 132, 103–155.
- Search Google Scholar
- Export Citation
Azam, M. and V. Bhatt (2015): “Like father, like son? Intergenerational educational mobility in India,” Demography, 52, 1929–1959.
- Search Google Scholar
- Export Citation
Azomahou, T. T. and E. Yitbarek (2020): “Intergenerational Mobility In Education: Is Africa Different?” Contemporary Economic Policy, e12495.
- Search Google Scholar
- Export Citation
Baah, J. A. and J. Eshun (2020): “Intergenerational Education Mobility and Globalization: Evidence from Ghana,” Journal of Economics, Management and Trade, 47–57.
- Search Google Scholar
- Export Citation
Becker, G. S. and N. Tomes (1979): “An equilibrium theory of the distribution of income and intergenerational mobility,” Journal of political Economy, 87, 1153–1189.
- Search Google Scholar
- Export Citation
Bhattacharyya, S. and N. Mamo (2021): “Natural Resources and Conflict in Africa: What Do the Data Show?” Economic Development and Cultural Change, 69, 903–950.
- Search Google Scholar
- Export Citation
Black, S. E. and P. J. Devereux (2010): “Recent developments in intergenerational mobility,” Tech. rep., National Bureau of Economic Research.
- Search Google Scholar
- Export Citation
Bütikofer, A., A. Dalla-Zuanna, K. G. Salvanes, et al. (2018): “Breaking the links: Natural resource booms and intergenerational mobility,” NHH Dept. of Economics Discussion Paper, 19.
- Search Google Scholar
- Export Citation
Cavalcanti, T., D. Da Mata, and F. Toscani (2019): “Winning the oil lottery: The impact of natural resource extraction on growth,” Journal of Economic Growth, 24, 79–115.
- Search Google Scholar
- Export Citation
Chetty, R., N. Hendren, P. Kline, and E. Saez (2014): “Where is the land of opportunity? The geography of intergenerational mobility in the United States,” The Quarterly Journal of Economics, 129, 1553–1623.
- Search Google Scholar
- Export Citation
Chetty, R., and N. Hendren (2018): “The impacts of neighborhoods on intergenerational mobility II: County-level estimates,” The Quarterly Journal of Economics, 133(3), 1163-1228.
- Search Google Scholar
- Export Citation
Cockx, L. and N. Francken (2016): “Natural resources: a curse on education spending?” Energy Policy, 92, 394–408.
- Search Google Scholar
- Export Citation
Collier, P. and A. Hoeffler (2005): “Resource rents, governance, and conflict,” Journal of conflict resolution, 49, 625–633.
- Search Google Scholar
- Export Citation
Corak, M. (2006): “Do poor children become poor adults? Lessons from a cross-country comparison of generational earnings mobility,” Research on economic inequality, 13, 143–188.
- Search Google Scholar
- Export Citation
Corak, M. (2013): “Income inequality, equality of opportunity, and intergenerational mobility,” Journal of Economic Perspectives, 27, 79–102.
- Search Google Scholar
- Export Citation
Corden, W. M. and J. P. Neary (1982): “Booming sector and de-industrialisation in a small open economy.” University of Stockholm, Institute for International Economic Studies, Reprint Series, 204, 825–848.
- Search Google Scholar
- Export Citation
Cust, J. and J. T. Mensah (2020): “Natural Resource Discoveries, Citizen Expectations and Household Decisions”, World Bank Policy Research Working Paper; No. 9372.
- Search Google Scholar
- Export Citation
Daude, C. and V. Robano (2015): “On intergenerational (im) mobility in Latin America,” Latin American Economic Review, 24, 9.
- Search Google Scholar
- Export Citation
Engzell, P. and F. C. Tropf (2019): “Heritability of education rises with intergenerational mobility,” Proceedings of the National Academy of Sciences, 116, 25386–25388.
- Search Google Scholar
- Export Citation
Farzanegan, M. R. and M. Thum (2020): “Does oil rents dependency reduce the quality of education?” Empirical Economics, 58, 1863–1911.
- Search Google Scholar
- Export Citation
Fisher, E., R. Mwaipopo, W. Mutagwaba, D. Nyange, and G. Yaron (2009): “The ladder that sends us to wealth”: Artisanal mining and poverty reduction in Tanzania,” Resources Policy, 34, 32–38.
- Search Google Scholar
- Export Citation
Goderis, B. and S. W. Malone (2011): “Natural resource booms and inequality: theory and evidence,” Scandinavian Journal of Economics, 113, 388–417.
- Search Google Scholar
- Export Citation
Gradstein, M., & Ishak, P. W. (2020): “We Don't Need No Education: The Effect of Natural Resource Income Shocks on Human Capital”, IZA Discussion Papers, No. 14004.
- Search Google Scholar
- Export Citation
Gylfason, T. (2001): “Natural resources, education, and economic development”, European economic review, 45, 847–859.
- Search Google Scholar
- Export Citation
Hamoudi, A. and D. Thomas (2014): “Endogenous coresidence and program incidence: South Africa’s old age pension,” Journal of development economics, 109, 30–37.
- Search Google Scholar
- Export Citation
Harding, T., Stefanski, R., & Toews, G. (2020): “Boom goes the price: Giant resource discoveries and real exchange rate appreciation”, The Economic Journal, 130(630), 1715-1728.
- Search Google Scholar
- Export Citation
Hausermann, H., D. Ferring, B. Atosona, G. Mentz, R. Amankwah, A. Chang, K. Hartfield, E. Effah, G. Y. Asuamah, C. Mansell, et al. (2018): “Land-grabbing, land-use transformation and social differentiation: Deconstructing “small-scale” in Ghana’s recent gold rush,” World Development, 108, 103–114.
- Search Google Scholar
- Export Citation
Heckert, J., A. Pereira, C. Doss, E. C. Myers, and A. Quisumbing (2021): “Structural Transformation and Gendered Transitions to Adulthood among Rural Youth: Cross-National Evidence from Low-and Middle-Income Countries,” The Journal of Development Studies, 57, 614–634.
- Search Google Scholar
- Export Citation
Henn, S. J. and J. A. Robinson (2021): “Africa’s Latent Assets,” National Bureau of Economic Research, Tech. rep., No. w28603
- Search Google Scholar
- Export Citation
Hertz, T., T. Jayasundera, P. Piraino, S. Selcuk, N. Smith, and A. Verashchagina (2008): “The inheritance of educational inequality: International comparisons and fifty-year trends,” The BE Journal of Economic Analysis & Policy, 7(2).
- Search Google Scholar
- Export Citation
Horn, M. M. K. (2011): “Giant Oil and Gas Fields of the World,” http://www.datapages.com/AssociatedWebsites/GISOpenFiles/HornGiantFields.aspx.
- Search Google Scholar
- Export Citation
Howell, J. (2019): “Neighbourhood effects in cross-Atlantic perspective: A longitudinal analysis of impacts on intergenerational mobility in the USA and Germany,” Urban Studies, 56, 434–451.
- Search Google Scholar
- Export Citation
Keen, D. (2012): “Greed and grievance in civil war,” International Affairs, 88, 757–777.
- Search Google Scholar
- Export Citation
Khan, T., T. Nguyen, F. Ohnsorg, and Schodde (2016): “From Commodity Discovery to Production,” World Bank Policy Research Working Paper No. 7823, 1–23.
- Search Google Scholar
- Export Citation
Kim, D.-H. and S.-C. Lin (2017): “Human capital and natural resource dependence,” Structural Change and Economic Dynamics, 40, 92–102.
- Search Google Scholar
- Export Citation
Kretzmann, S. and I. Nooruddin (2005): “Drilling into debt: An Investigation into the Relationship Between Debt and Oil,” Oil Change International, in http://priceofoil.org.
- Search Google Scholar
- Export Citation
Leamer, E. E., H. Maul, S. Rodriguez, and P. K. Schott (1999): “Does natural resource abundance increase Latin American income inequality?” Journal of development Economics, 59, 3–42.
- Search Google Scholar
- Export Citation
Lei, Y. H. and G. Michaels (2014): “Do giant oilfield discoveries fuel internal armed conflicts?” Journal of Development Economics, 110, 139–157.
- Search Google Scholar
- Export Citation
Loayza, N., A. Mier y Teran, and J. Rigolini (2013): “Poverty, inequality, and the local natural resource curse,” World Bank Policy Research Working Paper, 6366.
- Search Google Scholar
- Export Citation
Marlet, E. G. S. (2020): “The Mining Industry in Sub-Saharan Africa: Impacts on Employment, Migration and Remittances,” University of Washington, Ph.D. thesis.
- Search Google Scholar
- Export Citation
Maurer, S. E. (2019): “Oil discoveries and education provision in the Postbellum South,” Economics of Education Review, 73, 101925.
- Search Google Scholar
- Export Citation
Minex Consulting Datasets (2019): FERDI Study Major Discoveries Since 1950., FERDI Study Major Discoveries Since 1950
- Search Google Scholar
- Export Citation
Neidhöfer, G., J. Serrano, and L. Gasparini (2018): “Educational inequality and intergenerational mobility in Latin America: A new database,” Journal of Development Economics, 134, 329–349.
- Search Google Scholar
- Export Citation
Nimubona, A.-D. and D. Vencatachellum (2007): “Intergenerational education mobility of black and white South Africans,” Journal of Population Economics, 20, 149–182.
- Search Google Scholar
- Export Citation
Pegg, S. (2010): “Is there a Dutch disease in Botswana?” Resources Policy, 35, 14–19.
- Search Google Scholar
- Export Citation
Ross, M. (2006): “A Closer Look at Oil, Diamonds, and Civil War,” Annual Review of Political Science, 9, 265–300.
- Search Google Scholar
- Export Citation
Ross, M. L. (2004): “What do we know about natural resources and civil war?”, Journal of peace research, 41(3), 337-356.
- Search Google Scholar
- Export Citation
Rutstein, S. O. and S. Staveteig (2014): “Making the demographic and health surveys wealth index comparable,” DHS Methodological Reports No. 9. Rockville, Maryland: ICF International. Available at http://dhsprogram.com/pubs/pdf/MR9/MR9.pdf.
- Search Google Scholar
- Export Citation
Saavedra, M. and T. Twinam (2020): “A machine learning approach to improving occupational income scores,” Explorations in Economic History, 75, 101304.
- Search Google Scholar
- Export Citation
Sachs, J. and A. Warner (1995): “Natural Resource Abundance and Economic Growth,” Tech. rep., National Bureau of Economic Research, Cambridge, MA, Working Paper 5398
- Search Google Scholar
- Export Citation
Sachs, J. D. and A. M. Warner (2001): “The curse of natural resources,” European Economic Review, 45 (4-6), 827–838.
- Search Google Scholar
- Export Citation
Séri R.S. (2021) “How giant discoveries of natural resources impact sovereign debt ratings in developing and emerging countries?”, Études et Documents, n°10, CERDI.
- Search Google Scholar
- Export Citation
Smith, B. and S. Wills (2018): “Left in the dark? Oil and rural poverty,” Journal of the Association of Environmental and Resource Economists, 5, 865–904.
- Search Google Scholar
- Export Citation
Stijns, J.-P. (2006): “Natural resource abundance and human capital accumulation,” World development, 34, 1060–1083.
- Search Google Scholar
- Export Citation
Toews, G. and P.-L. Vezina (2017): “Resource discoveries and FDI bonanzas: An illustration from Mozambique,” OxCarre Research Paper, 199.
- Search Google Scholar
- Export Citation
Torche, F. (2014): “Intergenerational mobility and inequality: The Latin American case,” Annual Review of Sociology, 40, 619–642.
- Search Google Scholar
- Export Citation
Tsui, K. K. (2011): “More Oil, Less Democracy: Evidence from Worldwide Crude Oil Discoveries,” Economic Journal, 121, 89–115.
- Search Google Scholar
- Export Citation
Van Der Ploeg, F. (2011): “Natural resources: Curse or blessing?” Journal of Economic Literature, 49, 366–420.
- Search Google Scholar
- Export Citation
Van Der Ploeg, F. and S. Poelhekke (2017): “The Impact of Natural Resources: Survey of Recent Quantitative Evidence,” Journal of Development Studies, 53, 205–216.
- Search Google Scholar
- Export Citation
von der Goltz, J. and P. Barnwal (2019): “Mines: The local wealth and health effects of mineral mining in developing countries,” Journal of Development Economics, 139, 1–16.
- Search Google Scholar
- Export Citation
Weber-Fahr, M. (2002): “Treasure Or Trouble?: Mining in Developing Countries,” World Bank.
- Search Google Scholar
- Export Citation
Witter, S. and M. Jakobsen (2017): “Choices for spending government revenue: New African oil, gas, and mining economies,” Tech. rep., WIDER Working Paper., No. 2017/150.
- Search Google Scholar
- Export Citation
WorldBank (2020): “World Development Indicators’,” https://databank.worldbank.org/source/world-development-indicators.
- Search Google Scholar
- Export Citation
Young, A. (2012): “The African growth miracle,” Journal of Political Economy, 120, 696–739.
- Search Google Scholar
- Export Citation
Zabsonre, A., M. Agbo, and J. Some (2018): “Gold exploitation and socioeconomic outcomes: The case of Burkina Faso,” World Development, 109, 206–221.
- Search Google Scholar
- Export Citation

