Politics

Back during the 2008 presidential campaign, we coined a term called tipping-point states. A tipping-point state was one that might potentially put Barack Obama (or John McCain) over the top to his 270th electoral vote, thereby winning him the presidency. States like Colorado, Virginia, Florida and Ohio were ranked highly by this metric.

The same calculation can be applied to this year’s Senate election. Which states are most likely to make the difference between Republicans controlling the Senate by exactly one seat, and falling exactly one seat short of doing so?

Unless there is significant movement in another state that defies the current consensus of polls, the answers are Washington, West Virginia, and California.

Take California, for example. Our program runs 100,000 simulations each night. In 14,102 of the simulations that we ran tonight, Republicans finished with exactly 50 seats; in 8,682 of them, they finished with exactly 51.

Republicans won California in just 4 percent of the simulations when they won 50 seats, but in 21 percent of the simulations when they won 51 seats. The difference between these two figures (21 percent and 4 percent) is 17 percent. In essence, this number — 17 percent — reflects the percentage of the time that California made the difference between Republicans controlling the Senate and not.

Ranking higher by this metric was West Virginia. Republicans won the Mountain State in 62 percent of the simulations when they won 50 seats, but 93 percent of the time that they won 51 seats. The difference — 31 percent — is how often West Virginia was the tipping point state.

But the most important state of all was Washington. Republicans won Washington 88 percent of the time that they won the Senate by a single seat, but only 41 percent of the time that they fell one seat short. The difference there is 47 percent. Washington makes the difference between success and failure almost half the time. Any polling in the state, Read more…

I mentioned in passing in last night’s post that surveys that use automated scripts rather than live interviewers — what are sometimes called ‘robopolls’ — have shown more Republican-leaning results this year. Given how sensitive forecasts can be to fairly minor variations in the polling, it is worth going into more detail on this.

One of the various adjustments that our models make is to identify and correct for “house effects” — that is, persistent differences in the partisan lean of polls issued by a particular research firm. If, for instance, a particular company’s polls are, on average, 4 points more favorable to Democrats than the consensus of pollsters in the states that they’ve surveyed, we’ll pull most (although not all) of those 4 points back out of their survey in calculating our polling average.

Our process for calculating house effects evolves somewhat over the course of an election cycle — there are things we can do now, when we have almost 4,000 polls in our database, that the data wasn’t robust enough to support earlier in the year. But the basic method is to fit a regression model with a lot of dummy variables, one set representing each pollster and another set representing each election in each particular month (for instance, polls of the Connecticut Senate race in September, 2010). Different voter groups (e.g. likely voters and registered voters) are treated as constituting separate polls — so, for instance, we have a Zogby-registered voters house effect and a separate Zogby-likely voters one. All House, Senate, gubernatorial and generic ballot polls are included; races that have been surveyed by more polling firms receive a larger weight in the calculation.

The other tricky bit is in figuring out what the “right” answer should be. For instance, say that Rasmussen Reports polls are 5 points more favorable to Republicans, on average, than polls from Siena College. Do we adjust the Rasmussen polls to match the Siena ones, or the other way around? The answer, of course, is somewhere in between. Specifically, we calculate a weighted average from all the polling firms in our universe, where the weights are based mostly on pollster quality. The idea is to estimate what our polling average would show in a particular state or district if every polling firm in our universe conducted an infinite number of polls there.

Looking specifically at the universe of likely voter polls (not polls of registered voters or adults), what sort of house effect do the automated polling firms have? Read more…

Perhaps this year’s most intriguing gubernatorial contest is in Colorado, which had once looked safe for the Democratic nominee, Mayor John Hickenlooper of Denver. A series of recent polls, however, have shown sharp gains by Tom Tancredo, the former Republican Congressman, who is running on the ticket of the American Constitution Party.

Mr. Tancredo’s campaign had once seemed little more than a curiosity, his support starting out in the low 20s in polls, and having dropped to as low as 9 percent in a one survey in late August. Typically, once a third party candidate’s numbers decline into the single digits or the low teens, they have little hope of coming back.

Mr. Tancredo had the good fortune, however, of being on the ballot alongside an exceptionally marginal Republican nominee in Dan Maes. Mr. Maes had never earned more than 33 percent of the vote in any given survey. About a month ago, the standing of the two candidates crossed in the polls — perhaps as conservative Coloradans realized that Mr. Tancredo’s chances against Mr. Hickenlooper, while not strong, were no worse than Mr. Maes’s.

What all the pollsters agree upon is that Mr. Tancredo has continued to gain ground since then. They disagree, however, on exactly how much of a threat he represents to Mr. Hickenlooper.

