The consistency of estimates obtained through central-location sampling: analysis of the Partnership for a Drug-Free...

To evaluate the reliability and validity of estimates of drug use obtained through multiple-site central-location sampling, data from the Partnership Attitude Tracking Study (PATS), collected annually beginning in 1987, are tracked longitudinally and compared with data collected through traditional household interviews. Comparisons with demographic estimates from Current Population Surveys indicate that central-location sampling can provide a broadly representative sample of the adult population. However, there is some coverage bias in central-location sampling, most notably in the underrepresentation of adults who do not have high school diplomas. Sample estimates obtained from central-location sampling are consistent over time. Basic demographic characteristics of the samples--education, income, marital status, and area of residence--vary by less than 5% across the four waves of the tracking study. Prevalence estimates of self-reported drug use demonstrate an even higher degree of consistency over time. Comparisons of PATS and National Household Survey of Drug Abuse data and the Monitoring the Future data indicate a higher level of drug use and a lower level of perceived risk of occasional use among PATS respondents. While the trends in use and perceived risk reported in all studies are similar, we suggest that perceived threats to confidentiality and anonymity often result in significant underreporting of drug use, particularly in household surveys. Sample estimate differences are attributed to sampling and measurement error. Some discrepancy in prevalence estimates is associated with the fact that the PATS sample is not a true probability sample; as a result, the sampling error of the overall study cannot be estimated precisely. It is also likely that the difference in estimates between the studies is a product of the decreased measurement error of PATS methodology. In central-location sampling, respondents are completely anonymous and may feel more comfortable in providing honest answers about illicit activities such as drug use. The potential applications of this methodology are discussed.