Annexes

A. Sample

^{Table 20:}

Construction of Sample from raw IPUMS data

Notes: This table shows how we construct our final sample from the raw IPUMS data. Columns (1) and (2) give the country and census year, respectively. Columns (3) shows the initial number of observation in IPUMS data. Columns (4) gives the number of observations with available information on district. Columns (5) gives the number of observations with available information on educational attainment and district. Columns (6) gives the number of observations for individuals living with at least one relative and for which information on educational attainment and district is available. Columns (7) gives the number of observations for individuals living with at least one biological or step-parents and for which information on educational attainment and district is available. Columns (8) gives the number of observations for individuals aged 16-50 years old and born after 1950 living with at least one relative and for which information on educational attainment and district is available. Columns (9) gives the number of observations for individuals aged 16-50 years old and born after 1950 living with at least one relative and for which information on gender, educational attainment and district is available. Columns (10) gives the number of observations for individuals aged 16-50 years old and born after 1950 living with at least one relative and for which information on residency (urban or rural areas), gender, educational attainment and district is available. Columns (11) gives the number of observations for individuals aged 16-50 years old and born after 1950 living with at least one relative and for which information on residency (urban or rural areas), gender, educational attainment, district, and other control variables is available. Columns (12) gives the final sample with immediate older generation used as reference group for individuals. Columns (13) gives the final sample with biological or step-parents (baseline) used as reference group for individuals. Columns (14) gives the census/survey used in the final sample.

B. Summary statistics of Educational IM

Table 21:

Summary statistics on educational IM

C. Additional stylized facts on mineral discoveries

^{Table 22:}

Summary statistics of mineral discoveries, IPUMS countries, 1950–2000

^{Table 23:}

Compostion of minerals in each metal category

D. Additional stylized facts on country-level educational IM

E. Additional stylized facts on district-level educational IM

^{Table 24:}

District-Level Primary IM by country

Notes: This table shows the average conditional district-level educational IM by country. Columns (2)-(5), (3)-(6), and (4)-(7) give the number of districts, the average educational IM, and the coefficient of variation of education IM, for each country, respectively.

