Trump and the Dow (Blog Post)

Jump to navigation Jump to search

Donald Trump By Gage Skidmore (Flipped).jpg
Blog Post
Title Trump and the Dow (Blog Post)
Author Ed Egan
Series Election 2016
Content status Published
Publication date
Image Donald_Trump_By_Gage_Skidmore_(Flipped).jpg
©, 2016


The Article

Some recent articles suggest that a Donald Trump presidency would be bad for the U.S. economy, or at least bad for investors in the markets. One way to address whether this is true is to examine the relationship between Mr. Trump’s poll numbers and the market.

We took Donald Trump’s daily poll numbers from the Real Clear Politics National Average Poll data and the close-of-day price for the Dow Jones Industrial Average from the 19th of July, 2016– the date that Mr. Trump officially became the Republican party nominee – to the 5th of October, 2016. We then estimated the relationship between them using Ordinary Least Squares.

Trump rises, the Dow falls

The results are striking: A one percent increase in Donald Trump’s poll numbers are associated with a 65 point decrease in the Dow.

At the time of writing, the Dow is at 18,281.03. A 65 point decrease is therefore just over a 1/3rd of a percent drop in the Dow.

TrumpAndTheDow BakerInsititute.png

Donald Trump is currently polling at 43.8%, versus Hillary Clinton at 48.3%. If Mr. Trump were to poll at 50%, then these numbers suggest that the Dow would have a very bad day, losing more than 2.2% of its value and falling back below 17,900. Likewise, if Mr. Trump were to drop 6% in the polls, these numbers suggest that the Dow might surge to a new record high.

Statistical significance and explanatory power

The correlation between Donald Trump’s poll numbers and the performance of the Dow is highly statistically significant. The odds of finding this correlation by chance are less than one in ten thousand.

Donald Trump’s poll numbers explain 38% of the variation in the Dow. This is surprising, as many, many factors affect the Dow’s closing price. It suggests that the markets are really concerned with the outcome of the 2016 presidential election.

Correlation doesn’t imply causality

As any econometrician will tell you, correlation does not imply causality. All of the following, and more, are possible:

  • Forward causality: A rise in Donald Trump’s poll number causes a drop in the Dow.
  • Reverse causality: A fall in the Dow causes a rise in Donald Trump’s poll numbers.
  • Third-omitted variable issue: Something else changing in America is causing both the Dow to fall and Mr. Trump’s poll numbers to rise.
  • Construct validity issue: Donald Trump’s poll numbers may reflect something else besides his likelihood of winning the U.S. presidential election. The Dow may not be representative of other markets, or of the economy more generally.

Robustness checks

We included a range of other explanatory variables, including time trends, total U.S. debt and the 10-year bond rate, in a multivariate analysis. The effect of Mr. Trump’s poll numbers remained highly statistically significant. However, the magnitude of the decrease in the Dow was reduced to around 45 points for every 1% increase in his poll numbers.

The Dow is an index of 30 large publicly traded companies. It accounts for around 20% of the U.S. economy and historically it has correlated with U.S. GDP rather well. However, it might have been better to use the S&P 500, which accounts for around 75% of the U.S. economy. The S&P 500 closed at 2,160.77 yesterday. We found that a one percent increase Donald Trump’s poll numbers was associated with a 5 point (0.23%) drop in the S&P 500. Again, these results were highly statistically significant.

The margin of error for polls is typically around 3.5%. This would suggest a non-linear effect as the candidates close to within a few points of each other. We did check to see if there were significant non-linear effects within the bounds of our data and there were not. We should also point out that it is extremely unlikely that the Dow would surge to above 21,000 if Donald Trump polled zero. Arriving at this number requires extrapolation well beyond the bounds of our data.

Finally, it is tempting to attribute these results to the market’s reaction to differences in policies or leadership between the candidates. There are other explanations. For example, the risks of a Clinton presidency are fairly well known (America has had one, albeit with a different Clinton) but the risks of a Trump presidency are unknown. One interpretation consistent with our results is that the market fears the unknown.