Research

Abstract: In defined-benefit pension systems such as U.S. Social Security, retirement benefits depend on a history-dependent transformation of past earnings: the system condenses a worker’s entire earnings trajectory into the Average Indexed Monthly Earnings (AIME), typically by averaging the top 35 years of earnings. This earnings-summarization rule embeds both redistribution and behavioral incentives, yet its quantitative effects are difficult to characterize because AIME is a highly nonlinear, non-smooth function of lifetime earnings histories. To address this challenge, I develop a structural life-cycle model of labor supply, retirement, and earnings risk, solved using a novel deep neural network architecture that directly learns the mapping from full earnings histories to retirement benefits.  Using the estimated model, I evaluate alternative benefit rules that vary the number of counted earnings years—top 20, top 5, and lifetime-average earnings. Reducing the number of counted years modestly raises consumption and redistributes gains toward workers with volatile or intermittent earnings, while lifetime averaging reverses these patterns, sharply increasing labor supply and retirement ages and amplifying inequality by exposing benefits to low-earning years. These results show that the structure of the earnings-summarization formula plays a central role in shaping redistribution, insurance, and work incentives in public pension systems.

Abstract: What is the optimal design of the tax and transfer system for families who face labor market shocks and have children over their life cycle? In this paper, I develop and estimate a rich microeconometric life-cycle model and solve for the optimal child-dependent tax function. The model features endogenous family labor supply, consumption-saving, and the development of children’s cognitive ability in the family. First, I utilize the estimated model to evaluate the long-run effects of making the expanded child tax credits of the American Rescue Plan permanent. It has a modest negative effect on labor supply and increases children’s test scores at age 18 by 2 percent of a standard deviation. The ex-ante welfare gain from this policy is 14 percent higher than an unconditional cash transfer that costs the same. Next, I solve for the optimal child-dependent taxes using flexible linear spline functions. The optimal tax policy of families with one or two children consists of a large guaranteed income and earning subsidies with a high-tax phase-out region along with child tax credits at the top of the income distribution. For families without children, the optimal policy is an EITC-type transfer to low-income families with a drastic increase in marginal taxes of middle and high-income families compared to the status quo. The optimal tax policy enhances the social welfare by increasing non-working time of mothers, lowering consumption inequality, increasing consumption smoothing, and improving children’s abilities.

Abstract: This paper investigates the determinants of students’ choice of location for higher education using two complementary surveys of university entrants in Iran. Higher university quality and the opportunity to study in their preferred major are the most motivating factors to study in another city. The main discouraging factors are high living and travel costs and limited dormitory availability. Female students express lower willingness to migrate for education, reinforcing existing gender gaps in access to higher-quality universities. They also report receiving less parental support to do so, and they perceive economic costs as the primary reason for their parents’ opposition—consistent with weaker incentives to invest in daughters’ human capital. However, we find no evidence that girls attribute their parents’ disapproval to social control or conservative gender attitudes. In evaluating commuting and city characteristics for the location of higher education, safety is especially important to female students. Finally, a rank-ordered logit model based on hypothetical university choice scenarios is estimated to understand the willingness to relocate based on city attributes. The results show high willingness to pay for the opportunity to study in preferred majors, live in university-provided dorms, and enjoy large metropolitan amenities when deciding the location of higher education.

Abstract:  We examine the causal impact of natural disasters on access to higher education and develop a structural equality-of-opportunity framework for designing fair admission policies for affected regions. First, using district-level university entrance exam data linked with detailed records of earthquakes and floods in Iran, we estimate the causal impact of natural disasters on student performance through a stacked difference-in-differences design. We find that major disasters significantly reduce average exam scores—by about 2 % in the year following natural disasters—with larger and more persistent effects for stronger events and poorer and more rural regions. Then, motivated by Iran’s existing quota system—which, similar to policies in countries such as Turkey and China, provides limited compensation for affected students—we embed these estimated causal effects to design an optimal seat-allocation policy that restores fairness across regions. The proposed policy grants additional seats to affected regions to compensate for the predicted impact of disasters on students’ exam performance. Utilizing an estimated model of the university admission system in Iran taking into account the equilibrium effects, we show that, under a welfare-based equality-of-opportunity criterion, the optimal compensatory policy corre- sponds to around 8 % seat allocation for affected areas, effectively restoring parity without excessive overcompensation.

