We further developed the previous research report conducted by Claude Code by utilizing Claude Code

Background

Voting behavior in democratic societies is closely related to socioeconomic factors. Recent research in political economy has focused on the impact of economic uncertainty on voter political participation. In particular, the VIX index (fear index), which indicates financial market volatility, is widely recognized as a proxy indicator of economic anxiety, and its potential influence on voter turnout in elections has been theoretically suggested.

Research Objectives

This study aims to empirically examine the correlation between voter turnout and the VIX index in the United States. Specifically, we clearly distinguish between presidential and midterm elections to test the following hypotheses.

Hypothesis 1: When social anxiety increases, political interest under democratic systems increases, leading to higher voter turnout
Hypothesis 2: During periods when the VIX index is high, economic anxiety increases and amplifies social anxiety
Hypothesis 3: Based on the above two hypotheses, there exists a positive correlation between voter turnout and the VIX index

Methodology

Data

The following data were used for analysis.

Voter Turnout Data: We used voter turnout data based on Voting Eligible Population (VEP) published by the United States Elections Project. The analysis covered 18 elections in total: 9 presidential elections (1992-2024) and 9 midterm elections (1990-2022).

VIX Index Data: We used the average values for October-November of election years obtained from the CBOE Volatility Index (VIX) and Federal Reserve Economic Data (FRED). This period was determined to be optimal for capturing economic uncertainty immediately before elections.

Analytical Methods

This study applied the following statistical methods.

Descriptive Statistics: We calculated basic statistics for voter turnout and VIX index for each election type (presidential elections and midterm elections).
Correlation Analysis: We used Pearson’s correlation coefficient to evaluate the linear relationship between VIX index and voter turnout. By analyzing presidential and midterm elections separately, we clarified differences in effects by election type.
First-Difference Analysis: We calculated changes from the previous election (first differences) and examined correlations in time-series variation patterns. This attempted to reveal dynamic relationships that could not be captured by level values alone.
Regression Analysis: We used simple linear regression models to quantify the direction and magnitude of the VIX index’s impact on voter turnout.

Results

Descriptive Statistics

The average voter turnout in presidential elections was 59.52% (SD=4.72), and the average VIX index was 25.58 (SD=15.21). Meanwhile, the average voter turnout in midterm elections was 41.33% (SD=4.34), and the average VIX index was 20.90 (SD=6.50). Presidential elections showed approximately 18 percentage points higher voter turnout compared to midterm elections, and the VIX index also tended to be slightly higher.

Correlation Analysis

Presidential Elections: A correlation coefficient of r = 0.36 (p = 0.70, n.s.) was obtained between the VIX index and voter turnout. The coefficient of determination R² = 0.13 indicates that the VIX index explains approximately 13% of the variation in voter turnout, but statistical significance was not observed.

Midterm Elections: An extremely weak correlation of r = 0.03 (p = 0.98, n.s.) was observed between the VIX index and voter turnout. The coefficient of determination R² = 0.001 indicates that no substantial linear relationship exists between the two variables.

All Elections Combined: When presidential and midterm elections were integrated, the result was r = 0.30 (p = 0.76, n.s.), with no statistical significance observed.

First-Difference Analysis

Analysis using changes from the previous election yielded r = 0.36 (p = 0.70, n.s.) for presidential elections and r = -0.08 (p = 0.93, n.s.) for midterm elections. Neither showed statistical significance, and no relationship at the change level was confirmed.

Regression Analysis

Presidential Elections: The regression equation was “Voter Turnout = 56.60 + 0.11 × VIX” (SE = 4.67). This indicates that a 1-point increase in the VIX index corresponds to a 0.11 percentage point increase in voter turnout, but statistical significance was not observed.

Midterm Elections: The regression equation was “Voter Turnout = 40.91 + 0.02 × VIX” (SE = 4.71). This indicates that a 1-point increase in the VIX index corresponds to a 0.02 percentage point increase in voter turnout, but the effect size is extremely small and statistical significance was not observed.

Discussion

Key Findings

The empirical analysis of this study did not confirm a statistically significant correlation between voter turnout and the VIX index in the United States. This result does not support the initial hypotheses. Although a moderate correlation coefficient of r = 0.36 was observed in presidential elections, the constraint of sample size (n=9) results in insufficient statistical power, making it difficult to draw definitive conclusions.

Theoretical Implications

The results of this study suggest that the determinants of voting behavior are more complex than financial market volatility. While existing research has shown that economic anxiety influences voting behavior, the VIX index primarily reflects the expectations of financial market participants and may not adequately capture the economic anxiety of general voters.

The differences by election type are also important. The correlation coefficient being nearly zero in midterm elections suggests that voter motivations differ qualitatively between presidential and midterm elections. Presidential elections attract national attention and extensive media coverage, potentially making the relationship with economic anxiety relatively stronger.

Methodological Limitations

This study has several limitations. First, the sample size is limited (presidential elections n=9, midterm elections n=9), resulting in insufficient statistical power. Second, there is room for debate regarding whether the VIX index measurement period (October-November) is appropriate. Voter economic perceptions may be shaped by longer-term economic conditions rather than just pre-election conditions. Third, this study only examined correlations and did not identify causal relationships.

Future Research Directions

The following approaches would be useful for future research. First, using state-level or county-level data could increase sample size and improve statistical power. Second, conducting multivariate analysis using economic anxiety indicators other than the VIX index (unemployment rate, inflation rate, consumer confidence index, etc.) could provide more comprehensive understanding. Third, applying causal inference methods such as panel data analysis or instrumental variable methods would enable more rigorous verification.

Conclusion

This study empirically examined the relationship between voter turnout and the VIX index using U.S. election data from 1990 to 2024. The analysis did not confirm a statistically significant correlation between the two variables. Although a moderate positive correlation tendency was observed in presidential elections, the sample size constraint prevented reaching definitive conclusions.

This result suggests that the relationship between economic anxiety and voting behavior is a complex phenomenon that cannot be captured by simple linear relationships. The VIX index, which indicates financial market volatility, reflects investor expectations and may not directly predict the economic perceptions or political behavior of general voters.

Future research using larger datasets, diverse economic anxiety indicators, and rigorous causal inference methods is expected to provide deeper understanding of the impact of economic anxiety on voting behavior.

Data Sources

United States Elections Project (McDonald, 2024); CBOE VIX Index; Federal Reserve Economic Data (FRED)

Note

This study also serves as an empirical verification of AI-assisted research processes. Claude (Anthropic) was utilized for parts of data collection, statistical analysis, and report preparation.