Search

The regression discontinuity (RD) design is one of the most widely used nonexperimental methods for causal inference and program evaluation. Over the last two decades, statistical and econometric methods for RD analysis have expanded and matured, and there is now a large number of methodological results for RD identification, estimation, inference, and validation. We offer a curated review of this methodological literature organized around the two most popular frameworks for the analysis and interpretation of RD designs: the continuity framework and the local randomization framework. For each framework, we discuss three main topics: ( a) designs and parameters, focusing on different types of RD settings and treatment effects of interest; ( b) estimation and inference, presenting the most popular methods based on local polynomial regression and methods for the analysis of experiments, as well as refinements, extensions, and alternatives; and ( c) validation and falsification, summarizing an array of mostly empirical approaches to support the validity of RD designs in practice.

We present a practical guide for the analysis of regression discontinuity (RD) designs in biomedical contexts. We begin by introducing key concepts, assumptions, and estimands within both the continuity-based framework and the local randomization framework. We then discuss modern estimation and inference methods within both frameworks, including approaches for bandwidth or local neighborhood selection, optimal treatment effect point estimation, and robust bias-corrected inference methods for uncertainty quantification. We also overview empirical falsification tests that can be used to support key assumptions. Our discussion focuses on two particular features that are relevant in biomedical research: (i) fuzzy RD designs, which often arise when therapeutic treatments are based on clinical guidelines but patients with scores near the cutoff are treated contrary to the assignment rule; and (ii) RD designs with discrete scores, which are ubiquitous in biomedical applications. We illustrate our discussion with three empirical applications: the effect of CD4 guidelines for anti-retroviral therapy on retention of HIV patients in South Africa, the effect of genetic guidelines for chemotherapy on breast cancer recurrence in the United States, and the effects of age-based patient cost-sharing on healthcare utilization in Taiwan. We provide replication materials employing publicly available statistical software in Python, R and Stata, offering researchers all necessary tools to conduct an RD analysis.

This monograph, together with its accompanying first part Cattaneo, Idrobo and Titiunik (2020), collects and expands the instructional materials we prepared for more than $40$ short courses and workshops on Regression Discontinuity (RD) methodology that we taught between 2014 and 2022. In this second monograph, we discuss several topics in RD methodology that build on and extend the analysis of RD designs introduced in Cattaneo, Idrobo and Titiunik (2020). Our first goal is to present an alternative RD conceptual framework based on local randomization ideas. This methodological approach can be useful in RD designs with discretely-valued scores, and can also be used more broadly as a complement to the continuity-based approach in other settings. Then, employing both continuity-based and local randomization approaches, we extend the canonical Sharp RD design in multiple directions: fuzzy RD designs, RD designs with discrete scores, and multi-dimensional RD designs. The goal of our two-part monograph is purposely practical and hence we focus on the empirical analysis of RD designs.

Barseghyan and Molinari (2023) give sufficient conditions for semi-nonparametric point identification of parameters of interest in a mixture model of decision-making under risk, allowing for unobserved heterogeneity in utility functions and limited consideration. A key assumption in the model is that the heterogeneity of risk preferences is unobservable but context-independent. In this comment, we build on their insights and present identification results in a setting where the risk preferences are allowed to be context-dependent.

Decision tree learning is increasingly being used for pointwise inference. Important applications include causal heterogenous treatment effects and dynamic policy decisions, as well as conditional quantile regression and design of experiments, where tree estimation and inference is conducted at specific values of the covariates. In this paper, we call into question the use of decision trees (trained by adaptive recursive partitioning) for such purposes by demonstrating that they can fail to achieve polynomial rates of convergence in uniform norm, even with pruning. Instead, the convergence may be poly-logarithmic or, in some important special cases, such as honest regression trees, fail completely. We show that random forests can remedy the situation, turning poor performing trees into nearly optimal procedures, at the cost of losing interpretability and introducing two additional tuning parameters. The two hallmarks of random forests, subsampling and the random feature selection mechanism, are seen to each distinctively contribute to achieving nearly optimal performance for the model class considered.

