How labour migration raises human capital in sending communities over the long run

Taryn Dinkelman, Martine Mariotti

20 July 2016



Circular migration across national borders is a long-standing feature of labour markets in many developing countries, particularly in sub-Saharan Africa where close to 70% of all international migration from the region is between African countries (World Bank 2013). Yet we know little about how migration affects sending communities, and what we do know is limited to short run effects. In one recent study, Gibson and McKenzie (2014) analysed the short-run development effects of New Zealand’s seasonal worker programmes and found positive impacts on income, consumption, savings, asset ownership, and subjective standards of living for migrant households in sending communities. Could similar types of legal, time-limited migration improve economies in sending areas over the longer run? Not just for migrants, but for entire communities? Finding answers to these questions is not easy, partly because migration is a choice variable, and partly because of a general lack of migration data (Constant and Zimmerman 2013).

In new work, we investigate whether schooling investments in sending communities responded to fluctuations in legal migration flows between Malawi and South Africa (Dinkelman and Mariotti 2016a, 2016b). During the 20th century, most southern African countries were part of an organised system of contract migration to supply workers to mines in South Africa. Many institutional features of this historical migration resemble current-day contract migration to the Middle East and seasonal worker programs in the south Pacific. We analysed unexpected shocks to the option to migrate from Malawi that occurred in the 1960s and 1970s. We wanted to quantify how circular labour migration affected human capital profiles of the entire community – not just migrant households – decades after this migration ended.

Why might labour migration raise or lower investments in education?

Theory suggests that migration could increase or reduce investments in human capital. For example, remittances could increase school attendance through an income effect. However, migration might undermine schooling in rural communities if farming households use child labour to substitute for missing migrant workers. If children do not return to school after migrants come home, this could permanently reduce educational attainment. Malawi is a good context to look for both income and substitution effects. Workers were contractually obliged to send most of their earnings home – so sending communities received large income transfers from migrants; and children routinely worked on farms, making it technologically possible for children to substitute for missing migrants.

Migration shocks in Malawi

To study the net impact of labour migration on education, we had to address the two main issues that always arise in migration research:

  1. How to isolate the causal effects of migration; and
  2. How to find good data on migration, remittances, and relevant economic outcomes.

To address the first issue, we studied a unique event that generated large, relatively exogenous district-level shocks to the number of migrants, and amounts of money flowing back to Malawi. In Figure 1, we see a 200% surge in the number of migrants following the removal of a national labour quota for Malawian workers in 1967.[1] In some communities, over 20% of the working age male population was missing for up to two years at a time, the typical contract duration. In April 1974, a plane of returning miners crashed, killing over 70 workers. President Banda then recalled all miners to Malawi and banned recruiting indefinitely. Levels of migration to South African mines never recovered after the ban was lifted in 1977 because mining companies redirected their hiring policies away from foreign labour.

Figure 1 Annual employment of Malawians in South African mines, 1950-1994

Source: Dinkelman and Mariotti (2016a).

The beginning and end of the migration surge shown in Figure 1 brackets our ‘treatment’ – a concentrated period of shocks to migration and migrant earnings. Districts with pre-existing recruiting stations, established by the mining companies as early as the 1930s, had the greatest exposure to this treatment. Because having a recruiting station in your district made it easier to sign up for a mine job, when the option to migrate was no longer restricted in 1967, more men responded to these changes in districts with stations.

To address the lack of data that is common in migration work, we assembled several waves of Census data – some newly digitised – from 1931 to 1998 and matched this with archival data we found on district-specific flows of migrant money, and recruiting station locations. Because historical placement of these stations was not related to district characteristics that might have influenced education directly, we compared educational attainment among cohorts eligible for primary school during the migration shocks across districts with and without recruiting stations. We also constructed a second comparison of education gaps between districts with and without stations among older cohorts who would not have been eligible for primary school during the shocks.

