Venue: The Fuqua School of Business, Duke University, 1 Towerview Drive, Durham, NC 27708-0120
Presentation
That Instrument is Lousy! In Search of Agreement When Using Instrumental Variables Estimation in Substance Use Research
Economists have been studying the consequences (e.g., health problems, lower earnings, criminal activity) of using addictive substances (e.g., tobacco, alcohol, prescription drugs, illegal drugs) for many years. The primary statistical issue that must be addressed in this research is the potential endogeneity of the key regressor, substance use. Explanations abound for why substance use is likely to be endogenous in this context, but the end result is biased coefficient estimates if the models are estimated with standard single-equation methods such as OLS, probit, or negative binomial. Instrumental variables (IV) estimation is a powerful technique that can be used to overcome the endogeneity problem and derive consistent estimates. While economists are fairly unified on the need to address potential endogeneity bias and the value of IV estimation, the choice of intuitively sound and statistically reliable instruments is hotly debated during conference presentations and more subtly via journal referee reports. Because selecting valid (uncorrelated with the residual) and reliable (strongly correlated with the endogenous variable) instruments is challenging as well as critical for consistency, the present paper seeks to highlight the diversity of opinion about what constitutes credible instruments. A comprehensive review of the economics literature since 1990 identified over 40 papers that used IV techniques to estimate the effects or consequences of tobacco, alcohol, or other addictive drugs. Dividing the studies into three groups based on type of substance, we present a series of tables showing the range of instruments in each group and their relative popularity. While some instruments are fairly common across studies (e.g., beer taxes, cigarette taxes, marijuana decriminalization policies), the full list of published instruments is long and sometimes perplexing. This review article underscores the diversity of opinion among authors and referees concerning the choice and validity of instrumental variables, and highlights the varying degrees of rigor with which instrument reliability and validity have been assessed in the literature.