Polls suggest little change in gov race

Following last night’s release of the Suffolk poll, you may have read this morning about how Patrick is increasing his lead, or Baker is slipping, or some other story about movement in the polls.  The example here, from the Patriot-Ledger, is under the headline “Patrick’s poll numbers improved, Baker’s slipping,” and contains the following comparison.

Patrick garnered 41 percent of the vote in a Suffolk University/WHDH-TV poll of 500 registered voters conducted Thursday though Sunday. Republican Charles Baker got 34 percent, independent Treasurer Timothy Cahill of Quincy had 14 percent, and Green-Rainbow candidate Jill Stein garnered 4 percent. Six percent of voters remain undecided.

A post-primary poll by Rasmussen Reports last week had the race in a virtual tie between Patrick and Baker, and Cahill trailing far behind with only 5 percent of the 500 likely voters surveyed saying they planned to vote for him.

The headline suggests an interpretation which is problematic for the following reasons. 

1. Rasmussen’s results contained what he refers to as ‘leaners’.  While Rasmussen’s definition of ‘leaners’ is somewhat confusing, it appears to ask a follow-up horserace question to third-party supporters who say they are uncertain who they will  support (in addition to the usual question to undecided voters).  He does this to assess where weak third-party supporters might move if they move to a main-party candidate.  The “virtual tie” referred to above where Cahill draws only 5 percent is a result of this secondary process, not the initial horserace question, while Suffolk reports the basic horserace.  When you compare the initial horserace number (below) from Rasmussen and Suffolk, the race has not fundamentally changed between the three most recent polls.

2. Suffolk uses registered voters, whereas Rasmussen uses likely voters.  This difference is not insignificant.  Rasmussen is projecting 18 percent of the electorate will identify themselves as Republican, similar to the figures from CNN exit polls in 2006.  This is in comparison to the 11 percent of the electorate who are registered as Republicans in Massachusetts (see table).  Although Suffolk has not yet released their crosstabs, my guess is they are trying to match the registered voter universe.  Unfortunately, comparing party identification  to party registration is not direct, but I would guess from the figures below that Rasmussen has more Republicans (in either definition) in his sample than does Suffolk. 

These two factors make it difficult to do a one-to-one comparison between the polls.  But were we to look at them as a trend, noting the differences outlined above, we could at most allow ourselves to conclude that the fundamental numbers in the race are generally the same as they were three weeks ago.  

Steve Koczela is the president of the MassINC Polling Group.

CommonWealth Voices is sponsored by The Boston Foundation.

The Boston Foundation is deeply committed to civic leadership, and essential to our work is the exchange of informed opinions.  We are proud to partner on a platform that engages such a broad range of demographic and ideological viewpoints.