| || || ||
| ||ONLINE EXTRAS/What Americans Think|
Everyone Loves a Winner
By Richard Morin
Washington Post Polling Director
Monday, January 31, 2000
"Just win, Baby." Al Davis, the bad-boy owner of the Oakland Raiders, was talking football. But he just as well might have been talking politics, where candidates, political strategists and the media remain convinced that winning breeds winning in the presidential primaries.
Call it momentum. Call it the bandwagon effect. Or call it both, as many scholars do. With the presidential primary season hard upon us, everybody's talking about momentum: Who has it, who's losing it and whether it exists at all.
Let others ponder Big Mo. A new Washington Post-ABC News poll suggests there may be another kind of momentum: The benefit a candidate gets in early primaries simply because he or she is viewed as having the best chance of winning in November. Call it Little Mo. Or perhaps pre-Momentum momentum. Unlike Big Mo, it's not a function of winning. Little Mo shapes voter preference even before a single vote is cast. The latest Post-ABC News poll in New Hampshire suggests many New Hampshire voters decide, in part, which candidate to support in the primary on their assessment of which candidate can win in November.
That survey found that candidate preference was closely associated with whether or not voters believed that candidate had the best chance of winning in the fall. Nearly seven in 10 Democrats who said that Vice President Gore had the best chance of winning in November currently support Gore over former Sen. Bill Bradley of New Jersey. Conversely, among those who say Bradley has the best chance to take it all, nearly nine in 10 are voting for him.
The effect is likewise apparent among Republicans. Among those who believe John McCain has the best chance to win, the Arizona senator claimed nearly eight in 10 votes. Among those who believe Gov. George W. Bush has the best chance of beating the Democratic nominee, Bush had the clear advantage.
Oh pooh, the statistically savvy might say. That's the oldest error in the Stats 101 textbook. Little Mo doesn't influence voter preference. The "Winner Patina" is determined by the same factors that also motivate individuals to vote for one candidate over another. Being viewed as a winner is unimportant to vote choice; it merely is a byproduct of the factors that we already know push and pull voter preferences, such as perceptions of the candidates and positions on the issues, thoughtful critics might argue. Significant objections if true. As it happens, we can partially test these challenges. The latest Post-ABC poll conducted before the Iowa caucuses tested how voters saw the two front-runners in each party primary, asking if each was, among other things, a "strong leader," had the requisite experience to be president, understood the problems of ordinary people, and was not a "typical politician." In all, 10 candidate traits were measured.
The survey also asked respondents to name the issue that was most important to them in determining their vote. Taxes, health care, the economy, Social Security and Medicare reform led the list.
These data become the building blocks of a statistical model of the factors that are independently associated with a candidate. In theory, this model should tell us which variables independently predict whether, for example, a respondent is likely to be a Bush voter, or not. The model also will suggest the relative impact of each variable in predicting candidate preference. A friendly warning here: The statistically challenged may want to skip ahead a few paragraphs to the results, and call it math-a-magic.)
First, a statistical procedure called factor analyses distilled these 10 candidate traits into a smaller group of underlying factors. These factors are then included in a logistic regression model (more math-a-magic). Other independent variables included ones that identified each voter's top issue; the sex, age, education, ideology, party identification of respondents; and whether anybody in their household was a union member or a military veteran.
To this statistical stew was added an independent variable that indicated whether the respondent thought the candidate had the best chance of winning in November. If being viewed as a winner is truly associated with voter preference, this variable will emerge as statistically significant in the model. If Little Mo doesn't matter, the model will report that this variable doesn't help predict vote choice, all other factors in the model held constant.
The results, please.
The story these statistics tell is straightforward and dramatic: Being perceived as a winner is strongly related to vote choice, all other variables held constant. In the metric of the social sciences, the probability that this relationship is due to chance alone is less than .0001 next to nothing. Across each of the four candidates, someone who believes a specific candidate will win it all is about five to seven times more likely to be voting for that candidate in the party's primary even after such factors as issue preference and perceptions of the candidates are taken into account.
Of course a single, hastily constructed statistical model is inadequate to prove the existence or estimate the true impact of Little Mo on early primary voting. Other models using different or more highly refined variables might produce different results. But it's interesting, however, to note that other researchers have detected somewhat similar effects. Political scientist Larry Bartels, who literally wrote the book on political momentum, argued in the late 1980s that many voters use a simple winner/loser calculation to decide which candidate to support in primaries. He called them "strategic voters." Significantly, these voters will support a winner even if it sometimes means holding their noses and voting for a candidate they don't particularly like.
For these voters, the pre-nomination period thus becomes a process to select the candidate with the best shot of winning. It's not much of a stretch to suggest that the plethora of pre-election year horse race polls may help strategic voters settle on a likely winner and their choice even before the first primary votes are cast.
Archive: What Americans Think