TPMDC makes a notable catch--Marco Rubio, darling of insurgent conservative candidates everywhere, has not actually been embraced by the Florida tea parties:
In the midst of Sunday's heated Florida Republican Senate primary debate on Fox News Sunday, moderator Chris Wallace asked Marco Rubio a question that surprised many viewers up early on a Sunday to watch the festivities.
Wallace read Rubio a viewer email. "'Ask Marco Rubio why he refuses to be vetted by the Florida Tea Parties. I want to hear from Rubio or I will not vote for him,'" Wallace said. "We got this from a bunch of Tea Parties all over the state."
Behind the question is an interesting discovery: Despite carrying the torch for insurgent conservatives everywhere, Rubio actually has a problem connecting with the tea parties in his home state, according to several tea partiers I spoke with yesterday.
Wait-so tea parties actually have nothing to do with Rubio's success? But, at least they elected a guy from Massachusetts who stopped health care, and who has stayed true to tea party principles while in office! Oh wait...
Monday night, Brown announced that he would join four other Republicans in voting to block a GOP filibuster and move forward with a $15-billion jobs bill designed by Senate Majority Leader Harry Reid (D-Nev.).
Almost immediately, the political blogosphere exploded.
Cries of "letdown," "betrayal," "sellout," and "RINO" -- "Republican in name only" -- flew around Twitter. By late Tuesday afternoon, more than 4,200 people had left comments on Brown's Facebook page, most harshly negative. (And liberals engaged in some cyber-schadenfreude at the same time.)
Even the supposed tea-party success in the NY-23 special election, where conservative party candidate Doug Hoffman managed to push Republican nominee Dede Scozzafava out of the race, was actually engineered more by well-established groups like The Club for Growth than any new grassroots movement. The Club endorsed Hoffman early in the campaign (September 28th), and spent over $300,000 in support of Hoffman. This is several orders of magnitude beyond any material support offered by the tea-parties.
It is also worth noting that unlike the tea-parties, the Club for Growth has a long record of making huge impacts, including several big victories, in Republican primaries. Compared to the established success of the Republican primary-challenge machine, the tea parties are a new, and laughably ineffective addition.
OK, so elections where the Club for Growth isn't pulling the strings in the background haven't gone well for the tea partiers. But, at least they made a big impact, independent of existing right-wing infrastructure, on the health care debate with their town halls, right?
Despite conventional wisdom, polling indicates that the health reform plan actually increased in popularity last August during the tea-party assault on town halls. They were entirely ineffective at swaying popular opinion against the bill.
Even all of the protests, rallies and other grassroots enthusiasm around the tea-parties was clearly evident in 2008, after Sarah Palin's nomination, long before the term tea party was even coined. The anti-Obama rhetoric, and cries of socialism, was there too. About all tea parties have done is provide a long-standing right-wing political machine with a new image. This is actually a useful development for conservatives, since portraying their movement as based in grassroots energy, rather in than large corporate donations to the Club For Growth, plays better in the media. The new branding is undeniably a positive for the conservative movement, but really it is about all the tea parties have actually accomplished that existing right-wing infrastructure would not have achieved on its own.
Earlier today, I pointed out that the Massachusetts swung 18% against Martha Coakley during a time when national trends on Obama job approval, the health care bill, and the national House ballot remained static. From this evidence, I concluded that the main, though certainly not the only, cause of defeat in Massachusetts was a large gap in the quality of both the candidates and the campaigns.
Some people objected to that post, claiming that polling itself it stupid. The more sensible claims argued that I was comparing national trends to local ones. Even though that was precisely my point-campaign specific trends were more to blame than national trends--here is a look at local, Massachusetts polling that underscores the weak candidate theory:
PPP did not poll the health care bill in the two polls, but still hard to see pinning the late movement this primarily on President Obama. Granted, his 44-43 approval rating is very weak for a state like Massachusetts and helped set the table for a more difficult election.
