September 9 2009

The Bread and Peace Model Applied to the 2008 US Presidential Election

The 2008 Presidential Election Popular Vote Result: The official final vote count released on January 22 2009 by the Federal Election Commission gives John McCain 59,934,814 votes out of 129,391,711 votes cast for McCain and Barack Obama. McCain’s 2-party vote share is therefore 46.3% and Obama’s margin of victory is 7.4 percentage points.

On June 7 2008 I posted the essay “The Implications of the ‘Bread and Peace’ Model for the 2008 US presidential election outcome.” At the time I predicted a Republican 2-party vote share centered on 48.2% which implied that the odds of the Democrats taking the presidency were better than 3 to 1. The June 7 essay was published in the October 2008 issue of Public Choice.

A shorter, more casual essay discussing the Bread and Peace model’s predictions “Why Are the Democrats Likely to Win the 2008 US Presidential Election?” was published in translation to German as an op-ed piece in Welt am Sonntag on August 24 2008; “Das Brot-und-Frieden-Modell – Warum Obama gewinnt.”

The preliminary estimates of 2008q3 national income data released on October 30 2008 by the Bureau of Economic Analysis indicated that the US economy had weakened so much since the preceding summer that I updated my June 7 2008 election forecast on October 31 2008. Based on those preliminary estimates the Bread and Peace model predicted a 2-party vote share for McCain of 46.25%, implying Obama would win by a margin of 7.5 percentage points. Although the Bread and Peace model is designed to explain votes for president in terms of just two fundamental political-economic factors rather than to forecast optimally election outcomes, the October 31 prediction was extremely accurate: “October 31 2008 update of Presidential Vote Forecast.”

On August 28 2009 I re-estimated the Bread and Peace model including the 2008 election result (N=15 elections, 1952-2008) using the revised disposable personal income data for 2008 and earlier years released by the Bureau of Economic Analysis on August 27 2009. The coefficient estimates are nearly identical to those obtained for earlier samples:

Coefficient estimate:	α = 46.0 λ = .91	β₁= 3.6	β₂ = -.052		Adj R² = .86
(Std. error\|p-value):	(1.1\|.00) (0.05\|.00)	(0.55\|.00)	(0.01\|.00)		Root MSE = 2.19

The graph below shows the strong association of aggregate votes for president to average per capita real income growth in combination with US military fatalities suffered in Korea, Vietnam and (the comparatively small number) in Iraq. Those two fundamental factors explain postwar votes for president remarkably well. Note that the allegedly “ideological” elections of 1964 (when Lyndon Johnson trounced Barry Goldwater) and 1980 (when Jimmy Carter was routed by Ronald Reagan), which anchor the extremities of postwar election outcomes, are explained perfectly by the Bread and Peace objective fundamentals.

Here are the Stata do and Stata dta files used to estimate the regression and create the graph shown above. And here are a pdf of the Stata input program and output results and an XLS file giving the input data. A pdf of two Bread and Peace model graphs is here.

Background Literature

I first proposed that an over-the-term weighted-average of per capita real disposable personal income growth rates was the best way to quantify how the economy affects aggregate votes for president in my 1982 article in the American Politics Quarterly “President Reagan’s Mandate From the 1980 Elections: A Shift to the Right?”. The same single-variable set-up was used in chapter 6 of my 1987 Harvard University Press book The American Political Economy: Macroeconomics and Electoral Politics.

The first published write-up of the Bread and Peace model was in my 2000 article in Public Choice “Bread and Peace Voting in U.S. Presidential Elections.” Various properties of the model were also discussed in my 2007 paper “The Economy, the War in Iraq and the 2004 Presidential Election.” The basic Bread and Peace equation was also used to generate presidential vote predictions in memos circulated or in web site documents posted prior to the 1992, 1996, 2000 and 2004 elections.

I reviewed the broader literature on macroeconomic conditions and voting behavior in my 2006 essay for the Oxford Handbook of Political Economy “Voting and the Macroeconomy.”