Declining Hours Worked Among Entrepreneurs

In this paper, I show that over the past 35 years, hours worked by entrepreneurs have fallen substantially: by five hours per week more than for workers. This decline accounts for the bulk of the fall in total hours worked and is present in all available subgroups (gender, age, education, number of children, occupation, and industry). It is robust to adjusting for compositional effects and occurs without noticeable changes in the relative hourly income of entrepreneurs. The decline originates from the top of the hours distribution: the share of entrepreneurs working more than 45 hours has dropped significantly. I interpret these facts using a Roy model of occupational choice, augmented with an intensive labor supply margin. The model allows the marginal return of working an additional hour to depend on the level of hours. I estimate the model at two points in time and find that a fall in the relative marginal return at higher hours worked is key for explaining the drop in hours and the drop in the share of entrepreneurs. I show that changes in the market structure of the goods or services that entrepreneurs sell can account for this.

 A Nonhomothetic Price Index and Cost-of-Living Inequality

We derive a nonhomothetic generalization of all superlative price indices and document cost-of-living inequality in the United States. When necessities and luxuries are separable in the expenditure function, this generalization eliminates the need to estimate a complete demand model. Using CEX-CPI data from 1995 to 2020, we find no differences in average inflation rates across the expenditure distribution, but 2.5 times higher inflation volatility for the bottom decile than the top decile, stemming from a larger exposure to food, gasoline, and utilities. Our analysis challenges inequality measurements using group-specific homothetic price indices and suggests an income-effect bias in these estimations.

A Distributional PCE Price Index From Aggregate Data

This paper proposes a method to measure individual and aggregate changes in the cost of living when consumer behavior is nonhomothetic and microdata on consumption expenditures are not available. Aggregate prices and expenditure shares together with a single cross-sectional distribution of expenditures are sufficient to create a distribution of nonhomothetic cost-of-living indices with this approach. The cost-of-living indices derive from PIGL preferences, generalize the Törnqvist price index, and only contain one unknown parameter. Because PIGL preferences aggregate consistently, this parameter can be identified from aggregate data. Using US Personal Consumption Expenditure (PCE) data, we apply the method to obtain a nonhomothetic PCE price index covering 71 product groups for the period 1959 to 2023. This index reveals a 0.39 percentage point gap in average annual inflation rates between the poorest and richest ten percent since 1959 and a 1.9 percentage point gap throughout 2022, thus suggesting that poorer households are hit harder both in the long run and in the recent inflation surge.