The Becker-Murphy model of addiction effect is tested using a classic cross section household expenditure surveys forming panel and pseudo-panel data instead of usually used specific surveys. Special instrumentations, using age cohorts, are proposed and compared with classic ones. Unlike the traditional price – income instrumentations the proposed one gives in most cases results conforming to the theoretical predictions, showing the existence of addiction, and giving the realistic estimates of the inter-temporal rate of substitution. Moreover, the estimated rates of time preference for tobacco and alcohol are close to each other both when estimated separately or in a system.