# The Market Has a Vote Too

#### Abstract

With each presidential election in the United States comes speculation as to who will become our next president. Many try to predict the outcome and forecast the results of each election, but forecasts are imperfect; thus, outcomes remain unknown until Election Day. It is commonly thought that economic measurements, such as GDP growth and the unemployment rate, may influence voter behavior. If the economy is doing well under the current president, voters may be more likely to cast a ballot for the incumbent party candidate. Therefore, views of the economy’s performance during the administration of the current president may be especially important in the following election. We explore this idea with a primary goal of determining the impact of stock market performance on presidential election outcomes. Prior studies indicate that the relationship between the stock market and election outcomes is positive in terms of the incumbent party's vote share. Using a standard linear regression model and a binomial probability model, we examine the United States presidential elections from 1976 to 2016, along with the economic conditions surrounding them. The incumbent party candidate's vote share (as a percentage) serves as the dependent variable in the linear regression model. The probability model focuses on determining the likelihood that the incumbent party candidate's vote share is greater than that of the challenging party candidate's share. Explanatory variables include the change in the S&P 500, the percentage change in GDP (from quarter 2 to quarter 3), the change in the consumer price index, the change in the unemployment rate, and the change in the Consumer Confidence Index. Aside from GDP growth, the changes in the explanatory variable series span the three months prior to each election (i.e., August-October). Furthermore, we include an independent variable that indicates whether or not the incumbent party candidate is running for a second term. The models are estimated at the country-level and also at the state-level. Therefore, both national and state unemployment rates are utilized. To further examine the differences between states, the models are also estimated for “swing states” alone. Our findings appear largely consistent with previous literature and theory.

*This paper has been withdrawn.*

The Market Has a Vote Too

HUB 302-#187

With each presidential election in the United States comes speculation as to who will become our next president. Many try to predict the outcome and forecast the results of each election, but forecasts are imperfect; thus, outcomes remain unknown until Election Day. It is commonly thought that economic measurements, such as GDP growth and the unemployment rate, may influence voter behavior. If the economy is doing well under the current president, voters may be more likely to cast a ballot for the incumbent party candidate. Therefore, views of the economy’s performance during the administration of the current president may be especially important in the following election. We explore this idea with a primary goal of determining the impact of stock market performance on presidential election outcomes. Prior studies indicate that the relationship between the stock market and election outcomes is positive in terms of the incumbent party's vote share. Using a standard linear regression model and a binomial probability model, we examine the United States presidential elections from 1976 to 2016, along with the economic conditions surrounding them. The incumbent party candidate's vote share (as a percentage) serves as the dependent variable in the linear regression model. The probability model focuses on determining the likelihood that the incumbent party candidate's vote share is greater than that of the challenging party candidate's share. Explanatory variables include the change in the S&P 500, the percentage change in GDP (from quarter 2 to quarter 3), the change in the consumer price index, the change in the unemployment rate, and the change in the Consumer Confidence Index. Aside from GDP growth, the changes in the explanatory variable series span the three months prior to each election (i.e., August-October). Furthermore, we include an independent variable that indicates whether or not the incumbent party candidate is running for a second term. The models are estimated at the country-level and also at the state-level. Therefore, both national and state unemployment rates are utilized. To further examine the differences between states, the models are also estimated for “swing states” alone. Our findings appear largely consistent with previous literature and theory.