^{Table 25:}

District-Level Secondary and tertiary IM by country

^{Table 26:}

District-Level Primary IM by country and discovery

Notes: This table shows the average conditional district-level educational IM by country and district with or without mineral discovery. Columns (1) gives the country name. Columns (2) is "yes" for districts with discovery, and "No" otherwise. Columns (3) and (7) is "yes" if upward and downward IM is higher in districts with discovery than without discovery, respectively. Columns (4)-(8), (5)-(9), and (6)-(10) give the number of districts, the average educational IM, and the coefficient of variation of education IM, respectively.

^{Table 27:}

District-Level Secondary and tertiary IM by country and discovery

F. Baseline results with control variables

^{Table 28:}

Baseline results with control variables, primary education

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

^{Table 29:}

Baseline results with control variables, secondary and tertiary education

Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01

G. Validation of LIDO score and Wealth Index

These mineral resources include gold, silver, diamonds, emerald, ruby, iron, copper, coal, bauxite, cobalt, uranium, platinum and more.

https://www.unep.org/regions/africa/our-work-africa

https://minexconsulting.com/

See Collier and Hoeffler, 2005; Kretzmann and Nooruddin, 2005; Ross, 2004, 2006; Tsui, 2011; Van Der Ploeg, 2011; Keen, 2012; Lei and Michaels, 2014; Van Der Ploeg and Poelhekke, 2017; Smith and Wills, 2018; Harding et al., 2020 for more details on resource curse’s literature.

To further reinforce the returns to education channel, we also test whether mineral discoveries and productions affect the reallocation of individuals across the broad sectors, i.e., agriculture, manufacturing, and services, with population exposed to the mineral discoveries and productions more like to work in the manufacturing and services sectors. We also test a third channel of the provision of infrastructures but decide not to present them as we use an (imperfect) proxy––access to electricity and clean water––for the provisions of public goods. The results can be obtained upon request.

A higher intergenerational persistence implies lower intergenerational mobility, and inversely.

They also show that the regression coefficient representing the transmission of educational attainment from parent to child has decreased over the past 50 years, reflecting an improvement of mobility over time, while the IM’s correlation coefficient has not changed.

See other papers on developed countries (Black and Devereux, 2010, Corak, 2006, 2013; Chetty et al., 2014; Chetty and Hendren, 2018) and developing countries (Azam and Bhatt, 2015, Daude and Robano, 2015, Neidhöfer et al., 2018) for further discussions on the dynamics, disparities across countries and regions, and determinants of educational IM.

https://www.ipums.org/

The population of the 28 countries covered in our analysis represents around 75 percent of the total population in Africa. As such our results can be extrapolated to the continent.

We drop Burkina Faso (1985), Kenya (1979), Liberia (1974), Togo (1960, 1970) since they do not cover all local regions or do not have any identifier to match children to parents. Moreover, we have harmonized the countries boundaries and district names for countries such as Botswana, Burkina Faso, Ethiopia, Ghana, Guinea, Lesotho, Malawi, Mauritius, Morocco, Nigeria, South Africa, and Uganda. For Nigeria, data come from households’ survey rather than census surveys, therefore the number of observations is small as compared to other countries.

We assume that primary level of education is most of the time completed for individuals above 16 years, and secondary for individuals above 25 years.

According to Bhattacharyya and Mamo (2021), mineral discoveries are defined as giant if they generate an amount of at least US$500 million of revenue per annum for 20 years or more. They are qualified as major if they generate an annual revenue stream superior or equal to US$50 million over a shorter lifetime than in the case of giant discoveries.

For the use of relative educational IM measures, see, Hertz et al. (2008), Black and Devereux (2011); Chetty et al. (2014); Bütikofer et al. (2018); Chetty and Hendren (2018); Azomahou and Yitbarek (2020). These measures are based on continuous type variables such as years of schooling or rank based on years of schooling. We rather use data on education attainment and construct absolute measures of education IM, as they are more available than years of schooling, and therefore increase the coverage of our analysis and reduce the attrition biais. Moreover, as shown by Alesina et al. (2019), data on educational attainment are less subject to measurements errors and allow to identify a common reference group for children (e.g., parents without primary educatoin completed), as compared to years of schooling.

Our analysis does not cover the tertiary education exclusively given the few numbers of observations for the tertiary level.

By replacing biological/step parents in the last sentences with immediate older generation, we obtain our alternative measures of absolute educational IM including other relatives on top of the biological/step parents. We will use these alternative definitions of IM in the robustness section.

For instance, Hamoudi and Tomas (2014) show that lower-educated children are more likely to co-reside with elderly parents as they have a lower opportunity cost of providing elderly care. See also, Alesina et al. (2021) for discussion on cohabitation-selection issues.

We also verify that cohabitation rates are similar between rural and urban areas, across ages and districts with and without discoveries (available upon request).

We present the dynamics of country-level educational IM by district with and without discoveries in the Section 4 where we discuss the stylized facts on both educational IM and mineral discoveries and productions.

The differences are reported in parentheses.

For some districts, educational IM are either negative, close to zero, higher than one, or close to one, due to a small number of observations at district level. Moreover, we show that while country-level and district-level estimates of educational IM may differ, they are strongly correlated and provide a quite similar ordering of countries by educational IM.

For instance, in South Sudan, the probability of upward primary educational IM is between 35 and 52 percent for individuals living in the South against only less than 19 percent for those living in the North. In South Africa, the lowest and highest probability of upward primary educational IM are recorded in the North Central (between 84 and 101 percent) and in the West (between 101 and 117 percent) respectively.

One caveat is worth noting. The differences in the mean differences between districts with and without discoveries in Table 26 and Table 27 (district level) as compared to Figure 9 (country level) are mainly explained by the differences in the specification of the conditional IM. For the former, we use district-fixed effects whereas for the latter we use country-fixed effects. While the specifications with the district-fixed effects better capture the invariants characteristics at the district level (the most disaggregated level), those with the country-fixed effects offer insighful comparisons across countries, cohorts, gender, and residency.

We also test a third channel of the provision of infrastructures but decide not to present them as we use an (imperfect) proxy–– access to electricity and clean water––for the provisions of public goods. The results can be obtained upon request. Indeed, following a mineral discovery, large scale investments in mining infrastructures and other transport infrastructures related to the exploitation and transportation of the resources are needed. Moreover, the revenues generated by the resources offer an opportunity for the region and the country to address infrastructure gaps and enhance economic development. Among the types of infrastructures, the general or local government may increase their public spending in education through better access to all primary and secondary education for all aged-children (Witter and Jakobsen, 2017). Our findings show that for individuals exposed to the mineral discoveries and productions, the likelihood of having access to electricity and clean water increase by around 4 percent, and access to electricity and clean water is strongly and positively (negatively) correlated with the upward (downward) primary educational IM.

The variables used to construct the Wealth index include: (i) whether the household (HH) has at least one domestic servant, (ii) whether any HH member owns a dwelling unit, (iii) HH services and possessions such as drinking water, electricity, fuel cook, and their sources or types, (iv) characteristics of the dwelling such as characteristics of the floor, wall, and roof. The choice of variables is constrained by the availability of data for all countries included in the analysis.

We also use the minimum and maximum for the immediate older generation as a robustness check. Similarly, to the results presented for the biological/step parents, our findings remain unchanged. The results are available upon request.

