Last week, Suffolk University released its latest poll on the Senate race in Massachusetts. Not that there's a race at the moment. The only person running against Senator Scott Brown is Bob Massie, a nice guy who's political claim to fame is running for Lieutenant Governor on the Democrats' worst gubernatorial showing in the last 40 years (in 1996 against Weld). High profile candidates such as Governor Deval Patrick, Lieutenant Governor Tim Murray, and former U.S. Representative Joe Kennedy have all demurred on a run. Still, Suffolk decided to poll 500 registered voters on whether they (and a few other Democrats) could beat the incumbent. Kennedy came closest, losing 45-40. But, there is some reason to believe that he is actually running even with the incumbent - putting aside margin of error for the moment. The reason is rooted in a potential problem with the sample that was surveyed.
Polling is a difficult task, but done right it can be quite accurate. Polls do not dictate public opinion ("Push-polls" are not polls). Polls can and often do reflect public opinion. Polls can discourage people from supporting a candidate that they might otherwise support because they might seem unelectable. But, that unelectablility is not rooted in the polling; it is rooted in public opinion reported in polling. (Of course, I am talking about reputable pollsters here - and for this discussion Rasmussen is absolutely not reputable.) At this point, eighteen months out from the election, a poll gives us more of a sense of public opinion in general than an indication of how people will actually vote. Nevertheless, that general sense of opinion is not unrelated to how people will vote. Even in the best of efforts, there are almost always limitations with any particular poll. The one with Suffolk is buried, but it is there. It appears that Suffolk has oversampled suburban voters.
To understand this problem, one has to dig into the cross-tabs. Cross-tabs are the demographic details of the poll. How did people answer based on age or race or income or geography, etc.? Suffolk has given us four regions in Massachusetts into which all respondents are put (presumably based on residence). They are: Worcester/West; Northeast; Southeast/Cape Cod; Suffolk. No explanation of the areas is publicly given. (Here is another problem with many polls, discussed below - lack of transparency over methodology.) Since 8% of the respondents are from "Suffolk" one might conclude that it is either a proxy for the city of Boston (9.4% of the state population) or Suffolk County (at least 10%), but not Greater Boston (depending on how one looks at is, at least 30%, but as much as 60%). The fair assumption is that it is the county, since if it were only Boston it could have easily been labeled "Boston" and anything larger than Suffolk County would then been seriously underrepresented in the survey. That means the rest of the state is divided into three regions with no fine tuning for urban areas versus suburban areas. How do we know this? Because once the geographic areas are defined and respondent targets are set, those administering the survey will take whatever responses they can get to meet the goals. So, how many folks from Lawrence, Springfield, and Fall River were included in the sample? We don't know. But because we do know that it is easier to reach suburban, middle class folks and get them to participate in a survey, we can assume that plenty of folks in places like Swampscott, Dover, and Agawam were included in the sample. If the suburbs are overrepresented then this will skew the results, especially as it has been demonstrated that Senator Brown enjoys considerable popularity in the suburbs and Governor Patrick (for example, but it is true for the other Democrats as well) enjoys considerable popularity in the cities. Yet, Governor Patrick is losing significantly to the Senator. There are other potential reasons for this. One is that people are not supporting the Governor because he has said he will not run. Why support someone who won't run? People do consider such things. Enough people to influence the poll? I don't know. Another is that maybe people like the Governor where he is and like the Senator where he is. But, enough to have him losing by 15 points? Coakley lost to Brown by less than five points.
If we take the "Suffolk" area as a guide - we can conclude that urban voters are probably at least slightly underrepresented in this sample. Or are they? Here is the real problem: we don't know. Pollsters do need to adhere to some kind of ethics that compel complete transparency of their methodology. At least the ones who claim to be objective and release results to the general public as an - apparent or actual - public service. While polling tends to follow public opinion rather than the other way around, that is only through the diligence of many current pollsters. The problem lies not with garbage pollsters who never release anything worthwhile. It is the Rasmussen problem that is worth considering. Rasmussen has a fairly good track record with polls that are close to election day. However, it is unarguable at this point that Rasmussen's early polls are intended to form public opinion, not report it. Their early polls tend to oversample conservative voters and often ask opinion questions (rather than candidate questions). Until we get better information on who the Suffolk pollsters actually talked to, we can at least conclude this - Joe Kennedy is pretty close to Scott Brown - who is, after all, an incumbent. Not only is Kennedy down by only 5 points, but 13 percent are undecided. Even if all the undecideds in the Governor's match-up with Brown went to Patrick, he'd still lose. If that happened in the Kennedy match-up, he'd win in the kind of landslide election his uncle often enjoyed. The undecideds don't all move together in any direction during a campaign, but there is plenty of room for Kennedy to convince them. And that is assuming there is no bias in the sample against potential Kennedy supporters.
Here's the political moral of this poll: Joe Kennedy should run for Senate. If suburban voters are overrepresented in the Suffolk poll, then he is essentially tied with the incumbent eighteen months from election day in a race where no other Democrat, including the Governor, is coming close. Even if the poll is fairly representative, then he is pretty close to an incumbent who's been getting a lot of attention in a race in which no major Democrat has stepped into - and there are enough undecideds to move into his camp without even breaking off Brown supporters. The Democrats cannot pass on an opportunity to force Republicans to spend money in Massachusetts to defend a Senate seat. The president will win the state handily and the Democrats have other Senate seats (and House seats) that need attention. Kennedy will make it a race, reportedly has $3 million in the bank already leftover from his time in Congress, has name recognition money can't buy, and will help drain Republican coffers and thus prevent them from spending more money on the Virginia, New Mexico, Nevada, and Missouri races. The big question is this: does Kennedy have the guts to risk losing? Maybe now is the time for him to show everyone what he is really made of. The polling suggests he should seriously consider it.
Published by this author at The Big Idea.