Adjusting survey estimates for response bias - An application to trends inalcohol and marijuana use

Using retrospective self-report data on ages at first use of alcohol and ma rijuana from nine National Household Surveys on Drug Abuse conducted betwee n 1982 and 1995, we show that estimates of alcohol and marijuana incidence (initiation) during early adolescence decline with increases in the length of the time interval between the data collection and reference periods. The consistency of this finding by gender and across eight birth cohorts, surv eyed at different ages at interview and lengths of retention, supports an i nterpretation in terms of retrospective reporting bias. We apply an exponen tial decay model to adjust the estimates for response bias and use the mode l to show how bias distorts trends in alcohol and marijuana incidence betwe en 1961 and 1990. An analysis of changes in Lifetime-incidence and age-at-f irst-use reports of birth cohorts as the cohorts age suggests that forward telescoping accounts for most underreporting of early alcohol use and that intentional concealment accounts for most underreporting of marijuana use.