To take advantage of the information provided by these different measures, we created a composite variable of heavy drinking based on all three indicators (frequency of drinking 5+ in a day; frequency of subjective drunkenness; and the maximum amount, in standard drink equivalents, consumed in a single day, all within the last 12 months). Principle-axis factor analysis was conducted using our sample of current drinkers, that is, those who reported drinking at least one drink in the last 12 months, to derive a single factor accounting for 53% of the common variance, and which had factor loadings of 0.90, 0.60, 0.65 for number of 5+ days, frequency of drunkenness, and maximum daily volume, respectively. The composite measure was the factor score, initially categorized into four levels to capture substantive differences in heavy drinking, and also reflecting the skewed distribution of heavy drinking in the general population (Kerr and Greenfield, 2007): no or very low heavy drinking (40% of the sample), low (30%), moderate (20%), and high (10%). Table 1 shows how these four levels of the composite measure differ with
