Election Day Jitters

If you’d like to receive our market commentary by email every month, please click here.

In the past months, COVID-19 has dominated both headlines and markets, overshadowing the coming U.S. presidential election in November. We expect this to change as fall arrives. This month, we examine market behavior around presidential elections as a guide to when and how this year’s election may have a discernible market impact.

Fear of the unknown

Conventional market wisdom predicts “risk off” into presidential elections. This posits that market participants take “chips off the table” and rebalance into safer assets because the election is an uncertain event with potentially far-reaching and unpredictable policy consequences.

To test this theory, we gathered weekly S&P 500 data in the 16 weeks leading up to each of the past 15 presidential elections. Below we plot the cumulative weekly returns for the S&P 500 for each election year and superimpose them on the corresponding dates for the current 2020 election calendar. 

Source. Orthogonal, Reuters.

The chart reveals little at first glance. We note that the relatively steep drop shown in the data for the 2008 elections may be an outlier because those elections coincided with the Lehman crisis.

Next we calculate the average paths for our data, including and excluding the 2008 data, and plot these below.

Source. Orthogonal, Reuters.

The paths are similar, with a positive drift through the first or second week of September only to drop sharply in the four weeks leading up to the election itself. Excluding the 2008 data results in a much shallower, but still significant, drop.

A tangible effect

We make two general observations from these plots. First, the average return for the S&P 500 across the 16-week periods preceding the 15 elections we analyzed is only 0.27% (excluding 2008, 1.53%). The S&P 500 has an average annualized return of 10% in the long run, which means we would expect it to return roughly 3.3% over an average 16-week period. As such, it appears that S&P 500 returns are diminished on average during the 16 weeks leading up to a U.S. presidential election.

Second, the S&P 500 peaks ahead of a U.S. presidential election on average around the first or second week of September. (For the current election, this would be on or about September 18, 2020.) Thereafter, it falls for an extended period that may correspond to the derisking predicted by conventional market wisdom. It eventually stabilizes around 3 to 4 weeks ahead of election day.

Zeroing in

To analyze these general observations further, we tabulate historical returns to the S&P 500 beginning on the dates for each election year that are equivalent to the expected peak date for the current election cycle (September 18, 2020) for holding periods of 1 through 6 weeks.

Source. Orthogonal, Reuters.

The data suggests that if this year follows historical patterns, one option to reduce equity risk ahead of a potential pre-election derisking trend would be to reduce a portfolio’s S&P 500 exposure in September 2020 and, if appropriate, adjust it back up the next month. This has an information coefficient of 0.39, or 0.38 if 2008 is excluded, both of which are considered high, albeit the sample size is low.

The timing suggested by historical data also fits well with this year’s market calendar. The projected peak coincides with the September Federal Open Market Committee meeting, as marked in the plot below.

Source. Orthogonal, Reuters.

The minutes of the Federal Open Market Committee released last week suggest that at its September meeting the Fed is expected to announce a shift to average inflation targeting (an approach by which the Fed would tolerate periods of above-trend inflation to compensate for periods of below-target inflation) as well as return to open-ended quantitative easing. Both of these policies should be bullish for equities. As such, if these past patterns hold, a positive reaction to the September FOMC meeting sets up a good opportunity to derisk portfolios ahead of this year’s election.