This poll suggests a combination of factors. President Obama's approval rating dropped by a bit more than half the amount of Coakley's lead (19 vs. 35). Still, this would imply a primarily mixed result, where national trends and the differences between the candidates were roughly equal
However, assigning even that level of blame to national trends becomes difficult given that Coakley moved from a net favorable advantage of plus 29 on Brown, to a net disadvantage of negative 30. That is a 59 point swing against Coakley in terms of candidate and campaign quality, three times the drop in President Obama's favorable ratings.
****
Coakley lost last night by a little less than 5%. Available, empirical evidence suggests that last night's defeat would not have happened in such a poor, national political climate for Democrats (which cost Coakley at least 9%, and possibly as much as 15-19%). At the same time, she would not have lost without such a large difference in quality between the campaigns and the candidates, which cost Coakley at least 11% and more likely something around 18%.
There is not much supporting evidence for those looking to pin the blame entirely on the Obama administration and the national political environment. At worst, it was a roughly equal cause to the local campaign factors. It is much more likely that the local factors were the main (but, once again, certainly not the only) cause of defeat. But really, removing either problem would have prevented this disastrous result.
Think Martha Coakley lost mainly because of President Obama, the national Democratic position, and health care? Think again.
The following four charts measure trendlines since December 31st for the Massachusetts Senate race, the popularity of health care reform, the national congressional ballot, and President Obama's job approval. What they show is that Martha Coakley trended sharply downward during a time when opinions about the health care bill and the 2010 congressioanl vote were static, and job approval of President Obama was on the rise:
President Obama, job approval trend since December 31st
National Congressional Ballot, trend since December 31st
(Note: My numbers show a slightly better position for Democrats because I include Daily Kos polls)
Health care plan, favor / oppose trend since December 31st
Massachusetts Senate race, trend since December 31st
Martha Coakley went from a 13% lead to a 5% deficit since December 31st, for a swing of 18%. During that exact same time period, national approval of the health care bill stayed exactly the same, the national congressional ballot stayed exactly the same, and President Obama saw a 2% upswing in his approval rating.
The national political environment is worse for Democrats than it was in 2008. As I pointed out last night, the generic congressional ballot shows a 9% swing toward Republicans since Election Day, 2008. However, Martha Coakley's campaign dropped 18% independent of this trend.
There is a lot of blame to go around. However, the available evidence shows that the bulk of it rests on the candidate and the campaign. On that front, it is also worth considering the relative difference between the Scott Brown and Martha Coakley campaigns. That is, Coakley could have been slightly below average rather than terrible, while Scott Brown--and the conservative organizing behind him--might have been fantastic.
1. Why Scott Brown won There were three key factors to a Scott Brown victory: national trends, relatively lower Democratic turnout, and Scott Brown as the superior candidate.
National Trends. Nationally, Democrats down 9% from 2008. The current national House ballot shows Democrats ahead by 0.67%. In 2008, they won by 9.65%. Scott Brown needed a 26% swing from 2008, and got 9% of it from the national political environment.
Lower Democratic turnout. With the exception of Rasmussen, the generic ballot polls in the national House forecast are currently measuring "all adults" or "registered voters" instead of "likely voters." When they start measuring "likely voters," current evidence indicates they will probably find another 2-3% swing in favor of Republicans, based on a relative lack of Democratic enthusiasm.
The earliest polls on the campaign in January, UNH and Rasmussen, showed Coakley ahead by an average of 13%. So, two weeks ago, Scott Brown was just where the national political environment and relatively low Democratic enthusiasm would have put him.
Running a better campaign. Scott Brown made up the rest of the difference by being the superior candidate with the superior campaign. The final On Message, Rasmussen, Cross Target, PPP and Research 2000 polls showed, on average, Scott Brown with a net favorable rating 18% higher than Coakley's. That is the rest of the swing right there.
