REGRESSION, IV, MATCHING Treatment effect Boualem RABTA Center for World Food Studies (SOW-VU) Vrije Universiteit - Amsterdam.

Slides:



Advertisements
Similar presentations
Impact analysis and counterfactuals in practise: the case of Structural Funds support for enterprise Gerhard Untiedt GEFRA-Münster,Germany Conference:
Advertisements

The Simple Linear Regression Model Specification and Estimation Hill et al Chs 3 and 4.
Designing an impact evaluation: Randomization, statistical power, and some more fun…
Introduction to Propensity Score Matching
Hierarchical Linear Modeling: An Introduction & Applications in Organizational Research Michael C. Rodriguez.
The Simple Regression Model
Review of Identifying Causal Effects Methods of Economic Investigation Lecture 13.
What could go wrong? Deon Filmer Development Research Group, The World Bank Evidence-Based Decision-Making in Education Workshop Africa Program for Education.
Omitted Variable Bias Methods of Economic Investigation Lecture 7 1.
Instrumental Variables Estimation and Two Stage Least Square
PHSSR IG CyberSeminar Introductory Remarks Bryan Dowd Division of Health Policy and Management School of Public Health University of Minnesota.
Additional Topics in Regression Analysis
The Fundamental Problem of Causal Inference Alexander Tabarrok January 2007.
Impact Evaluation: The case of Bogotá’s concession schools Felipe Barrera-Osorio World Bank 1 October 2010.
Sampling and Experimental Control Goals of clinical research is to make generalizations beyond the individual studied to others with similar conditions.
© Institute for Fiscal Studies The role of evaluation in social research: current perspectives and new developments Lorraine Dearden, Institute of Education.
Violations of Assumptions In Least Squares Regression.
Chapter 8 Experimental Research
Design Experimental Control. Experimental control allows causal inference (IV caused observed change in DV) Experiment has internal validity when it fulfills.
Chap 20-1 Statistics for Business and Economics, 6e © 2007 Pearson Education, Inc. Chapter 20 Sampling: Additional Topics in Sampling Statistics for Business.
Quasi Experimental Methods I Nethra Palaniswamy Development Strategy and Governance International Food Policy Research Institute.
Propensity Score Matching and Variations on the Balancing Test Wang-Sheng Lee Melbourne Institute of Applied Economic and Social Research The University.
Instrumental Variables: Problems Methods of Economic Investigation Lecture 16.
Random Regressors and Moment Based Estimation Prepared by Vera Tabakova, East Carolina University.
Matching Estimators Methods of Economic Investigation Lecture 11.
Producing Data 1.
Properties of OLS How Reliable is OLS?. Learning Objectives 1.Review of the idea that the OLS estimator is a random variable 2.How do we judge the quality.
Non-Experimental Evaluations Methods of Economic Investigation Lecture 5 1.
Propensity Score Matching for Causal Inference: Possibilities, Limitations, and an Example sean f. reardon MAPSS colloquium March 6, 2007.
The Choice Between Fixed and Random Effects Models: Some Considerations For Educational Research Clarke, Crawford, Steele and Vignoles and funding from.
Application 2: Minnesota Domestic Violence Experiment Methods of Economic Investigation Lecture 6.
Generalizing Observational Study Results Applying Propensity Score Methods to Complex Surveys Megan Schuler Eva DuGoff Elizabeth Stuart National Conference.
Applying impact evaluation tools A hypothetical fertilizer project.
Non-experimental methods Markus Goldstein The World Bank DECRG & AFTPM.
Instrumental Variables: Introduction Methods of Economic Investigation Lecture 14.
Reliability and Validity Threats to Internal Validity Da Lee Caryl, Fall 2006.
Christel M. J. Vermeersch November 2006 Session V Instrumental Variables.
Lorraine Dearden Director of ADMIN Node Institute of Education
Using Propensity Score Matching in Observational Services Research Neal Wallace, Ph.D. Portland State University February
Randomized Assignment Difference-in-Differences
Regression Analysis1. 2 INTRODUCTION TO EMPIRICAL MODELS LEAST SQUARES ESTIMATION OF THE PARAMETERS PROPERTIES OF THE LEAST SQUARES ESTIMATORS AND ESTIMATION.
Bilal Siddiqi Istanbul, May 12, 2015 Measuring Impact: Non-Experimental Methods.
2. Main Test Theories: The Classical Test Theory (CTT) Psychometrics. 2011/12. Group A (English)
Chapter 14 Introduction to Regression Analysis. Objectives Regression Analysis Uses of Regression Analysis Method of Least Squares Difference between.
Lesson 2 Main Test Theories: The Classical Test Theory (CTT)
The Evaluation Problem Alexander Spermann, University of Freiburg 1 The Fundamental Evaluation Problem and its Solution SS 2009.
Alexander Spermann University of Freiburg, SS 2008 Matching and DiD 1 Overview of non- experimental approaches: Matching and Difference in Difference Estimators.
Experimental Evaluations Methods of Economic Investigation Lecture 4.
ENDOGENEITY - SIMULTANEITY Development Workshop. What is endogeneity and why we do not like it? [REPETITION] Three causes: – X influences Y, but Y reinforces.
Copyright © 2015 Inter-American Development Bank. This work is licensed under a Creative Commons IGO 3.0 Attribution-Non Commercial-No Derivatives (CC-IGO.
Looking for statistical twins
Lecture 6 Feb. 2, 2015 ANNOUNCEMENT: Lab session will go from 4:20-5:20 based on the poll. (The majority indicated that it would not be a problem to chance,
Constructing Propensity score weighted and matched Samples Stacey L
Basic Estimation Techniques
Basic Estimation Techniques
Impact evaluation: The quantitative methods with applications
Matching Methods & Propensity Scores
Matching Methods & Propensity Scores
Chapter 6: MULTIPLE REGRESSION ANALYSIS
Methods of Economic Investigation Lecture 12
Matching Methods & Propensity Scores
Identification: Instrumental Variables
Evaluating Impacts: An Overview of Quantitative Methods
Analysing RWE for HTA: Challenges, methods and critique
Linear Panel Data Models
Positive analysis in public finance
Instrumental Variables Estimation and Two Stage Least Squares
Violations of Assumptions In Least Squares Regression
Violations of Assumptions In Least Squares Regression
Advanced Tools and Techniques of Program Evaluation
Presentation transcript:

