Lifetime earnings in the US are highly unequal. This column examines a frequently overlooked factor to explain the persistence of lifetime earnings inequality: long-term differences in hours worked. Even among men who are very attached to the US labour market, lifetime differences in hours worked are substantial and correlate strongly with both lifetime earnings and lifecycle earnings growth. Through the lens of a human capital model, almost 20% of the variance in lifetime earnings for these men can be attributed to disparities in lifetime hours.
Lifetime earnings in the US are highly unequal. For instance, Guvenen et al. (2022) use administrative Social Security records to construct a measure of lifetime earnings by averaging earnings over ages 25 to 55. They find that among men in the US, lifetime earnings at the 75th percentile are 2.7 times larger than those at the 25th percentile. While many factors contribute to these differences, our recent research (Bick et al. 2024b) highlights an often overlooked aspect: labour supply, specifically long-term differences in hours worked and how they interact with human capital accumulation.
Data
Our empirical analysis is based on the National Longitudinal Study of Youth 1979 (NLSY79), a report on individuals born between 1957 and 1964. The dataset provides annual earnings and hours data from age 25 until at least age 55, offering, for the first time, a direct link between lifetime hours worked and lifetime earnings over this critical age range. In a companion paper (Bick et al. 2024a), we show that the lifetime earnings distribution in the NLSY79 closely aligns with the distribution found in administrative Social Security records for the same cohorts by Guvenen et al. (2022). The lifetime earnings distribution in the NLSY79 also matches cross-sectional profiles of earnings and hours worked for these cohorts in the Current Population Survey, the main labour force survey in the US.
Our analysis builds on previous work that abstracts from movements into and out of participation in the labour market. With this in mind, we exclude women from our analysis, as they are much more likely to experience non-participation spells, particularly around childbirth. Among men, we focus on those with high labour market attachment, specifically those who work at least 520 hours per year at every age from 25 to 55.
Facts on lifetime hours
Table 1 documents substantial variation in lifetime hours worked. Men at the 25th percentile of the lifetime hours distribution work an average of 2,155 hours per year, while men at the 75th percentile work 2,588 hours annually – a difference of over 400 hours per year, or 20%. This variation is driven primarily by differences in weekly hours worked rather than the number of weeks worked. Men at the 75th percentile work an average of 8.4 more hours per week in each year between 25 and 55 than those at the 25th percentile.
Table 1 The distribution of lifetime hours and its components
Notes: Individuals are sorted into percentiles according to their annualised lifetime hours. ‘Weeks per year worked’ and ‘hours per week worked’ are the average values for all individuals in a given percentile and the two adjacent percentiles. Note that the product of the two variables is slightly different than the average annualised lifetime hours for each percentile.
These large lifetime gaps in hours arise from both persistent life-cycle patterns and substantial year-to-year variation. In fact, short-term variation in hours worked is even greater than lifetime variation. For example, the variance of log hours worked annually is three times that of log lifetime hours. This suggests that differences in hours worked are influenced by both transitory and more permanent factors.
The impact of lifetime hours on earnings
Figure 1 shows that a worker’s lifetime hours are strongly correlated with their lifetime earnings. Panel (a) indicates that, unsurprisingly, men who work more also earn more. Perhaps more interestingly, earnings increase more than one-for-one with hours. For instance, men at the 75th percentile of lifetime hours work 20% more hours than men at the 25th percentile, yet their earnings are 35% higher. One explanation for this pattern is that men who work more also experience faster earnings growth over their lifecycle, as shown in panel (b). These findings suggest that working longer hours may have both direct and dynamic returns: increasing earnings in the current year and in future years.
Figure 1 Lifetime hours and earnings
Notes: The 3000 hours bin includes anyone with 3000 or more annualised lifetime hours.
Modelling hours and human capital
While these reduced form results demonstrate a strong correlation between lifetime hours and lifetime earnings, they cannot quantify the causal importance of hours worked. It could simply be that individuals who are more productive are also inclined to work longer hours, leading to higher earnings for reasons unrelated to hours worked.
To address this, we develop a model of human capital investment based on the Ben-Porath framework. The model incorporates differences in individual preferences for work and differences in human capital endowments, initial human capital and the ability to accumulate skills over time. For a simplified version, one can show in closed form that individuals who expect to work longer hours have a stronger incentive to invest in human capital, which in turn increases future earnings. Thus, heterogeneity in expected future hours leads to heterogeneity in human capital accumulation, and subsequently in lifetime earnings.
Our quantitative model allows for labour market risks, an important source of earnings inequality (e.g. Huggett et al. 2011, Guvenen et al. 2014), as well as permanent and transitory differences in the disutility of work. These features are necessary to replicate the lifecycle profiles of earnings and hours as well as their persistence over time. The transitory component of preference heterogeneity generates substantial cross-sectional variation in annual hours, while the permanent component is the dominant driver of lifetime hours. Interestingly, we find only a weak correlation between work preferences and initial human capital, with individuals who dislike working longer hours being only slightly more likely to have lower initial human capital (correlation = −0.15).
Lifetime hours: A new perspective on earnings inequality
Our calibrated model implies that about 20% of the variance in lifetime earnings is explained by differences in lifetime hours worked. Between one-third and one-half of this effect is due to human capital accumulation. Preference heterogeneity plays a key role: 90% of the impact of hours heterogeneity on lifetime earnings inequality is driven by differences in preferences for work. Moreover, the model demonstrates that the effect of hours heterogeneity on lifetime earnings inequality is driven largely by differences in lifetime hours, rather than annual hours. In other words, while annual hours vary significantly across individuals, it is the persistent differences in lifetime hours that have the most significant impact on lifetime earnings.
Policy implications
Our findings carry important policy implications. Since preference heterogeneity plays a significant role, we find that differences in initial human capital and learning ability matter less for lifetime earnings inequality than previously thought. As a result, public policies aimed at reducing disparities in human capital before labour market entry may have a more limited effect on reducing lifetime earnings inequality.
We use our calibrated model to assess the impact of restrictions on the length of the work week, a policy measure featured prominently in Europe. These policies are typically adopted with the goal of increasing employment (e.g. Garnero et al. 2022), but they may also affect the distribution of lifetime hours among existing workers and reduce their incentives to invest in human capital. To illustrate this, we simulate the impact of a regulation like that adopted in France in the early 2000s, which caps weekly working hours at 48. If implemented in the US, our model suggests that such a regulation would reduce both average earnings and inequality in lifetime earnings. However, the impact would not be evenly distributed. Workers who are willing to work longer hours – and who benefit most from additional human capital accumulation – would be disproportionately affected. These workers are found throughout the earnings distribution, not just at the top, reflecting the weak correlation between the disutility of work and human capital endowments.
Beyond highly attached workers
While our analysis focuses on highly attached men, the relationship between lifetime hours and earnings may be even more pronounced for workers with less consistent labour market attachment. In Bick et al. (2024b), we show that broadening the sample to include men who worked at least 520 hours in 20 of the 31 years between ages 25 and 55 strengthens the relationship between lifetime hours and both lifetime earnings and earnings growth. This suggests that our findings may represent a lower bound of the overall impact of lifetime hours on earnings inequality.
Source : VOXeu
