Sunday, January 24, 2010

Poll dancing

The Globe has a story today about the money spent on the Senate race, especially the cash that poured in during the last two weeks. In the end, remarkably, Scott Brown's campaign raised 10 times the money that they expected to collect -- so much that they couldn't spend it all.

How did the election change so much from the easy Democratic victory that most people -- including Martha Coakley herself -- anticipated in the days immediately after the primary? It's clear that the Democrat did not run a smart campaign, that people are frustrated with the health care debate and that the high unemployment rate has the country frightened.

I also think polls and the media reporting on them may have played a role. The first widely reported survey of voters after the holidays that I heard about was a Rasmussen poll released on Jan. 5, two weeks before the election, in which Coakley led, 50%-41%. Taking into consideration the lackluster Democratic campaign and the state of the nation, that sounded about right to me, but what really caught me was the news coverage of that poll. I heard and saw reports that led with this: "Brown and Coakley in a dead heat for senate seat."

While it's true that the margin of error in the poll was +/-4.5%, I was surprised to hear reports that kept saying the race was "virtually tied." Should the margin of error be read as though a worst-case scenario for one candidate and best case for the other? Isn't it just as likely to be the opposite, meaning the possibility that the leader is up double the total -- in this case, 18%?

Up until that moment I knew that Brown was doing better than expected and Coakley worse, but after hearing reports of that poll the drumbeat began: a Republican could actually win the seat held by Ted Kennedy for 46 years. From there on the talk was all about Brown's potential upset and the money started raining down. Coakley, too, received increased support -- much of it in the form of negative advertising from the DNC, which seemed to make matters worse.

The Jan. 12 Rasmussen poll showed the race considerably tighter: Coakley had dropped a point to 49%, but Brown had added six points, jumping to 47%. If you plot out the Wrentham Republican's rise to its endpoint, it is indeed meteoric: There's a virtually straight line exploding upward (41% to 47% to 52% on election day), whereas the Democratic AG's drop is less precipitous (51% to 50% to 47%) -- but, of course, that was enough.

How much did the electorate's perception of the closeness in race lead to its outcome? The reporting on that first poll said the race was close and it became close. After that, the reporting was that there could be an upset and there was one. This is not to say that the media was the sole, or even key, factor. As I said, there are a number of variables at work here, but I feel like the reporting on the polls did as much to drive the campaign as to react to it.

1 comment:

acf said...

I think a question could be asked as to the sources of the various polls. It appears that these polls, whether or not any are push polls, are actually part of the campaign for the seat. This is as opposed to well known pollsters such as Quinnipiac, or Suffolk. If any one of the little known polls had not shown the remarkable nature of the trend, we might not have heard about them, at all. Once the results of these lesser known polls, showing the extremeness of the momentum change were publicized, they triggered a firestorm of polls and ads, and fueled the interest that drove Brown's donations. Also, it is very difficult, given the uniqueness of each poll designer's parameters and style, to look at different polls at different points in the campaign, and draw any reasonable conclusion. It's almost comparing apples and oranges. Following a given pollster through a campaign is more likely to show a truer trend line.