The labor force participation of older Americans has been increasing since the 1990s. The tasks and characteristics of American work have also been changing in the past few decades, moving away from the routine and physical and towards the social and cognitive. If these shifts in the nature of work make working less unpleasant, then they may have contributed to the observed old-age labor supply increase. I measure the contribution of the changing nature of American work to the increase in older labor force participation, its impacts on the distribution of welfare at older ages, and implications for Social Security policy. Using the relationship in the Health and Retirement Study between occupation in one’s early 50s and later labor force participation, I find that 10–16% of the increase from 1990 to 2010 in the labor force participation of 60-to-69-year-old men can be explained by changes in occupation characteristics, while this amount is 5.8–9% for women. Exploiting differential changes in occupation characteristics across commuting zones and using the commuting zone’s predicted routineness in 1950 as an instrument, I confirm there is a causal relationship between occupation characteristics and old-age labor force participation. Estimating a structural model of male old-age labor supply with occupation differences, I find that the observed shifts in occupation characteristics led to welfare increases at all but the bottom quartile of lifetime income. Finally, I compare a policy that increases the Full Retirement Age to one that achieves similar savings but concentrates benefit reductions among higher earners. The former policy leads to large participation increases among men in the most physical occupations, while the latter does not. In my model, raising the retirement age is especially harmful for low-income workers because it induces those in the most physical jobs to continue working.

It has long been argued that efficient policies tend to provide larger legal entitlements to the rich than to the poor. This article shows how efficient legal rules can become even more skewed against the poor over time by sowing the seeds of their own vicious cycles. Repeated application over time of these rules can lead to increasingly adverse outcomes for the poor, which the article calls “policy snowballing." Consider a set of polluters choosing between locating in places with rich versus poor people and facing a strict liability rule for harm to earnings. Polluters will disproportionately locate in the poor area, where they face lower damages. That disproportionate share of polluters in the poor area can make it cheaper to harm the poor in the next period, making subsequent polluters locate yet more disproportionately in poor neighborhoods, driving down the poor’s earnings further. We identify the conditions for snowballing and explore its dynamics. When compensation for the harm is incomplete, policy snowballing can lead to spiraling income inequality. As a result, government transfers to the poor to compensate for the change in legal regime would be inadequate if calculated in a way that ignores the snowballing. The article raises the intriguing prospect that legal rules could generate state dependence in the legal costs of harm, and that efficient policymaking could be a contributing factor to increasing inequality over time.

Information about the income distribution in pre-industrial societies is sparse. We analyze labor income inequality in 18th century Murcia, a city in Mediterranean Spain. The historical income distribution of this region is relatively unknown, despite it having one of the highest urbanization rates in Europe in the pre-industrial era. We first use a census conducted in the 1750s which collected information on income and occupation. We then use this income information to conduct analyses of the income distribution in the 1730s and 1780s using censuses with information about the occupational distribution. We find large changes in the distribution of occupations across the censuses. We show that our results on inequality are sensitive to assumptions regarding household composition and within-occupation distribution of income, but not to the definition of household income.

We study generational change in the role of labor market behavior and marriage in determining the family income that individuals experience over their adult lives. Building on Altonji, Giraldo-Páez, Hynsjö, and Vidangos (2022), we estimate a model of individual earnings, marriage, divorce, fertility, and nonlabor income, where key parameters vary with birth year. For the 1935–44, 1945–62, and 1967–80 birth cohorts, we use the model to measure the dynamic responses of earnings, marital status, and family income to various labor market shocks, education differences, and permanent wage heterogeneity. For each cohort, we also provide gender-specific estimates of the contribution of education, permanent wages, labor market shocks, spouse characteristics, spouse wage shocks, and marital histories to the variance of lifetime family income. For both the dynamic responses and the variance decompositions, we isolate the importance of effects on marriage probabilities and on spouse characteristics (sorting). We find that gender asymmetries are substantially smaller for more recent cohorts. The decline reflects the increase in the labor supply of married women as well as other changes. We also find that own characteristics have become increasingly important in the determination of lifetime family income for women, while variation in spouse characteristics has become less important. The opposite is true for men. Gender differences in the sources of inequality in lifetime family income have narrowed.

We estimate a dynamic model of the family income individuals experience over their adult lives. We use the model to measure the dynamic responses of marital status, earnings, and family income to various labor market shocks, education, and permanent wage heterogeneity. We also provide gender-specific estimates of the contribution of education, permanent wages, labor market shocks, spouse characteristics, spouse wage shocks, and marital histories to the variance of family income by age and over a lifetime. For both the dynamic responses and the variance decompositions, we isolate the importance of effects on marriage probabilities and spouse characteristics (sorting). Marital status has a much larger effect on family income for women than men, while labor market shocks to men are more important than shocks to women. Marital sorting plays a major role in the return to education and permanent wages, especially for women. Marriage probabilities are less important. An individual’s own education and the permanent wage component account for 28.0% and 12.6% of the variation in lifetime family income for women, but 36.2% and 26.4% for men. Marital sorting on education and the wage components substantially increases the family income variance, especially for women. Random variation in marital histories accounts for 25.9% of the variance in lifetime family income for women and 7.5% for men but for only a modest part of the variation in lifetime family income per adult equivalent.