TOWARD ANEWPOLITICALMETHODOLOGY: Microfoundations and ART

Analyses of the persuasive effects of media exposure outside the laboratory have generally produced negative results. I attribute such nonfindings in part to carelessness regarding the inferential consequences of measurement error and in part to limitations of research design. In an analysis of opinion change during the 1980 presidential campaign, adjusting for measurement error produces several strong media exposure effects, especially for network television news. Adjusting for measurement error also makes preexisting opinions look much more stable, suggesting that the new information absorbed via media exposure must be about three times as distinctive as has generally been supposed in order to account for observed patterns of opinion change.