Skip to content
Licensed Unlicensed Requires Authentication Published by De Gruyter September 19, 2014

Exogenous Treatment and Endogenous Factors: Vanishing of Omitted Variable Bias on the Interaction Term

Olena Y. Nizalova and Irina Murtazashvili

Abstract

Whether interested in the differential impact of a particular factor in various institutional settings or in the heterogeneous effect of policy or random experiment, the empirical researcher confronts a problem if the factor of interest is correlated with an omitted variable. This paper presents the circumstances under which it is possible to arrive at a consistent estimate of the mentioned effect. We find that if the source of heterogeneity and omitted variable are jointly independent of policy or treatment, then the OLS estimate on the interaction term between the treatment and endogenous factor turns out to be consistent.


Corresponding author: Olena Y. Nizalova, PSSRU/CHSS, University of Kent, George Allen Wing, Cornwallis Building, Canterbury, Kent, Ct2 7NF, Phone: +44 1227 824966, E-mail:

Acknowledgements

This paper has benefited from valuable comments and suggestions of Jeffrey Wooldridge. We are also thankful to Tom Coupe, Soiliou Namoro, Jean-Francois Richard, Peter Schmidt for helpful comments and discussions. The authors acknowledge financial support from a UCSUR Steven Manners Research Development Award (University of Pittsburgh).

Appendix

The popular econometric textbook by Greene (2007) derives the following general result. Suppose the correct specification of the regression model for all observations stacked together is

(7)y=iγ1+Vγ2+Wγ3+ε, (7)

where i is a n×1 vector of ones. Premultiplying equation (7) by matrix M=Ii(ii)–1i′, where I is an n×n identity matrix, yields a demeaned version of the original model:

(8)y˜=V˜γ2+W˜γ3+ε˜, (8)

where Z˜ denotes mean-differenced Z for any Z. [9] Further, suppose we do not include W into our regression (7) and, therefore, estimate y˜=V˜γ2+u, where u=W˜γ3+ε˜. We make a standard assumption that E(ε*)=0. Then, we can modify the omitted variable formula from Greene (2007) to report the probability limit of γ˜2:

(9)plim(γ^2)=γ2+plim(V˜V˜)1V˜W˜γ3+plim(V˜V˜)1V˜ε˜. (9)

References

Banerjee, Abhijit, Esther Duflo, Rachel Glennerster, and Cynthia Kinnan. 2014. “The Miracle of Microfinance? Evidence from a Randomized Evaluation.” Retrieved September 1, 2014. http://econ-www.mit.edu/files/5993.Search in Google Scholar

Banerjee, Abhijit, Shawn Cole, Esther Duflo, and Leigh Linden. 2007. “Remedying Education: Evidence from Two Randomized Experiments in India.” The Quarterly Journal of Economics 122 (3), 1235–1264.Search in Google Scholar

Blank, Rebecca M. 1991. “The Effects of Double-Blind versus Single-Blind Reviewing: Experimental Evidence From The American Economic Review.” American Economic Review 81 (5): 1041–1067.Search in Google Scholar

Blau, Francine D., Janet M. Currie, Rachel T. A. Croson, and Donna K. Ginther. 2010. “Can Mentoring Help Female Assistant Professors? Interim Results from a Randomized Trial.” American Economic Review 100 (2): 348–352.10.1257/aer.100.2.348Search in Google Scholar

Ferber, Marianne A., and Michelle Teiman. 1980. “Are Women Economists at a Disadvantage in Publishing Journal Articles?” Eastern Economic Journal 6(3–4): 189–194.Search in Google Scholar

Glewwe, Paul, Michael Kremer, and Sylvie Moulin. 2009. “Many Children Left Behind? Textbooks and Test Scores in Kenya.” American Economic Journal: Applied Economics 1 (1): 112–135.10.1257/app.1.1.112Search in Google Scholar

Greene, William H. 2007. Econometric Analysis (6th ed.). Upper Saddle River, NJ: Prentice Hall.Search in Google Scholar

Published Online: 2014-9-19
Published in Print: 2016-1-1

©2016 by De Gruyter