Tiago Tavares

xhdfe is a Stata command for linear regression with high-dimensional fixed effects. It targets the same partialled-out least-squares estimator as reghdfe on the overlapping specifications studied here, while moving the main absorption workload into a compiled C++ backend with multithreaded CPU execution and optional NVIDIA CUDA acceleration. The command preserves a Stata e-class interface while supporting multiple absorbed fixed effects, clustered and robust standard errors, heterogeneous slopes in the tested specifications, backend choices, saved fixed effects, residuals, fitted values, and stored results. We validate xhdfe on two large wage regressions using Portuguese matched employer-employee data. The command reproduces reghdfe coefficients and standard errors within tight numerical tolerances and delivers speedups above 100-fold with CPU execution and 200-fold with CUDA acceleration in the fastest configurations. Compared with alternative high-dimensional fixed-effects implementations, xhdfe records the shortest runtimes among the implementations and configurations reported here. The development history also provides a case study of AI-assisted software development. Agentic tools supported coding and review, while conventional numerical tests and reproducible benchmarks assessed the resulting changes.

This paper develops a theory of financing and governing transformative search: costly search for a transformative opportunity whose timing, payoff, and implementation are uncertain. The central object is the runway-attainability wedge, the gap between desired runway and the runway that can be financed while preserving authority over search. The decomposition separates this wedge into financing-frontier and governance shortfalls. Liquidity is valuable because it buys discovery time and, after discovery, implementation capacity. The decision criterion is whether a financing architecture preserves attainable runway and search incentives, not how much capital it raises.

Emerging market economies often exhibit aggregate consumption that is more volatile than aggregate income, contrary to predictions of standard macro models based on consumption smoothing. We explore whether heterogeneity in access to financial services can explain this excess consumption volatility. We extend the standard small-open-economy RBC model by incorporating hand-to-mouth and unconstrained households alongside procyclical firm entry, and estimate the model using data for advanced, emerging, and low-income economies.

Managerial occupations represent a significant and expanding segment of the US labor force, while good managerial practices enhance production efficiency. We study how competition among firms for managerial services affects managerial compensation and, through firms' demand for other factors of production, the compensation of high-skill workers and the skill premium.

A salient feature of emerging economies is that government spending is procyclical while labor taxes move countercyclically. We account for this behavior in a small open economy in which the government conducts fiscal policy optimally and can commit to future policies. The presence of an informal sector widens the set of parameters under which distorting labor taxes are negatively correlated with output by amplifying fluctuations in the tax base.

Highly indebted developing economies commonly also hold large external reserves. I show that fiscal adjustments induced by sovereign default can generate strong demand for reserves when taxation is distortionary, while reserves also modify debt-restructuring negotiations. A calibrated model produces recovery-rate schedules increasing in reserves and replicates large positions in both reserves and debt.

How much do credit constraints contribute to investment slumps during financial crises? For the Greek crisis that erupted in 2010, we find that tightened credit constraints contributed to about half of the observed collapse in investment rates. A dynamic investment model with borrowing constraints subject to an aggregate collateral shock can account for the observed decline.

Using a randomized online information intervention in Mexico, we show that failing to account for delays in death reports lowers perceived contagion risk and intended compliance with social distancing. An equilibrium model incorporating the behavioral response illustrates how reporting delays affect the evolution of an epidemic.

Labor-market distortions deteriorate substantially around sovereign default episodes. I evaluate the roles of labor taxes and working-capital financing costs in a dynamic equilibrium default model. Both mechanisms help reproduce the observed labor wedge and the substantial employment declines surrounding financial crises.