Yurinskii's coupling is a popular tool for finite-sample distributional approximation in mathematical statistics and applied probability, offering a Gaussian strong approximation for sums of random vectors under easily verified conditions with an explicit rate of approximation. Originally stated for sums of independent random vectors in $\ell^2$-norm, it has recently been extended to the $\ell^p$-norm, where $1 \leq p \leq \infty$, and to vector-valued martingales in $\ell^2$-norm under some rather strong conditions. We provide as our main result a generalization of all of the previous forms of Yurinskii's coupling, giving a Gaussian strong approximation for martingales in $\ell^p$-norm under relatively weak conditions. We apply this result to some areas of statistical theory, including high-dimensional martingale central limit theorems and uniform strong approximations for martingale empirical processes. Finally we give a few illustrative examples in statistical methodology, applying our results to partitioning-based series estimators for nonparametric regression, distributional approximation of $\ell^p$-norms of high-dimensional martingales, and local polynomial regression estimators. We address issues of feasibility, demonstrating implementable statistical inference procedures in each section., Comment: 37 pages

This paper discusses the R package lpcde, which stands for local polynomial conditional density estimation. It implements the kernel-based local polynomial smoothing methods introduced in Cattaneo, Chandak, Jansson, Ma (2022) for statistical estimation and inference of conditional distributions, densities, and derivatives thereof. The package offers pointwise and integrated mean square error optimal bandwidth selection and associated point estimators, as well as uncertainty quantification based on robust bias correction both pointwise (e.g., confidence intervals) and uniformly (e.g., confidence bands) over evaluation points. The methods implemented are boundary adaptive whenever the data is compactly supported. We contrast the functionalities of lpcde with existing R packages, and showcase its main features using simulated data.

Dyadic data is often encountered when quantities of interest are associated with the edges of a network. As such it plays an important role in statistics, econometrics and many other data science disciplines. We consider the problem of uniformly estimating a dyadic Lebesgue density function, focusing on nonparametric kernel-based estimators taking the form of dyadic empirical processes. Our main contributions include the minimax-optimal uniform convergence rate of the dyadic kernel density estimator, along with strong approximation results for the associated standardized and Studentized $t$-processes. A consistent variance estimator enables the construction of valid and feasible uniform confidence bands for the unknown density function. We showcase the broad applicability of our results by developing novel counterfactual density estimation and inference methodology for dyadic data, which can be used for causal inference and program evaluation. A crucial feature of dyadic distributions is that they may be "degenerate" at certain points in the support of the data, a property making our analysis somewhat delicate. Nonetheless our methods for uniform inference remain robust to the potential presence of such points. For implementation purposes, we discuss inference procedures based on positive semi-definite covariance estimators, mean squared error optimal bandwidth selectors and robust bias correction techniques. We illustrate the empirical finite-sample performance of our methods both in simulations and with real-world trade data, for which we make comparisons between observed and counterfactual trade distributions in different years. Our technical results concerning strong approximations and maximal inequalities are of potential independent interest., Comment: Article: 21 pages, 3 figures. Supplemental appendix: 75 pages

The Regression Discontinuity (RD) design is a widely used non-experimental method for causal inference and program evaluation. While its canonical formulation only requires a score and an outcome variable, it is common in empirical work to encounter RD analyses where additional variables are used for adjustment. This practice has led to misconceptions about the role of covariate adjustment in RD analysis, from both methodological and empirical perspectives. In this chapter, we review the different roles of covariate adjustment in RD designs, and offer methodological guidance for its correct use.

