Mark Blumenthal | January 10, 2010
Topics: Divergent Polls , Likely Voters , Martha Coakley , Massachusetts , Scott Brown , Turnout
We have two new polls out in Massachusetts on the January 19 special election to fill Ted Kennedy's Senate seat, and their results could not be more different. The new survey conducted Saturday through Wednesday last week by the University of New Hampshire Survey Center on behalf of the Boston Globe shows Democrat Martha Coakley leading by 17 percentage points (53% to 36%), while a new automated poll conducted on Thursday and Friday by Public Policy Polling (PPP) shows a dead heat, with Brown one point ahead (48% to 47%). A third survey conducted on Monday by Rasmussen Reports has Coakley ahead by nine (50% to 41%).
The disparity of the results is likely to provoke the usual angst about inconsistent polls, debates about past pollster accuracy and the customary conspiracy theories about intentional bias. Forgive me if I don't join in, because as different as these results seem to be, I think the discrepancies actually add up to a consistent and important finding on the state of voter preferences this past week.
Here are three things to keep in mind about polls on the special election:
Turnout Will Matter -- The big spread in results among the polls, and differences apparent within two of them, are all consistent in supporting one finding: The lower the turnout, the better the odds for Scott Brown. These differences indicate that the voters most interested and most likely to vote are Republican, while Democrats are more blase.
Consider the differences in the table below from within Globe/UNH and Rasmussen surveys. Both show a dead even race among the most interested and certain voters, while Coakley leads by huge double-digit margins among all other voters.
Those differences mean the overall results reported by any poll are going to be very sensitive to the "tightness" of the screen or likely voter model used. The more restrictive the screen, the closer the result. My assumption is that the "if you do not intend to vote...please hang up" automated methodology employed by PPP produced an effectively tighter screen and, thus, a likely voter sample closer to the "certain" or "extremely interested" subgroups of the Boston Globe and Rasmussen polls.
Pollsters can't predict turnout - I have yet to see any poll or statistical model that can predict voter turnout with precision, especially in an oddly timed special election like the one in Massachusetts. What pollsters try to do is monitor self reported enthusiasm and interest as compared to previous, comparable contests and try to calibrate their screens and models appropriately (although there is much debate among pollsters about the accuracy of those calibrations and their necessity).
The bigger challenge in predicting turnout, however, has to do with something more fundamental: The size and makeup of the electorate will depend on decisions not yet made by those who may or may not vote on January 19. How many will become more interested and decide to vote over the next 9 days? I'm not sure any poll or methodology can predict that with confidence.
Keep in mind that as of this past week, most Massachusetts voters assumed that Coakley would win in a walk. According to Globe/UNH poll, nearly three quarters (74%) of Massachusetts voters believe Coakley will win, while only 11% say the same about Brown. In that sense, news of a narrowing race could work to Coakley's advantage if it convinces Democrats that their votes are needed and that Ted Kennedy's seat could be lost to the Republicans without their help.
Turnout differences complicate trend tracking - The big spread in these poll results complicates our ability to spot trends. For example, PPP's Tom Jensen last night noted that they fielded their poll on Thursday and Friday, while the Globe/UNH poll was fielded in the first part of last week (Saturday through Wednesday). The earlier start to the Globe poll, he wrote yesterday, "could make a diff[erence] when things are moving fast." That's true in theory but difficult to evaluate in this case because we have to assume we are comparing an apple (the Globe/UNH results) to an orange (PPP) in terms of their likely voter samples.
Now that we have more than five polls released for this race, we should have our tracking chart posted (along with the tracking table, probably later tonight), but be forewarned: The small number of polls and the big "house effects" among them mean that we will really need to limit ourselves to same-pollster comparisons to evaluate trends over the last week. Coakley lead by an average of 29 percentage points on three surveys conducted before the primary last year, but leads by an average of 8 point on the three surveys conducted this past week. So we will see narrowing of the margin between the trend lines on our chart. Has Brown continued to gain over the last week? To answer that questions, we will need o watch tracking polls conducted next week by the same pollsters in the field this week.
Do we have a clear picture today of who will win on January 19 and by how much? Probably not, but we do have a sense of the dynamics that will ultimately determine the outcome.
And one last thought for those covering and commenting on this race: please spare us the cliche about the outcome depending on which campaign's "troops" do the best job turning out their supporters. Field organizations can make a difference, especially when contests are close, but the discrepancies in enthusiasm we are seeing are unrelated to canvassing and phone banking. Conservative Republicans are angry and ready to walk on hot coals if necessary to register their discontent with the direction of government. If he enthusiasm gap narrows, it will be because Democrats come to believe that Martha Coakley shares their priorities, Scott Brown threatens those priorities and the outcome of the election is in doubt.
Update: Via Twitter, Alex Lundry notes that the Globe Poll tests independent Joseph L. Kennedy (no relation to the famous family), while the PPP poll does not. What's interesting about that is that the presence of a "Kennedy" on the ballot appears to cos Republican Brown more support than Democrat Coakley . Also, for what it's worth, roughly 90% of those who support "Kennedy" (4 of his 5 percentage points) have not yet "definitely decided on a candidate, and about the same number (90%) are voters that are less than "extremely interested" in the Senate race.
Update 2: Nate Silver reviews some of the other differences between the three polls.
Update 3: Our chart is now live: