Volume 8, Issue 1

In this study, we propose simple test statistics for identifying the source of spatial dependence in spatial autoregressive models with endogenous weights matrices. Elements of the weights matrices are modelled in such a way that endogenity arises when the unobserved factors that affect elements of the weights matrices are correlated with the unobserved factors in the outcome equation. The proposed test statistics are robust to the presence of endogeneity in the weights and can be used to detect spatial dependence in the dependent variable and/or the disturbance terms. The robust test statistics are easy to calculate as computationally simple estimations are needed for their calculations. Our Monte Carlo results indicate that these tests have good size and power properties in finite samples. We also provide an empirical illustration to demonstrate the usefulness of the robust tests in identifying the source of spatial dependence.

When are asymptotic approximations using the delta-method uniformly valid? We provide sufficient conditions as well as closely related necessary conditions for uniform negligibility of the remainder of such approximations. These conditions are easily verified by empirical practitioners and permit to identify settings and parameter regions where pointwise asymptotic approximations perform poorly. Our framework allows for a unified and transparent discussion of uniformity issues in various sub-fields of statistics and econometrics. Our conditions involve uniform bounds on the remainder of a first-order approximation for the function of interest.

This note investigates the numerical performance of an existing asymptotic test for the null hypothesis of equality between the average treatment effect (ATE) and the group fixed-effect (FE) estimands based on the standardized difference between ATE and FE estimators. It shows that this test has a size distortion. This distortion has implications to empirical economic research. It can lead to erroneously confirm the relevance of heterogeneous responses to policy interventions.

Using a sequential conditional independence assumption, this paper discusses fully nonparametric estimation of natural direct and indirect causal effects in causal mediation analysis based on inverse probability weighting. We propose estimators of the average indirect effect of a binary treatment, which operates through intermediate variables (or mediators) on the causal path between the treatment and the outcome, as well as the unmediated direct effect. In a first step, treatment propensity scores given the mediator and observed covariates or given covariates alone are estimated by nonparametric series logit estimation. In a second step, they are used to reweigh observations in order to estimate the effects of interest. We establish root-n consistency and asymptotic normality of this approach as well as a weighted version thereof. The latter allows evaluating effects on specific subgroups like the treated, for which we derive the asymptotic properties under estimated propensity scores. We also provide a simulation study and an application to an information intervention about male circumcisions.

In regression discontinuity designs (RD), for a given bandwidth, researchers can estimate standard errors based on different variance formulas obtained under different asymptotic frameworks. In the traditional approach the bandwidth shrinks to zero as sample size increases; alternatively, the bandwidth could be treated as fixed. The main theoretical results for RD rely on the former, while most applications in the literature treat the estimates as parametric, implementing the usual heteroskedasticity-robust standard errors. This paper develops the “fixed-bandwidth” alternative asymptotic theory for RD designs, which sheds light on the connection between both approaches. I provide alternative formulas (approximations) for the bias and variance of common RD estimators, and conditions under which both approximations are equivalent. Simulations document the improvements in test coverage that fixed-bandwidth approximations achieve relative to traditional approximations, especially when there is local heteroskedasticity. Feasible estimators of fixed-bandwidth standard errors are easy to implement and are akin to treating RD estimators as locally parametric, validating the common empirical practice of using heteroskedasticity-robust standard errors in RD settings. Bias mitigation approaches are discussed and a novel bootstrap higher-order bias correction procedure based on the fixed bandwidth asymptotics is suggested.

We develop a test for a restricted functional form of a mean regression when a complex distributional model for all variables is estimated. The test statistic is an average squared deviation from the estimated hypothesized function of the form implied by the estimated parametric model, and is asymptotically distributed as a mixture of χ 2 distributions. The test is easy to implement using numerical derivatives, and it performs well in samples of typical size. We illustrate the test using data on labor market characteristics of US young men.

We consider a difference-in-differences setting with a continuous outcome. The standard practice is to take its logarithm and then interpret the results as an approximation of the multiplicative treatment effect on the original outcome. We argue that a researcher should rather focus on the non-transformed outcome when discussing causal inference. The first step should be to decide whether the time trend is more likely to hold in multiplicative or level form. If the former, it is preferable to estimate an exponential model by Poisson Pseudo Maximum Likelihood, which does not require statistical independence of the error term. Running OLS on the log-linearised model might instead lead to confounding distributional and mean changes. We illustrate the argument with a simulation exercise.

We replicate eight influential papers to provide empirical evidence that, in the presence of heterogeneous treatment effects, OLS with fixed effects (FE) is generally not a consistent estimator of the average treatment effect (ATE). We propose two alternative estimators that recover the ATE in the presence of group-specific heterogeneity. We document that heterogeneous treatment effects are common and the ATE is often statistically and economically different from the FE estimate. In all but one of our replications, there is statistically significant treatment effect heterogeneity and, in six, the ATEs are either economically or statistically different from the FE estimates.

I analyze properties of misspecified discrete choice models and the efficacy of Huber-White (sometimes called ‘robust’) standard errors. The Huber-White correction provides asymptotically correct standard errors for a consistent estimator from a misspecified model. There is little justification for using Huber-White standard errors in discrete choice models since misspecification usually leads to inconsistent estimators. I derive necessary and sufficient conditions for consistency of the maximum likelihood estimator of any potentially misspecified random utility model (e.g. conditional logit). I also derive (easily satisfied) sufficient conditions for consistent estimation of the sign of the data generating parameter. It follows the researcher can consistently test the sign (or nullity) of the parameter from the data generating process using the (possibly) misspecified conditional logit. I investigate small sample properties of the Huber-White estimator via a simulation study and find the correction provides little to no improvement for inferences.

This paper provides a review of methodological advancements in the evaluation of heterogeneous treatment effect models based on instrumental variable (IV) methods. We focus on models that achieve identification by assuming monotonicity of the treatment in the IV and analyze local average and quantile treatment effects for the subpopulation of compliers. We start with a comprehensive discussion of the binary treatment and binary IV case as for instance relevant in randomized experiments with imperfect compliance. We then review extensions to identification and estimation with covariates, multi-valued and multiple treatments and instruments, outcome attrition and measurement error, and the identification of direct and indirect treatment effects, among others. We also discuss testable implications and possible relaxations of the IV assumptions, approaches to extrapolate from local to global treatment effects, and the relationship to other IV approaches.