I exploit the abrupt closure of rural rail stations after Côte d’Ivoire’s 2011 post-presidentialelection crisis to study how losing access to outside labor markets affects rural living standards. Using nationally representative microdata, I find sizable declines in per-capita spending and consumption in rural areas exposed to rail closures relative to nearby controls. These declines coincide with a labor reallocation from higher-paying nonfarm employment into contributing-family work. I rationalize these patterns using a Roy model with heterogeneous nonfarm returns and sectoral mobility costs. The estimated nonfarm returns reveal strong heterogeneity, implying that similar transport investments can generate different welfare outcomes depending on which workers gain access to nonfarm jobs. Counterfactual simulations show that spatially targeted reopenings of a subset of rural rail stations are economically effective as they can achieve nearly the same welfare gains as full network restoration. These findings help inform infrastructure policy when budget constraints limit expansion.

We propose sharp testable implications and tests to jointly assess the random assignment, exclusion, and monotonicity assumptions in judge leniency designs. Our procedures accommodate various data scenarios in which the number of defendants handled by a judge may be either small or large, and allow for discrete or continuous instrumental variables. When the validity of the design is rejected, a variant of the marginal treatment effect can be identified under weaker assumptions. We apply our test to the Philadelphia court data studied by Stevenson (2018) and demonstrate that it outperforms non-sharp joint tests by significant margins in simulation studies.

Standard enforcement theory treats detection and penalties as substitutes that matter only through the expected fine. We show they are not interchangeable when violations are durable states: detection clears violations, while penalties only price them. We study Colombia’s 2013 Law 1610, which sharply raised penalties for labor violations. In departments more exposed to pre-reform minimum-wage underpayment, event-study estimates show that the depth of violations fell after the reform while their incidence did not. We estimate a search-and-matching model with imperfect enforcement and compare the two instruments: matched to the same reduction in aggregate underpayment, penalties reduce both the number and the depth of violations without job losses, whereas detection cuts incidence by 11 percentage points, more than twice the reduction under penalties, while leaving a negatively selected pool of deeper violations and raising unemployment. Equal expected fines need not produce equal enforcement outcomes.

Judge-IV designs increasingly invoke average monotonicity when strong monotonicity is hard to defend. We show that average monotonicity does not preserve a stable target population: the response types receiving positive weight switch discretely when a judge’s treatment propensity crosses the assignment-weighted average. With heterogeneous treatment effects, the estimand remains a proper weighted average but may represent a different target population after an arbitrarily small design change. A standardized boundary-distance diagnostic flags fragile judges; in Philadelphia bail data, three of eight judges, covering a non-negligible share of cases, are fragile. Target-population stability is therefore an empirical object.