24 , 2016-05 , Graduate School of Economics and Osaka School of International Public Policy (OSIPP) Osaka University
This paper analyzes the association between time preference and smoking behavior whilecontrolling the e_ect of past smoking behavior using a dynamic panel analysis, and I compare the results in a static model. I use nationwide micro panel data from Japan; the data include time preference, demographic variables, and smoking status. The results of the estimation are as follows: (i) As in the previous literature, without controlling for the e_ect of past smoking behavior, I find a statistically significant association between time preference and smoking behavior. (ii) Using a dynamic panel analysis, I find no association between time preferenceand smoking behavior; however, the result changes when I use another definition for a current smoker. Additionally, I find an e_ect of past smoking behavior on current smoking behavior. (iii) The association between time preference and smoking behavior significantly decreases when the e_ect of past smoking behavior is controlled. The result is robust for a change in thedefinition of current smoker, addressing the di_erence of unobserved individual heterogeneity between those who never smoked and former smokers, the di_erent e_ect of persistence according to the level of cigarette consumption, the sample selection problem, and the use of a quasi-hyperbolic model.