Do investors demand a risk premium for divided government because of the policy uncertainty of gridlock? In the April 2013 preliminary draft of their paper entitled “Divided Governments and Asset Prices”, Elvira Sojli and Wing Wah Tham investigate the effect of divided government on asset prices by comparing U.S. stock market performance in years of divided and undivided government. They define divided government (in the U.S.) as one party holding the Presidency while the other controls one or both houses of Congress. To isolate the effects of divided government, they account for the states of four variables widely used to predict stock market returns: dividend-price ratio; credit default spread, the difference between BAA and AAA corporate bond yields; term spread, the difference between 10-year U.S. Treasury note and 3-month U.S. Treasury bill (T-bill) yields; and, deviation of the 3-month T-bill yield from its one-year moving average. To determine causality, they investigate stock market futures reactions to betting market predictions of divided versus undivided government during four election nights. Using monthly U.S. stock market returns, data for the four stock return predictors and U.S. Treasury bill yields during 1926 through 2011 (encompassing 43 elections resulting in 23 undivided and 20 divided governments), and high-frequency election outcome betting data (from Intrade) and U.S. stock market futures on biennial election nights during 2004 through 2010, they find that:
- The average annual excess return (relative to the T-bill yield) of the broad value-weighted U.S. stock market is 11.5% (3.3%) under undivided (divided) government. The difference of 8.2% is statistically and economically significant.
- After accounting for the states of the four predictive variables specified above, expected returns are about 1.4% per year higher for a divided than an undivided government after controlling for party holding the Presidency.
- Returns unexplained by the four predictors are 8% lower for divided than undivided government, but the party holding the Presidency accounts for this difference (returns are substantially higher when the President is a Democrat).
- High-frequency interaction of election outcome betting and U.S. stock market futures on election nights indicates 1.4%, 1.9% and 1.4% respective decreases in the the Dow Jones Industrial Average, the NASDAQ 100 Index and S&P 500 Index for emergence of divided government.
- In corroboration, emergence of the hung parliament during 2010 UK elections indicates 1.2% and 1.1% respective depreciations in GBP/USD and GBP/EUR exchange rates.
- Further analysis shows that divided government affects expected stock returns through the mechanism of economic policy uncertainty.
In summary, evidence suggests that there is an annual risk premium of about 1% for divided governments.
One interpretation of findings is that the election of divided governments drives stock prices down, with subsequent reversion.
Cautions regarding findings include:
- Any effect of divided versus undivided government appears to be small compared to the effect of party holding the Presidency (or, alternatively, party holding the Senate as in “Party in Power and Stock Returns”).
- There are arguably decisive/indecisive degrees of controlling houses of Congress based on size of majority.
- Sample and controlled subsample sizes are quite small, with “normality” of statistical environments arguable.
- Analyses of the party in power of different elective branches on the same set of stock market return data are susceptible to data snooping bias, such that the strongest indication likely overstates out-of-sample expectations.
