Conference Agenda

Highly educated individuals live longer, on average. Using an overlapping-generations model calibrated for Austria, I investigate the influence of skill-dependent mortality on public pension designs with an aging population. I focus on pension reforms which seek financial balance and avoid redistribution, such as notional defined contribution systems. Simulations show that ignoring the skill-dependence in mortality in standard notional defined contribution systems, where pension benefits adjust automatically, has negligible macroeconomic and welfare impacts. However, ignoring skill-dependence in mortality in the design of skill-dependent retirement age reforms leads to pro-skill biases. In realistic cases, high-skilled households could gain the equivalent of 9.5% of lifetime consumption and other households lose 4.0% of lifetime consumption. Reforms of special pension regimes or policy actions aiming at reducing early retirement, which implicitly depend on skill, should thus be designed with skill-dependent mortality in mind.

This paper studies the efficiency cost of pay-as-you-go pension systems that finance legacy transfers through implicit taxation embedded in pension contributions. The return on contributions is typically average income growth rather than the market return on capital. Because the gap between these returns compounds over the life cycle, early-career contributions have a higher implicit tax rate. I develop a sufficient-statistics framework in which the deadweight loss depends on the age profile of implicit tax rates. Deadweight loss is minimised when the implicit tax rate is constant, mirroring the logic of tax smoothing. I then propose an implementation within a notional defined contribution design: contributions are credited at a higher market-based rate of return, while a lower fraction of contributions is credited to notional accounts and the remainder treated as a pure tax. Finally, I quantify cross-country differences in pension formula parameters that drive life-cycle variation in implicit tax rates.

In defined-benefit pension systems such as U.S. Social Security, retirement benefits depend on a history-dependent transformation of past earnings: lifetime earnings are condensed into Average Indexed Monthly Earnings (AIME), typically averaging the top 35 years. This rule embeds redistribution and work incentives, yet its effects are hard to quantify because AIME is a nonlinear, non-smooth function of earnings histories. I develop a structural life-cycle model of labor supply, retirement, and earnings risk, solved with a novel deep neural network that learns the mapping from full earnings histories to benefits. Evaluating alternative rules—top 20, top 5, and lifetime averaging—I find that counting fewer years modestly raises consumption and redistributes toward workers with volatile earnings, while lifetime averaging increases labor supply and retirement ages and amplifies inequality. The earnings-summarization formula is central to redistribution, insurance, and incentives.

This paper uses a stylized overlapping-generations model to examine the effect of borrowing constraints on the economic implications of Social Security. In this framework, Social Security provides partial insurance against income risk that is uninsured due to incomplete markets. I find that when borrowing consistent with life cycle behavior is allowed in this framework, the micro- and macroeconomic effects of a downsizing in Social Security are considerably smaller than when borrowing is prohibited. I also find that the key mechanism behind this result is labor supply: with endogenous borrowing, households are able to exploit increasing labor productivity in early life to better self-insure against income risk.