Forecasting the 2016 election economy: it turns out consumers really do vote their wallets

 - by New Deal democrat

For the last 7 months, I have been engaged in a real-time experiement to see if I could use economic indicators to forecast the outcome of the 2016 Presidential election well in advance.

In previous installments,

• I've examined the best coincident measures of the Election Day economy found by Nate Silver.
• I've examined the correlation between housing permits, a long leading economic indicator, and those coincident indicators.
• I've examined the "Bread and Peace" model which makes use of real disposable personal income as a decisive economic indicator (in the past 20 years, it has not fared so well, especially in 2012).
• I've compared 160 years of election returns with NBER recession dates to conclude that whether or not the economy is in recession in the 3rd Quarter of the election year has forecast the outcome of the popular vote 75% of the time.   
• I have made a preliminary forecast that the democratic nominee is likely to win the election, based on the general positivity of the long leading indicators through October 2015.

One of my favorite economic indicators, however, is real retail sales, measured in the aggregate and per capita.  And lo and behold it has one of the best correlations of all with Presidential election results.  It turns out consumers really do vote with their wallets. It is considerably better than using income, probably because it is the acid test of the level of confidence average people have about the economy. As a result, how much of what is in their wallets they are wiling to spend is an excellent predictor of their voting preferences in Presidential elections.

Specifically, during the last 60 years, the following statements have always been true and have collectively always accurately predicted the outcome of the Presidential elections.

1. If there is no consumer spending recession, defined as at least 2 consecutive quarters' decline in real retail sales per capita, during a Presidential term, the candidate of the incumbent party wins the Presidential election.
2. If there is a consumer recession, and consumers are spending less per capita at the end of the third quarter of the election year than they were 3 years previously, then the candidate of the incumbent party loses the election.
3. If there is a consumer recession during the final two years of the Presidential term, the candidate of the incumbent party loses, even if consumers are spending more at the end of the third quarter of the election year than they were 3 years previously.
4. If there is a consumer recession during the first two years of the Presidential term, but the consumer has completely recovered (which has always happened), the candidate of the incumbent party wins the election.

Here is a table giving the breakdown of real retail sales per capita, ranked in order of strength from Q3 of the first year of the Presidential term through Q3 of the final year of the term, a total of 12 quarters.  I begin with the third quarter of the President's first year in office since voters historically excuse the President for responsibility for the economy until s/he has been in office at least 6 months.

The last two columns indicate whether or not a spending recession, defined as two consecutive quarters of a decline in real consumer spending per capita, occurred during the Presidential term.  If so, I indicate whether that spending recession ended, and if so, when:

Year	Real  retail sales per capita	Incumbent  party vote %	Spending recession?	Recovery from spending recession?
1964	11.8	61.3	None	---
1972	11.4	61.1	Yr1 m10	Yr3 m1
2012	7.7	52.0	None	---
2000	7.4	50.3	Yr4 m3	No
1996	6.6	54.7	None	---
1956	5.1	57.8	Yr1 m3; Yr 3 m10	Yr3 m3, No
2004	4.8	51.2	None	---
1984	4.8	59.3	Yr1 m8	Yr3 m6
1988	4.6	53.9	None	---
1968	2.2	49.6	Yr2 m3	Yr4 m11
1960	-2.1	49.9	None	---
1976	-2.8	49.0	Yr1 m7	No
2008	-5.9	46.3	Yr2 m1	No
1980	-6.1	44.7	Yr2 m12	No

 In general, the stronger the growth in real consumer spending during the 3 year period leading  up to the Presidential election, the higher the percentage of votes for the candidate of the incumbent party.  Further, every time consumer spending growth was negative, the incumbent party's candidate lost.  Every time the consumer spending growth was over 4%, the incumbent party candidate won.

At present were are 9 quarters into the 12 quarter period for the 2016 election.  Real retail sales per capita are up 3.2% since Q3 2013.  If they continue at that pace through Q3 of 2016, they will be up 4.3%, forecasting a win by the Democratic candidate.

Secondly, when consumer spending growth during the three years leading up to the Presidential election was weakly positive (i.e., 4% or less), there was always at least 1 consumer spending recession.  Whether and when the recovery from the downturn happened is vital. If it happened early in the term, the candidate won.  If it happened late in the  term, specifically including the last year before the election the candidate lost.  [Note that while in 1956, such a recession was ongoing at the time of the elction, the overall spending increase for the term was +5.1%.] Probably a weighted average where periods closer to the election get more weiighting than earlier periods would be an improvement in the model, but it is beyond my scope.

In 2015, real retail sales per capita did decline ever so slightly from Q3 to Q4 in 2015.  A one quarter decline is not unusual at all.  If, however, it continues to decline, that will be consistent with a scenario similar to that in 2000, where the candidate of the incumbent party, Al Gore, lost the election although he eked out a slight victory in the popular vote.

To summarize, if consumers vote their wallets this year,  the democratic nominee is likely to win, provided there is no spending recession over the next 8 months.