Search Right-Wing Genius's Blog

Tuesday, October 28, 2014

What 2010 Can Tell Us about How Accurate the Polls Are a Week from Election Day

With polls showing Republicans in good shape as Election Day approaches, one may wonder: Are these polls accurate? In the battle for control of the U.S. Senate, we can look to the previous midterm election for some empirical data to help answer that question before we actually see the results on election night. It's especially helpful because most of the states with competitive Senate races this year--including New Hampshire, North Carolina, Georgia, Louisiana, Arkansas, Iowa, Kansas, Colorado and Alaska--all had Senate races in 2010 as well. So, how were the Senate candidates in those states doing in the polls one week out from Election Day four years ago?
 
The short answer is: In all of the aforementioned states except Alaska (where incumbent Sen. Lisa Murkowski (R) was campaigning as a write-in candidate against Republican Joe Miller and Democrat Scott McAdams), the Republican candidate for U.S. Senate led his/her Democratic opponent in all or most of the polls heading into the final week of the campaign. Were the polls correct? Well, sort of...
 
In New Hampshire, Republican Kelly Ayotte had a nine-point lead over her Democratic adversary, then-Congressman Paul Hodes, in the RealClearPolitics Average of polls on October 26, 2010 (one week before Election Day 2010). Ayotte continued to expand her lead and ended up routing Hodes, 60% to 37%, in the general election. This year, the Senate race in New Hampshire is much closer: As of today, the incumbent, Sen. Jeanne Shaheen (D), holds a lead of 2.2 percentage points in the RCP Average over her challenger, former Sen. Scott Brown (R-MA). If Brown finishes strong the way Ayotte did in 2010, then he may yet return to the U.S. Senate.

Similarly, in Iowa, the Senate race in 2010 wasn't competitive, unlike the one this year. Four years ago, Sen. Charles Grassley (R) was such a heavy favorite to win re-election that there wasn't much polling of the race, at least not by independent (unaffiliated) outfits. A Des Moines Register poll conducted in late October nailed Grassley's margin of victory over former U.S. Attorney Roxanne Conlin (D): 31 percentage points. This year, the Register's latest poll found state Sen. Joni Ernst (R) ahead of Rep. Bruce Braley (D) by just one percentage point, 47% to 46%. That comports with the current RCP Average of polls, which has Ernst leading Braley by 1.7 percentage points.

Same story in Georgia: The Senate race in 2010 wasn't competitive at all; this year it's a nail-biter. However, there was something unique about the Georgia Senate race four years ago. Sen. Johnny Isakson (R) handily defeated his Democratic challenger, State Labor Commissioner Mike Thurmond, 58% to 39%, but polls overestimated his margin of victory. All the publicly released poll results in late October that year showed Isakson winning by anywhere from 21 to 30 percentage points. This year, the polls are showing a virtual tie between businessman David Perdue (R) and Points of Light CEO Michelle Nunn (D); Perdue's lead over Nunn in the RCP Average today is one-half a percentage point. Because this race will go to a runoff if no candidate wins an outright majority of the vote in the general election, and because all the polls are indicating that is exactly what will happen, further analysis of this race seems premature right now.

Admittedly, the dynamics of an open race are so different from a race with an incumbent in it that trying to use one to predict what will happen in the other may not be empirically sound. So, what races can we look to for an apples-to-apples comparison? Try Arkansas, where Sen. Blanche Lincoln (D) was probably the most vulnerable incumbent in the U.S. Senate in 2010, and she lost in a landslide to then-Congressman John Boozman (R), who won Lincoln's seat with 58% of the vote. (Lincoln received 37%, almost exactly her poll position in the RCP Average--37.8%--on October 26, 2010.) Late deciders must have broken for Boozman, who stood at 54.5% in the RCP average one week out. If history repeats itself this year, then the lone remaining Democrat in Arkansas's congressional delegation, Sen. Mark Pryor (D), will soon be replaced by Rep. Tom Cotton (R), who today leads Prior by five percentage points in the RCP Average, 46.8% to 41.8%.

