Abstract:We revisit the problem of estimating the local average treatment effect (LATE) and the local average treatment effect on the treated (LATT) when control variables are available, either to render the instrumental variable (IV) suitably exogenous or to improve precision. Unlike previous approaches, our doubly robust (DR) estimation procedures use quasi-likelihood methods weighted by the inverse of the IV propensity score - so-called inverse probability weighted regression adjustment (IPWRA) estimators. By properly choosing models for the propensity score and outcome models, fitted values are ensured to be in the logical range determined by the response variable, producing DR estimators of LATE and LATT with appealing small sample properties. Inference is relatively straightforward both analytically and using the nonparametric bootstrap. Our DR LATE and DR LATT estimators work well in simulations. We also propose a DR version of the Hausman test that can be used to assess the unconfoundedness assumption through a comparison of different estimates of the average treatment effect on the treated (ATT) under one-sided noncompliance. Unlike the usual test that compares OLS and IV estimates, this procedure is robust to treatment effect heterogeneity.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to estimate the Local Average Treatment Effect (LATE) and the Local Average Treatment Effect on the Treated (LATT) in the presence of control variables. Specifically, the paper proposes a new doubly robust (DR) estimation method, using Inverse Probability Weighted Regression Adjustment (IPWRA) to estimate these effects. This method aims to overcome some limitations of existing methods, such as the deficiencies of non - parametric conditional mean estimation and the AIPW method, while ensuring that the estimation results have good small - sample properties and are easy to implement. ### Background and Problem of the Paper In economics and econometrics, the Instrumental Variables (IV) estimation method is often used for causal effect estimation. However, in practical applications, the instrumental variables may not be randomly assigned, which will cause the simple IV estimator to no longer consistently estimate LATE. By introducing control variables, the instrumental variables can be made to be randomly assigned within certain sub - groups, thereby satisfying the unconfoundedness assumption. However, how to effectively estimate LATE and LATT in the case of introducing control variables remains a challenge. ### Research Objectives The main objectives of the paper are to propose a new doubly robust estimation method that can: 1. **Handle non - randomly assigned instrumental variables**: By introducing control variables, make the instrumental variables be randomly assigned within certain sub - groups. 2. **Improve the robustness and precision of estimation**: Use the Inverse Probability Weighted Regression Adjustment (IPWRA) method to ensure that the estimation results also have good properties in small samples. 3. **Simplify the estimation process**: Provide a simple and easy - to - implement method while ensuring the double - robustness of the estimation results. ### Method Overview The method proposed in the paper is based on the following steps: 1. **Estimate the propensity score of the instrumental variable**: First, estimate the propensity score \(\eta(X) = P(Z = 1|X)\) of the instrumental variable \(Z\), using a standard parametric model (such as the logit model). 2. **Estimate the conditional expectation of the treatment variable**: Estimate the conditional expectations \(\rho_0(X)\) and \(\rho_1(X)\) of the treatment variable \(W\) in the \(Z = 0\) and \(Z = 1\) sub - groups respectively, using the inverse - probability - weighted logit model. 3. **Estimate the conditional expectation of the outcome variable**: Estimate the conditional expectations \(\mu_0(X)\) and \(\mu_1(X)\) of the outcome variable \(Y\) in the \(Z = 0\) and \(Z = 1\) sub - groups respectively, using the inverse - probability - weighted Quasi - Maximum Likelihood Estimation (QMLE). 4. **Calculate LATE and LATT**: Use the above - mentioned estimation results to calculate the doubly robust estimators of LATE and LATT. ### Formula Representation - **Propensity Score of the Instrumental Variable**: \[ \eta(X)=P(Z = 1|X) \] - **Conditional Expectation of the Treatment Variable**: \[ \rho_0(X)=E(W|X,Z = 0),\quad\rho_1(X)=E(W|X,Z = 1) \] - **Conditional Expectation of the Outcome Variable**: \[ \mu_0(X)=E(Y|X,Z = 0),\quad\mu_1(X)=E(Y|X,Z = 1) \] - **Doubly Robust Estimation of LATE**: \[ \hat{\tau}_{\text{LATE}}=\frac{\hat{E}[\mu_1(X)] -

Doubly Robust Estimation of Local Average Treatment Effects Using Inverse Probability Weighted Regression Adjustment

Double-robust and efficient methods for estimating the causal effects of a binary treatment

A Stable and More Efficient Doubly Robust Estimator

The covariate-adjusted residual estimator and its use in both randomized trials and observational settings

Inverse probability of treatment weighting with generalized linear outcome models for doubly robust estimation

Using generalized doubly robust estimator to estimate average treatment effects of multiple treatments in observational studies

Robust Estimation of the Weighted Average Treatment Effect for A Target Population

Doubly robust estimation of average treatment effect revisited

High-dimensional Model-assisted Inference for Local Average Treatment Effects with Instrumental Variables

Non-parametric Replication of Instrumental Variable Estimates Across Studies

Double robust estimation of partially adaptive treatment strategies

Debiased Inverse Propensity Score Weighting for Estimation of Average Treatment Effects with High-Dimensional Confounders

Robust inference on the average treatment effect using the outcome highly adaptive lasso

An Alternative Doubly Robust Estimation in Causal Inference Model

The Finite Sample Performance of Instrumental Variable-Based Estimators of the Local Average Treatment Effect When Controlling for Covariates

High-dimensional model-assisted inference for treatment effects with multi-valued treatments

A nonparametric super-efficient estimator of the average treatment effect

Bounded, efficient and multiply robust estimation of average treatment effects using instrumental variables

Double Robust Bayesian Inference on Average Treatment Effects

Estimating Average Treatment Effects with Continuous and Discrete Covariates: The Case of Swan-Ganz Catheterization

Efficient Covariate Balancing for the Local Average Treatment Effect