HOME

Douglas Hibbs

31 July 2006

(edited-updated 17 December 2007)

Bread and Peace Voting in US Presidential Elections:

The Economy, the War in Iraq and the Vote for Bush in 2004

Background

Incumbent George W. Bush achieved a relatively narrow re-election victory in 2004. Yet unlike his victory as the Republican challenger in the 2000 election when he won the Electoral College vote but received fewer popular votes than the Democrat’s candidate Al Gore, in 2004 Bush attracted a 51.2% majority of the total Bush/Cheney v. Kerry/Edwards popular vote.

My analysis of how development of the economy and the war in Iraq affected the 2004 election outcome is based on research reported in my article “Bread and Peace Voting in U.S. Presidential Elections” (Public Choice, 2000). The ‘Bread and Peace model’ assumes that postwar American presidential elections should for the most part be interpreted as a sequence of referendums on the White House party’s economic record. My research showed that politically relevant economic performance is best measured by a weighted-average of quarterly growth rates of per capita real disposable personal income, computed from the election quarter back to the first full quarter of each presidential term. Growth of per capita real disposable personal income is probably the broadest single aggregate measure of changes in voters’ economic well-being, in as much as it includes income from all market sources, is adjusted for inflation, taxes, government transfer payments and population growth, and tends to move with changes in unemployment.

The only additional factors I found that significantly affected votes for President in the postwar era were the discretionary US military interventions in the Korean and Vietnamese civil wars. My research indicated that the electoral penalties exacted by Korea and Vietnam fell almost wholly on the party of the President initiating the commitment of US forces (the “war party,” in both those cases the Democrats), and were proportionate to the cumulative numbers of American military fatalities at the dates of the 1952 and 1968 presidential elections, respectively. Applying the Bread and Peace model to the 2004 election, I place the American invasion of Iraq (but not Afghanistan) on the same footing as US involvement in Korea and Vietnam: An unprovoked hostile deployment of American armed forces in a foreign conflict never sanctioned by a formal Congressional declaration of war.

The Bread and Peace Equation

Quantitative estimates of the effects on votes for President of per capita real income growth and the cumulative number of American military fatalities in Korea, Vietnam and most recently Iraq were obtained by estimating the following equation:

where

· Vote is the percentage share of the two-party vote going to the candidate of the incumbent party

· R is per capita disposable personal income deflated by the Consumer Price Index. is the annualized quarter-on-quarter rate of growth, . At the election quarter (j=0) the weighting parameter is scaled down to 1/3 because of the within-quarter date of presidential elections (the first Tuesday following the first Monday of November)

· KIA is the cumulative number of American military fatalities (in 1000s) in Korea, Vietnam and Iraq during the presidential terms preceding the 1952, 1968, 1976 and 2004 elections. (Technical information about how KIA in Korea and Vietnam were determined to affect the 1952, 1968 and 1976 elections, but not the 1972 election, is given in the appendix to my Public Choice 2000 article.)

Note that the Bread and Peace model is designed to explain voting outcomes in terms of political-economic fundamentals rather than to predict elections using pre- election poll data on voter sentiments, preferences and the like. Such attitudinal variables are themselves generally affected by objective fundamentals and for that reason supply no insight into the ultimate causes of voting behavior.

Estimates, Fits and Predictions

Table 1 shows nonlinear-least-squares estimates of the Bread and Peace equation for presidential elections spanning 1952-2004. The model was fit using the latest (July 2006) data on personal incomes from the Bureau of Economic Analysis and consumer prices from the Department of Labor, and data on US military fatalities in Korea, Vietnam and Iraq originating with the Department of Defense. According to the coefficient estimates in Table 1 each percentage point of growth in per capita real disposable personal income sustained over the presidential term boosts the in-party candidate’s vote share by 3.6 percentage points above a benchmark constant of approximately 46 percent. And hostile deployment of US armed forces in unprovoked, discretionary wars depresses the incumbent’s vote share by about 0.3 percentage points per 1000 American military fatalities.

Table 1. Bread and Peace Equation Estimates
Incumbent Vote Share	1952 – 2004	N = 14 elections
	R^2 = .866	Adj R^2 =.826	Root MSE =2.42
	Coef. Estimate	Std. Error	t-ratio
Constant ()	46.2	1.24	37.3
Real Income Growth ()	3.61	0.615	5.87
Weighting parameter ()	0.914	0.058	15.9
Cumulative KIA ()	-0.307	0.078	-3.95

Expressed on an annual basis, the weighted average per capita real income growth rate during Bush’s first term was 1.72%. US military fatalities in Iraq stood at 1.13 thousand at the end of October 2004. The within-sample prediction (regression fit) of Bush’s two-party vote share from the Bread and Peace model is therefore 52.08%, which gives a prediction error of -0.84%: ; . The model estimates indicate that the Iraq war made only a small dent in the vote for Bush – depressing his two-party share by around 1/3 of a percentage point. However, if casualties continue to mount all the way up to the next election, Iraq could have decisive effect on the 2008 outcome, particularly if on economic grounds alone the election would likely be close. But in 2004 economics dominated the fundamental sources of Bush’s re-election.

