SIM RILE KFRILE LRILE PLR FKLR EELR K J
SIM RILE KFRILE LRILE PLR FKLR EELR K J
0.00 0.25 0.50 0.75 1.00
Figure 6.5: Proportion of Times Predicting Together, Extended Model. The figure shows how often the indices predict the same government as a proportion of all the governments which were predicted by at least one of the indices in the pair.
study.
at this point – what does this contrast between the index of similarity and ideology measures mean and what could be a plausible explanation in this context? On the one hand, one can think of purely mechanical explanations. The end result that we want out of these measures is the same – an estimate of how different any two parties are from each other in terms of their political profiles. All of the measures were based on the manifesto data set. If the beginning is the same and the end result different, then something must happen along the way. For the index of similarity the way is rather short – we get an estimate of difference straight from the data. For the ideology measures, however, we need to downscale the data to a single dimension and we need to make certain assumptions about this dimension and this down-scaling. It is therefore not surprising that something important might get lost on the way. Differences that are there in the original data and that are relevant for party interaction in coalition formation seem to be captured better by the index of similarity, which is much simpler than any of the ideology measures. Occam’s Razor is rather sharp here – the simplest measure is the best.
On the other hand, coming back to the distinction between programmatic heterogeneity and ideological polarisation mentioned above (Franzmann 2011) (see section 2.2.3), one might also suspect that the index of similarity, which strictly focuses on programmatic differences, and the measures of ideological position, which focus on certain abstractions of these differences, actually measure slightly different things and are thus suitable for different purposes – the index of similarity gives us just the difference between two parties. A measure of ideological position, however, aims to tell us also something substantive about the locations of parties separately. Franzmann (ibid.) locates the first measure more on the supply side of politics and the latter more on the demand side. One could expect that manifestos matter more to parties themselves and ideological dimensions, which are used to represent and communicate polarisation, matter more in the domain of the electorate or the society at large. The latter refer to the arena, where parties articulate social divisions and create and communicate heuristics, which would enable their voters and supporters to easily identify them in a certain region of a political space. The complete political profiles of parties do not enter this picture as they are, but only as imperfect summaries. In this context it is also not surprising that the index of similarity, which captures programmatic profiles in more detail, performs better where these profiles can matter more – where parties have to find as much overlap in their issue positions as possible in order to come up with a commonly agreed governing agenda.
CEUeTDCollection
Chapter 7
Comparing Measures for Change in the Political Profiles of Parties
If parties and voters did not change their positions between elections in reaction to events and government performance, there would be little reason for more than one election.
– Russell Dalton and Ian McAllister,Random Walk or Planned Excursion? Continuity and Change in the Left-Right Positions of Political Parties
It does not matter whether we confine ourselves to idealized constructs of party behaviour or if we look at how actual parties evolve from one election to the next, there is one thing we would agree on in any case – that the political profiles of parties should and do change over time. However, if we only focus on what we empirically know about party policy change, then we rather see a lack of consensus.
As political scientists we should all hope that there is regularity in observable political change, but the empirical research we have until now is hesitant to allow such a conclusion. We can suspect that political change might be related to parties’ changing electoral fortunes, changes in party leadership, rival parties, public opinion or wider social conditions. And indeed there is research, which shows all of these connections, but there is also an equal amount of research that does not show them or that conditions them on other factors. At best there is consensus over what phenomena should be related to programmatic change, but there is much less consensus on what actually is and how.
The only thing that all research on political change almost without exception agrees on is that the data that is to be used is the manifesto data set and that the measure that is to be used is the RILE left-right index. This is the case for the overwhelming majority of such studies over the last two decades and there are only a few exceptions where a different data source or a different measure derived from the manifesto data set have been implemented. Although there are good reasons for why this data set is the best source of data for the analysis of party change (see Meyer 2013, pp.
31-CEUeTDCollection
35), there are few excuses for why the RILE index should be the only option. This fact alone makes a comparison of different measures of party left-right position that are based on the manifesto data set and the index of similarity an insightful endeavour in this particular context. Especially keeping in mind that all of these measures are supposed to be improvements over the RILE index. Tradition is not a very good excuse for the continuing use of a measure that has been extensively criticised and for which a broad range of alternatives have been proposed.
The structure of the current chapter will follow the basic template of the ones that preceded.
After a review of the state of research into party policy change and its caveats, a benchmark model that can be used for the comparison of the indices is outlined. Like the models for coalition formation or party system polarisation, it does not include all the nuances and intricacies that have been the focus of previous research, as this would not be feasible in the context of a single study with an alltogether different purpose. Instead, it will include the most important variables and associations that have been suggested. This general model is used to compare the measures for party change, focussing on both the overall fit of the model, which tells us how good the description of reality provided by that model is compared to other models, and the associations that are indicated, as this will allow us to compare the models to previous research and to each other substantively.