There are a number of comments that appear repeatedly (indeed, nearly every time) the topic of a given poll--or recent spate of polls--arises, whether as a dedicated diary or as a comment within a thread. In most cases, these comments attempt to dismiss the relevance of the poll (especially if the poll is not favorable to our candidate). While I believe that a healthy degree of skepticism is warranted regarding any poll, I also believe it important that we not unquestioningly perpetuate a series of assertions that are simply not true.
In particular, I would like to address three of the most common false (or perhaps better said, unproven) assertions:
- National polls are meaningless (since it is the electoral college that determines the winner of the election).
- All polls will skew to the right, since, it is asserted, pollsters do not call cell phones, and younger voters (who tend strongly toward Obama) often have only cell phones and no land lines, and will thus be unrepresented.
- Certain polls (especially Fox News and Zogby) can safely be ignored due to bias in the case of Fox News, or incompetence in the case of Zogby.
Let me offer rebuttals to all three of these arguments.
- While it is true that it is possible for a candidate to lose the popular vote but win in the electoral college, this in fact occurs so rarely that it is misguided in the extreme to discount the relevance of national polls. Aside from the 2000 election (which will receive further consideration below), the last time a candidate was elected president without winning the popular vote was in 1888(when even John McCain was still a youngster). In other words, the 2000 election excepted, every president elected has won the popular vote for the last 120 years.
Regarding the 2000 election: As is now common knowledge, had all of Florida's votes--including those votes accurately tallied during the initial recount as well as so-called "undervotes" and "overvotes"--been counted, Gore would have won the plurality of the Florida vote and thus the election. Moreover, had it not been for the butterfly ballot fiasco, Gore's margin would have been padded further. In other words, had voters' intentions (which is all a poll can show) been accurately calculated, there can be little doubt that the winner of the nationwide popular vote would have been elected president. It has been 120 years, then, since a legitimately-elected president lost the popular vote. Accordingly, not only is it inaccurate to claim that national vote is "meaningless," but, to the contrary, it might be argued that the national vote is an extraordinarily reliable indicator of who will win the election.
- In some cases, pollsters do indeed contact voters via cell phones. Gallup, among the most often-maligned pollster on these pages, makes this clear in the explanation of their survey methods; so too does the Associated Press. Clearly, then, the oft-voiced claim that pollsters don't call cell phones because it is "illegal" for them to do so is simply false.
Unfortunately, other polling organizations do not make clear in their statements of methodology whether they poll cell-phones or not. Absent any hard evidence , I see no reason to make the assumption that they exclude cell phones, especially if they use random-digit dialing.
The above information does not, of course, address questions regarding habits of cell-phone users (whether, for example, they are less likely than land-line users to answer a call from an unfamiliar number). Such considerations, however, are at this point mere anecdotal conjecture, and don't strike me as particularly reliable in any assessment of the merits of a given poll.
- Counterintuitive though it may be Fox News's polls do not seem to have an inordinate "house effect"--they skew approximately 1.5 points to the right. In fact, during the final days of the 2004 election, they were among only a small handful of polling outfits to place Kerry in the lead, by two points.
While Zogby's "interactive" internet-based polls are absurd, and little (if any) better than drawing numbers out of one's ass, his standard telephone polling is highly accurate. In the estimation of pollster.com, the Zogby "house effect" is only a single point (to the right, even though he is a Democrat himself), and his popular-vote predictions in both 2000 (Gore 49, Bush 48) and 2004 (Bush 49.4, Kerry 49.1) were extremely close to the eventual totals.
So let's continue to think critically about polls and assess their merits, hopefully in the service of refining our strategies on how to get Obama elected. But let's make sure that when we do so, we are dealing with the facts (as near as they can be determined), and not repeating inaccurate claims or accepting them as true simply because we've seen them repeated so many times.