Given their relative numerical impact, ranked in terms of importance, the factors were: 1) Brown is the superior candidate, 2) national trends, 3) relatively lower Democratic turnout.
Feel free to postulate whatever unprovable, subjective and / or anecdotal explanation you like. Make sure that this explanation fits into whatever preconceived notions you have about politics. It is what everyone is doing, these days--now you can be cool, too!
****
What happens to health care, and what the election means, can be found in the extended entry.
This projects to a 70% chance of a Brown victory (and thus a 30% chance of a Coakley victory).
That is a bit better for Coakley than most other forecasts, mainly because my methodology discounts trendlines over the final two weeks. While that may not hold up in this specific case, my research indicates that discounting trends over the last 15 days produces about 20% more accurate results than regression analysis which emphasizes more recent results. Across 144 instances, one occasion of greater error in a special election won't change those findings much.
We will find out soon enough. This is an open thread on the Massachusetts special election. Results thread will appear at 8 p.m. In the comments, please post links to election returns websites.
There is a lot of useless information floating around about the Massachusetts Senate campaign. Here is some information that should actually be helpful:
No exit polls. There will be no exit polls tonight.
Anecdotal reports on turnout mean nothing. Barring a comparison to:
Past turnout levels;
At exactly the same point in the day;
Across several dozen, randomly selected precincts;
And only if that selection of precincts is representative of the state as a whole;
Then, and only then, will all those reports you are hearing about turnout mean anything at all.
There is no such systematic turnout report. Since there is no such systematic comparison, the turnout reports you hear favoring one side or the other are not solid information on how the election is going. As mikegehrke wrote over twitter:
New fake exit poll: 48-52 gonna be a squeaker! Turnout at some guys precinct in Wooster looks moderate.
That goes for absentee ballot reports, too. Barring a statewide comparison of absentee ballot reporting, the same argument in bullet point #2 applies here.
Polls close at 8 p.m.. Results will start coming in at 8 p.m., but really won't start coming in for at least 30 minutes after that.
Following the results live. Still haven't found a site that lists them town by town, but the Boston Globe and The Boston Channel are good bets. If you have a better one, please post it in the comments.
What are Coakley's chances?About 20%, averaged across the forecasting methodologies I have found to be the most accurate in the past.
In my experience, most people prefer anecdotal, or even entirely subjective, information that supports their perspective. Grounding in empirical observation, facts, and solid reasoning is frustrating rare. Consider that when watching the pre- and post-election spin during the rest of the day.
Results are soon pending in the special election to replace Senator Ted Kennedy. Once a guaranteed Democratic victory, the race has become surprisingly competitive due to a bad national environment and a lackluster campaign run by Democrat Martha Coakley. In fact, several polls have put Republican Scott Brown in the lead, striking panic amongst the Democratic establishment.
Interpreting incomplete results can be difficult if one is not familiar with how different areas in a state vote. Senator John McCain, for instance, led the vote in Virginia during much of election night; this was because deep-red rural Virginia reported first. After Democratic strongholds in Northern Virginia began posting, Barack Obama quickly pulled away (he ultimately won by 6.30%). Because Massachusetts is rarely competitive outside of gubernatorial elections, geographic unfamiliarity probably extends to even most politically active folk.
I have created a map indicating what a tied election would probably look like:
I want to reiterate something Nate Silver wrote last night: although Scott Brown is the favorite, he is by no means a 100% lock to win. This is the case across all forecasting models:
On Pollster.com, Charles Franklin looks at 18 forecasting models for the campaign, all of which show Brown ahead. However, a lead in a forecasting model, even across 18 forecasting models, is still not a lock.
Franklin says that the standard estimate for Pollster.com shows Brown ahead by 6.2% (although the chart at Pollster.com says 6.9%). In the 55 closest (under 17.0% estimate) statewide elections from 2008-2009, there were 5 instances where Pollster.com's standard Loess regression estimate missed the final results by more than 6.2%, and 4 instances where it was missed the final results by more than 6.9%. That would give Martha Coakley a 4% chance to win if the lead is 6.9%, and a 5% chance to win if the lead is 6.2%.
