Search

We offer straightforward theoretical results that justify incorporating machine learning in the standard linear instrumental variable setting. The key idea is to use machine learning, combined with sample-splitting, to predict the treatment variable from the instrument and any exogenous covariates, and then use this predicted treatment and the covariates as technical instruments to recover the coefficients in the second-stage. This allows the researcher to extract non-linear co-variation between the treatment and instrument that may dramatically improve estimation precision and robustness by boosting instrument strength. Importantly, we constrain the machine-learned predictions to be linear in the exogenous covariates, thus avoiding spurious identification arising from non-linear relationships between the treatment and the covariates. We show that this approach delivers consistent and asymptotically normal estimates under weak conditions and that it may be adapted to be semiparametrically efficient (Chamberlain, 1992). Our method preserves standard intuitions and interpretations of linear instrumental variable methods, including under weak identification, and provides a simple, user-friendly upgrade to the applied economics toolbox. We illustrate our method with an example in law and criminal justice, examining the causal effect of appellate court reversals on district court sentencing decisions., NeurIPS 2020 Workshop on Machine Learning for Economic Policy

Consider a normal location model $X \mid \theta \sim N(\theta, \sigma^2)$ with known $\sigma^2$. Suppose $\theta \sim G_0$, where the prior $G_0$ has zero mean and unit variance. Let $G_1$ be a possibly misspecified prior with zero mean and unit variance. We show that the squared error Bayes risk of the posterior mean under $G_1$ is bounded, uniformly over $G_0, G_1, \sigma^2 > 0$.

This paper proposes empirical Bayes shrinkage methods. Compared to common shrinkage methods, we do not assume that the unknown parameters are independent from the known standard errors. This prior independence assumption is both theoretically tenuous and often empirically rejected. We instead model the conditional distribution of the parameter given the standard errors as a location-scale family. This assumption leads to a family of methods that we call CLOSE. We establish that (i) CLOSE is rate-optimal for squared error Bayes regret up to logarithmic factors, (ii) squared error regret control is sufficient for a class of economic decision problems, and (iii) CLOSE is worst-case robust. We illustrate our method with an empirical application to the Opportunity Atlas and Creating Moves to Opportunity (Chetty et al., 2018; Bergman et al., 2019). For the decision problem of selecting high mobility Census tracts in Bergman et al. (2019), CLOSE selects Census tracts that are more economically mobile than the standard shrinkage method. This estimated gain is larger than the gain of using the standard method relative to selecting tracts uniformly at random.

Many economic settings involve an outcome $Y$ that is weakly positive but can equal zero (e.g. earnings). In such settings, it is common to estimate an average treatment effect (ATE) for a transformation of the outcome that behaves like $\log(Y)$ when $Y$ is large but is defined at zero (e.g. $\log(1+Y)$, $\mathrm{arcsinh}(Y)$). This paper argues that ATEs for such log-like transformations should not be interpreted as approximating a percentage effect, since unlike a percentage, they depend arbitrarily on the units of the outcome when the treatment affects the extensive margin. Intuitively, this dependence arises because an individual-level percentage effect is not well-defined for individuals whose outcome changes from zero to non-zero when receiving treatment, and the units of the outcome implicitly determine how much weight the ATE places on such extensive margin changes. We further establish that when the outcome can equal zero, there is no treatment effect parameter that is an average of individual-level treatment effects, unit-invariant, and point-identified. We discuss a variety of alternative approaches that may be sensible in settings with an intensive and extensive margin, including (i) expressing the ATE in levels as a percentage (e.g. using Poisson regression), (ii) explicitly calibrating the value placed on the intensive and extensive margins, and (iii) estimating separate effects for the two margins (e.g. using Lee bounds). We illustrate these approaches in three empirical applications.

We study nonparametric identification and estimation of causal effects in centralized school assignment. We characterize the full set of identified treatment effects in common school choice settings, under unrestricted heterogeneity in individual potential outcomes. This exercise highlights two points of caution for practitioners: We find that lack of overlap poses a challenge to regression-based estimators; we also find that, asymptotically, regression-based estimators that aggregate across many treatment contrasts put zero weight on treatment effects identified from regression-discontinuity (RD) variation, when the mechanism allows for both RD and lottery-based variation. Due to the complex interplay between heterogeneous causal effects and school choice algorithms, we recommend empirical researchers clearly decompose aggregate causal effect estimates by sources of variation in these settings. Lastly, we provide estimators and accompanying asymptotic results for causal contrasts identified by RD variation in school choice., Presented at SOLE 2021

Long-term outcomes of experimental evaluations are necessarily observed after long delays. We develop semiparametric methods for combining the short-term outcomes of experiments with observational measurements of short-term and long-term outcomes, in order to estimate long-term treatment effects. We characterize semiparametric efficiency bounds for various instances of this problem. These calculations facilitate the construction of several estimators. We analyze the finite-sample performance of these estimators with a simulation calibrated to data from an evaluation of the long-term effects of a poverty alleviation program.

Artificial Neural Networks (ANNs) can be viewed as nonlinear sieves that can approximate complex functions of high dimensional variables more effectively than linear sieves. We investigate the performance of various ANNs in nonparametric instrumental variables (NPIV) models of moderately high dimensional covariates that are relevant to empirical economics. We present two efficient procedures for estimation and inference on a weighted average derivative (WAD): an orthogonalized plug-in with optimally-weighted sieve minimum distance (OP-OSMD) procedure and a sieve efficient score (ES) procedure. Both estimators for WAD use ANN sieves to approximate the unknown NPIV function and are root-n asymptotically normal and first-order equivalent. We provide a detailed practitioner's recipe for implementing both efficient procedures. We compare their finite-sample performances in various simulation designs that involve smooth NPIV function of up to 13 continuous covariates, different nonlinearities and covariate correlations. Some Monte Carlo findings include: 1) tuning and optimization are more delicate in ANN estimation; 2) given proper tuning, both ANN estimators with various architectures can perform well; 3) easier to tune ANN OP-OSMD estimators than ANN ES estimators; 4) stable inferences are more difficult to achieve with ANN (than spline) estimators; 5) there are gaps between current implementations and approximation theories. Finally, we apply ANN NPIV to estimate average partial derivatives in two empirical demand examples with multivariate covariates.