Polls or Pols? the Real Driving Force Behind Presidential Nominations

Article excerpt

Since the publication in 1971 of David Broder's influential book The Party's Over, pundits and scholars have written of the decline of political parties. Although observers now recognize a party "revival" of sorts in Congress, within the electorate, and in the recruitment and funding of congressional candidates, most regard presidential nominations as driven by momentum, money, and the luck of the state-by-state sequence of contests. Few analysts see signs of party influence. We hope to put another nail in the coffin of the party-decline thesis by arguing that party elites have regained a large measure of control in presidential nominations.

Polls vs. Pols

In his recent essay "Forecasting Presidential Nominations," William Mayer shows that polls taken before the start of presidential primary voting predict candidate primary vote shares extremely well from 1980 to 2000. We have collected data that show the same thing. Across all contested nominations since 1980, the final Gallup poll before the Iowa caucus explains around 90 percent of what happens in the state-by-state voting. (In technical terms, the correlation between pre-Iowa polls and delegates won in the primaries is r=.93.)

Taken at face value, this correlation seems to show that a candidate's pre primary public support reliably determines the outcomes of presidential primaries. But our observation of nominations has led us to believe that party insiders also play a big role. To test this possibility, we developed a measure of insider support for presidential candidates by tallying all publicly reported endorsements in a broad range of publications, including local and national newspapers, political magazines and newsletters, news wires, and Internet sources. The measure covers all candidates in contested nominations from 1980 to 2000 and is based on endorsements made before the Iowa caucus.

Some of the endorsements, which averaged roughly 300 to 400 per contest, were more important than others. In 2000, for example, Michael Jordan endorsed Bill Bradley and Bill Clinton endorsed Al Gore. To capture such differences, we sorted endorsers into categories--celebrity, state legislator, incumbent president--and asked knowledgeable coders to rate the significance of people in each category. We then weighted our endorsement measures according to these judgments so that each candidate's score could be interpreted as his or her percentage share of politically important endorsements. The correlation between the share of delegates won in the primaries and the weighted endorsements is .89, which is nearly as high as that between poll share and delegate share. Moreover, a multiple regression analysis shows that polls and endorsements are almost equally important predictors of primary outcomes: each 1 percent of pre-Iowa poll share adds 0.68 percent of delegate share; each 1 percent of pre-Iowa endorsements adds 0.60 percent of delegate share.

These results, however, are ambiguous. The party insiders who make endorsements do so against a backdrop of nearly continuous polls about voter preferences. What if these endorsers simply follow these polls? They would have several incentives to do so. They might want to choose a poll frontrunner to settle on a winner as quickly as possible to avoid intra-party bloodshed. They might simply want to pick a popular candidate. Some endorsers, wishing to ingratiate themselves with the likely winner, might jump on the poll leader's bandwagon. In any of these scenarios, public support, as measured in polls, would be the real cause of primary success, and endorsements would be secondary.

But an entirely different scenario is equally tenable. Political insiders know that pre-Iowa polls may reflect little more than name recognition--and perhaps confused recognition at that. In the current cycle, supporters of John Edwards joke nervously that their candidate's poll standing may be partly due to confusion with the popular psychic from the television show "Crossing Over. …