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/001/2022/160/article-A001-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully

Collapse
Expand

Is Education Neglected in Natural Resources-Rich Countries? An Intergenerational Approach in Africa

Author:

Jean-Marc B. Atsebi

Rasmané Ouedraogo

, and

Regina S. Séri

Volume/Issue:: Volume 2022: Issue 160

Publisher:: International Monetary Fund

ISBN:: 9798400218071

ISSN:: 1018-5941

Pages:: 75

DOI:: https://doi.org/10.5089/9798400218071.001

Keywords
I. Introduction
II. Review of literature
III. Data sources and IM index
IV. Stylized facts on educational IM and mineral discoveries and production
V. Empirical methodology
VI. Results
- A. Baseline results
  - 1) Educational primary IM
  - B. Dynamic effects of mineral discoveries and productions on educational IM
VII. Transmission channels
- A. Income channel: parents working in mining sector
- B. Returns to education: Wealth index and LIDO score
VIII. Robustness checks
IX. Sensitivity
X. Conclusion and Policy implications
References
Annexes

View raw image

^{Figure 1:}

Location of mineral discoveries, by size, for all African countries, 1950–2019

View raw image

^{Figure 2:}

Cohabitations rates

View raw image

^{Figure 3:}

Educational IM by birth cohorts

View raw image

^{Figure 4:}

Educational IM by birth cohorts and gender

View raw image

^{Figure 5:}

Educational IM by birth cohorts and residency

View raw image

^{Figure 6:}

Ranking: Country-level educational IM

View raw image

^{Figure 7:}

District-level educational IM

View raw image

^{Figure 8:}

District-level educational IM in South Africa

View raw image

^{Figure 9:}

Dynamics of country-level IM by district with and without discoveries and cohort

View raw image

^{Figure 10:}

Dynamics of Unconditional IM around the first discovery and start of production

View raw image

^{Figure 11:}

Number of mineral discoveries for all African countries, 1950–2019

View raw image

^{Figure 12:}

Educational IM by country and gender

View raw image

^{Figure 13:}

Educational IM by country and residency

View raw image

^{Figure 14:}

Dynamics of IM by districts with and without discoveries, cohorts, and gender

View raw image

^{Figure 15:}

Dynamics of IM by districts with and without discoveries, cohorts, and residency

View raw image

^{Figure 16:}

Gaps of IM by districts with and without discoveries for each country

View raw image

^{Figure 17:}

Correlations between LIDO score and Wealth Index

View raw image

^{Figure 18:}

Correlations between LIDO score, Wealth Index, and PPP GDP per capita

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
Country	Year	Nall	Ndistrict	Neduc	Nliverelative	Nlivebiop	Nage	Nsex	Nurban	Ncont	NIMiog	NIMbiop	Inc.
Benin	1979	331 049	331 049	246 094	96 925	80 434	22 566	22 566	0	0	0	0	no
Benin	1992	498 419	498 419	435 827	212 711	196 722	50 851	50 851	50 851	40 875	38 869	38 689	yes
Benin	2002	685 467	685 467	612 658	317 380	250 883	84 593	84 593	84 593	48 913	48 651	48 606	yes
Benin	2013	1 009 693	1 009 693	911 604	512 405	412 923	152 398	152 398	152 398	94 195	93 987	93 975	yes
Total		2 524 628	2 524 628	2 206 183	1 139 421	940 962	310 408	310 408	287 842	183 983	181 507	181 270
Botswana	1981	97 238	97 238	73 096	32 554	25 225	9 359	9 359	0	0	0	0	no
Botswana	1991	132 623	132 623	113 172	49 711	41 258	16 624	16 624	16 624	13 877	13 058	13 037	yes
Botswana	2001	168 676	168 676	159 446	76 312	63 327	29 167	29 167	0	0	0	0	no
Botswana	2011	201 752	201 752	190 510	84 347	69 536	33 686	33 686	0	0	0	0	no
Total		600 289	600 289	536 224	242 924	199 346	88 836	88 836	16 624	13 877	13 058	13 037
Burkina Faso	1985	884 797	884 797	484 993	0	0	0	0	0	0	0	0	no
Burkina Faso	1996	1 081 046	1 081 046	803 642	377 219	327 148	112 834	112 834	0	0	0	0	no
Burkina Faso	2006	1 417 824	1 417 824	1 244 906	609 130	543 338	127 886	127 886	127 886	103 987	103 605	103 518	yes
Total		3 383 667	3 383 667	2 533 541	986 349	870 486	240 720	240 720	127 886	103 987	103 605	103 518
Cameroon	1976	736 514	736 514	605 857	224 570	210 439	40 560	40 555	0	0	0	0	no
Cameroon	1987	897 211	897 211	763 744	312 705	301 224	63 193	63 193	63 193	55 113	52 612	52 357	yes
Cameroon	2005	1 772 359	1 766 211	1 536 785	769 108	712 995	240 864	240 864	240 864	207 506	205 996	205 902	yes
Total		3 406 084	3 399 936	2 906 386	1 306 383	1 224 658	344 617	344 612	304 057	262 619	258 608	258 259
Egypt	1986	6 799 093	6 799 093	5 421 801	2 825 392	2 552 405	1 297 832	1 297 829	1 297 829	1 131 225	1 100 397	1 098 502	yes
Egypt	1996	5 902 243	5 902 243	4 453 785	2 126 960	2 027 351	1 185 312	1 185 312	1 185 312	1 090 368	1 083 995	1 083 953	yes
Egypt	2006	7 282 434	7 282 434	5 739 722	2 553 381	2 450 443	1 610 591	1 610 591	1 610 591	1 515 467	1 514 369	1 514 364	yes
Total		19 983 770	19 983 770	15 615 308	7 505 733	7 030 199	4 093 735	4 093 732	4 093 732	3 737 060	3 698 761	3 696 819
Ethiopia	1984	3 404 306	3 400 221	2 735 271	1 209 735	1 149 841	228 972	228 971	228 971	196 636	175 867	175 365	yes
Ethiopia	1994	5 044 598	5 044 598	4 201 617	2 015 604	1 979 879	566 246	566 246	566 246	539 043	524 827	524 590	yes
Ethiopia	2007	7 434 086	7 434 086	1 097 614	544 065	514 140	154 345	154 345	154 345	141 002	140 034	140 024	yes
Total		15 882 990	15 878 905	8 034 502	3 769 404	3 643 860	949 563	949 562	949 562	876 681	840 728	839 979
Ghana	1984	1 309 352	1 309 352	1 050 813	545 036	545 036	187 288	187 288	0	0	0	0	no
Ghana	2000	1 894 133	1 894 133	1 730 902	727 288	671 959	243 122	243 122	243 122	212 320	208 074	208 010	yes
Ghana	2010	2 466 289	2 466 289	2 262 894	1 091 326	1 018 943	400 015	400 015	400 015	360 219	359 524	359 489	yes
Total		5 669 774	5 669 774	5 044 609	2 363 650	2 235 938	830 425	830 425	643 137	572 539	567 598	567 499
Guinea	1983	457 837	457 837	364 823	106 728	105 679	30 830	30 815	30 749	30 444	29 418	28 999	yes
Guinea	1996	729 071	729 071	553 173	246 286	207 001	89 851	89 851	89 851	66 128	63 700	63 495	yes
Guinea	2014	1 050 916	1 050 916	951 617	539 972	444 014	184 293	184 293	184 293	126 181	125 892	125 830	yes
Total		2 237 824	2 237 824	1 869 613	892 986	756 694	304 974	304 959	304 893	222 753	219 010	218 324
Kenya	1969	659 310	659 310	659 310	273 058	263 394	20 959	20 959	0	0	0	0	no
Kenya	1979	1 033 769	1 033 769	854 251	0	0	0	0	0	0	0	0	no
Kenya	1989	1 074 098	1 074 098	829 247	364 258	355 117	115 571	115 571	115 571	110 185	106 171	106 095	yes
Kenya	1999	1 407 547	1 407 547	1 191 268	487 645	482 717	160 587	160 587	160 587	156 553	154 897	154 879	yes
Kenya	2009	3 841 935	3 841 935	3 402 695	1 717 135	1 593 028	507 075	507 075	507 075	462 754	456 403	456 195	yes
Total		8 016 659	8 016 659	6 936 771	2 842 096	2 694 256	804 192	804 192	783 233	729 492	717 471	717 169
Lesotho	1996	187 795	187 776	165 945	88 666	79 967	39 728	39 728	39 728	34 241	33 503	33 496	yes
