professional class 1 are the most likely to report drinking on most days. Once again
there is a significant interaction between cohort and social class such that, for the
1970 cohort, social class is less strongly associated with drinking frequency. A further
model was estimated (but not shown here) including the three way interaction
between gender, social class and cohort, but this higher-order interaction was not
found to be significant. In summary, these results suggest that drinking behaviour is
becoming less socially differentiated, so that for the 1970 cohort gender and social
class are less strongly associated with frequency of drinking than for the 1958 cohort.
Heavy drinking
Using the definition of ‘heavy drinking’ defined above, i.e. more than 21 units of
alcohol per week for men and more than 14 units of alcohol per week for women, a
further set of logistic regression models were estimated to investigate the links
between cohort, gender, social class and heavy drinking. As can be seen from Table
4b below, women are less likely to be heavy drinkers than men, (even when a lower
weekly threshold is used for women). There is relatively little association between
social class and heavy drinking although those with no social class assigned are
significantly less likely to be heavy drinkers than other groups. In contrast to the model
reported above, cohort is not significantly linked to heavy drinking (p>0.3). However,
there is a significant interaction between gender and cohort such that women in the
1970 cohort have an elevated probability of being heavy drinkers. For comparison
with the previous model, the interaction between cohort and social class is also
included in the table below, however this was also not found to be significant.
16