Six polling firms have been active in Colorado within the last two weeks. Four of them — Public Policy Polling, Pulse Opinion Research (a subsidiary of Rasmussen Reports), Magellan Strategies, and SurveyUSA — use automated Read more…

After a day that contained some scary polling for incumbent Democrats, today’s indicators were more equivocal.

Republicans strengthened their position in a couple of districts that received fresh polling from The Hill. In particular, John Spratt, the longtime Democratic incumbent in South Carolina’s 5th congressional district, was shown 10 points behind the Republican, Mick Mulvaney. Because the district had not received polling in some time, the poll has a lot of influence on Mr. Spratt’s forecast. The model now gives him just a 12 percent chance of holding his seat, a sharp decline from 53 percent yesterday.

The chances for two other Democrats, John Salazar in the Colorado 3rd district, and Baron Hill in the Indiana 9th, also dropped on The Hill’s polling.

But the same set of polls contained good news for other Democrats whom it tested, like Leonard Boswell in the Iowa 3rd district, and the two Democratic incumbents in the Dakotas, Earl Pomeroy and Stephanie Herseth-Sandlin, although both Mr. Pomeroy and Ms. Herseth-Sandlin are still rated as underdogs in the model.

Another Democrat to see her odds improve today was Colleen Hanabusa in the Hawaii 1st district, who was given a 5-point lead in a new poll that ordinarily has a strong Republican lean. Ms. Hanabusa is one of two Democrats favored to knock off a Republican Read more…

Sharron Angle, the Republican candidate in Nevada, has been improving her position in our forecast in recent days, and for the first time since the spring has better than a three-in-four chance to win her race against Harry Reid, according to the FiveThirtyEight model.

Ms. Angle’s odds improved today on the strength of a CNN poll showing her 4 points ahead of Mr. Reid. The lead, while slight, is her largest since January in a poll from a firm other than Rasmussen Reports. Moreover, CNN’s poll, unlike some others, included Nevada’s unique option for none-of-these-candidates in their poll, as well as the name of Scott Ashijan, who is running under the banner of the Tea Party on the Nevada ballot, although he lacks the support of most Tea Party affiliated groups. A separate version of CNN’s poll, which tested solely Ms. Angle and Mr. Reid, found the Republican with a slightly larger lead of 6 points, and with over 50 percent of the vote.

The polling has been a little bit sparse in Nevada relative to some other states; only CNN and Rasmussen Reports, which also now shows Ms. Angle with a 4-point lead, have released surveys in the past 10 days. Still, at least on the basis of the public polling, it is reasonable to consider the race as leaning toward Ms. Angle. Read more…

Pretty much every time that we issue a House forecast — like last night, for instance, when we projected that Republicans’ gains were most likely to be on the order of 52 seats — we point out that the forecast has a lot of uncertainty in it.

I realize that these reminders can seem pedantic and abstract — or that we can seem to be simply covering our butts in case something goes wrong. So in this post I want to put a little flesh on the caveats. Read more…

While there was an interesting set of polls in Senate and U.S. House races today, the polling load was lighter in gubernatorial contests, with most surveys confirming current assumptions about the races.

That’s bad news for California’s Meg Whitman, the Republican who appears increasingly likely to lose a once-tight race against Jerry Brown. Indeed, that assumption will become more entrenched after three new polls today.

The surveys, from Suffolk University, Fox News, and Public Policy Polling, gave Mr. Brown leads of 8, 9 and 11 points, respectively. While this is the first time that Suffolk has polled California, the Fox News and Public Policy Polling surveys show Mr. Brown having expanded his lead by several points, as other recent surveys have also shown.

As a result of these polls, our forecasting model now gives Ms. Whitman just a 6 percent chance of winning, down from 10 percent yesterday.

At this point — especially with Mr. Brown having at least 50 percent of the vote in several of the surveys — Ms. Whitman’s chances probably rely on there being systematic errors in the polling, rather than her being able to do anything in particular in her last week on the campaign trail to narrow her deficit.

There’s an argument, however — which also has implications for Carly Fiorina, the Republican Senate candidate — that some of the polls suggest a heavier Democratic turnout than it is realistic to expect in this political environment.

For instance, based on an analysis by Real Clear Politics’ Sean Trende, the partisan split in current California polls ranges from 6 to 14 points, favoring Democrats — while the average split is 11 points. By contrast, Democratic turnout outpaced Republican turnout by 12 percent in an exit survey of 2008 general election voters. The argument is that it may be wrong for the Democratic turnout advantage to have decreased by only one point (from 12 points to 11) from 2008 when it will probably be down by more than that elsewhere in the country because of the Republican enthusiasm advantage.