Lesotho	2006	180 208	180 208	171 947	85 473	77 758	37 556	37 556	37 556	32 983	32 851	32 850	yes
Total		368 003	367 984	337 892	174 139	157 725	77 284	77 284	77 284	67 224	66 354	66 346
Liberia	1974	150 256	150 256	127 442	0	0	0	0	0	0	0	0	no
Liberia	2008	348 057	348 057	294 517	126 770	118 977	47 631	47 631	47 631	44 199	44 012	44 007	yes
Total		498 313	498 313	421 959	126 770	118 977	47 631	47 631	47 631	44 199	44 012	44 007
Malawi	1987	798 669	798 669	658 449	222 672	220 229	49 617	49 617	49 617	48 293	45 303	45 291	yes
Malawi	1998	991 393	991 393	826 197	292 284	286 039	77 453	77 453	77 453	73 746	72 671	72 668	yes
Malawi	2008	1 341 977	1 341 977	1 156 748	497 097	492 609	106 570	106 570	105 741	102 459	101 882	101 877	yes
Total		3 132 039	3 132 039	2 641 394	1 012 053	998 877	233 640	233 640	232 811	224 498	219 856	219 836
Mali	1987	785 384	784 096	581 806	243 229	227 034	78 724	78 724	0	0	0	0	no
Mali	1998	991 330	991 330	737 487	340 903	318 695	117 063	117 063	117 063	98 261	96 126	95 816	yes
Mali	2009	1 451 856	1 451 856	1 285 750	741 784	648 243	209 408	209 408	209 408	157 779	156 175	155 877	yes
Total		3 228 570	3 227 282	2 605 043	1 325 916	1 193 972	405 195	405 195	326 471	256 040	252 301	251 693
Mauritius	1990	106 710	106 710	101 646	54 683	52 074	24 727	24 727	24 727	22 367	21 738	21 736	yes
Mauritius	2000	119 695	119 695	114 499	56 350	53 366	28 650	28 650	28 650	25 829	25 539	25 539	yes
Mauritius	2011	126 332	126 332	121 383	54 820	51 921	29 191	29 191	29 191	26 478	26 431	26 430	yes
Total		352 737	352 737	337 528	165 853	157 361	82 568	82 568	82 568	74 674	73 708	73 705
Morocco	1982	1 012 873	1 012 873	948 008	546 732	511 677	178 034	178 034	0	0	0	0	no
Morocco	1994	1 294 026	1 294 026	1 293 467	835 569	783 915	338 124	338 124	0	0	0	0	no
Morocco	2004	1 482 720	1 482 720	1 482 716	928 290	864 824	442 157	442 157	0	0	0	0	no
Morocco	2014	3 341 426	3 340 830	3 340 830	1 939 870	1 803 530	935 976	935 976	933 662	808 379	808 249	808 242	yes
Total		7 131 045	7 130 449	7 065 021	4 250 461	3 963 946	1 894 291	1 894 291	933 662	808 379	808 249	808 242
Mozambique	1997	1 551 517	1 551 517	1 248 483	515 184	472 203	146 861	146 861	146 861	120 325	118 632	118 617	yes
Mozambique	2007	2 047 048	2 047 048	1 616 853	718 962	664 244	197 113	197 113	197 113	164 534	163 418	163 380	yes
Total		3 598 565	3 598 565	2 865 336	1 234 146	1 136 447	343 974	343 974	343 974	284 859	282 050	281 997
Nigeria	2006	83 700	83 700	82 740	46 326	44 662	10 436	10 436	10 436	9 802	9 738	9 708	yes
Nigeria	2007	85 183	85 183	84 123	47 618	45 235	10 557	10 557	10 557	9 761	9 715	9 696	yes
Nigeria	2008	107 425	107 425	105 944	62 622	60 823	15 199	15 184	15 142	14 469	14 323	14 306	yes
Nigeria	2009	77 896	77 896	77 666	41 179	39 621	9 690	9 673	9 673	8 920	8 883	8 876	yes
Nigeria	2010	72 191	72 191	59 173	30 890	29 756	11 240	11 240	11 240	10 584	10 392	10 377	yes
Total		426 395	426 395	409 646	228 635	220 097	57 122	57 090	57 048	53 536	53 051	52 963
Rwanda	1991	742 918	742 918	0	0	0	0	0	0	0	0	0	no
Rwanda	2002	843 392	843 392	629 146	277 153	264 800	107 446	107 446	107 446	103 899	102 745	102 725	yes
Rwanda	2012	1 038 369	1 038 369	938 201	485 203	463 875	150 671	150 671	150 671	144 338	143 843	143 826	yes
Total		2 624 679	2 624 679	1 567 347	762 356	728 675	258 117	258 117	258 117	248 237	246 588	246 551
Senegal	1988	700 199	700 199	527 680	207 551	181 977	76 093	76 093	0	0	0	0	no
Senegal	2002	994 562	994 562	911 891	477 330	440 205	192 555	192 555	192 555	168 266	167 086	166 999	yes
Senegal	2013	1 245 551	1 016 023	908 310	578 764	477 271	257 347	257 347	257 347	186 875	186 345	186 044	yes
Total		2 940 312	2 710 784	2 347 881	1 263 645	1 099 453	525 995	525 995	449 902	355 141	353 431	353 043
Sierra Leone	2004	494 298	494 298	397 137	182 198	144 915	74 010	74 010	74 010	55 065	53 064	52 469	yes
Total		494 298	494 298	397 137	182 198	144 915	74 010	74 010	74 010	55 065	53 064	52 469
South Africa	1996	3 621 164	3 621 164	3 083 346	1 385 515	1 255 578	621 444	621 444	621 435	562 421	538 506	537 039	yes
South Africa	2001	3 725 655	3 725 655	3 353 684	1 509 080	1 366 352	728 007	728 007	727 870	656 452	650 438	650 174	yes
South Africa	2007	1 047 657	1 047 657	842 103	384 555	347 943	197 500	197 500	197 500	179 236	177 447	177 401	yes
South Africa	2011	4 418 594	4 418 594	3 845 633	1 494 142	1 355 066	777 948	777 948	766 491	699 181	697 461	697 347	yes
South Africa	2016	3 328 793	3 328 793	3 023 034	1 345 876	1 186 490	674 872	674 872	674 872	603 780	591 480	590 681	yes
Total		16 141 863	16 141 863	14 147 800	6 119 168	5 511 429	2 999 771	2 999 771	2 988 168	2 701 070	2 655 332	2 652 642
South Sudan	2008	542 765	542 765	542 333	300 590	273 961	71 220	71 220	71 220	60 552	60 416	60 399	yes
Total		542 765	542 765	542 333	300 590	273 961	71 220	71 220	71 220	60 552	60 416	60 399
Sudan	2008	5 066 530	5 066 530	3 902 071	2 048 229	1 936 664	792 810	792 810	792 810	714 673	699 070	697 097	yes
Total		5 066 530	5 066 530	3 902 071	2 048 229	1 936 664	792 810	792 810	792 810	714 673	699 070	697 097
Tanzania	1988	2 310 424	2 310 424	1 916 737	683 484	679 664	183 596	183 596	0	0	0	0	no
Tanzania	2002	3 732 735	3 732 735	3 123 724	1 245 172	1 186 155	366 594	366 594	366 594	336 996	333 681	333 660	yes
Tanzania	2012	4 498 022	4 498 022	3 918 823	1 763 397	1 661 165	503 981	503 981	503 981	466 289	464 823	464 806	yes
Total		10 541 181	10 541 181	8 959 284	3 692 053	3 526 984	1 054 171	1 054 171	870 575	803 285	798 504	798 466
Togo	1960	13 759	13 759	13 758	5 005	4 238	0	0	0	0	0	0	no
Togo	1970	23 680	23 680	23 672	12 267	11 999	712	712	0	0	0	0	no
Togo	2010	584 859	584 859	517 900	120 225	115 641	33 730	33 730	33 730	32 274	31 926	31 889	yes
Total		622 298	622 298	555 330	137 497	131 878	34 442	34 442	33 730	32 274	31 926	31 889
Uganda	1991	1 548 460	1 529 024	1 226 290	450 737	442 411	131 819	131 819	131 819	124 996	121 888	121 856	yes
Uganda	2002	2 497 449	2 497 449	2 042 838	847 255	838 411	213 799	213 799	213 799	208 499	207 355	207 284	yes
Uganda	2014	3 506 546	3 506 546	3 145 894	1 600 477	1 506 609	400 450	400 450	400 450	374 435	373 919	373 908	yes
Total		7 552 455	7 533 019	6 415 022	2 898 469	2 787 431	746 068	746 068	746 068	707 930	703 162	703 048
Zambia	1990	787 461	787 461	664 239	304 994	304 281	106 751	106 751	106 751	106 406	104 920	104 920	yes
Zambia	2000	996 117	996 117	825 110	417 749	341 108	145 247	145 247	145 247	98 694	97 967	97 933	yes
Zambia	2010	1 321 973	1 321 973	1 028 628	537 693	465 478	177 428	177 428	0	0	0	0	no
Total		3 105 551	3 105 551	2 517 977	1 260 436	1 110 867	429 426	429 426	251 998	205 100	202 887	202 853
Zimbabwe	2012	654 688	654 688	587 748	222 825	204 736	65 518	65 518	65 518	60 166	59 615	59 599	yes
Total		654 688	654 688	587 748	222 825	204 736	65 518	65 518	65 518	60 166	59 615	59 599

