Econometric Evaluation of Socio-Economic Programs by Giovanni Cerulli provides an excellent introduction to estimating average treatment effects from observational data. This book provides thorough introductions to the models and estimators implemented in teffects, etregress, and etpoisson and provides many examples using these commands and some similar commands written by the author.
After presenting an overview of the topic, the book delves into methods based on conditional independence. Next, it considers methods that drop conditional independence, thereby allowing for endogenous treatment effects. Finally, the book covers more advanced topics, including local average treatment effects and regression-discontinuity designs.
The author provides a nice mix of intuition, mathematics, and Stata examples. Professors, graduate students, and researchers will find this book useful in the classroom and for self-study in preparation for research projects.
AN INTRODUCTION TO THE ECONOMETRICS OF PROGRAM EVALUATION
Introduction
Statistical Setup, Notation, and Assumptions
Identification Under Random Assignment
A Bayesian Interpretation of ATE Under Randomization
Consequences of Nonrandom Assignment and Selection Bias
Selection on Observables and Selection on Unobservables
Selection on Observables (or Overt Bias) and Conditional Independence Assumption
Selection on Unobservables (or Hidden Bias)
The Overlap Assumption
Characterizing Selection Bias
Decomposing Selection Bias
The Rationale for Choosing the Variables to Control for
Partial Identification of ATEs: The Bounding Approach
A Guiding Taxonomy of the Econometric Methods for Program Evaluation
Policy Framework and the Statistical Design for Counterfactual Evaluation
Available Econometric Software
A Brief Outline of the Book
References
METHODS BASED ON SELECTION ON OBSERVABLES
Introduction
Regression-Adjustment
Regression-Adjustment as Unifying Approach Under Observable Selection
Linear Parametric Regression-Adjustment: The Control-Function Regression
Nonlinear Parametric Regression-Adjustment
Nonparametric and Semi-parametric Regression-Adjustment
Matching
Covariates and Propensity-Score Matching
Identification of ATEs Under Matching
Large Sample Properties of Matching Estimator(s)
Common Support
Exact Matching and the “Dimensionality Problem”
The Properties of the Propensity-Score
Quasi-Exact Matching Using the Propensity-Score
Methods for Propensity-Score Matching
Inference for Matching Methods
Assessing the Reliability of CMI by Sensitivity Analysis
Assessing Overlap
Coarsened-Exact Matching
Reweighting
Reweighting and Weighted Least Squares
Reweighting on the Propensity-Score Inverse-Probability
Sample Estimation and Standard Errors for ATEs
Doubly-Robust Estimation
Implementation and Application of Regression-Adjustment
Implementation and Application of Matching
Covariates Matching
Propensity-Score Matching
An Example of Coarsened-Exact Matching Using cem
Implementation and Application of Reweighting
The Stata Routine treatrew
The Relation Between treatrew and Stata 13’s teffects ipw
An Application of the Doubly-Robust Estimator
References
METHODS BASED ON SELECTION ON UNOBSERVABLES
Introduction
Instrumental-Variables
IV Solution to Hidden Bias
IV Estimation of ATEs
IV with Observable and Unobservable Heterogeneities
Problems with IV Estimation
Selection-Model
Characterizing OLS Bias within a Selection-Model
A Technical Exposition of the Selection-Model
Selection-Model with a Binary Outcome
Difference-in-Differences
DID with Repeated Cross Sections
DID with Panel Data
DID with Matching
Time-Variant Treatment and Pre-Post Treatment Analysis
Implementation and Application of IV and Selection-Model
The Stata Command ivtreatreg
A Monte Carlo Experiment
An Application to Determine the Effect of Education on Fertility
Applying the Selection-Model Using etregress
Implementation and Application of DID
DID with Repeated Cross Sections
DID Application with Panel Data
References
LOCAL AVERAGE TREATMENT EFFECT AND REGRESSION-DISCONTINUITY-DESIGN
Introduction
Local Average Treatment Effect
Randomization Under Imperfect Compliance
Wald Estimator and LATE
LATE Estimation
Estimating Average Response for Compliers
Characterizing Compliers
LATE with Multiple Instruments and Multiple Treatment
Regression-Discontinuity-Design
Sharp RDD
Fuzzy RDD
The Choice of the Bandwidth and Polynomial Order
Accounting for Additional Covariates
Testing RDD Reliability
A Protocol for Practical Implementation of RDD
Application and Implementation
An Application of LATE
An Application of RDD by Simulation
References