Mode Effects on cigarette smoking estimates: Comparing CAPI and CATI responders in the 2001/02 Current Population Survey

There are several aspects of survey design that may affect cigarette smoking estimates, including questionnaire introduction, question ordering, and participation rates. We analyzed the Tobacco Use Supplement of the 2001/02 Current Population Survey (TUS-CPS). The CPS is a monthly area probability labor force survey of approximately 56,000 households conducted by the U.S. Census; a Tobacco Use Supplement (TUS) has been fielded periodically since 1992. There are two main data collection modes in the CPS, Computer Assisted Personal Interview (CAPI) and Computer Assisted Telephone Interview (CATI). Our analyses are restricted to adult (i.e. 18 years +) self-respondents (n= 184,947), 34.5% of which were interviewed in-person and 65.5% via telephone.

Overall, we found a significant mode effect; those interviewed in-person were more likely to report cigarette smoking than those interviewed via telephone (23.2% vs. 19.5%; p<.01). In-person interviewees had a 1.26 times greater odds of reporting smoking compared to telephone interviewees when controlling for gender, race/ethnicity, and age (AOR=1.26; 95% confidence interval: 1.23-1.28). The magnitude of the mode effect differed by demographic characteristics; Males (4.6%), Whites (4.0%) and Black/African Americans (5.2%) exhibited the largest mode effects, whereas females, those with a college education, Hispanics and those age 65+ exhibited small or negligible effects.

In sum, our findings suggests that researchers using the TUS-CPS should recognize the effect data collection mode can have on smoking rates for different subgroups, and should consider the appropriateness of excluding or including respondents by mode when conducting research on specific populations.