Aims and design: In order to assess the effects of survey modality on alcohol consumption estimates, data from two surveys using different interview modes (face-to-face and telephone) were compared on several alcohol measures.
Setting and participants: Face-to-face survey data were drawn from the 1990 National Alcohol Survey, while the telephone data came from the 1990 Warning Labels Survey. Both surveys used a probability sampling of the US adult general population in the 48 contiguous states.
Measurements: Measures of alcohol use derived from an identical "graduated frequencies" series included estimates of any drinking in the past 12 months, overall volume, and heavy (5+) drinking days.
Findings: Abstention rates did not differ by survey mode, nor did distributions of alcohol consumption by volume and reported frequency of drinking five or more drinks in a day. Multiple regression models including demographic-mode interaction terms were used to examine how mode effects might differ across demographic subgroups. Lower income respondents were under-represented in the telephone sample, and were associated with lower reports of volume and 5+ days, compared to respondents in the face-to-face mode.
Conclusions: The results suggest that although there are few differences in alcohol consumption estimates by interview mode, telephone samples may need to be supplemented or estimates adjusted by income level in order to attain equivalent results.