Abstract: This paper develops a structural life-cycle model to study how redistributive policies affect human capital formation through two competing channels: a resource channel that relaxes liquidity constraints and a dynamic discouragement channel that weakens incentives for effort when future support is anticipated. The model traces family decisions from a child’s birth through adolescence and into adulthood, incorporating parental time, educational expenditures, and the child’s endogenous study effort under evolving agency. Heterogeneous beliefs about college returns, calibrated to empirical evidence, mediate how families respond to redistributive transfers. Quantitative simulations reveal that the timing and perceived permanence of transfers critically shape human capital outcomes. When transfers target childhood and adolescence, average human capital rises by about 2.7 percent relative to the baseline, reflecting liquidity relief. When redistribution expands only in adulthood, human capital falls by 0.2 percent, consistent with dynamic discouragement. Combined policies yield intermediate gains of 1.4 percent, as incentive and resource effects partially offset each other. Under subjective expectations, redistribution produces greater heterogeneity—college enrollment declines and inequality in educational trajectories widens—even as average outcomes improve slightly. These findings highlight that redistribution is not neutral with respect to human capital: its long-run effects depend on when support is delivered and how households form expectations about educational returns.

Abstract: The rapid diffusion of Artificial Intelligence (AI) has introduced profound uncertainty about the future of work, yet little is known about how such expectations affect students’ educational decisions. This paper provides novel evidence using original survey data collected from Iranian high-school students during the national university application period. The survey elicits perceived effects of AI on future job prospects across disciplines, students’ confidence in these beliefs, and their willingness to pay (WTP) for perfect information about AI’s true labor-market impact. Over 70 percent of respondents expect AI to have a large or very large effect on graduate employment. Students consistently view computer-related fields as the main beneficiaries and physical sciences as the most negatively affected. Despite heterogeneity in views, students display high confidence and limited demand for information: the median WTP is only 10 percent of expected lifetime income, and few are willing to pay more than 25 percent. 

Abstract: Women in Iran now outnumber men in universities but remain largely absent from the labor market. This paper examines how such a paradox takes shape by studying expectations about women’s future labor supply among students at the time of university entry. I design and field an original survey that elicits subjective probabilities of employment after graduation, after marriage, and after childbirth, as well as attitudes toward family roles and marriage bargaining power. Female respondents express very high expectations of future work: they report on average a 81 percent probability of employment after graduation and 76 percent after childbirth—far above the actual labor-force participation of women in Iran. Male respondents expect substantially lower employment probabilities for their future wives, averaging 54 percent after marriage and 42 percent after childbirth. These male expectations are closer to realized outcomes, even though men and women report similarly gender-eqaul attitudes toward gender roles. The findings reveal a sharp expectation gap: women overestimate both their long-term attachment to the labor market and their control over labor-supply decisions within marriage, while men underestimate women’s aspirations but hold more realistic beliefs about social and institutional constraints. By documenting these differences before students enter the labor market, the paper shows that the “high-education, low-work” pattern of female employment originates not from lack of ambition but from misaligned expectations formed on campus.

Abstract: We study the heterogeneous effects of longer unemployment insurance (UI) duration over the life cycle and work history. Using linked administrative data from Iran that merge the universe of UI spells with tax, banking, and demographic records from the Iranian Welfare Database, we trace pre-unemployment contribution histories and post-exit outcomes for more than 400,000 recipients. Eligibility for UI benefits in Iran features several discontinuities in maximum duration based on months of prior contributions. These thresholds create multiple quasi-experimental settings that we exploit using regression-discontinuity design. Our analysis examines how extended UI benefits affect post-unemployment earnings, re-employment, and consumption across age and experience groups. Younger workers, with lower assets and higher career incentives, are expected to display stronger consumption-smoothing responses and weaker moral hazard, while older or long-tenure workers may reduce job search or delay re-entry. We further document bunching in contribution histories around eligibility thresholds, providing evidence on behavioral responses to UI design. By combining rich administrative linkages with multiple natural experiments, this study provides new evidence on how the insurance and incentive effects of UI vary over the life cycle, offering guidance for age- and career-contingent optimal UI policy design.