not be available. This makes it particularly well suited to the German case where price
variation for both goods is extremely limited. Moreover, the lack of price variation is
a frequent obstacle to survey data based analyses of consumers’ behavior, not unique
to the case of alcohol and tobacco. Instrumental variables approaches, similar to
the one proposed here, might therefore serve as a promising modeling strategy if
interdependence in consumption is analyzed and if sufficient price data for estimating
a conventional demand system is not available.
Our estimation results suggest that tobacco and alcohol are consumed as com-
plements. This result rests on a positive effect from the consumption of tobacco to
consumption of alcohol that is found in the data. From a policy perspective, this
can be interpreted as follows: if the government could achieve a reduction in smoking
or in the inclination to smoke by any anti-drug policy, this would also decrease the
propensity to consume alcohol. Thus, there would be no unintended side-effects in
form of an increased (ab)use of alcohol to compensate for the reduced level of nicotine
intake. Even the reverse, i.e. a reduction in the consumption of both drugs, seems to
be the consequence. Yet, this result is statistically firm only for males. For females
this effect seems to be much smaller and might even be non-existent.
22