Working Paper Cnr-Ceris, N° 04/2014

Identification and Estimation of Treatment Effects in the Presence of Neighbourhood Interactions

Giovanni Cerulli
CNR - National Research Council of Italy
CERIS - Institute for Economic Research on Firm and Growth
Via dei Taurini 19, 00185 Roma, ITALY
Mail: Indirizzo e-mail protetto dal bots spam , deve abilitare Javascript per vederlo
Tel.: 06-49937867

ABSTRACT: This paper presents a parametric counter-factual model identifying Average Treatment Effects (ATEs) by Conditional Mean Independence when externality (or neighbourhood) effects are incorporated within the traditional Rubin’s potential outcome model.

As such, it tries to generalize the usual control-function regression, widely used in program evaluation and epidemiology, when SUTVA (i.e. Stable Unit Treatment Value Assumption) is relaxed. As by-product, the paper presents also ntreatreg, an author-written Stata routine for estimating ATEs when social interaction may be present. Finally, an instructional application of the model and of its Stata implementation through two examples (the first on the effect of housing location on crime; the second on the effect of education on fertility), are showed and results compared with a no-interaction setting.

Keywords: ATEs, Rubin’s causal model, SUTVA, neighbourhood effects, Stata command.

JEL Codes: C21, C31, C87

-------

An early version of this paper was presented at CEMMAP (Centre for Microdata Methods and Practice), University College London, on March 27th 2013. The author wishes to thank all the participants to the seminar and in particular Richard Blundell, Andrew Chesher, Charles Manski, Adam Rosen and Barbara Sianesi for the useful discussion. This version of the paper has been presented at the Department of Economics, Boston College, on November 12th 2013.

The author wishes to thank all the participants to the seminar and in particular Kit Baum, Andrew Beauchamp, Rossella Calvi, Federico Mantovanelli, Scott Fulford and Mathis Wagner for their participation and suggestions.