We study the spending profile of workers who experience both a positive transitory income shock (lump-sum severance pay) and a negative permanent income shock (layoff). Using de-identified expenditure and employment data from Brazil, we show that workers increase spending at layoff by 35 percent despite experiencing a 14 percent long-term loss. We find high sensitivity of spending to cash-on-hand across consumption categories and for several sources of variation, including predictable income drops. A model with present-biased workers can rationalize our findings, and highlights the importance of the timing of benefit disbursement for the consumption-smoothing gains of job displacement insurance policies.

This paper provides stark evidence of hysteresis—the failure of an effect to reverse itself as its underlying cause is reversed—in energy demand. We estimate that half of the 23% reduction in residential electricity use caused by a 9-month-long policy that was imposed on millions of Brazilians has persisted for at least 12 years. We examine the implications of our finding by extending the traditional welfare analysis of corrective policies to allow for hysteresis. Our estimate highlights that failing to take hysteresis into account could severely bias the welfare evaluation of policies aimed at reducing (long-run) energy demand.

We measure the effects of firm policies on racial pay differences in Brazil. Non-Whites are less likely to be hired by high-wage firms, explaining about 20 percent of the racial wage gap for both genders. Firm-specific pay premiums for non-Whites are also compressed relative to Whites, contributing another 5 percent for that gap. A counterfactual analysis reveals that about two-thirds of the underrepresentation of non-Whites at higher-wage firms is explained by race-neutral skill-based sorting. Non-skill-based sorting and differential wage setting are largest for college-educated workers, suggesting that the allocative costs of discriminatory hiring and pay policies may be relatively large in Brazil.