Total All		130 727 972	130 466 872	104 306 886	48 454 385	45 000 794	18 160 723	18 160 667	16 214 531	14 459 893	14 263 922	14 252 719	61/82

(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
Country	Year	Nall	Ndistrict	Neduc	Nliverelative	Nlivebiop	Nage	Nsex	Nurban	Ncont	NIMiog	NIMbiop	Inc.
Benin	1979	331 049	331 049	246 094	96 925	80 434	22 566	22 566	0	0	0	0	no
Benin	1992	498 419	498 419	435 827	212 711	196 722	50 851	50 851	50 851	40 875	38 869	38 689	yes
Benin	2002	685 467	685 467	612 658	317 380	250 883	84 593	84 593	84 593	48 913	48 651	48 606	yes
Benin	2013	1 009 693	1 009 693	911 604	512 405	412 923	152 398	152 398	152 398	94 195	93 987	93 975	yes
Total		2 524 628	2 524 628	2 206 183	1 139 421	940 962	310 408	310 408	287 842	183 983	181 507	181 270
Botswana	1981	97 238	97 238	73 096	32 554	25 225	9 359	9 359	0	0	0	0	no
Botswana	1991	132 623	132 623	113 172	49 711	41 258	16 624	16 624	16 624	13 877	13 058	13 037	yes
Botswana	2001	168 676	168 676	159 446	76 312	63 327	29 167	29 167	0	0	0	0	no
Botswana	2011	201 752	201 752	190 510	84 347	69 536	33 686	33 686	0	0	0	0	no
Total		600 289	600 289	536 224	242 924	199 346	88 836	88 836	16 624	13 877	13 058	13 037
Burkina Faso	1985	884 797	884 797	484 993	0	0	0	0	0	0	0	0	no
Burkina Faso	1996	1 081 046	1 081 046	803 642	377 219	327 148	112 834	112 834	0	0	0	0	no
Burkina Faso	2006	1 417 824	1 417 824	1 244 906	609 130	543 338	127 886	127 886	127 886	103 987	103 605	103 518	yes
Total		3 383 667	3 383 667	2 533 541	986 349	870 486	240 720	240 720	127 886	103 987	103 605	103 518
Cameroon	1976	736 514	736 514	605 857	224 570	210 439	40 560	40 555	0	0	0	0	no
Cameroon	1987	897 211	897 211	763 744	312 705	301 224	63 193	63 193	63 193	55 113	52 612	52 357	yes
Cameroon	2005	1 772 359	1 766 211	1 536 785	769 108	712 995	240 864	240 864	240 864	207 506	205 996	205 902	yes
Total		3 406 084	3 399 936	2 906 386	1 306 383	1 224 658	344 617	344 612	304 057	262 619	258 608	258 259
Egypt	1986	6 799 093	6 799 093	5 421 801	2 825 392	2 552 405	1 297 832	1 297 829	1 297 829	1 131 225	1 100 397	1 098 502	yes
Egypt	1996	5 902 243	5 902 243	4 453 785	2 126 960	2 027 351	1 185 312	1 185 312	1 185 312	1 090 368	1 083 995	1 083 953	yes
Egypt	2006	7 282 434	7 282 434	5 739 722	2 553 381	2 450 443	1 610 591	1 610 591	1 610 591	1 515 467	1 514 369	1 514 364	yes
Total		19 983 770	19 983 770	15 615 308	7 505 733	7 030 199	4 093 735	4 093 732	4 093 732	3 737 060	3 698 761	3 696 819
Ethiopia	1984	3 404 306	3 400 221	2 735 271	1 209 735	1 149 841	228 972	228 971	228 971	196 636	175 867	175 365	yes
Ethiopia	1994	5 044 598	5 044 598	4 201 617	2 015 604	1 979 879	566 246	566 246	566 246	539 043	524 827	524 590	yes
Ethiopia	2007	7 434 086	7 434 086	1 097 614	544 065	514 140	154 345	154 345	154 345	141 002	140 034	140 024	yes
Total		15 882 990	15 878 905	8 034 502	3 769 404	3 643 860	949 563	949 562	949 562	876 681	840 728	839 979
Ghana	1984	1 309 352	1 309 352	1 050 813	545 036	545 036	187 288	187 288	0	0	0	0	no
Ghana	2000	1 894 133	1 894 133	1 730 902	727 288	671 959	243 122	243 122	243 122	212 320	208 074	208 010	yes
Ghana	2010	2 466 289	2 466 289	2 262 894	1 091 326	1 018 943	400 015	400 015	400 015	360 219	359 524	359 489	yes
Total		5 669 774	5 669 774	5 044 609	2 363 650	2 235 938	830 425	830 425	643 137	572 539	567 598	567 499
Guinea	1983	457 837	457 837	364 823	106 728	105 679	30 830	30 815	30 749	30 444	29 418	28 999	yes
Guinea	1996	729 071	729 071	553 173	246 286	207 001	89 851	89 851	89 851	66 128	63 700	63 495	yes
Guinea	2014	1 050 916	1 050 916	951 617	539 972	444 014	184 293	184 293	184 293	126 181	125 892	125 830	yes
Total		2 237 824	2 237 824	1 869 613	892 986	756 694	304 974	304 959	304 893	222 753	219 010	218 324
Kenya	1969	659 310	659 310	659 310	273 058	263 394	20 959	20 959	0	0	0	0	no
Kenya	1979	1 033 769	1 033 769	854 251	0	0	0	0	0	0	0	0	no
Kenya	1989	1 074 098	1 074 098	829 247	364 258	355 117	115 571	115 571	115 571	110 185	106 171	106 095	yes
Kenya	1999	1 407 547	1 407 547	1 191 268	487 645	482 717	160 587	160 587	160 587	156 553	154 897	154 879	yes
Kenya	2009	3 841 935	3 841 935	3 402 695	1 717 135	1 593 028	507 075	507 075	507 075	462 754	456 403	456 195	yes
Total		8 016 659	8 016 659	6 936 771	2 842 096	2 694 256	804 192	804 192	783 233	729 492	717 471	717 169
Lesotho	1996	187 795	187 776	165 945	88 666	79 967	39 728	39 728	39 728	34 241	33 503	33 496	yes
Lesotho	2006	180 208	180 208	171 947	85 473	77 758	37 556	37 556	37 556	32 983	32 851	32 850	yes
Total		368 003	367 984	337 892	174 139	157 725	77 284	77 284	77 284	67 224	66 354	66 346
Liberia	1974	150 256	150 256	127 442	0	0	0	0	0	0	0	0	no
Liberia	2008	348 057	348 057	294 517	126 770	118 977	47 631	47 631	47 631	44 199	44 012	44 007	yes
Total		498 313	498 313	421 959	126 770	118 977	47 631	47 631	47 631	44 199	44 012	44 007
Malawi	1987	798 669	798 669	658 449	222 672	220 229	49 617	49 617	49 617	48 293	45 303	45 291	yes
Malawi	1998	991 393	991 393	826 197	292 284	286 039	77 453	77 453	77 453	73 746	72 671	72 668	yes
Malawi	2008	1 341 977	1 341 977	1 156 748	497 097	492 609	106 570	106 570	105 741	102 459	101 882	101 877	yes
Total		3 132 039	3 132 039	2 641 394	1 012 053	998 877	233 640	233 640	232 811	224 498	219 856	219 836
Mali	1987	785 384	784 096	581 806	243 229	227 034	78 724	78 724	0	0	0	0	no
Mali	1998	991 330	991 330	737 487	340 903	318 695	117 063	117 063	117 063	98 261	96 126	95 816	yes
Mali	2009	1 451 856	1 451 856	1 285 750	741 784	648 243	209 408	209 408	209 408	157 779	156 175	155 877	yes
Total		3 228 570	3 227 282	2 605 043	1 325 916	1 193 972	405 195	405 195	326 471	256 040	252 301	251 693
