When Christine O’Donnell won Delaware’s Republican Senate primary two weeks ago, some headlines screamed that it was a nightmare for the G.O.P. Indeed, unless polling is unusually inaccurate in Delaware, Ms. O’Donnell is unlikely to win in November, and her nomination significantly reduced the likelihood of a Republican takeover of the Senate. Two other races that once looked like tossups — in California and Washington — have also broken in the Democrats’ direction.
But Democrats have a nightmare of their own: What if the Republicans expand the Senate playing field by putting two new states, Connecticut and West Virginia, into play? Were they to win those states, Republicans could lose California, Delaware and Washington and still take claim of the Senate. And new polling suggests they could do just that.
We’ll start in West Virginia, which is the clearer of the two opportunities. A new Rasmussen Reports poll out this morning gives the Republican, John Raese, a 2-point lead over the Democrat, Gov. Joe Manchin III. This comes on the heels of a Public Policy Polling survey released two weeks ago that gave Mr. Raese a 3-point advantage.
The obvious conclusion is that West Virginia is a tossup, maybe even leaning in Mr. Raese’s direction. The only caution I’d urge is that this is an unusual race that would probably benefit from more polling: voters in the Mountain State like Mr. Manchin, but they do not like the agenda of President Obama or the Congressional Democrats — as Mr. Manchin, who yesterday called for repeal of parts of the Democrats’ health care bill, has surely become aware. Having a number of polls to evaluate might give us a better read on how West Virginians are planning to resolve this dilemma.
Three prior Rasmussen polls had shown Mr. Manchin with small-but-steady leads of 5 points to 7 points. For those who believe that pollsters look toward others for guidance, it might seem a little weird that their poll shifted in Mr. Raese’s direction only after Public Policy Polling had published their result. But it’s not hard to imagine Mr. Raese’s having gained ground in a late-developing race in which Republicans just made a million-dollar ad buy — a ton of money in a state with exceptionally cheap television markets.
There are also two new polls out in Connecticut: Quinnipiac gives the Democratic attorney general, Richard Blumenthal, a 3-point advantage over Linda McMahon — down from 6 points earlier this month — while Rasmussen shows Mr. Blumenthal’s lead having fallen to 5 points, from 9.
Our forecasting model — which has not yet incorporated these two new polls — has generally been quite skeptical of Ms. McMahon’s chances. There are two reasons for this, one of which seems prudent to me and the other of which I’m a little wary of.
The first reason is that there are few undecideds in the race — Quinnipiac shows 4 percent of the electorate having yet to make a decision, and Rasmussen only 3 percent. While I’d caution against an overly literal interpretation of the term “undecideds” — some voters who call themselves undecided probably aren’t really so, and some voters who say they have made up their minds can change them — our research shows unambiguously that fewer undecided voters means less volatility and uncertainty in the outcome of a race. Thus, even if Mr. Blumenthal’s lead is only 4 points — the average of the two polls released this morning — it may be a rather solid 4-point lead.
The other factor is that, in addition to the polling, the model also builds in a regression analysis that forecasts the result in each state based on various nonpolling factors. Although the weight assigned to the regression analysis is fairly small, it is somewhat larger in Connecticut than in other states because only two nonpartisan pollsters, Rasmussen and Quinnipiac, have been active there.
One factor the regression model looks at is fund-raising, and it looks at one particular type of fund-raising — contributions raised from individual contributors. Ms. McMahon, whose campaign is largely self-financed, has received hardly any individual contributions; the regression model would produce a different result if she were given credit for her own cash.
But individual contributions are often a good proxy for grass-roots support, and a candidate who bypasses the routine of building fund-raising networks may also miss an opportunity to build volunteer networks, which can be useful in motivating turnout. Moreover, self-financed candidates have had a fairly poor track record on Election Day in recent years. Ms. McMahon is an unusual candidate running an unusual campaign in an unusual cycle, so we’ll have to see how this all plays out.
One other reason for Democrats to be only somewhat worried about the Connecticut polling — as opposed to incredibly worried — is that Rasmussen polling’s has had somewhat Republican-leaning results throughout this cycle, and Quinnipiac has as well recently after having switching over to its likely voter model. As Connecticut has now taken on a pivotal role in the battle for control of the Senate, we will probably see some additional polling firms enter the state, and they may show slightly better results for Mr. Blumenthal. In the meantime, an internal poll released by Mr. Blumenthal’s campaign gives him a 12-point advantage — although since internal polls “cheat,” on average, by about 5 or 6 points in their candidate’s direction, it is not necessarily incompatible with the independent polling.
Mr. Blumenthal has seen a lead that was once quite formidable — well into the double digits — dwindle. Some people take this to mean that there will be continued momentum in Ms. McMahon’s direction. That conclusion is probably unwarranted: while this deserves a longer exposition, there is not a lot of evidence that “trendlines” have predictive power in general elections (primaries are another matter). That is, that a candidate has gained ground in the past may not make it any more likely, statistically speaking, that she will continue to do so in the future. Still, now and then you encounter a race that just seems to have a mind of its own, and this could be one of them.
We will have our official Senate forecast out later this week.