Validity of measured party positions on European integration: Assumptions, approaches, and a comparison of alternative measures * Leonard Ray Department of Political Science, Louisiana State University, 240 Stubbs Hall, Baton Rouge, LA 70803, USA Abstract A variety of methodologies have been employed in order to measure the positions of political parties on the issue of European Integration. This paper compares four types of measures of party position: reputational, textual, behavioral, and self reported in- dicators. These indicators are assessed in terms of their content validity, convergentvalidity, and, construct validity. All of these approaches yield plausible measures which correlate reasonably well with each other, although care must be taken with the creation of indices from the Comparative Manifesto Project dataset. However, these indicators are not interchangeable, and the results of substantive research on EU politics are dependent, to an alarming extent, on the choice of indicators. Careful attention should be paid to the choice of measures, and the theoretical justification of any measure used. Future researchers would be well advised to test whether their results are robust across alternative measures of party positions. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: European integration; Expert surveys; Measurement; Parties; Validity 1. Introduction: validity of party locations (assuming a political space) This article seeks to establish the validity of a num- ber of alternative methodologies for the measurement of party locations on the EU issue. Four main ap- proaches to the measurement of the location of polit- ical actors will be discussed. These are reputational, textual, behavioral, and self reported indicators. The validity of these indicators is assessed in terms of three of the tests of validity described by Bollen (1989); content validity, convergent validity, and con- struct validity. A review of measures of party positions on the European dimension reveals wide variation in the underlying conceptualization of the EU political space, and substantial differences in the actual actors (voters/parties/leaders) whose locations are being measured. Most of the widely used indicators correlate highly with each other, suggesting that they all capture part of the same underlying reality. However, when re- lationships with other variables are examined, these measures yield different empirical results, suggesting that any measurement error may not be random, but is systematically related to theoretically important * Upon publication, replication data and details of the analyses will be available online at: http://www.lsu.edu/faculty/lray2. E-mail address: lray2@lsu.edu 0261-3794/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.electstud.2006.03.008 Electoral Studies 26 (2007) 11e22 www.elsevier.com/locate/electstud