Mauritius	1990	106 710	106 710	101 646	54 683	52 074	24 727	24 727	24 727	22 367	21 738	21 736	yes
Mauritius	2000	119 695	119 695	114 499	56 350	53 366	28 650	28 650	28 650	25 829	25 539	25 539	yes
Mauritius	2011	126 332	126 332	121 383	54 820	51 921	29 191	29 191	29 191	26 478	26 431	26 430	yes
Total		352 737	352 737	337 528	165 853	157 361	82 568	82 568	82 568	74 674	73 708	73 705
Morocco	1982	1 012 873	1 012 873	948 008	546 732	511 677	178 034	178 034	0	0	0	0	no
Morocco	1994	1 294 026	1 294 026	1 293 467	835 569	783 915	338 124	338 124	0	0	0	0	no
Morocco	2004	1 482 720	1 482 720	1 482 716	928 290	864 824	442 157	442 157	0	0	0	0	no
Morocco	2014	3 341 426	3 340 830	3 340 830	1 939 870	1 803 530	935 976	935 976	933 662	808 379	808 249	808 242	yes
Total		7 131 045	7 130 449	7 065 021	4 250 461	3 963 946	1 894 291	1 894 291	933 662	808 379	808 249	808 242
Mozambique	1997	1 551 517	1 551 517	1 248 483	515 184	472 203	146 861	146 861	146 861	120 325	118 632	118 617	yes
Mozambique	2007	2 047 048	2 047 048	1 616 853	718 962	664 244	197 113	197 113	197 113	164 534	163 418	163 380	yes
Total		3 598 565	3 598 565	2 865 336	1 234 146	1 136 447	343 974	343 974	343 974	284 859	282 050	281 997
Nigeria	2006	83 700	83 700	82 740	46 326	44 662	10 436	10 436	10 436	9 802	9 738	9 708	yes
Nigeria	2007	85 183	85 183	84 123	47 618	45 235	10 557	10 557	10 557	9 761	9 715	9 696	yes
Nigeria	2008	107 425	107 425	105 944	62 622	60 823	15 199	15 184	15 142	14 469	14 323	14 306	yes
Nigeria	2009	77 896	77 896	77 666	41 179	39 621	9 690	9 673	9 673	8 920	8 883	8 876	yes
Nigeria	2010	72 191	72 191	59 173	30 890	29 756	11 240	11 240	11 240	10 584	10 392	10 377	yes
Total		426 395	426 395	409 646	228 635	220 097	57 122	57 090	57 048	53 536	53 051	52 963
Rwanda	1991	742 918	742 918	0	0	0	0	0	0	0	0	0	no
Rwanda	2002	843 392	843 392	629 146	277 153	264 800	107 446	107 446	107 446	103 899	102 745	102 725	yes
Rwanda	2012	1 038 369	1 038 369	938 201	485 203	463 875	150 671	150 671	150 671	144 338	143 843	143 826	yes
Total		2 624 679	2 624 679	1 567 347	762 356	728 675	258 117	258 117	258 117	248 237	246 588	246 551
Senegal	1988	700 199	700 199	527 680	207 551	181 977	76 093	76 093	0	0	0	0	no
Senegal	2002	994 562	994 562	911 891	477 330	440 205	192 555	192 555	192 555	168 266	167 086	166 999	yes
Senegal	2013	1 245 551	1 016 023	908 310	578 764	477 271	257 347	257 347	257 347	186 875	186 345	186 044	yes
Total		2 940 312	2 710 784	2 347 881	1 263 645	1 099 453	525 995	525 995	449 902	355 141	353 431	353 043
Sierra Leone	2004	494 298	494 298	397 137	182 198	144 915	74 010	74 010	74 010	55 065	53 064	52 469	yes
Total		494 298	494 298	397 137	182 198	144 915	74 010	74 010	74 010	55 065	53 064	52 469
South Africa	1996	3 621 164	3 621 164	3 083 346	1 385 515	1 255 578	621 444	621 444	621 435	562 421	538 506	537 039	yes
South Africa	2001	3 725 655	3 725 655	3 353 684	1 509 080	1 366 352	728 007	728 007	727 870	656 452	650 438	650 174	yes
South Africa	2007	1 047 657	1 047 657	842 103	384 555	347 943	197 500	197 500	197 500	179 236	177 447	177 401	yes
South Africa	2011	4 418 594	4 418 594	3 845 633	1 494 142	1 355 066	777 948	777 948	766 491	699 181	697 461	697 347	yes
South Africa	2016	3 328 793	3 328 793	3 023 034	1 345 876	1 186 490	674 872	674 872	674 872	603 780	591 480	590 681	yes
Total		16 141 863	16 141 863	14 147 800	6 119 168	5 511 429	2 999 771	2 999 771	2 988 168	2 701 070	2 655 332	2 652 642
South Sudan	2008	542 765	542 765	542 333	300 590	273 961	71 220	71 220	71 220	60 552	60 416	60 399	yes
Total		542 765	542 765	542 333	300 590	273 961	71 220	71 220	71 220	60 552	60 416	60 399
Sudan	2008	5 066 530	5 066 530	3 902 071	2 048 229	1 936 664	792 810	792 810	792 810	714 673	699 070	697 097	yes
Total		5 066 530	5 066 530	3 902 071	2 048 229	1 936 664	792 810	792 810	792 810	714 673	699 070	697 097
Tanzania	1988	2 310 424	2 310 424	1 916 737	683 484	679 664	183 596	183 596	0	0	0	0	no
Tanzania	2002	3 732 735	3 732 735	3 123 724	1 245 172	1 186 155	366 594	366 594	366 594	336 996	333 681	333 660	yes
Tanzania	2012	4 498 022	4 498 022	3 918 823	1 763 397	1 661 165	503 981	503 981	503 981	466 289	464 823	464 806	yes
Total		10 541 181	10 541 181	8 959 284	3 692 053	3 526 984	1 054 171	1 054 171	870 575	803 285	798 504	798 466
Togo	1960	13 759	13 759	13 758	5 005	4 238	0	0	0	0	0	0	no
Togo	1970	23 680	23 680	23 672	12 267	11 999	712	712	0	0	0	0	no
Togo	2010	584 859	584 859	517 900	120 225	115 641	33 730	33 730	33 730	32 274	31 926	31 889	yes
Total		622 298	622 298	555 330	137 497	131 878	34 442	34 442	33 730	32 274	31 926	31 889
Uganda	1991	1 548 460	1 529 024	1 226 290	450 737	442 411	131 819	131 819	131 819	124 996	121 888	121 856	yes
Uganda	2002	2 497 449	2 497 449	2 042 838	847 255	838 411	213 799	213 799	213 799	208 499	207 355	207 284	yes
Uganda	2014	3 506 546	3 506 546	3 145 894	1 600 477	1 506 609	400 450	400 450	400 450	374 435	373 919	373 908	yes
Total		7 552 455	7 533 019	6 415 022	2 898 469	2 787 431	746 068	746 068	746 068	707 930	703 162	703 048
Zambia	1990	787 461	787 461	664 239	304 994	304 281	106 751	106 751	106 751	106 406	104 920	104 920	yes
Zambia	2000	996 117	996 117	825 110	417 749	341 108	145 247	145 247	145 247	98 694	97 967	97 933	yes
Zambia	2010	1 321 973	1 321 973	1 028 628	537 693	465 478	177 428	177 428	0	0	0	0	no
Total		3 105 551	3 105 551	2 517 977	1 260 436	1 110 867	429 426	429 426	251 998	205 100	202 887	202 853
Zimbabwe	2012	654 688	654 688	587 748	222 825	204 736	65 518	65 518	65 518	60 166	59 615	59 599	yes
Total		654 688	654 688	587 748	222 825	204 736	65 518	65 518	65 518	60 166	59 615	59 599