I have addressed a version of this argument before and I do not find it that convincing.

Indeed, I would expect Republicans to narrow the 7-point partisan disadvantage they endured nationwide in 2008 — perhaps to about even. But, this is almost certain to vary from region to region. There are several reasons to think the shift could be less (or there could even be a shift toward Democrats) in California:

• There is an initiative to legalize marijuana on the state’s ballot, which could draw younger and more liberal-leaning voters to the polls;
• It is easy to vote by mail in California, which increases the convenience factor and usually results in higher turnout;
• There has been a long-term trend away from Republican identification in California because of demographic changes in the state’s electorate. Party registration figures show that both Republicans and Democrats have fewer registered voters in the state than they did in 2008, but Republicans have endured a slightly steeper decline (whereas Democratic party registration is off slightly more nationwide);
• The gubernatorial and Senate candidates in California have spent in excess of $200 million on advertising and voter turnout efforts;
• There could be regional factors in play, as Democratic polling numbers have also held up relatively well in Washington and Oregon, two other Western states.

But say that you still find implausible a shift from a 12-point Democratic turnout advantage to an 11-point turnout advantage (still a shift in the Republican direction!) The other problem is that it’s not at all certain that the Democratic turnout advantage was in fact 12 points in 2008; it could have been larger (or smaller).

The 12-point estimate is based on an exit poll. This exit poll had a large sample size — about 2,300 people — but it is still just a poll, and subject to various errors. For instance, there are a relatively large number of non-English speaking voters in California, who may sometimes elude pollsters, and who tend to have a Democratic affiliation. And exit pollsters, because they station themselves only at some fraction of a state’s many polling places, use cluster sampling techniques — which can introduce error even when the sample sizes are large. People sometimes treat exit polling data as though it’s ‘hard’ evidence, but it’s not immune from the uncertainties intrinsic to all forms of polling.

At some point, if all the other polls are showing a different demographic or partisan split than the exit polls, it becomes likely that the exit poll, and not the other surveys, is the outlier. Also, these comparisons are not always apples-to-apples in the first place, because pollsters have many different ways of asking about a voter’s partisan registration or identification.

Can there be errors in polling? Yes, of course. And as often as not, when the polling is in error, most or all of the surveys miss in the same direction.

But our model accounts for the fact that there can be these sorts of errors in polling: that’s essentially where Ms. Whitman’s 6 percent chances, and Ms. Fiorina’s 8 percent chances, stem from.

About 8,500 people were interviewed in polls of California within the past two weeks. The margin of sampling error on an 8,500-person sample is only about 1 point.

Our model, however, assumes that the true margin of error of our polling average is quite a bit higher than this — about 5 points. Why? Because sampling error (the error that stems from the fact that you’re only picking a randomly-selected subset of the whole population to interview) is just one way that pollsters can make mistakes. Error can also result from difficulties in drawing a truly random sample, from faulty assumptions in your turnout model, from excessive demographic weighting, from last-minute changes in voter preferences, and from other factors.

Perhaps these other types of errors in the suveys are causing against Ms. Fiorina and Ms. Whitman in California. For this reason, we can’t conclude that the Democratic candidates are totally safe.

But we can conclude that they’re pretty safe: that the pollsters would have to have an unusually bad night to get the winner wrong — especially in the case of Ms. Whitman, who is about 7.5 points behind Mr. Brown in the polling average.

I don’t think the argument about the partisan splits in the California polls is persuasive enough to suggest that there is an especially large chance of mistakes in the California polling. I can also think of arguments one could make that the California polls are especially unlikely to show such a large error — for instance, because much of the voting there has already taken place (by mail), and because the polling is quite consistent there on balance, as compared with other states. I don’t know that those arguments are terribly persuasive either. But if you held a gun to my head and forced me to take the long side of the odds that my model is offering in a few states, Ms. Whitman’s race would not be one of the ones I’d pick. Ms. Firoina’s might be a closer call, because of the unusually large amount of resources that national Republican committees are putting into the state at this late hour on her behalf.

But the fact is that California is a really tough state for Republicans, and these two races are likely to be blemishes on what is otherwise a really fine night for them.

We have been noting for several weeks that, with Republican chances having dwindled from slim to virtually none in a few East Coast Senate contests like Connecticut, Delaware, and the New York special election, their prospects of the party winning a 51st Senate seat on Nov. 2 instead boiled down to winning either California or Washington (while sweeping the other competitive contests). Today, they have gotten bad news in one of those races.

