Aldo Paparo & Lorenzo De Sio (2017) PTV gap as a new measure of partisanship: a panel-data, multi-measure validation showing surprising partisanship stability,
Contemporary Italian Politics, 9:1, 60-83, DOI: 10.1080/23248823.2017.1289733
Aldo Paparo, interviewed by Andrea Maccagno
(English translation by Elisabetta Mannoni)
The basic concept your paper starts from is party identification. What does it mean?
CISE Interviews are aimed at spreading CISE research activities, which yield scientific publications in national and international journals. Their format, as an interview to a young CISE intern, allows to present publication contents in a simple form, overcoming difficulties of technical language and often complex statistical tools.
The idea of party identification is crucial for research on politics and political behavior. It was introduced in the volume The American Voter (by Campbell, Converse, Miller, Strokes). The basic idea is that many voters tend to identify themselves with a specific party, according to a set of features, such as membership of a particular social group or any formative experience during the period of political socialization. Such identification is, therefore, partly ascriptive: if one happens to have Democratic parents and to live her adolescence in a mainly Democratic context, one will tend to vote Democratic, on a quite regular basis. In the United States this concept is operationalized through a scale drawn up in accordance to a fundamental survey question on whether the respondent would define herself as democratic, republican or independent. Yet, the same thing cannot be easily realized in Europe, because of the multi-party systems of European countries. In the 70s an attempt was made to find a similar measure, based on the question: “Is there a party you feel more attached to?”. Here the idea was that of attachment to a certain party, but it still was kind of tricky. On this basis, we tried to develop a new measurement which could fit both the American and the European context. Indeed, our project is a transatlantic comparative one. We conducted first some research on the United States, which encouraged us in this direction, and then decided to extend our analyses to European multiparty systems.
This paper analyzes the Italian context between the end of 2012 and the beginning of 2013, thus focusing on a multi-party system in a particularly turbulent period. Why did you choose it?
Choosing a turbulent period makes the validation of the new measurement we propose more reliable. As for the Italian context, we chose it because in Italy there is a very strong multipartism, and traditionally an almost-extreme pluralism. Just think that our panel [study within which the same respondents are interviewed more than once, Ed.] considers four waves and at the beginning Scelta Civica was not born yet, coming only afterwards. Or think of the Five-Star Movement, which in the first wave was around a 4% and in the last one registered a 25%. So it is an interesting challenge to test a measurement aiming at capturing long lasting attitudes. The Italian case in the abovementioned period lets us make a particularly severe test on the new measurement adopted.
Tell us what the PTV and the PTV-gap you introduced are, and the reasons why you decided to do so.
PTVs (propensity-to-vote) are questions first introduced in the Netherlands in the 80s, asking the respondent how likely it is that she will vote for a given party in the future. The question is posed for all the parties in the system. The respondent will give different scores, generally from 0 to 10, independent of one another. Introducing this kind of measurement helped better understanding how things such as party competition, overlaps of electorates from different parties, and parties potential catchment areas work. Thus, PTVs are nothing but a measurement, for each party, of respondents’ propensity to vote for that party. Yet, to measure whether and to what extent a respondent identifies herself with a given party, there’s the need to introduce the gap, that is the difference in the scores she attributes to her favorite party (the one with the highest PTV) and to her second-best one. The PTV-gap might be 10 if one party has PTV=10 and all the others have PTV=0, or it can even be 0 when the two most favorite parties are given the same PTV score. The PTV-gap can be defined, therefore, as the difference between the PTV highest score and the PTV second-highest one. However, it’s important to understand the minimum threshold for a respondent to be considered as identified with a given party. In the paper we have just published, we tested several potential thresholds and in the end we assessed a PTV-gap = 2 as a convenient threshold.
To sum up, which hypotheses backed this research?
The research was aimed at testing the validity of the new measurement by comparing it to the party-closeness measure, traditionally used for the European multi-party systems. To validate our innovative measure, we expect the performances observed for the two different measurements to be at least similar. We compare them on two dimensions, both extremely connected to the notion of party identification: stability over time and ability to predict voting behavior.
What are the results of your research? Are they consistent with the initial expectations? Can the PTV-gap be considered as a good measure for party identification?
The results made us satisfied with the way our measurement worked. In fact, for what concerns stability over time we found that both the standard measure and ours show a higher stability compared to the vote, in line with party identification theory (according to which the vote is more volatile, as more subject to short-term dynamics). This is an interesting result, as it disproves some Dutch research dating back to the 70s, advising against the use of the party identification theory in Europe (as then vote choice was more stable than party identification). Our data show, on the contrary, and consistently with the theory, that there is much more turbulence around the vote than party identification.
With reference to stability, the two partisanship measurements show definitely similar performances: around 70% for vote; higher than 80% for the traditional measurement, and just below 80% for our new measurement. As for the ability to predict voting behavior, we found out that the two measurements are again quite similar, but ours is less affected in its dynamics by vote choice. Hence, on this aspect, our measurement works even better than the traditional one.
As a consequence, these results are extremely promising in terms of future research. The PTV-gap seems to be a potentially good measurement as for party identification and, since the PTVs can easily be used both for multiparty systems and two-party systems, like the United States – we think it may open the path to significant progress in terms of comparative research involving Europe and the United States.
Finally, the quantitative nature of the new measurement will allow the use of more sophisticated statistical tools. At this stage, our future horizon is to convalidate this measurement for other countries, too – especially the United States, for which we have already found some preliminary and encouraging results.