The Monkey Cage

To answer this question, I rely on an Washington Post/ABC Poll from October 29 - November 1, 2007. (Thanks to Jon Cohen for sharing these data, which are here.) This poll asked whether respondents felt favorably or unfavorably toward each of the candidates. Jon kindly broke down this item by Democratic and Republican respondents. For each candidate, I generated separate favorability scores for Democratic respondents and Republican respondents, where the score is the percent favorable minus the percent unfavorable. So, for example, 78% of Democrats viewed Hillary Clinton favorably, and 18% viewed her unfavorably, so her net favorability score is 78-18=60. Her score among Republicans is 19-79=-60.

Then I computed the “distance” between Democratic and Republican respondents by subtracting the two favorability scores, and taking the absolute value. This becomes the “polarization score.” For Hillary, 60-(-60)=120. The maximum here would be 200, i.e., if every Democratic loved Hillary and every Republican hated her. The minimum is 0, i.e., if Republicans and Democrats felt the same.

What are the results?

Clinton 120
Obama 76
Giuliani 67
Thompson 59
Edwards 54
Romney 49
McCain 45
Huckabee 23

Of these candidates, Clinton is the most polarizing by a significant margin. However, Obama is actually the second most polarizing. If Democrats wanted a “post-polarization” candidate, then Edwards seems to better fit the bill. Among the Republicans, Giuliani is the most polarizing and Huckabee the least.

However, not even Huckabee should start proclaiming himself a “uniter not a divider.” The candidates with the lowest polarization score were also the candidates with the highest percentage of respondents who couldn’t rate them. For example, compare Clinton to Huckabee: 4% did not evaluate Clinton, but 49% did not evaluate Huckabee.

This suggests that as the campaign goes on and voters become more familiar with all of the candidates, the polarizing impact of even the lesser-knowns will increase. (This is likely what more recent data would find.) Political figures do have the ability to affect polarization (see Jacobson), but polarization in American politics has other fundamental causes (see McCarty, Poole, and Rosenthal) .

Addendum: In response to Aaron’s question below, I re-computed the polarization score in a way that assumes the following: the respondents in each party who could not rate the candidates would come to have preferences with the same ratio of favorable to unfavorable responses to each candidate. Democratic respondents’ favorability score for each candidate is thus calculated as: 100*(% fav/(% fav + % unfav)), and similarly for Republican respondents. Subtract the two scores, take the absolute value, and multiply by 2 to get polarization scores on the same 0-200 interval. Here are the results:

Clinton 124
Obama 86
Thompson 84
Giuliani 73
Romney 71
Edwards 64
McCain 52
Huckabee 45

Note that everyone’s score goes up, as I originally suggested they would. Clinton and Obama are still the most polarizing Democrats. Thompson, Giuliani, and Romney are the most polarizing Republicans. The less polarizing candidates are Edwards, McCain, and (again) Huckabee. Of course, as a “prediction” of what would happen if every respondent had an opinion, these results should be taken with a significant grain of salt, since this exercise depends on the assumption noted above.