- by New Deal democrat
How well do economic conditions predict the outcome of Presidential elections? Can we forecast the most important of those conditions in advance? If so, what can we say even now about 2016?
That's what I am examining in this series, Forecasting the 2016 Election Economy.
Fortunately, I don't have to start from scratch. As I noted in my inaugural post, most significantly, in 2011, Nate Silver undertook an examination of 43 economic variables, and their potency in predicting the election outcomes, going back all the way to the 1950s. Additionally, James Surowiecki has suggested that growth in real personal income in Q2 of the election year is the most determinative economic fact. Also, Professor of Economics Douglas Hibbs has used what he calls:
I'll look at that model, which has a generally good record, but missed badly in 2000 and 2012, in a later post.
In this post, I still start by looking at 5 of the variables identified by Nate Silver as being the most dispositive. To cut to the chase, in order of their predictive importance, here's the top 20 variables he found most predictive of election results, in order of their value:
Below, I have taken those 5 indicators, and charted them all the way back to the 1952 election, or the start of the series when applicable, as to their YoY change (q/q change for ISM) during Q2 and Q3 of each election year. All of these are adjusted for inflation when relevant:
For reasons of space limitation, for now I omitted the series for real GDP, real final sales, real aggregate payroll growth, and real disposable personal income per capita.
The top 5 variables identified by Silver mostly give me the same result: if they were consistent, Humphrey would have won in 1968, and Ford in 1976. Both Humphrey and Ford had objectively great economic numbers to run on. Further, while poor economic performace as in 1980 and 2008 are a death sentence for the incumbent party, mediocre numbers are less dispositive. If they were, Clinton in 1992, Bush in 2004, and Obama in 2012, all would have lost.
That being said, the single variable with the best record is the change in the unemployment rate. It has correctly forceast 13 of the 15 elections since 1956. If the unemployment rate is rising in Q2 and Q3 of the election year, the incumbent party is going to lose. If it is declining, with two exceptions, (Humphrey in 1968, Gore in 2000) the incumbent party is going to win.
Meanwhile, Surowiecki's measure of real personal income calls for Clinton and Obama to lose their re-election bids, and also misses 1968 and 1976.
How well do economic conditions predict the outcome of Presidential elections? Can we forecast the most important of those conditions in advance? If so, what can we say even now about 2016?
That's what I am examining in this series, Forecasting the 2016 Election Economy.
Fortunately, I don't have to start from scratch. As I noted in my inaugural post, most significantly, in 2011, Nate Silver undertook an examination of 43 economic variables, and their potency in predicting the election outcomes, going back all the way to the 1950s. Additionally, James Surowiecki has suggested that growth in real personal income in Q2 of the election year is the most determinative economic fact. Also, Professor of Economics Douglas Hibbs has used what he calls:
... the Bread and Peace model to explain presidential voting outcomes. The model claims that just two fundamental variables systematically affected post-war aggregate votes for president: (1) weighted-average growth of per capita real disposable personal income over the term, and (2) cumulative US military fatalities due to unprovoked, hostile deployments of American armed forces in foreign wars.