Total All		130 727 972	130 466 872	104 306 886	48 454 385	45 000 794	18 160 723	18 160 667	16 214 531	14 459 893	14 263 922	14 252 719	61/82

	(1)	(2)	(3)	(4)	(5)	(6)
	Obs.	Mean	Sd	Obs.	Mean	Sd
	Panel (A) Primary			Panel (B) Secondary and tertiary
	(I) Upward mobility
	(a) Biological or step-parents
IM (Mean)	9,258,374	0.508	0.500	12,447,352	0.204	0.403
IM (Min)	10,727,953	0.550	0.497	13,228,050	0.221	0.415
IM (Max)	9,258,374	0.508	0.500	12,447,352	0.204	0.403
(b) Immediate older generation
IM (Mean)	9,280,274	0.507	0.500	12,462,921	0.204	0.403
IM (Min)	10,811,989	0.552	0.497	13,275,780	0.222	0.415
IM (Max)	9,132,978	0.505	0.500	12,370,537	0.204	0.403
(II) Downward mobility
(a) Biological or step-parents
IM (Mean)	4,994,345	0.099	0.299	1,805,367	0.438	0.496
IM (Min)	3,524,766	0.064	0.245	1,024,669	0.384	0.486
IM (Max)	4,994,345	0.099	0.299	1,805,367	0.438	0.496
(b) Immediate older generation
IM (Mean)	4,983,648	0.097	0.296	1,801,001	0.435	0.496
IM (Min)	3,451,933	0.062	0.241	988,142	0.375	0.484
IM (Max)	5,130,944	0.104	0.305	1,893,385	0.449	0.497

	(1)	(2)	(3)	(4)	(5)	(6)
	Obs.	Mean	Sd	Obs.	Mean	Sd
	Panel (A) Primary			Panel (B) Secondary and tertiary
	(I) Upward mobility
	(a) Biological or step-parents
IM (Mean)	9,258,374	0.508	0.500	12,447,352	0.204	0.403
IM (Min)	10,727,953	0.550	0.497	13,228,050	0.221	0.415
IM (Max)	9,258,374	0.508	0.500	12,447,352	0.204	0.403
(b) Immediate older generation
IM (Mean)	9,280,274	0.507	0.500	12,462,921	0.204	0.403
IM (Min)	10,811,989	0.552	0.497	13,275,780	0.222	0.415
IM (Max)	9,132,978	0.505	0.500	12,370,537	0.204	0.403
(II) Downward mobility
(a) Biological or step-parents
IM (Mean)	4,994,345	0.099	0.299	1,805,367	0.438	0.496
IM (Min)	3,524,766	0.064	0.245	1,024,669	0.384	0.486
IM (Max)	4,994,345	0.099	0.299	1,805,367	0.438	0.496
(b) Immediate older generation
IM (Mean)	4,983,648	0.097	0.296	1,801,001	0.435	0.496
IM (Min)	3,451,933	0.062	0.241	988,142	0.375	0.484
IM (Max)	5,130,944	0.104	0.305	1,893,385	0.449	0.497

Class of mineral categories	Composition
Gold	Gold
Bulk	Bauxite, Coal, Iron ore, Phosphate, Potash
Precious	Diamond, Emerald, PGE, Platinum,
Base Metal	Copper, Lead, Nickel, Z
Other	Andalusit, Chromium, Cobalt, Flourine, Graphite, Lithium, Manganese, Niobium, Rare earth, Tantalum, Tanzanite, Tin, Tungsten, Vanadium
Mineral Sands	Mineral sands, Zircon
Uraniu	Uranium

Topics

Countries

Series

About

Resources

Abstract

I. Introduction

II. Review of literature

A. General effects of natural resources

B. Natural resources and education

C. Intergenerational mobility in education and its determinants

III. Data sources and IM index

A. Data sources

1) IPUMS data

2) Mineral discoveries data

B. Construction of educational IM

C. Cohabitation-selection issues

D. Stylized facts on educational IM in Africa

1) Trends of IM by decade

2) Educational IM at country and district levels

a. Country-level educational IM

b. Country-level educational IM by gender

c. Country-level educational IM by residency

d. Mapping of district-level educational IM across Africa

IV. Stylized facts on educational IM and mineral discoveries and production

V. Empirical methodology

VI. Results

A. Baseline results

1) Educational primary IM

2) Educational secondary and tertiary IM

B. Dynamic effects of mineral discoveries and productions on educational IM

VII. Transmission channels

A. Income channel: parents working in mining sector

B. Returns to education: Wealth index and LIDO score

VIII. Robustness checks

A. Use of alternative control groups (only countries/districts with discoveries)

B. Use of alternative structures of fixed effects and time variable

C. Use of alternative time window around the mineral discoveries or productions

D. Use of alternative time exposition to the mineral discoveries or productions

E. Use of alternative IM definitions

F. Use of all mineral discoveries and productions

G. Use of conflicts as additional control variables

IX. Sensitivity

A. Depending on the African regions

B. Depending on the size of mineral discoveries

C. Depending on the gender

D. Depending on the urban-rural living area

X. Conclusion and Policy implications

References

Annexes

A. Sample

B. Summary statistics of Educational IM

C. Additional stylized facts on mineral discoveries

D. Additional stylized facts on country-level educational IM

E. Additional stylized facts on district-level educational IM

F. Baseline results with control variables

G. Validation of LIDO score and Wealth Index

Same Series

Other IMF Content

Other Publishers

Asian Development Bank

Inter-American Development Bank

The World Bank

Share Link

Table of Contents

Headings

Figures

Metrics

Foresee