Do electoral results come as a surprise because our empirical models are limited?

Recent elections in many European countries have seemed less predictable, as party systems fragment and new parties challenge the established ones on a range of issues. Looking at the recent case of Finland, Zhen Im, Hanna Wass, Heikki Hiilamo and Timo Kauppinen argue that political scientists need to develop new models for mapping multiple, changing issues for electoral competition.

Picture: Baptiste Valthier/Pexels.

The 2019 Finnish parliamentary elections on 14 April 14 threw up a big surprise. The radical right Finns’ Party (FP) came second, trailing the centre-left party – the Social Democratic Party (SDP) – by only 0.2 percentage points, thus consolidating its position as one of the most influential parties in Finland. In the two months prior to the elections SDP led FP by 10 percentage points. This result is even more surprising if we consider that FP split in June 2017, with more moderate members splintering off to form the Blue Reform party. Why did scholars and commentators once again miscalculate FP’s support? We argue that existing models underestimate the dynamic construction of political issue combinations as well as the role of more fragmented and intersecting social groups.

From previous research, we know that parties determine how different issues combine in specific elections. Although voters may hold relatively stable views on key issues over their life cycle, the relationship between them and relative importance of different issue positions may vary across elections. The correlations between issues and the campaign are determined by parties as they seek to mobilise different parts of the voting population.

In the New York Times, Barry and Lemola argued that FP’s electoral success could be explained by its position on the environment. The party successfully campaigned against climate protection policies which were promoted by all other parties during the campaign. FP opposed such policies to revive the division between the rural ‘folk’ and urban ‘elites’. It tied environmental protections to higher and unfair economic costs borne by the rural ‘folk’, and to the loss of traditional lifestyles such as driving your own car and eating meat. The party framed environmental issues along rural-urban lines.

FP’s campaign against the environment was surprising for two reasons. First, environmental protection has turned from a positional issue in the 1980s to a valence issue in recent years (that is, voters are seen to be united about the benefits of environmental protection, differing only on how important they think it is). Although parties have access to a greater range of potential issue combinations in recent times, many scholars did not anticipate that political entrepreneurs such as the FP, and to a lesser extent the Gilets Jaunes, could have reinstated environmental and climate protection as a positional issue. Second, political entrepreneurs such as FP have successfully related environmental protection to both economic and social issues. In the 1980s, environmental protection correlated primarily with social issues. FP’s recent campaign demonstrated that as political entrepreneurs they could recognise societal divisions to be exploited in an election. One manifestation of this may be the popularity of FP on the semi-rural fringes of the major urban regions, where car-based mobility is the norm.

FP’s success in voter mobilisation can also be attributed to the fact that voters are no longer exclusive members of big, homogenous and all-encompassing social groups such as social class. Instead, voters belong to multiple smaller social groups. For instance, a voter could be a highly educated, professionally qualified, middle-aged, woman, but on a part-time contract who is working in a routine occupation vulnerable to automation. These intersectionalities mean that voters may find they may have multiple and, on occasions, conflicting interests. With numerous and intersecting social groups, political parties face a choice of a substantial number of possible electorates to mobilise. Interestingly, a quick analysis on the Finnish election results shows that FP also gained support from areas with relatively high socio-economic status.

With more fragmented constituencies, parties may highlight the interests of multiple smaller group(s) which they seek to mobilise. They may do so by coupling different issues together. Seemingly distinct issues yesterday could therefore become related issues tomorrow. Environmental issues were previously framed together with civil liberty issues by the New Left and Green parties in the 1970s and 1980s. In the late 2000s and early 2010s, immigration and civil liberty issues were strung together by radical right parties. During the same period, environmental issues developed into valence ones. Turning to the recent Finnish elections, we find ourselves back to the square one – the potential emergence of a new environmental issue dimension that correlates with both economic and social issues.

From this perspective, it is less helpful to consider successive elections as stable iterations of issue combinations which only develop slowly over time. Rather, political entrepreneurs and parties ‘pick and combine’ issues according to the electorate they seek to mobilise in a specific election. If parties mobilise similar segments of society over successive elections, we should then see repeated iterations of similar issue combinations. Otherwise, parties may break from previous frames when they try to appeal to other segments of society. When this happens, we may observe sudden and ‘surprising’ new issue combinations as seen in the recent Finnish elections. Seemingly incoherent positions on uncorrelated issues today may therefore appear as coherent positions on correlated issues tomorrow.

In sum, we know that parties are responsible for structuring issue spaces through issue framing. We might, however, have underestimated how innovative parties can be in combining issues. We might therefore also be less aware of how the contemporary issue space is fluid over successive electoral contexts. As uncomfortable as it sounds, we may be eating the dust of political parties and political entrepreneurs because they recognise exploitable social divisions better than us as political scientists. The crucial challenge then is to develop more nuanced models to better understand the dynamic permutations of issue space in electoral campaigns.

For future models to fully capture fluid issue combinations across elections, they should address three concerns. First, models should be sufficiently dynamic across time. They should be able to accommodate changes in the number of issue dimensions, and changes in how important these issues are. Second, models should be open to how different issues may relate and combine with other issues. Issues that seem unrelated to experts may appear coherent to voters. Our models should thus aim to detect how voters perceive the relationship between different issues. Third, models should overcome the challenge of mapping voter and party positions in a multidimensional non-orthogonal issue space. This challenge also extends to the difficulties of estimating the interaction effects of multiple issue dimensions on party choice in standard regression models, especially when more than two issue dimensions exist.

We do not claim to have ready solutions to these empirical challenges. Instead, we suggest that answers could be derived from empirical practices in other disciplines and fields. Recent developments in dynamic factor analysis, as well as cluster analysis, such as correspondence analysis and latent class analysis, seem promising. [1]

This post represents the views of the authors and not those of Democratic Audit.

About the authors

Zhen Im is a researcher at the Department of Social Research at the University of Helsinki.

Hanna Wass is a Professor of Political Science at the University of Helsinki.

Heikki Hiilamo is Professor of Social Policy at the University of Helsinki.

Timo Kauppinen is Research Manager at the Finnish National Institute for Health and Welfare (THL).

[1] For a discussion of dynamic factor analysis, see https://www.rug.nl/research/portal/files/2887795/05F10.pdf. For an example of correspondence analysis, see Jakob Horneber. (2019) ‘Assessing Cleavage Dynamics. An analysis of the German political system between 2013 and 2017’. Paper presented at the 5th Leuven-Montréal Winter School on Elections, 3/2019, KU Leuven. For an example of latent class analysis, see Zhen Im. (2019) ‘Who votes for whom? Using a latent profile analysis approach to examining political preferences and party support’. Paper presented at the 5th Leuven-Montréal Winter School on Elections, 3/2019, KU Leuven.