In this post, I still start by looking at 5 of the variables identified by Nate Silver as being the most dispositive. To cut to the chase, in order of their predictive importance, here's the top 20 variables he found most predictive of election results, in order of their value:
Below, I have taken those 5 indicators, and charted them all the way back to the 1952 election, or the start of the series when applicable, as to their YoY change (q/q change for ISM) during Q2 and Q3 of each election year. All of these are adjusted for inflation when relevant:
| Y ear | ISM Mfg | Change Nonfarm payrolls | Change unemplymt rate | Real personal income | Change Emp/Pop ratio |
|---|---|---|---|---|---|
| 1956 | 1.0/0.4 | 3.9/2.5 | -0.2/0.0 | 5.6/4.1 | 1.2/0.5 |
| 1960 | -5.9/-4.4 | 1.9/1.4 | 0.1/0.2 | 2.9/3.2 | 0.1/0.1 |
| 1964 | 9.3/13.2 | 2.6/3.0 | -0.5/-0.5 | 5.8/6.4 | 6.1/5.5 |
| 1968 | 5.6/2.9 | 3.2/3.4 | -0.2/-0.3 | 5.1/5.5 | 0.8/0.3 |
| 1972 | 9.8/12.3 | 3.2/4.3 | -0.2/-0.4 | 5.1/6.3 | 0.4/0.6 |
| 1976 | 9.2/4.7 | 3.6/3.2 | -1.3/-0.8 | 4.5/4.3 | 1.6/1.2 |
| 1980 | -17.6/-6.5 | 0.8/-0.3 | 1.6/1.8 | -0.7/-0.8 | -0.1/-0.8 |
| 1984 | 9.2/3.0 | 5.0/4.9 | -2.7/-2.0 | 7.7/8.3 | 2.7/1.9 |
| 1988 | 6.9/6.2 | 3.2/3.2 | -0.8/-0.5 | 4.5/0.5 | 0.8/0.7 |
| 1992 | 4.0/2.3 | 0.3/0.6 | 0.6/0.7 | 3.2/3.3 | -0.4/-0.2 |
| 1996 | 0.7/0.8 | 2.0/2.2 | -0.2/-0.4 | 4.3/4.4 | 0.3/0.7 |
| 2000 | 3.1/0.7 | 2.5/2.1 | -0.4/-0.2 | 6.1/6.5 | 0.2/-0.1 |
| 2004 | 10.8/8.6 | 1.1/1.5 | -0.5/-0.7 | 0.2/3.5 | 0.0/0/4 |
| 2008 | -0.9/-13.8 | -0.1/-0.7 | 0.8/1.3 | -0.3/-1.2 | -0.5/-0.9 |
| 2012 | 3.0/0.9 | 1.7/1.6 | -0.9/-1.0 | 3.7/3.8 | 0.6/0.7 |
For reasons of space limitation, for now I omitted the series for real GDP, real final sales, real aggregate payroll growth, and real disposable personal income per capita.
The top 5 variables identified by Silver mostly give me the same result: if they were consistent, Humphrey would have won in 1968, and Ford in 1976. Both Humphrey and Ford had objectively great economic numbers to run on. Further, while poor economic performace as in 1980 and 2008 are a death sentence for the incumbent party, mediocre numbers are less dispositive. If they were, Clinton in 1992, Bush in 2004, and Obama in 2012, all would have lost.
That being said, the single variable with the best record is the change in the unemployment rate. It has correctly forceast 13 of the 15 elections since 1956. If the unemployment rate is rising in Q2 and Q3 of the election year, the incumbent party is going to lose. If it is declining, with two exceptions, (Humphrey in 1968, Gore in 2000) the incumbent party is going to win.
Meanwhile, Surowiecki's measure of real personal income calls for Clinton and Obama to lose their re-election bids, and also misses 1968 and 1976.
Note that when a trend has markedly accelerating or decelerating during the election year, the voters have reacted to that. In 2000, it was clear by the end of Q3 that the economy was decelerating markedly. The reverse was true in 2004.
Looking at Silver's most valuable indicators, in general good growth will lead to re-election of the incomumbent, and usually re-election of the incombent party. A downturn, unsurprisingly, results in the incoumbent or their party being ousted. Mediocre growth leads to more incomsistent results, although the change in the unemployment rate stands out as having the most promise.
That being said, none of even the best economic variables have an infallible record. The economiy is not always the dominant issue in an election. For lack of a better word, the economy can be trumped by "moral" issues -- winning or losing a war, a big increase or decrease in crime, Watergate and the Nixon pardon, The Most Expensive Blow Job in History, and even in 2004 the spectre of gay marriage.
In my next post, I will look at one of the long leading economic indicators, and what it can already tell us about employment and unemployment next year.
