Jo E. Stryker

Date of Award


Degree Type


Degree Name

Doctor of Philosophy (PhD)



First Advisor

Robert Hornik


Two primary research hypotheses were tested concerning aggregate effects of news media on aggregated health behaviors over time for four health behaviors: marijuana use, seatbelt use, beef consumption, fruit consumption. Several measures of seatbelt use and fruit consumption were used. The first primary hypothesis sought to establish any evidence of news media impact on behavior, and tested for effects using two different operationalizations of media coverage. The first operationalization distinguished between PRO and CON coverage. PRO coverage consisted of stories emphasizing positive aspects of performing the healthy behavior, while CON coverage consisted of stories emphasizing negative aspects of performing the healthy behavior. The second operationalization measured any media stories containing references to performing the behavior (the general behavioral media measure, or GBM). The second hypothesis proposed that media messages emphasizing the positive (PRO) and negative (CON) aspects of performing the healthy behavior would be more strongly associated with behavior change than would the more general behavioral media coverage measure (GBM) (Hypothesis 2A). It was further proposed that if there were very low levels of CON media, the PRO measure should still offer greater prediction than the general measure (Hypothesis 2B). Two methods, distributed lagged regression analysis and ideodynamic models, were used to test hypotheses. iii In sum, there was substantial support for Research Hypothesis 1, that trends in media coverage could explain a significant portion of the variation in trends in behavioral outcomes. Considering any measure of media coverage, any measure of behavior, and any method of analysis, there was at least one significant media/behavior association for each behavior. The conviction with which claims of causal inference could be made was varied. There was less convincing evidence supporting the second set of research hypotheses, that PRO/CON (or PRO only in the absence of CON) coverage would better predict behavior change than the GBM measure. These hypotheses could only be considered if there was any evidence of an association between media coverage and the behavioral measure. Of the five significant media/behavior relationships, four of them provided support (in varying degrees) for the superiority of the more refined media measure(s).