Why do newspapers and broadcast organizations publish polls? Well, it’s information, it’s news, sort of manufactured news, but still of interest to consumers. But I suspect the bigger reason is marketing.
Whenever your poll is quoted, your news organization’s name gets mentioned. It enhances your reputation and puts you on the shopping list for web- and channel-surfers and maybe even potential subscribers.
Is this marketing plan upset if your poll is wrong?
This question came to mind as I read this column by political science professor Todd Eberly on National Review’s website, which raises some intriguing questions about two recent polls by the Washington Post/ABC and Newsweek. Their most recent polls made news because they showed Sen. Barack Obama opening up a big lead.
With these two polls, the political junkies’ much-watched Real Clear Politics average of polls jumped up to 8.1 percent at this writing. For a couple of weeks, Obama’s lead had been hovering at around 4-5 percent on RCP. (In fact, since the National Review column, the LA Times/Bloomberg and CBS/New York Times polls were released showing Obama with leads of 9 and 14 percent respectively, boosting his margin even higher.)
Looking at the WaPo/ABC poll and the Newsweek poll, Eberly discovers that both polls assume that far more Democrats will turn out to vote on Election Day than Republicans. Newsweek’s weighting assumes the following turnout scenario:
- Democrats, 40 percent
- Republicans, 27 percent
- Independents, 30 percent
Which means the Democratic turnout edge will be 13 percent. WaPo/ABC assumes a Democratic turnout advantage of 9 percent.
Such ratios would represent a profound shift in party identification. According to Eberly’s research, Democrats had only a slight turnout advantage in most of the past few elections:
- 2006: 2 points
- 2004: Tied
- 2000: 4 points
- 1996: 4 points
Using the historical data and assuming a high-end turnout differential of four points, Eberly claims the two polls under examination would actually show Obama with a four-point lead — a lead consistent with the recent results published by a couple of independent pollsters including John Zogby. If one assumes 2006 is the most relevant point of comparison because of its recency, Obama’s lead would only be two points.
What this means to the presidential campaigns is not my concern here. You’ll get arguments on one side saying Obama has excited a massive wave of new voters that will leave historical models in the dust. You’ll get arguments on the other side saying these news organizations are biased and the bias extends to how they weight their data. Dunno.
What interests me is the business impact. Doesn’t it hurt the Washington Post, ABC or Newsweek if their polls turn out to be wrong?
For whatever reason, these three news organizations have attached their names to polls that have adopted assumptions starkly different from the actual data of the recent past. It’s a roll of the dice. If they’re wrong, and if their error derives from making unusual, unprecedented assumptions that don’t prove out, the effect on their reputations will outlast the polls’ effects on the news cycle. Won’t they?
The news business is in trouble. Circulation over the past few years has dropped like a rock for most daily newspapers, as have TV news ratings, in part because the mainstream media aren’t trusted implicity the way they used to be. If these polls turn out to be way wrong, and if the turnout assumptions turn out to be the underlying reason, political insiders, bloggers and commentators will zing them publicly, as will their competitors in the political research business. Is it prudent for these news organizations to gamble with their reputations this way?