Colorado was one state where the pollsters really missed the mark in 2010. A week before Election Day that year, Weld County District Attorney Ken Buck (R) led Sen. Michael Bennet (D), who had been appointed to the seat in 2009, by one percentage point in the RCP Average. By Election Day, however, that lead had grown to three percentage points. Buck appeared to have the momentum going into the election, but Bennett managed to pull out one of the tightest victories that year, winning election to a full term by less than 2% of the vote. (Buck received 46.4% of the vote, significantly less than his 49.3% standing in the RCP average on Election Day, while Bennett took 48.08%, slightly more than the 46.3% the RCP average of polls had him winning. What probably happened is that a lot of voters who were going to vote for Buck ended up voting for one of the third-party or "Independent" candidates on the ballot, or maybe they just didn't vote.) This year, Rep. Cory Gardner (R) has a slightly larger lead over incumbent Sen. Mark Udall (D), 3.3 percentage points in the RCP Average, and he hasn't made a lot - or really any - of the stupid mistakes Buck made four years ago. Republicans are also determined this year not to get caught off guard again by the Democrats' ground game. Suppose for a minute, though, that this race plays out from here just like the 2010 contest did. That would mean Gardner would continue to gain in the polls and head into Election Day with a lead of between five and six percentage points in the RCP average. If the actual election results then showed a four-point swing towards Udall, then Gardner would still win. It's also worth noting that Udall is polling much lower than Bennett was four years ago. However, there are also more "undecided" voters than there were at this point in the 2010 race, so while Udall has more ground to make up than Bennett did in the last week of the campaign, there are also more potential late deciders who could swing this race.
 
In North Carolina, the polls were much more accurate in 2010. Incumbent Sen. Richard Burr (R) led his Democratic challenger, then-North Carolina Secretary of State Elaine Marshall, by 10.7 percentage points in the RCP Average on October 26th, and he won re-election by a twelve-point margin, 55% to 43%. (One interesting thing to note is that, in that race, undecided voters appear to have split evenly between the two candidates, something we don't usually see in a race where one candidate has a comfortable lead in the polls and is expected to win.) That may be good news for Sen. Kay Hagan (D), who is currently clinging to a one-percentage-point lead over her Republican challenger, State House Speaker Thom Tillis (R). Unlike Marshall in 2010, however, Tillis appears to be chipping away at the incumbent's lead, which was 3.8 percentage points in the RCP Average at the beginning of this month. But Hagan may be helped by the Libertarian candidate in the race, Sean Haugh, who was not a factor in 2010.

In Louisiana four years ago, Democrats tried to pick off incumbent Sen. and probable prostitute patron David Vitter (R) with Rep. Charlie Melancon (D). In another state where a candidate must receive a majority of the vote to win, Vitter avoided a runoff, winning 57% of the vote to Melancon's 38%. One week before Election Day, Vitter led Melancon by exactly sixteen percentage points in the RCP Average. This year, it's Republicans who are trying to pick off a Democratic incumbent, and let's just say it's going to a runoff; the latest polls indicate that Sen. Mary Landrieu (D) will likely receive a plurality of the vote in the "jungle primary" next week but not even close to the 50% needed to avoid a runoff, which would likely be won by Rep. Bill Cassidy (R). In a one-on-one matchup, Cassidy leads Landrieu by 4.5 percentage points in the RCP Average. 
 
Although there were also Senate races in Kansas and Alaska in 2010, those races were so different from the ones in those states this year that I don't think it's useful to look at them. Also, while there was a Senate race in Kentucky four years ago, I don't really consider the race between incumbent Sen. Mitch McConnell (R) and Kentucky Secretary of State Alison Lundergan Grimes (D) to be "competitive" at this point; McConnell has a small but consistent lead, and Grimes is showing no traction whatsoever.

What about the "generic ballot"? Republicans have a six-point advantage among likely voters in the Washington Post/ABC News poll out this week. That's even better than they were doing four years ago, writes FOX News digital politics editor Chris Stirewalt:
Double, double, toil and trouble - Struggling with female voters and young voters, Democrats fare worse in the latest WaPo/ABC poll than they did at the same point ahead of the disastrous 2010 election and at the same level as the punishing 1994 midterm elections that cost them both houses of Congress. Without any significant increase in Democratic intensity since the previous polling cycle, the auguries are getting dire for the president’s party.
Among registered voters, Democrats lead Republicans in the poll, 47% to 44%, but that's "identical to the difference recorded at this point in 2010." That poll may actually be good news for Democrats when compared with the results of a recent Wall Street Journal/NBC News/Annenberg survey, in which 52% of likely voters in the survey said they wanted the election to produce a Republican-led Congress, while 41% favored Democratic control.

"By historical measures, an 11-point lead on the question of which party should control Congress is large," Janet Hook wrote for the WSJ. "Republicans held a seven-point lead on the question at this point in the 2010 election in a Journal/NBC survey, which used a different method to determine which voters were most likely to cast ballots."

Finally, if you're one of those statistics-mongers who just can't get enough data and analysis, or if you're looking for a number to put on the GOP's odds of taking over the U.S. Senate this cycle, then consider this:

Today, One week out from Election Day 2014, Nate Silver’s infamous FiveThirtyEight gives the GOP a 64.6% chance of winning the Senate, and "the Upshot" (a similar thing affiliated with the New York Times) says there’s a 70% likelihood of the same. The Washington Post’s Election Lab, meanwhile, forecasts that Republicans will see a net gain of seven Senate seats and projects a 93% chance Republicans take the upper chamber. Make of this what you will; I'm going back to running my law practice.

No comments:

Post a Comment