On 538, Coakley is given a 25% chance to win, and Brown an estimated advantage of 2.2%. That squares with my estimate of Nate's error rate. Based on Nate's 2008 results and a deficit of 2.2%, Coakley would have a 24% chance to win.
And finally, my model gives Coakley a 35% chance to win (although, in this specific case, I actually think it is less, around 31%, given the recent trendline).
4%, 5%, 25%, 31%, 35%--none of these are great chance for Coakley, but they are still chances. This campaign has not reached the 10% range, at which point the odds of victory would be reduced to zero (at least in an election pitting a Democrat vs. a Republican).
Another point of hope for Coakley is that no polling was conducted yesterday. Brown's support is very new, and thus very soft. As such, it is possible there has been some movement back in her direction since Sunday night. Public opinion does not follow physical laws, and just because a candidate was trending upward does not mean that candidate will continue to trend upward.
My best estimate, based on the available data, is that Brown will win by between 0.9% and 2.7%. However, no matter which way you look at it, Coakley does still have a chance. Not a good chance, but a chance none the less.
Based on that model, I feel pretty confident Scott Brown will not win by as much as 6.9%, as Pollster.com is currently estimating. Unfortunately, I feel pretty confident Scott Brown will win anyway. The polling is converging and, as such, so are the electoral forecasts.
Further, the error in my model will probably favor Brown at this point. This is because the polls from the second to last week of the campaign are more favorable to Coakley than the polls over the final week. A report by NCPP showed that polls taken in the second to last week of the campaign are about 15% less accurate than polls taken during the final week of the campaign. While I have not yet tested to see if applying this negative 15% weight to polls taken during the penultimate week of the campaign would produce even more accurate results, I suspect it would.
So, I think it will be closer than many are forecasting, but I still think Scott Brown will win. While a Coakley win is not out of the question, at this point it is not very likely.
Charles Franklin of Pollster.com, whose polling analysis I respect greatly, has posted an article claiming:
But no matter how you slice the data, the only reasonable conclusion is that Scott Brown has moved from well behind to a lead somewhere between 4 and 11 points.
I disagree, for two reasons:
Pollster.com methodology over-emphasizes recent polls. There is empirical evidence that Franklin's methodology overemphasizes recent polls (that is, polls in the final week of the campaign over polls in the second and third weeks). In this case, that means Pollster.com is probably overstating the rapid shift toward Brown.
As I noted on Friday, here is a comparison of the 2008 average error rates from final estimation to final vote margin for 538 (which weights polls on a variety of factors, including recentness), Pollster.com (whose Loess method pretty much only weights polls on recentness) and a simple, 15-day mean which does not weight polls based on recentness (or really, anything at all):
Error rates, final predicted margin to final vote margin, 52 closest campaigns, 2008
Pollster
538
Simple 15-day mean
Mean error
2.76
2.88
2.56
Median error
2.14
2.16
1.68
By not placing any extra weight on more recent polling, the simple, 15-day mean actually produced results about 22% more accurate on the median than more recentness-focused methodologies. This wasn't a fluke, either. Looking back to 2004, across 143 campaigns, including older polls at equal weight did not harm the accuracy of the simple mean:
Mean error rate, various date ranges, simple polling mean, 143 campaigns, 2004-2009
30-day
25-day
20-day
15-day
10-day
Mean error
2.63
2.60
2.56
2.54
2.59
All of this makes it a good bet that the Pollster.com trendline is overstating the trend toward Brown. It doesn't mean that the trend doesn't exist, just that it is smaller than the trend posited by Franklin.
