Fast and wild: Bootstrap inference in Stata using boottest

Abstract

The wild bootstrap was originally developed for regression models with heteroskedasticity of unknown form. Over the past 30 years, it has been extended to models estimated by instrumental variables and maximum likelihood and to ones where the error terms are (perhaps multiway) clustered. Like bootstrap methods in general, the wild bootstrap is especially useful when conventional inference methods are unreliable because large-sample assumptions do not hold. For example, there may be few clusters, few treated clusters, or weak instruments. The package boottest can perform a wide variety of wild bootstrap tests, often at remarkable speed. It can also invert these tests to construct confidence sets. As a postestimation command, boottest works after linear estimation commands, including regress, cnsreg, ivregress, ivreg2, areg, and reghdfe, as well as many estimation commands based on maximum likelihood. Although it is designed to perform the wild cluster bootstrap, boottest can also perform the ordinary (nonclustered) version. Wrappers offer classical Wald, score/Lagrange multiplier, and Anderson–Rubin tests, optionally with (multiway) clustering. We review the main ideas of the wild cluster bootstrap, offer tips for use, explain why it is particularly amenable to computational optimization, state the syntax of boottest, artest, scoretest, and waldtest, and present several empirical examples.

The Impact of Automated Reminders on Credit Outcomes: Results from an Experimental Pilot Program

“…One challenge in clustering standard errors at the counselor level is that the number of observations per counselor is not balanced and the number of clusters is relatively small (25). To account for these issues, we use a wild bootstrapping approach, which performs well in cases where the number of clusters is small and the size of clusters is heterogeneous (Roodman et al ). The results of these models are very similar to the results of the models without clustered standard errors; the only notable change is that the TOT estimate for 30‐day payment delinquencies becomes significant at the 5% level.…”

How cluster‐robust inference is changing applied econometrics

“…Perhaps surprisingly, it can often be computed remarkably quickly using the Stata routine boottest; see Roodman et al. ().…”

“…The computations for the AR statistic can take advantage of the tricks that make the WCR bootstrap so fast (see section ), but the ones for t β cannot do so; see Roodman et al. (). Thus, in large samples, it can be much faster to bootstrap the cluster‐robust AR statistic than the cluster‐robust IV t statistic.…”

Does Individualizing the Labor Contract Hurt Women?

“…Unless otherwise noted, the bootstrap procedure is applied here in the final specification of each regression, using the Stata module boottest, with the null hypothesis imposed (Roodman et al. ). The regression tables report significance levels using this procedure on the indicator variable, as well as adjusted standard errors.…”