the correlation between the propensity to drink and parental smoking behavior is
positive, we find a significantly, negative correlation between the propensity to smoke
and parental drinking habits.
With regard to our control variables the results of the reduced forms exhibit a
trend of a decreasing inclination to smoke and drink over time as well as a lower
propensity to consume tobacco and alcohol for women compared to men. Moreover,
results indicate a significant negative correlation of the propensity to drink or smoke
with having grown up with at least one parent compared to individuals having grown
up with other persons. Yet, we find a significant positive (but diminishing) correlation
with age. Parental education has a significantly negative effect on the propensity to
smoke and the number of children at parents home as well as the parental marital
status are significant only for the inclination to smoke as well.
When including other, potentially endogenous variables, we find no significant cor-
relation of the inclination to drink and smoke with income; see Table 7 in Appendix B
for reduced form estimation results. Yet, the own education level is significantly and
negatively correlated as well as being married compared to being single. With regard
to the working status results exhibit that unemployed, full time workers as well as per-
sons doing military or community service have a larger propensity to drink and smoke
than individuals not participating at the labor market whereas pupils are less inclined.
Individuals working part-time or being in vocational training show a higher propen-
sity to smoke than non-participating individuals. Moreover, the higher the number of
children the higher is the inclination to smoke. Again, in our final specification we
omit these variables because the causal direction is not clear a priori.
5.2 Structural Model Results
Table 2 reports the results for the structural equations (1) and (2). Regarding γa
and γc, the parameters of primary interest, estimates exhibit that (i) smoking signifi-
cantly increases the propensity to drink but, in contrast, that (ii) drinking significantly
decreases the propensity to smoke. In conclusion, the first result argues in favor of
complementarity between smoking and drinking while the latter seems to indicate that
drinking and smoking were substitutes. The latter, apparently, mirrors the reduced
form result that exhibits a negative correlation between the propensity to smoke and
parental drinking habits.
16