Institut for Statskundskab

What explains the dynamics of citizens’ satisfaction with democracy? An integrated framework for panel data

Publikation: Bidrag til tidsskrift/Konferencebidrag i tidsskrift /Bidrag til avisTidsskriftartikelForskningpeer review

  • Ann Kristin Kölln, Uppsala University
  • ,
  • Kees Aarts, University of Groningen

Literature on political support broadly offers three micro-level models: socio-economic status, democratic process evaluations, and political performance evaluations explain people's differences in satisfaction with democracy. While tests show that these explanations complement each other, we do not know how. We combine for the first time all three models into one common longitudinal framework by explicitly considering aspects of time. We argue that relatively stable factors, such as socio-economic status, only explain general levels, whereas more time-sensitive factors, such as evaluations, explain differences between citizens at specific points in time. The results of latent growth curve modelling applied to nine-wave panel data support our general hypothesis of a common longitudinal framework. These results also show that economic evaluations play a prominent role as do some (but not all) electoral results. The findings have theoretical and methodological implications, and they offer a new perspective on the meaning of ’satisfaction with democracy’.

OriginalsprogEngelsk
Artikelnummer102271
TidsskriftElectoral Studies
Vol/bind69
ISSN0261-3794
DOI
StatusUdgivet - feb. 2021

Bibliografisk note

Funding Information:
Kölln acknowledges funding from Aarhus University Research Foundation (ref. AUFF-E2017-7-13), Riksbankens Jubileumsfond (ref. RIK18-1267:2), and the Swedish Research Council for Health, Working Life and Welfare (ref. 2013–2692).

