We examine the relationship between election uncertainty, economic policy uncertainty, and financial market uncertainty in a prediction-market analysis, covering seven US presidential election campaigns. We argue theoretically that changes in the incumbent party re-election probability should be a key driver of changes in policy uncertainty. Consistent with this theory, we find that a large portion of changes in financial uncertainty in the final stages of election campaign seasons is explained by changes in the probability of the incumbent party getting re-elected. Our findings suggest that the incumbent-party election probability, derived from prediction markets, is an important measure of economic policy uncertainty in the days leading up to US elections.
We investigate how election uncertainty, measured directly from prediction markets, impacts policy uncertainty and financial uncertainty. Pástor and Veronesi (2013) propose a policy uncertainty channel to explain how political uncertainty impacts on market uncertainty. They apply the economic policy uncertainty (EPU) measure of (Baker et al., 2016) in empirical analysis, subsequently evidencing support for their theoretical model. While (Pástor and Veronesi, 2013) highlight the importance of EPU on financial uncertainty, it is not clear how election uncertainty impacts EPU and financial uncertainty. In order to examine the role of election uncertainty, we measure election uncertainty (EU) directly as a separate variable to EPU. We then consider the impact of daily changes in outcome probabilities and election uncertainty in the run up to US presidential elections on economic policy uncertainty and financial market uncertainty. Towards this end, we use a panel vector autoregressive (VAR) system which includes both election uncertainty, EPU, and two measures of financial uncertainty. We analyze the system using Granger causation analysis, impulse response analysis, and forecast error variance decomposition (FEVD). As measures of election uncertainty, we use daily futures prices from US presidential election prediction markets around seven US elections. To capture financial market uncertainty, we use a conditional variance forecast model from (Bekaert and Hoerova, 2014), and also use this measure to extract a variance premium variable from the VIX.
Our main finding is that changes in the election probability of the party of incumbency have a dominant role in how election uncertainty impacts both a leading measure of policy uncertainty and financial uncertainty.
We consider our research design and concomitant findings provide an important contribution alongside previous studies. In addition to Pástor and Veronesi (2013), there have been a number of studies on election uncertainty. For instance, Knight (2006) and Mattozzi (2008) consider the impact of daily changes in election-outcome probabilities on daily changes in stock prices through a partisan lens, whereby differing candidate policies are expected to benefit individual companies or industries. Other research examines the impact of election uncertainty on market volatility. For instance, Goodell and Vähämaa (2013) look at the impact of election uncertainty on implied volatility at the monthly level. More recently, Kelly et al. (2016) measure the impact of political uncertainty in an event study using a range of options measures.
Our contribution relative to existing work is that we measure the level of election uncertainty directly from daily prediction markets and perform analysis at the daily frequency in the run up to US presidential elections. This allows us to construct a system panel VAR with the daily variables of interest. We use daily data from Monday on the week before the election through to the Wednesday after the election (eight trading days) for seven US elections covering the period 1992–2016.1