The out-of-sample prediction of the 2004 election result is almost as good as the within-sample prediction/fit. Estimation over the sample range 1952-2000 yields coefficients nearly identical to the full sample estimates () and yields an out-of-sample prediction of 52.16% for Bush’s vote, implying a prediction error of -0.92%. George Bush’s narrow 2004 victory is then very well accounted for by the political-economic fundamentals in the Bread and Peace model. Table 2 reports actual and predicted vote shares for all elections in the postwar sample generated by the estimates in Table 1, along with election period values of the real income growth and KIA independent variables. Contrary to the line of argument in William Nordhaus’ June 2006 Quarterly Journal of Political Science article “Electoral Victory and Statistical Defeat? Economics, Politics and the 2004 Presidential Election,” the data in Table 2 indicate that 2004 is among the better explained postwar elections.

Table 2. Candidates, Votes, Predictions and Performance
In-Party v. Out-Party Candidates	Election Year	Incumbent % Vote Share	Predicted % Share	Regression Error	Weighted-avg. Real Income Growth	Cumulative KIA (1000s)
Stevenson v. Eisenhower	1952	44.60	45.90	-1.30	2.40	29.260
Eisenhower v. Stevenson	1956	57.76	56.65	1.11	2.89	0.00
Nixon v. Kennedy	1960	49.91	49.28	0.63	0.85	0.00
Johnson v. Goldwater	1964	61.34	61.43	-0.09	4.21	0.00
Humphrey v. Nixon	1968	49.60	48.23	1.37	3.02	28.896
Nixon v. McGovern	1972	61.79	59.27	2.52	3.62	0.00
Ford v. Carter	1976	48.95	50.00	-1.05	1.08	0.414
Carter v. Reagan	1980	44.70	44.82	-0.12	-0.39	0.00
Regan v. Mondale	1984	59.17	60.16	-0.99	3.86	0.00
GHW Bush v. Dukakis	1988	53.94	54.41	-0.47	2.27	0.00
GHW Bush v. Clinton	1992	46.55	47.58	-1.03	0.38	0.00
Clinton v. Dole	1996	54.74	49.98	4.76	1.04	0.00
Gore v. GW Bush	2000	50.27	54.75	-4.49	2.36	0.00
GW Bush v. Kerry	2004	51.24	52.08	-0.84	1.72	1.130

A Refined Bread and Peace Equation: Scaling KIA to Population

The US population grew from 158 million at time of the 1952 election to 295 million at the time of the 2004 election. A sensible refinement of the original Bread and Peace model would be to scale the KIA variable by population size. Doing so yields a slight improvement to the model’s fit statistics. As shown in Table 3 below, the Adjusted R² increases from .826 to .836 and the Root Mean Square Error decreases from 2.42 to 2.35.

Table 3. Bread and Peace Equation Estimates with KIA scaled to population
Incumbent Vote Share	1952 – 2004	N = 14 elections
	R^2 = .874	Adj R^2 =.836	Root MSE =2.35
	Coef. Estimate	Std. Error	t-ratio
Constant ()	46.3	1.20	38.5
Real Income Growth ()	3.55	0.595	6.03
Weighting parameter ()	0.908	0.057	16.0
Cumulative KIA per million population()	-0.050	0.012	-4.15

The coefficient estimates in Table 3 imply KIA and (especially) real income growth effects that are very similar to those of the original Bread and Peace model. The within-sample prediction of Bush’s 2004 two-party vote share is 52.23% and so the prediction error about -1%: ; . Like the original equation, the Bread and Peace model with military fatalities scaled to population indicates that the Iraq war had minor impact on the 2004 vote – reducing the two-party share going to Bush/Cheney by as little as 1/5 of a percentage point.

The Figure below graphs the strong connection of votes for President to real income growth over the term. (The regression line in blue is based on Table 3 estimates; Table 2 estimates would have shown nearly the same relation.) Cumulative US military fatalities at the times of the 1976 and 2004 elections were too small (414 and 1130, respectively) to exert much influence. The big KIA effects were in 1952 (Korea) and 1968 (Vietnam). In both cases the high fatality levels (29,260 or 197 per million population in Korea and 28,896 or 152 per million in Vietnam) most likely deprived the in-party Democrat candidates of victory. However, as I mentioned earlier, things may be different in 2008. By the end of July 2006 American fatalities in Iraq had reached 2540, a US exit strategy had not yet materialized, and the accumulation of American body-bags was showing no sign of slowing down.

The only postwar presidential election results not well accounted for by the Bread and Peace model are 1996 and 2000. The vertical deviations of the 1996 and 2000 outcomes from the regression line are noticeably larger in those non-war years than in all the others. (See also the regression errors in Table 2.) A partisan of the Bread and Peace like myself model might be tempted to conjecture that idiosyncratic influence of candidate personalities took especially strong form in 1996 and 2000 – with the ever charming Bill Clinton looking especially attractive when pitted against the darkly foreboding Bob Dole in 1996, and the unfailingly wooden Al Gore paling by comparison to the affable George Bush in 2000.

Here you may obtain the Stata program and Stata data file used to generate statistical results discussed at this web page. A somewhat longer analysis of the same issues is given in my paper “The Economy, the War in Iraq and the 2004 Presidential Election.” This study is targeted partly on William Nordhaus’ article “Electoral Victory and Statistical Defeat? Economics, Politics, and the 2004 Presidential Election.” in the Quarterly Journal of Political Science, which in turn was based mainly on the implications for the 2004 election result of various equations proposed by Nordhaus’ Yale economics colleague Ray Fair. My paper was submitted to and ultimately rejected by the QJPS. The QJPS submission, review, re-submission and rejection history is available here.