Measuring Exposure to Direct-to-Consumer Advertising: A Validation Study in the Context of Cancer-Related Treatment Advertising

Abstract

This research examines two recurrent conceptual issues of measuring media exposure in survey research—content specificity of survey items and whether exemplars should be provided to aid recall. In two population-based surveys, we evaluated three candidate measures of cancer patients’ self-reported exposure to cancer-related direct-to-consumer advertising (CR-DTCA); these measures varied in content specificity and provision of ad exemplars. All three measures performed equally well in terms of internal consistency, convergent, nomological, and discriminant validity. Increased content specificity or ad exemplars did not improve performance of the exposure measures. Participants were able to extrapolate from ad exemplars to report their exposure to broad categories of CR-DTCA. The briefest of the three measures posed the lowest level of survey costs and was deployed successfully for mailed and internet-based survey administration. We discussed future directions for application of these findings in DTCA research for other illness and for media exposure research more generally.

Notes

¹ This study regarded the survey responses as interval instead of ordinal because earlier scholarship has repeatedly noted that treating ordinal level data as interval is common in the communication literature and is not a serious violation based on Monte Carlo studies (Asher, 1983; Baker, Hardyck, & Petrinovich, 1966; Hayes, 2005). Of course, they are not arguing one should therefore routinely ignore the conceptual distinctions between interval and ordinal data and implications on analytic methods. To assess if the substantive conclusions would differ dramatically, we reanalyzed the data using non-parametric procedures appropriate for ordinal variables (Spearman rank correlation) and found that there was no difference in the conclusions about the internal consistency of Measure I items in Study 1 and Measure I – III items in Study 2 described below. We therefore present only the results treating the measures as interval.

²We did not report Cronbach's alpha for these nomological criterion variables because the items comprising each index are conceptualized as “causal or formative indicators” rather than “effect indicators” of the underlying concept (Bollen & Lennox, 1991). This is consistent with earlier research that described indices for health media exposure (Lee, 2009; Romantan et al.,2008), scanning about cancer treatment (Lewis, Gray, Freres, & Hornik, 2009), and discussion with doctors (Lewis et al.). The Cronbach's alpha statistic would not be appropriate for these indices (Streiner, 2003). We have instead examined the stability of these indices over time by analyzing the test-retest correlations.

³ To obtain sufficiently reliable estimates of means within each ad, we randomly assigned two-thirds of the study population to receive Measure II.

⁴ Only data of the overall survey duration was available; the response time at the item level was not available. By design, respondents receiving ad exemplars will take longer to complete the surveys because they had to view the entire duration of video ads (about 30 seconds each) and print ads (at least 10 seconds). All the respondents received Measure I items; therefore, these response times would only be appropriate for comparing Measure II items with Measure III items. It was not possible to compare response times between these items and Measure I.