The Gendered Impacts of Income Fluctuations on Household Departure, Labor Supply, and Human Capital Decisions: Evidence from Kyrgyzstan

Abstract

How do fluctuations in income affect labor supply decisions, and how do their effects differ by gender? This study analyzes data from a thirteen-year rolling panel in Kyrgyzstan spanning 2004–16. It addresses the endogeneity of fluctuations in income to labor supply decisions by employing shift share instruments that exploit region-level changes over time in growth rates of different sources of revenue and production costs. Estimating a household fixed effects model, the study finds that reductions in income relative to the median spur departure from the household (for example, due to migration), with smaller impacts on women than men. However, women’s labor supply at the origin is affected significantly more, with short-term increases in hours of employment and declines in home production and other activities. Reductions in income also fuel temporary migration for both genders, with larger effects for men, and widen the gender gap in pursuit of non-compulsory education.

HIGHLIGHTS

• Declines in income spur household departure, with larger impacts on men than women.

• Women are not always “left behind” following shocks; like men, they respond through changes in labor supply and livelihood decisions.

• At the origin, women face significantly greater increases in workloads than do men.

• Declines in income widen the gender gap in pursuit of non-compulsory education, favoring men.

• Policies that support women’s ability to control income can promote domestic work sharing and ensure income generation empowers women.

Keywords:

• Income fluctuations
• employment
• human capital
• temporary migration
• gender
• Central Asia

JEL Codes:

• J60
• O15
• J16

ACKNOWLEDGMENTS

This research was part-funded by the CGIAR Research Program on Policies, Institutions, and Markets led by IFPRI and IFPRI's Collaborative Research and Capacity Strengthening Program for Enhancing Agricultural Productivity and Food and Nutrition Security in Central Asia. We are grateful to Kamiljon Akramov, Kate Ambler, Alan de Brauw, Brian Dillon, Cheryl Doss, Bilge Erten, Sylvan Herskowitz, Melissa Hidrobo, Jessica Leight, Roman Mogilevskii, Valerie Mueller, Agnes Quisumbing, and Kanat Tilekeyev for comments and helpful discussions.

SUPPLEMENTAL DATA

Supplemental data for this article can be accessed online at https://doi.org/10.1080/13545701.2022.2101680. Replication files can be found at https://doi.org/10.7910/DVN/IYTPXL.

Correction Statement

This article has been republished with minor change. These changes do not impact any of the empirical analyses or any other academic content of the article.

Notes

1 See detailed description in Damir Esenaliev, Antje Kroeger, and Susan Steiner (2011).

2 KIHS data do not provide information on the district or community of residence of households, or GPS coordinates; the only geographic identifier is the oblast, and an indicator for a household residing in a rural or urban area. As a result, data such as community-level migration rates are neither available nor can they be constructed.

3 Incomplete income data from 2003 led us to drop that year. Additionally, we drop household-years with three or fewer months of data, and up-weight observations from household-years with 4–11 months.

4 The Kyrgyz. som–USD exchange rate was 45.95 on June 1, 2010 (National Bank of Kyrgyzstan 2019).

5 We do not estimate models controlling for past household departure.

6 Results are similar when we do not allow an individual to re-enter the sample after exiting – and thus only consider the first departure (Online Appendix Table A3); one in ten returned home.

7 Household fixed effects capture, for example, household ethnic composition as well as the (pre-shock) quality and characteristics of household migrant networks.

8 For household level outcomes, we omit household fixed effects and individual level controls.

9 Following John Bellows and Edward Miguel (2009), we estimated that selection on unobservables would have to be 2.31 (6.14) times greater than selection on observables for women (men) to explain away the entire effect of the income shock. This is computed by comparing columns 1 and 2 in Panel A of Table 3, which yields −0.030 / (−0.043 + 0.030) = 2.31 for women, and −0.043 / (−0.050 + 0.043) = 6.14 for men.

10 While our analysis allows multiple potential departures by the same individual following a return to the household, similar results hold when restricting attention to the first departure (see Online Appendix Table A3).

Additional information

Funding

This work was supported by the CGIAR Research Program on Policies, Institutions, and Markets led by IFPRI and IFPRI’s Collaborative Research and Capacity Strengthening Program for Enhancing Agricultural Productivity and Food and Nutrition Security in Central Asia.

Notes on contributors

Katrina Kosec

Katrina Kosec is Senior Research Fellow at the International Food Policy Research Institute and Lecturer at Johns Hopkins University. Her research focuses on the linkages between governance, public investment, gender, and economic shocks.

Jie Song

Jie Song is a PhD student at the University of California, Berkeley. Her research interests include development, behavioral economics, and political economy.

Hongdi Zhao

Hongdi Zhao is a PhD student in applied economics and management at Cornell University. Her research interest is the intersection of development economics and machine learning.

Brian Holtemeyer

Brian Holtemeyer is Research Analyst at the International Food Policy Research Institute. His research interests include data science, cluster computing, econometrics, and computable general equilibrium modeling.