Objective: To assess the effect of mode of administration in alcohol survey
s (telephone vs face-to-face interviews), prevalence rates of self-reported
harms due to alcohol were compared for two datasets with equivalent measur
es. Method: Two national alcohol surveys were used: the 1990 Warning Labels
Survey, in which random digit dialing was used to generate a sample of 2,0
00 adults interviewed by telephone, and the 1990 National Alcohol Survey (f
ace-to-face interviews), a probability sample of U.S. adults living in hous
eholds (N = 2,058). Both surveys included identical items on five areas of
alcohol-related harm, yielding one composite index of any harm reported in
the last 12 months that was compared between the two surveys for current dr
inkers. Results. After controlling for demographic characteristics and alco
hol use, the telephone survey yielded significantly higher rates of alcohol
-related health harm, work harm and "any harm" as compared to the in-person
survey. The interaction between heavier drinking (five or more drinks duri
ng 1 day, weekly or more often) and method of data collection was significa
nt for health harm and any harm. Respondents in the telephone survey who dr
ank 5+ less than weekly were more likely than those interviewed in person t
o report health harm due to alcohol use; those in the telephone survey who
drank 5+ weekly or more often were more likely to report any harm. Conclusi
ons: Possible explanations for differences between the surveys include anon
ymity and fewer social desirability issues associated with telephone survey
s, as well as potentially differing cognitive requirements in telephone ver
sus face-to-face interviews.