Sunday, October 17, 2010

Polling models - why do they vary?

More polling data is out in the wake of the Cahill debacle of the past two weeks.

Suffolk University poll - October 14
Patrick 46
Baker 39
Cahill 10
Stein 1

The results suggest that Patrick has indeed helped his own position by remaining above the Baker-Cahill fray. Patrick is enjoying an 11 point lead over Baker among female voters (male voters are essentially splitting between them) and has amassed a large lead in the urban core around Boston and liberal Western Massachusetts.

The bad news for Baker is that he simply is not getting much momentum and the polling suggests that Patrick is drawing enough of former Cahill adherents to maintain his lead.

Patrick continues to struggle though with high (44%) disapproval ratings. I think that he is heading toward a narrow victory due to him being a good campaigner with strong base support and Charlie Baker just hasn't caught fire.

Interestingly, there is an Opinion Dynamics poll out last week as well that is showing Baker with a five point lead. (37% to 32%) over Patrick However, I am waiting for the next Rasmussen poll as I find their methodology to be closest to the actual vote. They use a model that seems to be very reflective of reality and really gets at voter intensity and motivation as those are key turnout components.

One thing that all (except for the Opinion Dynamics poll) of these polls are showing is a consistent Patrick lead of 4 to 7 points. Consistency over time is an indication that the polling is indeed correct.

Any reputable poll has a model for turnout - in other words, an educated guess as to what the voter demographics and turnout rate will be on election day. This is drawn from past elections and current trends and is very open to different interpretations. For example if a pollster believes that rural and suburban independents will turn out in high numbers and urban voters will have an average turnout - advantage Republicans. If the model shows a high big city turnout the Democratic candidates will benefit. The polling reflects the assumptions that pollsters make. They do make mistakes of course as in a close race like this one, just a little turnout variation can produce results that are contradictory to the polls.

Who shows up on Nov. 2 in Massachusetts? Will it be the Tea Party activists looking to kick out incumbents or are public employee unions showing up to uphold the Democratic Party dominance? This is why the advertising is so intense along with phone banks. It's all about GOTV (get out the vote) at this stage of the game. Who has a better ground game in a close race like this will likely remain victorious.

My money is still on Patrick and I give him 60/40 odds on winning on Nov. 2- but Baker still has a window to catch him. However, it is closing by the day and Baker needs to get the Cahill distraction behind him and get back on track and on message. The endorsement he received from former Democratic AG Tom Reilly may be the start of this.

16 days to go - time is short!!


The Prof

No comments: