ABSTRACT
This study examines how large-scale, predominantly male emigration affects the education of girls staying in Tajikistan, the poorest post-Soviet state and one of the most remittance-dependent economies in the world. Using data from a three-wave household panel survey conducted in 2007, 2009, and 2011, this study finds that the net effect of migration on girls’ schooling turns from positive to negative with girls’ age. These results lend support to various conceptual channels through which the emigration of household members may affect girls’ education, including the relaxation of budget constraints, a change of the household head, and an increase in household work. At the practical level, the results imply that migration can be detrimental to women’s empowerment and cast doubt on whether emigration is an appropriate long-term development strategy for Tajikistan.
ACKNOWLEDGMENTS
We thank Christiane Timmerman, Jean-Philippe Platteau, and the participants of the UNU-WIDER workshops in Namur and Helsinki and the 14th Biannual EACES Conference in Regensburg for many useful comments and suggestions. Financial support from UNU-WIDER is gratefully acknowledged.
Notes
1 Remittances in Tajikistan were equivalent to 29 percent of its GDP in 2015 (World Bank Citationn.d.).
2 The theoretical model of household decision making, originally formulated by Gary S. Becker (Citation1965), supports this consideration. In the framework of household decision making, the model argues that adult household members decide on the schooling of children to maximize household utility. Typically, girls are taken out of school if their contribution to household chores or farmwork is expected to produce higher benefits than further education.
3 A similar transmission of the workload of mothers onto their daughters has been discovered in the case of maternal illness in Ethiopia (Dinku, Fielding, and Genç Citation2017). While maternal illness reduces the time children spent in play, girls tend to be more engaged in domestic work than boys.
4 In fact, the response rate is 100 percent for all variables, except satisfaction with household finances, where information is not available in 0.46 percent of cases.
5 In designing the dummy variable, we checked whether, in some cases, girls dropped out of school but later returned. We found that 4 percent of all girls had left education at some point in time and later re-entered.
6 The information on receipt of remittances is available for current migrants in all waves and for returned migrants in the 2011 wave. However, it is not available for the returned migrants in the 2007 and 2009 waves. In our analyses, we use the information on receipt of remittances from current migrants only. The dummy variable remittances captures the receipt of remittances from at least one international labor migrant.
7 Note that our reference group – children from nonmigrant households – includes children whose parent died (which could also affect school attendance). Parental death information is only available for the year 2007; it suggests that twelve children (0.80 percent of the sample) had lost their mother prior to the interview and forty-six children (3.55 percent of the sample) had lost their father. Given that parental death information is available only for the first wave of the survey, we cannot control for it in our longitudinal analysis.
8 The age difference of 15 years was chosen because this is the minimum difference between the age of a typical parent and a child. Our results do not change substantially if the age difference is reduced to 10 years or if the cousins/uncles are excluded from this category.
9 Note that we do not include the child’s age as an individual control, as the age effect will be captured by year-fixed effects.
10 There are children living in households that have both current and returned migrants (3 percent) or parent and sibling migrants (2 percent).
11 Although our dependent variable is binary, the fixed-effects OLS estimation (linear probability model) is the only feasible option; the logit and probit models do not easily accommodate fixed effects.
12 Because our surveys contain no information on the decision-making practices of the household head, we cannot test the latter argument empirically and only discuss it at the conceptual level.
13 National Strategy of Educational Development 2012–2020 declares educational enhancement as one of the priority goals of Tajikistan (National Strategy of Educational Development Citation2012).
Additional information
Funding
Notes on contributors
Kseniia Gatskova
Kseniia Gatskova is Postdoctoral Researcher at the Leibniz Institute for East and Southeast European Studies, Regensburg. She holds a PhD in sociology from the University of Konstanz. Her main research interests are in the areas of migration and integration, social inequality, and post-Soviet transformation studies.
Artjoms Ivlevs
Artjoms Ivlevs is Associate Professor of Economics at the Bristol Business School, University of the West of England (UWE Bristol) and Research Fellow at the Institute of Labor Economics (IZA, Bonn). His research interests include migration, subjective well-being, corruption, and education, with a particular focus on the postsocialist economies of Central and Eastern Europe.
Barbara Dietz
Barbara Dietz is an associated researcher at the Leibniz Institute for East and Southeast European Studies, Regensburg and Research Fellow at the Institute of Labor Economics (IZA, Bonn). Her research interests include migration and integration of immigrants, with a particular focus on Central European and post-Soviet societies.