We introduce an Attention Overload Model that captures the idea that alternatives compete for the decision maker's attention, and hence the attention that each alternative receives decreases as the choice problem becomes larger. Using this nonparametric restriction on the random attention formation, we show that a fruitful revealed preference theory can be developed and provide testable implications on the observed choice behavior that can be used to (point or partially) identify the decision maker's preference and attention frequency. We then enhance our attention overload model to accommodate heterogeneous preferences. Due to the nonparametric nature of our identifying assumption, we must discipline the amount of heterogeneity in the choice model: we propose the idea of List-based Attention Overload, where alternatives are presented to the decision makers as a list that correlates with both heterogeneous preferences and random attention. We show that preference and attention frequencies are (point or partially) identifiable under nonparametric assumptions on the list and attention formation mechanisms, even when the true underlying list is unknown to the researcher. Building on our identification results, for both preference and attention frequencies, we develop econometric methods for estimation and inference that are valid in settings with a large number of alternatives and choice problems, a distinctive feature of the economic environment we consider. We also provide a software package in R implementing our empirical methods, and illustrate them in a simulation study.

Binned scatter plots, or binscatters, have become a popular and convenient tool in applied microeconomics for visualizing bivariate relations and conducting informal specification testing. However, a binscatter, on its own, is very limited in what it can characterize about the conditional mean. We introduce a suite of formal and visualization tools based on binned scatter plots to restore, and in some dimensions surpass, the visualization benefits of the classical scatter plot. We deliver a comprehensive toolkit for applications, including estimation of conditional mean and quantile functions, visualization of variance and precise quantification of uncertainty, and formal tests of substantive hypotheses such as linearity or monotonicity, and an extension to testing differences across groups. To do so we give an extensive theoretical analysis of binscatter and related partition-based methods, accommodating nonlinear and potentially nonsmooth models, which allows us to treat binary, count, and other discrete outcomes as well. We also correct a methodological mistake related to covariate adjustment present in prior implementations, which yields an incorrect shape and support of the conditional mean. All of our results are implemented in publicly available software, and showcased with three substantive empirical illustrations. Our empirical results are dramatically different when compared to those obtained using the prevalent methods in the literature.

We introduce the Stata package Binsreg, which implements the binscatter methods developed in Cattaneo, Crump, Farrell and Feng (2023a,b). The package includes seven commands: binsreg, binslogit, binsprobit, binsqreg, binstest, binspwc, and binsregselect. The first four commands implement point estimation and uncertainty quantification (confidence intervals and confidence bands) for canonical and extended least squares binscatter regression (binsreg) as well as generalized nonlinear binscatter regression (binslogit for Logit regression, binsprobit for Probit regression, and binsqreg for quantile regression). These commands also offer binned scatter plots, allowing for one- and multi-sample settings. The next two commands focus on pointwise and uniform inference: binstest implements hypothesis testing procedures for parametric specifications and for nonparametric shape restrictions of the unknown regression function, while binspwc implements multi-group pairwise statistical comparisons. These two commands cover both least squares as well as generalized nonlinear binscatter methods. All our methods allow for multi-sample analysis, which is useful when studying treatment effect heterogeneity in randomized and observational studies. Finally, the command binsregselect implements data-driven number of bins selectors for binscatter methods using either quantile-spaced or evenly-spaced binning/partitioning. All the commands allow for covariate adjustment, smoothness restrictions, weighting and clustering, among many other features. Companion Python and R packages with similar syntax and capabilities are also available.

This handbook chapter gives an introduction to the sharp regression discontinuity design, covering identification, estimation, inference, and falsification methods.

We introduce the \texttt{Stata} (and \texttt{R}) package \texttt{rdmulti}, which includes three commands (\texttt{rdmc}, \texttt{rdmcplot}, \texttt{rdms}) for analyzing Regression Discontinuity (RD) designs with multiple cutoffs or multiple scores. The command \texttt{rdmc} applies to non-cumulative and cumulative multi-cutoff RD settings. It calculates pooled and cutoff-specific RD treatment effects, and provides robust bias-corrected inference procedures. Post estimation and inference is allowed. The command \texttt{rdmcplot} offers RD plots for multi-cutoff settings. Finally, the command \texttt{rdms} concerns multi-score settings, covering in particular cumulative cutoffs and two running variables contexts. It also calculates pooled and cutoff-specific RD treatment effects, provides robust bias-corrected inference procedures, and allows for post-estimation estimation and inference. These commands employ the \texttt{Stata} (and \texttt{R}) package \texttt{rdrobust} for plotting, estimation, and inference. Companion \texttt{R} functions with the same syntax and capabilities are provided.

