Developing Valid Knowledge Scales

central variables in research on mass political behavior, but insufficient attention has been given to the implications of alternate approaches to the measurement of political knowledge. Two issues are considered here: (1) survey protocol regarding don't responses and (2) item format. Data from the 1998 NES Pilot and a survey conducted in the Tallahassee metropolitan area are used to assess alternate survey protocols. or decades, students of mass politics have assessed how much citizens know about politics, how people acquire and process political information, whether political behavior varies as a function of differences in knowledge levels, and whether various psychological mechanisms enable deficiencies in political knowledge to be overcome. Virtually all research on public opinion, voting behavior, and media effects confronts one or more of these questions. Because of the importance of knowledge as an analytical construct, several authors have explored issues in the measurement of political knowledge. This research has produced compelling evidence that political awareness is best represented with data from survey batteries that measure factual political knowledge. Less insight has emerged, however, regarding the implications of how those knowledge batteries are constructed. Therefore, my purposes in this article are to identify some of the key issues in the construction of knowledge batteries and to demonstrate the consequences associated with use of alternate measurement procedures. Two issues in the measurement of political knowledge are addressed: (1) survey protocol regarding don't responses and (2) item format. For both, possible changes in measurement procedures are discussed. Following this discussion, several potential liabilities associated with use of new procedures are considered. In the last section of this article, I demonstrate how the choice of knowledge measures affects hypothesis testing. Throughout this article, alternate measurement procedures are evaluated using data from two split-ballot surveys, the 1998 NES Pilot, and a survey conducted in late 1998 and early 1999 in the Tallahassee metropolitan area.