Estimating addicts' price response of heroin: A panel data approach based on re-interviewed sample

Drug abuse inflicts considerable harm on users, non-using persons and on society, and a variety of means to curtail consumption of illegal drugs have been adopted. The consumption of drugs differ from consumption of most other goods in that it involves addiction. Inter-temporal models are needed to encompass this aspect. The data for this study have been collected through interviews with heroin injectors attending a needle exchange service in Oslo, and the respondents have undergone a second interview about one year after the first. Four regression models will be considered: two are static panel data models and two are cross-section models with lagged or leaded drug consumption as additional regressors. Each model comes in two versions, one for non-dealers and one for dealers of heroin. Despite our relatively small sample, we obtain negative and statistically significant price responses and positive and significant income responses for nearly all the models and specifications applied. The results from the two classes of models reflect the same picture although the absolute values of the elasticities vary. For the price elasticity, dealers obtain values in the range of [-0.25, -1.55] and non-dealers in the range of [- 0.72, -1.83]. Somewhat surprisingly, we obtain low estimates for the habit component in the panel data model, but higher for non-dealers than for dealers.