Challenges in comparing cross-country responses in voting advice applications.

The last years saw the release of several cross-country data sets generated by Voting Advice Applications (VAAs) - websites that "match" a respondent's position on various political attitudes with those of political parties or candidates. As they hold the information of a high number of respondents on a large number of political attitudes, these datasets are promising sources of data for political and social scientists. Yet, as with all cross-country data, we should first establish if they meet the requirements of good data quality and construct equivalence. Using data from the 2019 euandi VAA, I employ Multiple Correspondence Analysis (MCA) to study the behavior of 6 items that form an economic left-right scale. Doing so for 28 countries, I find that not only the quality of the data differs between countries, but the underlying structure as well. This means that for some countries the data is of too low a quality and thus not useful for cross-country analysis. While this does not invalidate VAA data in general, it does limit it. In addition, this paper illustrates the use of MCA as a pre-processing technique for establishing the quality of both VAA data and cross-country data in general. [ABSTRACT FROM AUTHOR]