Abstract
Data were analyzed from a multistage probability household survey of over 600 adults, ages 18-40 from the city of Chicago conducted during 2001-2002. The survey employed audio computer-assisted self-interviews to obtain information about drug use. To investigate race/ethnicity differences in reporting validity, drug test results were compared with self-reports of past month drug use for cocaine, marijuana, and a combined indicator of both substances. The main indicators of validity were self-report sensitivity and concordance. Possible theoretical models accounting for potential cultural differences in reporting validity were discussed. Survey variables reflecting these potential explanations were examined as potential mediators of race/ethnicity differences in validity and as direct correlates of validity. Socioeconomic status was identified as one potential mediator. With this exception, race/ethnicity differences suggesting lower levels of marijuana and cocaine concordance for African Americans as compared with Whites were sustained after controlling for potential mediators. Methodological implications for epidemiological and health disparities research are discussed.