Modeling public opinion over time and space: Trust in state institutions in Europe, 1989-2019

Abstract

Political trust is crucial for state legitimacy, and yet its dynamics remain understudied. We present time series of political trust among European societies between 1989 and 2019. Relying on Bayesian non-linear multilevel models for ordinal responses, we estimated levels of trust in 27 European countries with data from 12 cross-national survey projects. To improve the quality of estimates and correct for the differences in survey sample representativeness, we applied poststratification by sex, age, and education. We describe trajectories of political trust in Europe finding support for the thesis about trendless fluctuations in trust rather than any clear long-term tendency, with substantial differences between countries and European regions. In line with prior research, we find larger differences in political trust by education level than by sex or age. In discussing the results, we outline opportunities for further research enabled by our methods and the dataset obtained through our models’ predictions.

Political trust as a cause and consequence of democracy: Longitudinal analysis of European data

“…To measure country-year levels of political trust in Europe between 1991 and 2019, both overall, and by education and age groups, we use poststratified estimates created by jointly analyzing data from 13 crossnational survey projects with hierarchical Item Response Theory models. These models were applied to survey variables referring to trust in national parliaments, political parties, and justice systems as indicators of political trust (see Kolczynska et al, 2020, for more details about the data and models, and the Online Supplement for the list of source data sets). We use estimates of levels of political trust overall, as well as by education and age.…”

“…whereỹ jt1 is the mean trust value and s jt1 is the standard deviation of trust for country j at time t both obtained from the trust prediction model of Kolczynska et al (2020).…”

Modeling public opinion over time: A simulation study of latent trend models

“…Common approaches in models estimating dynamic public opinion include various kinds of ordinal models (e.g. Caughey and Warshaw, 2015;Solt, 2020;Kołczyńska et al, 2020), binary models applied to dichotomized data (e.g. McGann, 2014;Claassen, 2019), and linear models applied to ordinal data treated as continuous (e.g.…”

“…The most common strategy of correcting sample representativeness is weighting the data, and alternatives include multilevel regression and poststratification (Gelman and Little, 1997). For a poststratification approach applicable within our modeling framework, see Kołczyńska et al (2020). Since our simulations do not involve non-response or individual-level covariates, we leave the relative merits of weights and poststratification to future research.…”

“…Prior studies proposed solutions primarily relying on discrete autoregressive (AR) models (Caughey and Warshaw, 2015;Claassen, 2019;Solt, 2020). The paper by Kołczyńska et al (2020) used penalized splines. Gaussian processes (GP) provide yet another alternative, that -to the best of our knowledge -have not yet been applied or studied in this context.…”