REGRESSION, IV, MATCHING Treatment effect Boualem RABTA Center for World Food Studies (SOW-VU) Vrije Universiteit - Amsterdam

The problem Consider a dependent variable y, a treatment variable u and a vector of characteristics x (observables). The issue is to estimate correctly how for given observed x, variation in u impacts y. Examples: - Effect of school feeding programs (On: Enrollment, school attendance, School dropout, Child Nutritional Status, Learning..) - Impact of policy response to poverty patterns. - Failing to properly estimate the treatment effect could result in the costly implementation of ineffective programs or faulty public policies that block the development of real solutions for the problem that was meant to be addressed.

The main difficulty is due to the fact that variations in unobserved factors q, to the extent they correlate with variation in u might explain the effect on y, rather than variation in u itself. The main challenge is to disentangle the effect of the treatment u from that of unobserved q. In experimental studies : design the variation in u so as to break any correlation with q (e.g., randomization). In non-experimental studies : take u and q as given but try to account for their relation explicitly, on an a priori basis. Difficulties

OLS regression Consider the simple model The OLS estimator for is Under some assumptions OLS estimator is BLUE. BUT Omitted variable bias: OLS regression should include all influential variables. If we miss out an important variable it not only means our model is poorly specified it also means that any estimated parameters are likely to be biased. Systematic errors : Bias can also occur from error in measurements Correlation X with the errors: If X is correlated with the errors then the OLS estimator will be biased. … etc.

IV regression If X is correlated with the errors then the OLS estimator will be biased. We still can obtain an unbiased estimator by using instrumental variables (IV). The idea is to find a variable Z that is highly correlated with X (relevance) and independent of the errors (validity). (2 stages least squares) Problems with IV regression - How to find good instruments? - How many instruments?

Matching The method compares the outcomes of treated individuals with those of matched non-treated, where matches are chosen on the basis of similarity in observed characteristics. (counterfactual)

u is a binary variable, i.e., u = 1 if the subject has received the treatment, 0 otherwise. Let y 1 (resp. y 0 ) the outcome with (resp. without) treatment.

The treatment effect is the difference y 1 - y 0 We define the average treatment effect as ATE = E(y 1 -y 0 ) the average treatment effect for the treated ATT = E(y 1 -y 0 | T=1) and conditionally to x ATE x = E(y 1 -y 0 | x) ATT x = E(y 1 -y 0 | T=1, x)

In reality we may observe y 0 or y 1 but not both. Counterfactual: what would have happened to the treated subjects, had they not received treatment? Also, in observational studies it is not possible to randomize. We still be able to estimate ATE.

Matching The method compares the outcomes on treated with those on matched non treated, where matches are chosen on the basis of similarity in observed characteristics (x). It is motivated by the assumption that the only source of omitted variables or selection bias is the set of observed covariates. The conditional independence assumption becomes E(y i | x, u) = E(y i | x), i=0,1 which implies ATE = E(y 1 - y 0 ) = E(E(y 1 | x, T=0) - E(y 0 | x, T=1))

Other matching procedures Propensity score matching (estimate the probability of treatment) Nearest neighbor matching (compare with the most similar/nearest individual) Kernel matching (more general definition of proximity)

THANK YOU