Funding Information:
In this paper, our contribution is twofold. First, we combine all three strands of explanation in one longitudinal framework to theorise that different explanations predict different parts of citizens' dynamics in SWD. That is, unlike previous studies, we are explicitly theorizing the relationship between the micro-level explanations and aspects of time. Second, in further contrast with earlier work, in which repeated cross-sectional survey data was employed, we use nine-wave panel data from the Netherlands with a focus on intra-individual change over a long period of time (2007?2016). This approach yields the benefit of holding many possible confounding contextual variables constant while simultaneously allowing to trace changes over time and their potential origins. With these data we will for the first time be able to study exactly how prominent theories of political system support explain the overall dynamics in citizens? SWD.Specifically, we argue that the different micro-level explanations for SWD complement each other because they represent relatively stable and more time-varying factors. Relatively stable explanatory factors, such as individuals' socio-economic status, should explain mean group differences because of their over-time stability. They allow distinguishing, for example, between those citizens who are broadly satisfied and those who are not. Other prominent explanatory factors that are more time-varying, such as citizens' process and performance evaluations, should explain short-term differences between individuals at particular points in time, simply because they fluctuate more over time. Together, time-invariant predictors should explain general levels of SWD, and more variable factors should offer an explanation for short-term fluctuations, independent of the general trend. In other words, all three micro-level models of political support interact in a longitudinal framework and account for the dynamics in SWD once ?time? as a variable is explicitly considered.The proposition is based on the idea that common predictors of SWD show different levels of over-time stability and can be easily incorporated into a common longitudinal framework that distinguishes between citizens' levels, trends, and differences at particular points in time and thus around the general trend. More precisely, since some of the predictors are more or less time-invariant, such as socio-economic status, they should explain differences in levels of SWD. For example, under normal circumstances, an individual's personal characteristics such as gender or education do not change very much over a longer period of time. Given these characteristics' relative stability over time, they should contribute to the dynamics of SWD by explaining mean differences in levels. In support of this, previous research has shown that differences in socio-economic status are systematically related to levels of SWD, yet not to trends (Aarts et al., 2014, 226; see also Norris 2011 or ?nnud?ttir and Har?arson 2011). It suggests that socio-economic status should only explain variation in general levels of SWD.Based on these theoretical considerations, we propose to distinguish between aspects of time and specifically between three components in the dynamics of (individual) SWD: differences in general levels and trends, and differences at particular points in time and thus around the trend. All three components relate to different explanatory theories of political support and they jointly explain the over-time dynamics of SWD on the individual-level. This forms our general expectation, from which we can infer a number of specific hypotheses.In the absence of available indicators measuring directly process evaluations, they are regularly approximated through indicators of trust in representative institutions (e.g. Christmann and Torcal 2017; Grimes 2006; Hobolt 2012; Torcal and Trechsel 2016). The focus on representative institutions in this approximation reflects that process evaluations are about ?the quality of governance? and ?the will of the people?. The focus on the attitude of trust is justified by the assumed causal relationship between evaluations and trust (Kaina 2008): Positive evaluations lead to higher trust. Applied to theories of political system support, trust in representative institutions should thus result in higher levels of SWD because positively evaluated representative institutions are a sign that citizens believe ?the will of the people? is respected.There is a long-standing debate on the distinctiveness of trust and SWD as well as on the causal flow between them (see for example Gr?nlund and Set?l? 2007; Weber et al., 2017). Both concepts belong to the wider set of political support measures that can be arrayed on a scale from specific to diffuse support with increasing levels of abstraction (Easton 1965; Norris 1999). While Norris (1999) considers trust to be further towards the ?specific support? end of the scale than SWD, both causal directions between them seem plausible (Weber et al., 2017). In this paper, we are not evaluating any causal claims and simply follow the existing literature predicting that process evaluations, approximated with trust in representative institutions, explain variation in SWD.1 Model fit indices (CFI, TLI, and RMSEA) show that the data fit the assumed structure well. In addition, the p-value associated with the RMSEA above 0.05 indicates a particularly close-fitting model. Most importantly, all of this supports our general hypothesis about the dynamics of SWD consisting of different components that can be explained by different factors.Turning to our specific hypotheses tests, the first column in Table 1 reports the predicted increase in the intercept for an increase of two standard deviations (because of rescaling) in levels of education and income. Educational level has a statistically significant effect on respondents' common intercept of SWD. Education has a positive effect, where approximately every additional 3 educational levels (two standard deviations) increase the predicted initial score on SWD by 0.33 (p < .05) on the eleven-point scale; this corresponds to a third of a unit. It means that people of higher educational status have a higher level of SWD at the beginning of the period. This is in line with H1 even though the effect is not large. Contrary to our hypothesis, income has no statistically significant association.8 Although only one of our measures of socio-economic status show the expected association with SWD, the variance of the intercept provides additional support for the importance of socio-economic status. Comparing the intercept variance in this conditional model with the intercept variance in the basic model without covariates (see Table A.1 in Appendix 5) shows a reduction from 1.82 to 0.55; this corresponds to a decrease in variance of almost 70 percent, only explained by the two socio-economic variables. It means that differences in socio-economic status are indeed a good explanation or predictor of differences in citizens? general level of SWD.The second column in Table 1 further shows that citizens' socio-economic status does not explain their trend in SWD. Neither variable has a statistically significant effect on the latent slope and the variance in the slope does not decrease either, compared to the unconditional growth model. It supports H2 and means that individuals with different levels of education or income in 2007 did not follow different trends in SWD in the following nine years. In other words, someone's socio-economic status does not help us predict how the person's SWD will develop over the next decade. This is as might be expected, since there are no obvious theoretical reasons for such an effect to occur.In this paper, we combined several prominent micro-level models explaining SWD to one common longitudinal framework by explicitly considering aspects of time, and we validated this framework against unique panel data from the Netherlands 2007?2016. Previous literature had only shown that the models tend to complement each other but we did not know how. In the foregoing, we have shown how this might work. People's socio-economic status explains their general levels of SWD but not the trend, and process and performance evaluations explain differences at particular points in time only and thus around the trend. These findings have larger implications for future theoretical and empirical work on SWD and political support, more generally.K?lln acknowledges funding from Aarhus University Research Foundation (ref. AUFF-E2017-7-13), Riksbankens Jubileumsfond (ref. RIK18-1267:2), and the Swedish Research Council for Health, Working Life and Welfare (ref. 2013?2692).Earlier versions of this paper were presented to colleagues at the 2015 ECPR General Conference in Montr?al, the 2015 EPSA conference in Vienna, the 2016 Etmaal in Brussels, the workshop 'Measuring individual dynamics with political science data' at the University of Essex, and at Aarhus University's Political Behaviour research meeting. We are very grateful for to participants and especially discussants for their help and advice. For additional suggestions and comments, we would like to thank Anna Kern, Christopher Anderson, Dieter Stier, P?r Zetterberg, and Palle Svensson. We are also grateful to three anonymous reviewers for comments that greately improved the paper.

Publisher Copyright:
© 2020 Elsevier Ltd

Copyright:
Copyright 2021 Elsevier B.V., All rights reserved.

Se relationer på Aarhus Universitet Citationsformater

ID: 213184317