In this Element and its accompanying Element, Matias D. Cattaneo, Nicolas Idrobo, and Rocio Titiunik provide an accessible and practical guide for the analysis and interpretation of Regression Discontinuity (RD) designs that encourages the use of a common set of practices and facilitates the accumulation of RD-based empirical evidence. In this Element, the authors discuss the foundations of the canonical Sharp RD design, which has the following features: (i) the score is continuously distributed and has only one dimension, (ii) there is only one cutoff, and (iii) compliance with the treatment assignment is perfect. In the accompanying Element, the authors discuss practical and conceptual extensions to the basic RD setup.

Modern empirical work in Regression Discontinuity (RD) designs often employs local polynomial estimation and inference with a mean square error (MSE) optimal bandwidth choice. This bandwidth yields an MSE-optimal RD treatment effect estimator, but is by construction invalid for inference. Robust bias corrected (RBC) inference methods are valid when using the MSE-optimal bandwidth, but we show they yield suboptimal confidence intervals in terms of coverage error. We establish valid coverage error expansions for RBC confidence interval estimators and use these results to propose new inference-optimal bandwidth choices for forming these intervals. We find that the standard MSE-optimal bandwidth for the RD point estimator is too large when the goal is to construct RBC confidence intervals with the smallest coverage error. We further optimize the constant terms behind the coverage error to derive new optimal choices for the auxiliary bandwidth required for RBC inference. Our expansions also establish that RBC inference yields higher-order refinements (relative to traditional undersmoothing) in the context of RD designs. Our main results cover sharp and sharp kink RD designs under conditional heteroskedasticity, and we discuss extensions to fuzzy and other RD designs, clustered sampling, and pre-intervention covariates adjustments. The theoretical findings are illustrated with a Monte Carlo experiment and an empirical application, and the main methodological results are available in \texttt{R} and \texttt{Stata} packages.

The linear regression model is widely used in empirical work in Economics, Statistics, and many other disciplines. Researchers often include many covariates in their linear model specification in an attempt to control for confounders. We give inference methods that allow for many covariates and heteroskedasticity. Our results are obtained using high-dimensional approximations, where the number of included covariates is allowed to grow as fast as the sample size. We find that all of the usual versions of Eicker-White heteroskedasticity consistent standard error estimators for linear models are inconsistent under this asymptotics. We then propose a new heteroskedasticity consistent standard error formula that is fully automatic and robust to both (conditional) heteroskedasticity of unknown form and the inclusion of possibly many covariates. We apply our findings to three settings: parametric linear models with many covariates, linear panel models with many fixed effects, and semiparametric semi-linear models with many technical regressors. Simulation evidence consistent with our theoretical results is provided, and the proposed methods are also illustrated with an empirical application.

The linear regression model is widely used in empirical work in Economics, Statistics, and many other disciplines. Researchers often include many covariates in their linear model specification in an attempt to control for confounders. We give inference methods that allow for many covariates and heteroskedasticity. Our results are obtained using high-dimensional approximations, where the number of included covariates are allowed to grow as fast as the sample size. We find that all of the usual versions of Eicker-White heteroskedasticity consistent standard error estimators for linear models are inconsistent under this asymptotics. We then propose a new heteroskedasticity consistent standard error formula that is fully automatic and robust to both (conditional)\ heteroskedasticity of unknown form and the inclusion of possibly many covariates. We apply our findings to three settings: parametric linear models with many covariates, linear panel models with many fixed effects, and semiparametric semi-linear models with many technical regressors. Simulation evidence consistent with our theoretical results is also provided. The proposed methods are also illustrated with an empirical application.