Migration raised education of the next generation of workers

Figure 2 shows the result of comparing these two sets of education gaps. The figure plots the estimated difference in total years of schooling between recruiting station districts and districts without stations, for sets of five-year age-groups in 1998. Among older cohorts (those aged 45 to 59) who were age-ineligible for school at the time of the shocks, there were no differences in educational attainment between districts with high and low exposure to the migration shocks. However, age-eligible cohorts in recruiting station districts (ages 25 to 44 in 1998) gained between 0.12 and 0.18 more years of education, relative to cohorts of the same ages in districts without recruiting stations. This education gap is smaller (0.14 years) but still positive for the youngest cohorts (ages 20 to 24 in 1998). These are relatively large effects (4.6-6.9% gain), given that the average adult in our sample had 2.5 years of completed education.

Figure 2 Estimated differences in education by cohort and recruiting station status of district

Source: Dinkelman and Mariotti (2016a). The figure shows estimated interaction term coefficients (solid line) and 95% confidence intervals (dotted lines) from a regression of education on nine age group dummies and their interaction with the number of recruiting (WENELA) stations in the district. Controls include a female indicator, log population density in 1931, literacy rate in 1945, district and region fixed effects, and trend terms interacted with region fixed effects, baseline population density, and baseline literacy rates. The x-axis shows five-year age cohorts in 1998.

Our estimates of the long-run elasticity of education with respect to migration are somewhat smaller than existing positive estimates of the shorter-run effects of migration and remittances estimated in Asian and Latin American settings. In addition to the overall positive impacts, we found that the migration shock had smaller positive impacts on education during the labour expansion period, when migrants were still leaving Malawi each year and larger effects during the labour ban years, when all migrants had returned and collected their earnings. We also found that the long-run impacts were largest in districts without agricultural estates, a proxy for districts where child labour was least valuable, and where the opportunity costs of going to school were lowest. Importantly, because education was not valuable in mine work, the human capital improvements we found were not the result of children getting more education in order to find a mining job. Children in the next generation –both males and females – attained more education in the wake of the migration shocks, largely because of positive income effects from migrant earnings.


Instead of crowding out school enrolment, we found that international labour migration within Africa raised human capital profiles of future workers in sending areas. Our results are relevant for any country with an existing or historical bilateral guest worker program including, for example, the Gastarbeiterprogramm between Germany and Turkey, as well as all African countries affected by labour demand from South Africa’s mining industry. Modern programmes have many similarities with the organised mine migration – for example, limited-time work contracts, in-built circular migration flows and periodic labour bans. Our work shows that there may be scope for modern guest worker programmes to have positive, long-lasting impacts on human capital in sending regions, contributing to the development impacts of international labour migration.


Constant, A F and K F Zimmerman (2013) “The economics of circular migration”, in International Handbook of Migration, A Constant and K Zimmerman (eds), Edward Elgar Publishers, Ch 3: 55-74.

Dinkelman, T and M Mariotti (2016a) “The long run effects of labor migration on human capital accumulation in sending communities”, NBER Working Paper No 22049.

Dinkelman, T and M Mariotti (2016b) “The long run effects of labor migration on human capital accumulation in sending communities”, American Economic Journal: Applied Economics, Forthcoming (October).

Gibson, J and D McKenzie (2014) “The development impact of a best practice seasonal worker policy”, Review of Economics and Statistics, 96(2): 229-243.

The World Bank (2013) “Bilateral migration matrix 2013”.


[1] International labour recruitment for mine work was highly centralised and bureaucratic. The mining industry in South Africa recruited workers through the Witwatersrand Native Labour Association (or Wenela) which became The Employment Bureau of Africa (TEBA). 



Topics:  Labour markets Migration

Tags:  circular migration, guest workers, seasonal work, brain drain, brain gain, Malawi, South Africa, years of schooling, Africa, migration shocks, educational investment

Assistant Professor in the Economics Department, Dartmouth College; CEPR Research Affiliate

Senior Lecturer, Research School of Economics, Australian National University; Research Fellow, Stellenbosch University