By Robert Prechter | July 2012
Most statistical studies in the field of political science test whether A causes B. With respect to election outcomes, for example, researchers test things such as which party in power is better for the economy, or whether the state of the economy predicts which party will get elected. Sometimes the results seem to be meaningful, but when the time period of the test is changed, or the economic variable is changed, the apparent significance often disappears. Generally, results have not been very useful.
My colleagues at the Socionomics Institute and I set out to test something different and more difficult. Socionomic theory proposes that unconscious social mood regulates social actions. For example: In the field of finance, social mood regulates aggregate pricing in the stock market. In the field of politics, social mood is a strong regulator of which leaders are selected and how they are perceived. More specifically, when the trend in social mood is positive, investors tend to bid stock prices higher and voters tend to retain incumbent leaders; and when the trend of social mood is negative, investors tend to bid stock prices lower and voters tend to oust incumbent leaders. In other words, our theory is that C (social mood) causes A and B.
But C—social mood—is a hidden variable. We can’t measure it directly by hooking up electrodes to 200 million human brains. We can measure only the effects of social mood, as expressed in social actions. With this hurdle, how could we test our idea?
We first set out to test our hypothesis that the stock market should be a useful indicator of whether voters retain or reject an incumbent. Since pundits tend to credit or blame the predecessor for conditions during the first year of a new president’s term, we suspected that the performance of the stock market three years prior to the election would be the best duration for testing how voters would judge an incumbent. Nevertheless, for robustness, we tested stock market returns for durations of 1, 2, 3 and 4 years prior to each election.
To assure further robustness of our results, we tested the stock market’s performance against multiple measures of voting results: popular vote margin and percentage, electoral vote margin and percentage, straight win/loss, and landslides (both electoral and popular-vote); we tested various thresholds to judge landslide wins and losses; we tested percentage and lognormal changes; we tested stock market returns using both nominal and inflation-adjusted prices; we subjected the data to four different types of statistical tests; and we used all available data, going back to 1824 for popular vote and to George Washington’s reelection in 1792 for electoral vote.
We found that the stock market is significant predictor of election results in all cases. As surmised, the best result was the 3-year period prior to the election. One of our landslide tests obtained a perfect score.
But this result, while useful, did not answer our primary theoretical question. After all, maybe the stock market is just one of many conditions by which voters judge a president’s performance. In other words, perhaps there are many A’s, but A still causes B.
Conventional wisdom says, “It’s the economy, stupid.” But socionomic theory predicts that the stock market would be a better register of social mood than the economy. The reason is that investors can express social mood nearly instantaneously by buying or selling stock, whereas it takes time for business people to implement business plans formulated at the very same time.
So, we next set out to test how the stock market compared to the economy as a predictive variable. If it performed as well as the stock market, then we still had a useful result, but not one that is theoretically important.
To assure robustness, we tested the predictive ability of the so-called “big three” economic indicators: gross domestic product (GDP), the inflation rate, and the unemployment rate. We also tested the inflation-adjusted (real) GDP. We tested these variables individually and in combination with the stock market.
We found that the stock market outperformed every one of the economic variables, on all four time frames. In our tests, in fact, inflation and unemployment showed no significant ability to forecast election outcomes. GDP was a significant predictor, but its stand-alone predictive ability was weaker than DJIA’s, and its significance disappeared in the presence of DJIA, whereas DJIA remained significant in the presence of other economic variables.
This counter-intuitive result got us closer to our goal of demonstrating socionomic causality. After all, it would be difficult to claim in this instance that A caused B, i.e. that the stock market caused voters to retain or reject an incumbent president. Why would voters care more about their portfolios than about their jobs, their savings and their country’s economy?
Nevertheless, economic-voting theorists could still claim that for some unknown reason, voters care more about the stock market than about the economy, the inflation rate or the availability of jobs. In other words, that A still causes B.
So, we did one last test. If voters truly are voting in responses to their stock market gains and losses, then our results should be stronger in times when more voters owned stock. To test whether this is the case, we bifurcated our results.
In the first half of the 20th century, few Americans owned stock; estimates for different decades during that time range from 2% to 10% of the population, the latter number appearing briefly, near the 1920s stock market peak. In the second half of the 20th century, on the other hand, up to 50% of Americans households owned stock. Yet we found that the predictive value of the stock market was about the same in both periods. Ownership didn’t seem to matter.
For added robustness, we also tested the two centuries against each other. In the 19th century, most Americans were farmers, and almost no one owned stock. Yet the stock market remained a significant predictor of re-election outcomes in both the 19th and 20th centuries. In fact, the record for the 19th century is actually slightly better than that for the 20th!
Clearly, A cannot be causing B. Voters are not deciding to keep or reject incumbent presidents simply because they are making or losing money in the stock market.
The socionomic explanation holds: Voters in the aggregate are not responding to stock market changes, economic changes, inflation rates or the availability of jobs. Rather, they are voting in accordance with socionomic theory, which proposes that both investors and voters are acting in accordance with a hidden variable: social mood. In other words, C is causing both A and B.
We believe our study helps demonstrate that aggregate voting at the margin—swing voters—are not so much rationally weighing the potential value of each candidate but rather voting primarily based on how they feel. When a positive trend in social mood induces investors to push the stock market upward during the three years prior to an incumbent’s re-election bid, it also induces voters to credit the incumbent for their good moods and vote to retain him in office. When a negative trend in social mood induces investors to push the stock market downward during the three years prior to an incumbent’s re-election bid, it also induces voters to blame the incumbent for their bad moods and vote to reject him from office.
Our study has practical value for election forecasters: To improve long-range prediction of election outcomes, use social-mood indicators in your mix. We have some advice for politicians, too: If you want to beat an incumbent, make your initial bid for office when the stock market has fallen a long way. If you are already in office and the stock market has risen a long way, run for re-election; if it has fallen a long way, bow out gracefully, take a respite, and keep a humiliating defeat off your resume.■
To read “Social Mood, Stock Market Performance and U.S. Presidential Elections: A Socionomic Perspective on Voting Results” by Prechter, Goel, Parker and Lampert, search on the SSRN website or use this link: [http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1987160].