Pronatal Property Rights over Land and Fertility Outcomes: Evidence from a Natural Experiment in Ethiopia

Abstract

An exogenous policy change that ended the ability of rural Ethiopian households to affect the size or security of their land holdings through fertility decisions provides a natural experiment to explore the impact of land tenure institutions on fertility. Use of a difference-in-differences approach that uses aggregated data from censuses before (1994) and after (2007) the reform found large fertility effects, with rural women estimated to have reduced life-time fertility by one child due to the reform. Estimated effects on urban women or employment outcomes are not significantly different from zero and robustness checks show no evidence of spillovers or policy endogeneity.

Keywords:

• Property rights
• fertility
• natural experiment
• Ethiopia

Acknowledgement

We are indebted to Gebeyehu Belay, Solomon Haile, Zerfu Hailu, Seid Nuru, and Dessalegn Rahmato for in-depth discussions on land reforms in Ethiopia. We thank participants for comments at the World Bank’s Annual Research Conference on Land and Poverty and the Brown Bag Seminar in Development Economics at the Goethe University in Frankfurt. We are also grateful for the valuable comments of Alexandra Avdeenko, Albrecht Bohne, Markus Frölich, Dany Jaimovich, Heiner Schumacher, Matthias Schündeln, and Pia Unte. The data and codes (in STATA format) will be made available upon request from the authors. The views presented in this paper are those of the authors and do not represent those of the World Bank or its member countries.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary Material

Supplementary Materials are available for this article which can be accessed via the online version of this journal available at https://doi.org/10.1080/00220388.2021.2013465

Notes

1. Contamination of our control group by reform measures spilling over to non-reform areas would exert a downward bias on estimates obtained and thus not affect our substantive results.

2. The ASFR is the number of live births per 1.000 women in a given 5-year age group for a geographic area and year: i.e. $ASF{R}_a=\left(\frac{Numberofliverbithstowomeninspecifiedagegroup}{Numberofwomeninsameagegroup}\right)\times 1000$, where $a$ denotes 5-year age groups. The TFR is the number of children that would be born per woman in her lifetime if she were to pass through the childbearing years having births according to the current schedule of age-specific fertility rates, i.e. $TFR=5\times \sum \left(ASF{R}_a/1000\right)$.

3. While 83 per cent of our sample distri. cts include a DHS cluster, the fact that most (60 per cent) include only one cluster makes generation of reliable TFR and ASFR estimates difficult.

4. Clustering at the next highest (zone) level produces qualitatively equivalent results.

5. As the administrative data had missing information on three districts, the analysis is missing six cluster-time observations.

6. Expressed in elasticity terms at mean values, the effect ranges from 5 per cent to 12 per cent for the entire sample and from 15 per cent to 7 per cent in the rural sample. Once all control variables are included coefficients lose significance in both samples.

7. As per the 2000 DHS, 82.6 per cent of 15–48 old women reported to know and 10.2 per cent to have ever used modern contraceptive methods.

8. The approach of Altoni et al. (2005) is a special case where observables and unobservables contribute equally to explaining the outcome variable, implying $R_{\mathrm{m}\mathrm{a}\mathrm{x}}=\tilde{R}+\left(\tilde{R}-\dot{R}\right)$ This results in ${\delta }^0=\frac{\tilde{\beta }}{\dot{\beta }-\tilde{\beta }}$. or which $\beta =0$. in. terpreted as the relative degree of selection on unobservables compared to observables that would be needed to explain away the estimated treatment effect. If the difference between $\dot{R}$ nd. $\tilde{R}$ s. greater than 30 per cent, the value of $\delta$. har would result in $\beta =0$ il. l be higher than that of under Oster’s (2019) assumptions.