schools, and even motorists. Analyzing the effect of such bans is of great interest and
research has been carried out for other countries, where bans have already become
implemented; see for example Wasserman et al. (1991), Chaloupka & Grossman (1996)
and Tauras & Chaloupka (1999). These studies come to the conclusion that smoking
restrictions in fact do reduce tobacco consumption. In this paper, we condition on
the effectiveness of anti-drug policies and address the question of whether a reduced
tolerance against smoking will ultimately lead to less overall drug usage or whether it
may encourage smokers to turn to drugs other than tobacco.
From the perspective of economic theory, in order to answer this question one has
to determine whether tobacco and other drugs are consumed as substitutes or comple-
ment goods. Correspondingly, our analysis addresses this question using econometric
techniques. In doing this, we focus on the interdependence of tobacco and alcohol,
namely because these two substances are by far the most commonly consumed drugs
in Germany, and therefore are likely to be the most relevant ones. Moreover, data
concerning the consumption of alcohol and tobacco are available through several sur-
veys and are probably more reliable than data concerning the consumption of illicit
psychoactive substances.
2 Literature
The question of whether two goods are consumed as complements or substitutes lies
within the core of micro-economic consumer theory. The sign of the cross-price deriva-
tives derived from Hicksian demand functions precisely answers this question. Corre-
spondingly, the vast majority of econometric analyses that has so far addressed the
question of substitutability with respect to tobacco and alcohol is based on estimating
demand functions and calculating cross-price effects from estimated price-coefficients.3
Yet, the majority of empirical studies fail to distinguish Hicksian from Marshallian de-
mand properly.
Jones (1989), Florkowski & McNamara (1992) and Goel & Morey (1995) rely on
aggregate data at state or national level. Several more recent studies use survey data
at individual consumer level from several different countries; e.g. Jimenez & Labeaga
(1994), Decker & Schwarz (2000), Cameron & Williams (2001), Zhao & Harris (2004),
3 As an extension to basic consumer theory, lagged endogenous variables are often additionally
included as regressors in order to account for the addictive nature of both drugs and to test for
Becker & Murphy’s (1988) rational addiction hypothesis.