Estimates of effects of women in leadership positions on other women at the same organization can have incorrect signs and sizes because of endogenous selection bias. Similar effects can be produced for any underrepresented group.
Inference on quantile processes with a finite number of clusters
Randomization methods can be used for inference on entire quantile processes and related objects even if there are finitely many heterogeneous clusters. This includes processes arising from distributional difference-in-differences estimation.
Inference with a single treated cluster
April 2023, revise and resubmit at the Review of Economic Studies (2nd round)
A mild restriction on some but not all of the heterogeneity allows inference about a scalar parameter in research designs with a finite number of heterogeneous clusters where only a single cluster received treatment.
Reaching for gold: Frequent-flyer status incentives and moral hazard
Using data from a large airline, we document systematic changes in the behavior of frequent flyers as they progress towards elite status. The behavior is more pronounced for business travelers and if the home airport is a hub of the airline.
Permutation inference with a finite number of heterogeneous clusters
July 2022, Review of Economics and Statistics, forthcoming
Permutation inference can be adjusted to test conventional (non-sharp) hypotheses about the effect of a binary treatment even if you have finitely many arbitrarily heterogeneous clusters.
Placebo inference on treatment effects when the number of clusters is small
Journal of Econometrics 213(1) (2019), pp. 190-209
The intuitive notion of generating null distributions by assigning placebo treatments to untreated clusters can be formalized to conduct nearly exact inference about the lack of effect of a binary treatment.
Cluster-robust bootstrap inference in quantile regression models
Journal of the American Statistical Association 112(517) (2017), pp. 446-456
I provide a computationally convenient wild bootstrap for cluster-robust inference on the entire quantile regression process. This method is implemented as the cluster option in the quantreg package in R.
The effect of college education on mortality
with K. Buckles, O. Malamud, M. Morrill, A. Wozniak
We exploit exogenous variation in college completion induced by draft-avoidance behavior during the Vietnam War to examine the impact of college completion on adult mortality.
Stochastic equicontinuity in nonlinear time series models
Simple and easily verifiable conditions imply a strong form of stochastic equicontinuity in a wide variety of modern time series models. No mixing conditions are needed.
A simple test for regression specification with non-nested alternatives
Journal of Econometrics 166(2) (2012), pp. 247-254
The classical J test for non-nested regression models can be modified to test for the presence of the data-generating process among several non-nested alternatives.
Quantiles can capture systematic changes in the impact of cycles on the distribution of a time series and allow robust spectral estimation and inference when the dependence structure is not captured by the auto-covariance function.
