Not Just Minutiae and Methodology

Share on Facebook
Share on Twitter
Share on


The Boston Globe has the power to set the agenda at times in this town. On Sunday, they released the results of a poll of the mayor’s race that suggested it was tightening a bit but still favored the incumbent by a wide margin. The opposite result probably would have gotten a lot more casual observers focussed on the race.

David Bernstein of the Phoenix and the challenger’s campaign immediately took issue with the poll, noting that the sample appeared to consist of self-described ‘likely’ voters, not registered voters.

My read is a bit more ambiguous. The source document, the UNH Survey Center’s full report, is not precisely worded. It refers initially to a sample of “553 randomly selected Boston adults”, without any qualifiers like ‘registered voters’. But then a question (INT2) implies that each person being interviewed is a registered voter. The issue of whether the sample consisted of registered voters or not does not come up again.

UPDATE: OK I learned something today. There actually was a qualifier on the initial sample listing, it reads “553 randomly selected Boston adults (RDD)”. Turns out that RDD stands for Random Digit Dialing, which I think strengthens the argument that these were randomly selected residents of Boston, not a list of registered voters.

And rereading the questionnaire, there is no question regarding voter registration. It just pops up as an assumption in a question (INT2) without being established elsewhere.

A later question (LIKELY) then splits up those polled into Not Voting, Don’t Know, May Vote, Will Vote, and Definitely Voting. Some combination of the last three categories results in a sample of 438 voters being classified as “likely voters”. I can’t get the numbers to work to figure out how they made the transition from 553 to 438, but that may be my shortcoming.

Where Bernstein is correct is that, even if we assume that all respondents to the poll are registered voters, this assumes a voter turnout of 80%, which would be a miracle. If they aren’t registered voters, it assumes a turnout of impossible high proportions.

Given the Globe’s stature and the importance of this race, it would be nice to clear up the questions about methodology and figure out how valid these results are.