Mateus Morais

This paper shows that local grocery markets make cities more unequal. I combine the price of every barcode at every formal grocery store in Maceió, Brazil, with a quantitative spatial model in which a neighborhood's grocery stores, their assortments and prices, its rent and the incomes of its residents are determined jointly. Richer neighborhoods pay more for the same barcode, yet their grocery cost of living is lower because they can reach more stores. Giving every household access to every store in the city would be worth 12.3 percent of total consumption to the poorest quartile of households and 6.8 percent to the richest, reducing the ratio of the richest quartile's real consumption to the poorest's by 5.1 percent.

with Gustavo Gonzaga, Henrique Mota, and Rafael Cayres

We study the consequences of abolishing mandatory homologation in Brazil's 2017 Labor Reform. Homologation was a dismissal-triggered audit required for workers with at least twelve months of tenure and served as the primary enforcement channel for employer deposits into the FGTS, a mandatory severance savings account. Using matched employer–employee administrative data and a differences-in-differences design that compares workers at 11 versus 12 months of tenure before and after the reform, we find that unjustified dismissal rates at 11 months fell by 0.524 percentage points — about 16% of the pre-reform level. Effects are largest in small firms, low-inspection sectors, and among less-educated workers, consistent with non-compliance being more prevalent where oversight is weakest. We then estimate a search-and-matching model with learning and endogenous compliance, disciplined by the full tenure-hazard profile. Estimated on pre-reform data only, the model predicts that the month-11 spike disappears after homologation is removed — an out-of-sample prediction confirmed by the data. The model implies that routine inspections alone are insufficient to sustain compliance, so the reform increases evasion. Although job stability improves, within the model per-worker welfare falls by 0.92 monthly wages because the loss of enforced savings contributions outweighs the reduction in anticipatory dismissals. Overall, enforcement design matters as much as statutory firing costs.