This chapter explores two methods of bringing together the complementary strengths of general population surveys (GPS) and web surveys to develop improved estimates of overall quantities of drugs consumed — namely, raking and regression. Raking re-weights responses to a web survey in an attempt to make the weighted averages or totals match those of the general population. Regression is used to impute values for what GPS respondents might have said if they had been asked about the frequency of their drug use and the quantities consumed. This chapter applies these two methods to data captured by the EWSD and the countries’ respective GPS to measure total cannabis consumption in the general population. This is pertinent to the literature described in the EMCDDA (2012) study on the cannabis market in Europe and by van Laar et al. (2013), among others.
This publication is published as part of a collection of papers on web surveys: Monitoring drug use in the digital age: studies in web surveys (Insights 26).
Knowing the size of illicit drug markets is important for understanding their revenues and making informed policy decisions. While general population surveys (GPS) can provide an estimate of the number of potential users of such markets, many GPS are limited by only collecting information on the prevalence and frequency of drug use, missing detailed data on the quantity and value of drugs consumed. In contrast, web surveys are much cheaper to implement and are often constructed with a more specific focus. This chapter explores two statistical methods of bringing together the complementary strengths of GPS and web surveys to develop improved estimates of overall quantities of drugs consumed — namely, raking and regression. By doing such, it shows how the rich information collected by online surveys could potentially be extrapolated to larger populations. This chapter applies these two methods to data captured by the European Web Survey on Drugs (EWSD) and six European countries’ respective GPS to measure total cannabis consumption in the general population. While both of these methods have their own limitations, this chapter shows that they could still represent an improvement over making projections based on simple rules of thumb or having no estimates at all.