TL;DR
This paper provides the first comprehensive map of economic voting across 19 developed democracies over two decades, revealing that while the economy significantly influences vote choice, its impact varies considerably by country and over time. The authors develop a robust methodology to quantify this "economic vote," finding that incumbent parties can expect a median vote loss of 4.4% when economic evaluations decline by one unit.
What Problem Does This Paper Address?
The conventional wisdom among politicians and commentators is that elections are referenda on the economy. However, empirical evidence for the universality and consistent strength of this "economic vote" across different countries and time periods has been surprisingly mixed and difficult to compare. Previous studies often relied on disparate methodologies or limited datasets, making it hard to draw broad conclusions. This paper aims to bridge this gap by providing a systematic, comparable, and large-scale analysis of economic voting across a wide range of developed democracies.
Methods
The authors employ a novel, two-stage multi-level estimation strategy to address the challenges of cross-national and temporal comparability.
- Data Harmonization: They meticulously select 163 national surveys from 19 countries spanning 1980-2001, ensuring comparability by standardizing questions on vote choice and retrospective economic perceptions. They focus on retrospective evaluations (how the economy has changed) rather than prospective ones, due to data availability and theoretical arguments.
- Individual-Level Modeling: For each survey, they estimate a multinomial logistic model (MNL) of vote choice. This approach is preferred over simpler logistic regressions as it appropriately handles multi-party elections without artificial constraints. The MNL model estimates the probability of a voter choosing each party, allowing for the analysis of how these probabilities shift with changes in economic perceptions.
- Generating Substantive Effects: To quantify the "economic vote," they simulate the impact of a one-unit decline in economic perceptions on predicted vote probabilities for each individual in the sample. This method, which involves averaging simulated effects across all respondents, provides a more representative measure of the population's response than relying on an "average" individual.
- Cross-National and Temporal Analysis: The estimated economic vote magnitudes from each country and year are then pooled to analyze cross-national variation (using box-plots) and temporal trends (using time-series plots).
Datasets
The study utilizes a dataset of 163 national voter preference surveys from 19 developed democracies, collected between 1980 and 2001. These surveys include:
- National Election Studies
- EuroBarometer Studies
- EuroBarometer Studies (repeated for some countries)
A detailed list of these studies, including country and year, is provided in Appendix I.
Key Insights and Ideas
- Economic Voting is Real, But Not Universal: The study confirms that economic evaluations significantly impact vote choice in developed democracies, but its magnitude is far from uniform.
- The "Economic Vote" is a Distributional Phenomenon: The authors define economic voting not just by its overall size but by how economic perceptions shift vote probabilities among parties, particularly affecting incumbent parties.
- Incumbents Bear the Brunt: The data strongly suggest that voters tend to punish incumbent parties when the economy is perceived to be doing poorly.
- Methodological Rigor is Key: The paper highlights the importance of comparable survey questions and advanced statistical modeling (MNL) for robust cross-national and temporal analysis, moving beyond simpler correlational approaches.
- Temporal Trends Matter: The magnitude of the economic vote appears to have declined in recent decades, particularly after 1990, suggesting evolving voter behavior or political contexts.
Key Results
- Median Economic Vote Effect: The median impact of a one-unit decline in economic evaluations on incumbent Prime Ministerial (PM) parties is a 4.4% vote loss.
- Ubiquity of Punishment: Almost 90% of the estimated economic vote effects for PM parties are negative, indicating a strong tendency for voters to punish incumbents during economic downturns.
- Statistical Significance: While individual studies might show statistically insignificant results, the combined sample of 163 studies provides strong evidence for a statistically significant economic vote.
- Cross-National Variation: The magnitude of the economic vote varies dramatically across countries. Portugal shows the highest economic vote score (under -5%), while the US Presidential elections rank fourth among the 19 countries.
- Temporal Trend: A notable temporal trend shows a decline in the PM economic vote magnitude, from around 8% before 1988 to about 5% after 1990.
Limitations and Future Work
- Data Limitations: While the study uses a large dataset, the comparability of survey questions, even after harmonization, remains a challenge. The authors acknowledge that deviations in question wording were tolerated, and they investigate the potential impact of these differences.
- Focus on Retrospective Voting: The study primarily focuses on retrospective economic evaluations due to data availability. Future work could explore the interplay between retrospective and prospective evaluations more deeply.
- Explanation of Trends: The paper identifies temporal trends (e.g., a decline in economic voting) but states that a precise explanation for these trends is beyond the scope of the current essay. The authors suggest that their forthcoming work develops a "competency model" to explain both temporal and cross-national variations.
- Theoretical Explanations: While establishing the existence and variation of the economic vote, the paper sets the stage for future research to develop and test appropriate theories for explaining these cross-sectional and temporal patterns.
Summary updated