- A copy of my research statement is available here.
JOB MARKET PAPER
“The Effect of Subjective Longevity Expectations on Retirement” [view pdf]
Summary : The timing of retirement depends on a number of factors including the individuals’ longevity expectations. As these expectations cannot be directly observed, the extrapolation of past trends in actual mortality is used as a standard measure. However, this approach makes two strong assumptions: (1) the upward trend in life expectancy will continue in the future; (2) the individuals who were born in the same year have the same life expectancy regardless of their characteristics.
In this study, I relax these assumptions and develop a new measure of these expectations using self-reported survival probabilities in the Health and Retirement Study (HRS). Specifically, I estimate cohort-specific survival curves by education, health status, race and gender by following two-steps methodology that combines random effects order probit and non-linear least square estimation methods. The new measure reveals that white Americans born in the 1940s and 1950s have a shorter anticipated life expectancy than their counterparts born in the 1930s and, moreover the gap is bigger for low educated groups. This implies that younger cohorts, instead of delaying retirement, would retire slightly earlier than the cohorts who are already retired, contradicting the expectations of the Social Security Administration which assumes longer life expectancy and therefore later retirement for these younger cohorts.
To show how longevity expectations affect the retirement behaviors of the older Americans, I estimate a dynamic structural life cycle model of retirement, savings and claiming decisions in which future health status, survival and wages are uncertain. Using the Method of Simulated Moments and data for the 1930s birth cohort from the HRS, I estimate this model separately for these two different measures of longevity expectations and find that new measure provides better fit to the observed pattern of retirement than the standard measure. This finding does not only verify the relevancy of the new measure but also reveals that the added heterogeneity in life expectancy captured by the new measure provides important information to explain the retirement decision.
To show the effect of the observed divergence in expected longevity estimates between standard and new measure on the retirement behavior, using the estimated primitives of the model, I simulate the potential retirement behavior of the birth cohorts of 1940s and 1950s who have not fully retired yet. I find that the new measure points a retirement peak around early retirement age, while the standard measure used by Social Security administration predicts a retirement peak around full retirement age. These findings imply that federal programs such as Social Security are less redistributive than they might appear because low-educated individuals obtain their benefits for significantly fewer years than more educated individuals, especially in later birth cohorts.
“A Forward-Looking Age based on Subjective Longevity Expectations”
Abstract: The way in which individuals experience specific transitions in their lives can be influenced by their perception of aging associated with their characteristics. Often, studies that analyze these life-course transitions use chronological age, an inherently backward-looking measure, as a proxy for aging. Implicit to this approach is the notion that all groups regardless of their characteristics move through life-course stages in a chronological lockstep. In this paper, I develop a new age measure which allows us to capture the heterogeneity in the aging process of individuals with different characteristics. To do that, first I estimate remaining subjective life expectancy for different groups using the data on self-reported survival probabilities obtained from the HRS, and translate them into an index, measured in years, what I call a “forward-looking age”. In simple terms, two individuals have the same forward-looking age if they have the same remaining life expectancy. This allows us to make a simple and straightforward comparison of the aging of different groups. I find that there is substantial variation in the forward-looking age of individuals with different characteristics and this variation tends to increase with chronological age. In particular, I observe that education matters for both genders (i.e., individuals with higher education levels are younger in terms of forward looking age), but the magnitude of its effect is larger for women. Surprisingly, I also find that the forward-looking age of the younger cohort is always greater than the forward-looking age of the older cohort at any education level. This is in line with the recent literature that documents increasing mortality in younger cohorts and with my own results in my job market paper which shows a lower expected longevity for younger cohorts.
- Previously circulated as “A Forward-Looking Age based on Longevity Expectations”
“Models of Household Interaction and Their Effects on Retirement” with Miguel Poblete Cazenave
Abstract: Household behavior has been often modeled by relying on two different behavioral assumptions: either individuals act strategically to maximize their own utility (under Nash bargaining) or act cooperatively to maximize household utility (by selecting a point on the household Pareto frontier). Following Del Boca and Flinn (2012) we analyze the differences in retirement outcomes under these standard behavioral assumptions and under two model extensions: a “Constrained Pareto Optimal” equilibrium, where household members agree on a point on the Pareto frontier where both individuals receive a utility as least as high as the one they would obtain under Nash Bargaining, and an “Endogenous Household Interaction” model, where individuals follow a trigger strategy through time in which they cooperate as long as the future value of continuous cooperation is higher than the value of deviating one period and receiving the outcome of the Nash bargaining in the forthcoming periods. We estimate these four different structural models of household retirement using the Method of Simulated Moments and Panel Study of Income Dynamics (PSID) data from 2001 to 2009. We find evidence that the model extensions provide a better fit to the data.
OLDER WORKING PAPERS
“Is it Men or Women who Face a Glass Ceiling in Turkey” with Gökçe Uysal, 2012.
- Previously circulated as “Explanining the Gender Gap in Turkey Using Wage Structure Survey“.
“Childcare Decision and Female Labor Participation: Evidence From Turkey”, 2012. [view pdf]
- Winner of the Ibn Khaldun Prize of Middle East Economic Association for the Best Paper, 2008