Monday, October 30, 2006
Poll Fatigue
With the November Midterm elections fast approaching, it behooves us to better scrutinize the usefulness of polling as a tool to assess popular opinion.
Since the 1936 elections, public-opinion polling has emerged as a mainstay in American politics. Founded on a commitment to provide an accurate snapshot of public opinion, the science of mass polling has been perfected over the decades to discount external factors. Pollsters know well, for example, that approval ratings for Democrats tend to slip on the weekends because young, educated singles, reliably Democratic, are out partying (see Michael Schwartz’s The Opiate of the Electorate). An earthquake in Pakistan, a significant day on Wall Street, or a recent House scandal can likewise distort the findings of a poll. The difficulties in effective polling, in fact, are emblematic of the larger challenge in Social Sciences to subject human behavior to rigorous scientific research.
To cite but one example, a recent poll conducted by the Rasmussen Reports shows Bob Corker (R) leading his Democratic opponent, Harold Ford Jr. (D), by a single point in the 2006 Tennessee Senate race. Most experts consider this race a “Toss-Up”.
Some experts, however, suggest that Ford may not enjoy the support the polls imply. David Bositis, a Senior Political Analyst with the Joint Center for Political and Economic Studies, spoke today on NPR’s Morning Edition about a tendency for respondents to rate favorably a black candidate over a white candidate in order to appear socially tolerant. Known as the “Sociable Desirability effect”, some people express an intention to vote for the black candidate while planning to do otherwise. Bositis cited a 1989 exit-poll survey showing Doug Wilder, the gubernatorial candidate in Virginia, held a ten percent lead over his opponent among voters. He won, however, by less than one-half a percent, prompting an automatic recount. Ford’s recent gains in the polls may partially reflect voters’ impressions, not their intentions.
Mass polls are reliable for measuring broad trends in public opinion, but not for specific predictions, especially during an election season.
With the November Midterm elections fast approaching, it behooves us to better scrutinize the usefulness of polling as a tool to assess popular opinion.
Since the 1936 elections, public-opinion polling has emerged as a mainstay in American politics. Founded on a commitment to provide an accurate snapshot of public opinion, the science of mass polling has been perfected over the decades to discount external factors. Pollsters know well, for example, that approval ratings for Democrats tend to slip on the weekends because young, educated singles, reliably Democratic, are out partying (see Michael Schwartz’s The Opiate of the Electorate). An earthquake in Pakistan, a significant day on Wall Street, or a recent House scandal can likewise distort the findings of a poll. The difficulties in effective polling, in fact, are emblematic of the larger challenge in Social Sciences to subject human behavior to rigorous scientific research.
To cite but one example, a recent poll conducted by the Rasmussen Reports shows Bob Corker (R) leading his Democratic opponent, Harold Ford Jr. (D), by a single point in the 2006 Tennessee Senate race. Most experts consider this race a “Toss-Up”.
Some experts, however, suggest that Ford may not enjoy the support the polls imply. David Bositis, a Senior Political Analyst with the Joint Center for Political and Economic Studies, spoke today on NPR’s Morning Edition about a tendency for respondents to rate favorably a black candidate over a white candidate in order to appear socially tolerant. Known as the “Sociable Desirability effect”, some people express an intention to vote for the black candidate while planning to do otherwise. Bositis cited a 1989 exit-poll survey showing Doug Wilder, the gubernatorial candidate in Virginia, held a ten percent lead over his opponent among voters. He won, however, by less than one-half a percent, prompting an automatic recount. Ford’s recent gains in the polls may partially reflect voters’ impressions, not their intentions.
Mass polls are reliable for measuring broad trends in public opinion, but not for specific predictions, especially during an election season.
# posted by Jonathan Noble @ 4:49 PM 
Comments:
<< Home
Excellent observation re: the Ford race, and generally good overview of some key dilemmas in accurate polling.
Post a Comment
<< Home
