Working Papers

The Economic Costs of Compulsory Military Service 
Joint with Mohammad Hoseini     [ Manuscript ]  

Abstract. Compulsory military service remains widespread in many developing countries, yet its long-term consequences for human capital and earnings are not well understood. We develop a dynamic discrete choice model of schooling and career decisions to study the effect of conscription in Iran. The model replicates observed life-cycle patterns in the data and is validated using quasi-experimental variation from a temporary exemption reform. Conscription substantially reduces annual earnings in early adulthood, but the penalty narrows over the life-cycle. Counterfactual experiments suggest that the forgone labor market experience and low pay during service are the main mechanisms. Although compulsory service raises educational attainment, it does not translate into higher earnings, as those induced to study longer enter occupations where returns to education are limited.
We use GPU parallel programming in Julia to solve the dynamic discrete choice model in the paper. In case you're interested, older version of the code is available [here]

Intergenerational Consequences of Parental Employment Under Childcare Scarcity
Joint with Arash Nekoei and Josef Sigurdsson [New draft coming soon.]

Estimating Inequality with Missing Top Incomes: A Nonparametric Framework Integrating Auxiliary Data
Joint with Zahra Shamlou, Mohammad Hoseini, Djavad Salehi-Isfahani   [Manuscript]

Abstract. We propose a framework for estimating the income distribution when household surveys miss top incomes using auxiliary datasets that allow for nonparametric and machine-learning-based prediction of top incomes. In particular, traditional parametric approaches such as Pareto-tail fitting or the direct use of tax registries appear as special cases of this framework. Identification hinges on a coverage condition requiring that the auxiliary data include the households in the upper tail of the income distribution. We apply the method to estimating income inequality in Iran, combining household surveys with newly available administrative data on real estate transactions. Our estimates reveal that inequality is substantially higher than previously reported: the income share of the top one percent rises from about five to nine percent in some years, and the Gini coefficient increases by roughly five points. Moreover, we show that the coverage assumption fails for tax registries because high-income individuals are often absent from the tax base, a limitation that likely extends to many countries without comprehensive or effectively enforced tax systems.