Demand for Marijuana, Cocaine and Heroin: A Multivariate Probit Approach

This paper investigates factors affecting the demand for marijuana, cocaine and heroin in Australia using micro-unit data from a national survey. Accounting for cross-commodity correlation potentially induced by unobserved personal characteristics such as tastes and addictive personalities, we estimate a trivariate probit model where the participation decisions for all drugs are jointly modelled as a system with correlated error terms. The estimated correlation coefficients are high across all three drugs. The multivariate approach allows us to predict joint and conditional probabilities, unavailable from univariate models. Such results offer extra insights in designing educational programs and drug policies within a multi-drug framework.