Ever since its initial formulation from the Michigan School, party identification has been converted into a cornerstone for voting behavior research. A voluminous literature over the last decades has delved into the origins, meaning and theoretical implications of this construct. Part of this debate has dealt with the extent to which this concept, originally employed to explain political preferences and voting behavior in the American electorate, can be applied in European multi-party systems. The main problem of this research is that it has failed to produce a functional measure of this concept in European settings. This lack of measure has resulted either in the use of survey items that resemble the American measure, which however is not adequate for a multi-party context, or in the use of ‘closeness’ measures, which approximate to ideological proximity. However, party identification is not just the mirror image of ideology and this distinction should manifest itself in the chosen measure. The use of such problematic measures has resulted in findings that have generated great deal of criticism about whether this concept is actually applicable in European political systems. A typical example in this respect is the often-made argument that party identification appears to be so susceptible to short-term influences that it ceases to denote an enduring attachment to a political party or even a running tally of prior evaluations. In this paper, we present a measure that overcomes these problems and can be readily used with existing observational data for the measurement of party identification in Europe. This measure, which builds on a series of preference questions about all salient parties in a given political context, performs remarkably well in terms of reliability and face, convergent and construct validity. Regarding the first, our measure, which we call the Mean Party Differential (MPD), passes all test-retest tests, indicating that when measurement error is reduced, the construct denotes substantial over time continuity. With regard to the second, we show how MPD helps to isolate party identification from ideological orientations. Coming to convergent validity, our measure correlates well with all previous measures of party identification as well as with measures of vote choice continuity. With respect the construct validity, it predicts better than previous measures turnout and habitual vote choice and can also identify those people who have decided what to vote early before the campaign. These are all outcomes that are known to be related to this construct. Importantly, MPD builds on survey items that can be found in various European electoral studies, permitting its use in both longitudinal and comparative analyses of voting behavior.