That is the race in California, where several polls released within the past 24 hours show leads of 4 to 9 points for the Democrat, Barbara Boxer. Trendlines in these polls show little indication that Ms. Boxer’s lead has eroded; instead, it appears to have at least held steady and perhaps expanded by a point or so. As a result, the chances that the Republican, Carly Fiorina, would win are down to 8 percent from 16 percent yesterday. The forecast does not account for the fact that much voting takes place early in California, which can make it especially difficult to make up a late deficit there, although Republicans are pouring as many resources as possible into the state.

There has been little recent polling in Washington, where the model gives Dino Rossi a 15 percent chance of beating Patty Murray. That figure has been little changed over the past week.

Republicans did get good news today in some Senate contests. For instance, Kentucky: there, two new polls show Rand Paul with leads of 7 and 13 points over the Democrat Jack Conway, suggesting that attack advertisements released by Mr. Conway’s campaign may have backfired. Mr. Conway’s upset odds are reduced to just 4 percent from 12 percent.

Meanwhile, in Nevada, Sharron Angle’s chances of defeating Harry Reid are improved to 73 percent from 67 percent on the strength of a Rasmussen Reports survey showing her with a 4-point advantage, and an extremely strong third-quarter fundraising report. (This fundraising data, which we updated today, is a fairly minor component of our Senate model and had little effect on the other forecasts.)

Kentucky, however, was a state that was more relevant to the Democrats’ best-case scenario of holding on to as many as 56 or 57 seats than it was to the Republican one of winning a majority of the chamber. And while the Republicans would surely relish any victory over Mr. Reid in Nevada, Ms. Angle would represent something like their 47th or 48th senator, based on their relative likelihood of taking over the various Democratic-held seats — rather than their 51st.

As a result, Republican chances of winning a majority of the Senate are down to 14 percent in today’s forecast from 16 percent yesterday.

If Democrats were hoping for a late surge to improve their chances of retaining control of the House, there isn’t any evidence of it yet. Instead, Republicans have generally had the better of the polls in individual House districts released in the past 24 hours.

FiveThirtyEight’s forecast now projects the most likely composition of the House to be 231 Republicans and 204 Democrats. This is a one-seat improvement for the Republicans from yesterday’s forecast, and would mean that they’d gain a net of 52 seats over all.

There is uncertainty in the forecast: Democrats have a 20 percent chance of maintaining control of the House, essentially unchanged from a 21 percent chance yesterday. Much of this 20 percent probability reflects the potential for there to be systematic errors in the polling, as there were in years like 1998.

Since there are a very large number of competitive seats, relatively small anomalies in the polling could potentially affect the outcome of dozens of races. Although the Democrats’ overall position is poor, it is not yet so poor that it couldn’t be salvaged if they beat their polling averages by 2 or 3 points nationwide.

Still, such errors could also work in Republicans’ favor, potentially enabling gains in excess of 60 or even 70 seats. And much of the data released within the last day suggests that, if anything, they are strengthening their position. Republican candidates received encouraging polling numbers today in districts like the California 20th, the New York 20th, the Florida Eighth, the New Jersey Third, the Virginia Ninth, and the Idaho First, and their winning chances there were improved as a result.

Democrats did see a few relatively strong polls today, like in the Maryland First District and the Pennsylvania 10th. I’d caution against the notion that the “bottom is falling out” for Democrats, particularly given that their polling in Senate and governors’ races has been reasonably strong in the past several days, and since generic ballot polls show flat trends.

The upside case for Republicans, however, remains very high. And one gets the sense that the extremely large playing field in the House is doing more to stretch Democratic resources than Republican ones, particularly given that Republicans have sometimes been able to rely on fund-raising from outside groups to get a second wind in marginally competitive districts that they might otherwise have had to give up on. If pressed, I’d probably bet on the Republican side of the 52-seat over-under line that our model sets today.

It’s looking increasingly likely that Democrats will hold the Senate but not the House, although upsets in either chamber remain possible.

In recent days, I’ve heard a lot of speculation about why this is so. Common answers are that the personalities of the candidates matter more in the Senate (this is possibly true: partisan crossover is about 15 percent in Senate races, versus about 10 percent in the House, which might indicate that voters scrutinize the candidates more carefully), that the G.O.P. has a comparatively weaker set of candidates in the Senate (I think this is actually much more debatable than it seems), or some combination thereof.

These explanations are really missing the boat. Sometimes, the reason for something is so obvious that it becomes easy to miss; this may be one of those times. Read more…