Estimating the
Expected Vote for Buchanan in Palm Beach County, Florida:
An Alternative Model Based on Prior Support for Buchanan
Author:
Greg Thorson
University of Minnesota at Morris
Recently, a great deal of national attention has focused on the voting behavior of Palm Beach County, Florida. Democrats have alleged that the “butterfly” ballot used in the county resulted in a significant number of voters who intended to vote for the Democratic candidate Al Gore instead mistakenly casting votes for the Reform Party candidate Pat Buchanan. Indeed, Buchanan indicated on the November 9th edition of NBC’s Today Show that he did not believe all of the votes he received were intended for him and were probably intended for Gore. Buchanan even suggested that the margin might have been large enough to affect the outcome of the 2000 presidential race.
Not surprisingly, both major political parties have offered different explanations for the high levels of support for Buchanan in Palm Beach County. Democrats argued that the votes cast for Buchanan were intended for Gore and supported their claims by pointing to the high number of votes for Buchanan relative to his support levels in other counties. Republicans countered that Buchanan could have legitimately earned that many votes and cited the fact that the Reform Party has seen a higher than average increase in registrations in Palm Beach County.
Because the presidential race will likely be decided by a very small margin, the topic warrants further investigation. Other scholars (see Adams and Fastnow for an up-to-date list of analyses) have already developed models to investigate the question. Virtually all have concluded that Buchanan was unlikely to have received that many votes in Palm Beach County.
My model varies a bit from previous work in that I use different indicators and statistical methods to predict support for Buchanan in Palm Beach County. I test the model that the county level support for Buchanan can be predicted by examining two indicators. The first indicator is each county’s support for Buchanan in the 1996 Presidential Primary. In 1996, Buchanan won 18.1% of the vote in the Florida Presidential Primary. I speculate that each county’s support for Buchanan in the 1996 primary will be positively correlated with Buchanan’s support in the 2000 presidential general election.
The second indicator I use is the change in the number of Reform Party registrations between 1996 and 2000. Most of the counties in Florida experienced significant increases in the number of Reform Party registrations. Of course, this pattern of growth was not uniform. I predict that those counties that saw the most rapid growth in Reform Party registrations would also be more likely to support Buchanan. All of the county level data used in this analysis were obtained directly from the Florida Secretary of State’s web site.
I used Ordinary Least Squares Regression (OLS) to estimate the parameters. Due to the concerns over Palm Beach County, I omitted them during the parameter estimation stage. The key results are displayed in Table 1.
Table 1. Factors Predicting Support for Buchanan in Florida Counties, 2000 Presidential Election
R-Squared: .89
Adjusted R-Squared: .88
|
Variable |
Unstandardized Coefficient |
Standard Error |
Standardized Coefficient |
T-Score |
Significance |
|
Constant |
47.50 |
12.39 |
|
3.84 |
.000 |
|
Support for Buchanan in 1996 Primary |
0.02062 |
0.007 |
0.260 |
2.97 |
.004 |
|
Change in Reform Party Registration |
2.668 |
0.331 |
0.705 |
8.06 |
.000 |
Overall, the model is extremely predictive. The two variables predict 89% of the variance in support for Buchanan among counties in Florida. Both variables easily pass traditional tests of statistical significance.
For each increase in the number of Reform Party registrants, Buchanan could expect to gain approximately 2.7 votes. Additionally, Buchanan would be expected to pick up approximately 2 votes for each 100 cast for him in the 1996 Presidential Primary. An analysis of the standardized coefficients suggests that the change in the Reform Party Registration was indeed more predictive of support for Buchanan than the support he received in the 1996 Presidential Primary.
So what does the model predict for Palm Beach County? The model predicts that Palm Beach County voters would cast 898 votes for Buchanan. The 95th Confidence Intervals are 588-1207 votes. As previous scholars have suggested, it seems extremely unlikely that Buchanan would win 3407 votes in Palm Beach County.