It is likely that there are a lot of flawed polls on the campaign. There is a very good reason not to trust the available data (aka, the polling) on the Massachusetts special election. Later in the same article, Franklin notes:
Finally but significantly, we are seeing more pollster variation in this race than normal. If we look at the residuals around the trend estimates, past experience with 2004, 2006 and 2008 state and national contests has pretty consistently found that most of the polls (about 95%) fall within +/- 5 points of the trend estimate. Now that is an empirical observation, not a theoretical one.(...)
Only half of the current polls are inside +/-5 points of the linear trends.
Normally, 19 out of 20 polls group together, which also happens to be the percentage of polls that should group together given polling error rates. But in this case, the polls are not grouping together at all. Half of all polls on the special election are outliers. The odds of this occurring naturally are astronomical--how often does a 1 in 20 shot happen 5 times out of 10?.
There are two more likely explanations for the outlying data: a rapidly trending campaign, or massive error in the polls. Now as I explained above, apparent rapid movement in a Democratic vs. Republican campaign turns out to be illusory--or at least significantly overstated--more often than not. As such, this means the most likely explanation is that the polling data is itself f*cked heavily flawed.
If the polling data is heavily flawed, then we are potentially dealing with a situation that outlies from the 4-11% Scott Brown lead that Franklin posited based on existing data. In that scenario, is it more likely that Scott Brown is ahead by less than 4%, or that he is ahead by more than 11%? I agree that Scott Brown may well win--I actually give him a 48% chance at this point. However, the idea that he will win by more than 11% strains credulity. As Nate Silver notes, such a Scott Brown victory would require the electorate in Massachusetts to be more conservative than the country as a whole. Not bloody likely.
As such, one can be reasonable, and conclude based empirical observations of polling dynamics that Scott Brown is ahead by less than 4%, or even slightly behind. There is good reason to believe that much of the polling data (cough, Pajamas Media, cough) is screwy, and overstating the extent of Scott Brown's rise.
I rate this campaign as extremely close, and believe it will be decided by less than 2%. The forecasts currently showing the campaign as "lean Brown" are overstating the extent of Scott Brown's rise. This is both because a focus on late polls has historically overstated campaign trends, and because it is likely there is something screwy in the existing data.
Yesterday, we attended a Grow the Hope (GTH) house meeting with their Rapid Response team, Organizing for America (OFA 2.0), and the Carrots and Sticks Project. After the meeting everyone decided to stick around and make calls into MA to help elect Martha Coakley in the race for Ted Kennedy's former Senate seat. We took a moment before leaving to talk with Jon Randall, Maryland's 8th district liaison from OFA 2.0, and we put together this video, quickly detailing how to make phone calls online to support Coakley:
Based on a study I conducted of the 143 closet, statewide general elections from 2004-2009, I have been making some bold predictions in the Massachusetts Senate race of late. Even though virtually every other election forecaster has moved this campaign to "toss-up" or "lean Brown" arguing that it is still "lean Coakley."
According to that study, Coakley would still have a 67% chance to win. This is because the final, 15-polling average erred by at least 1.09% in 94 of the 143 elections in the study, and since there are equal chances the polls could be wrong in favor of either candidate. Given how few elections ever come so close to even odds, that is close enough for me to describe the election as a "toss-up."
However, this is a special election, not a general election. Even though some local election officials are seeing signs that turnout will be very high, perhaps it is best not to apply a general election study to this campaign, but instead to use a study of primary and special election results. As such, today I looked at the 37 closest (final polling margin within 18.50%) Presidential primaries in 2008, along with the two general, House special elections in 2009. The study compared Pollster.com's final estimates for those 39 campaigns with the simple polling mean of the last-15 days of those campaigns. Here are the results:
Primary and special election error, 39 campaigns
Pollster.com
Simple 15-day
Mean error
7.03
6.41
Median error
5.76
4.80
Closest
18
20
Two things jump out from this chart. First, the simple, 15-day polling mean again proved slightly more accurate. Second, the error rate from both methods is catastrophic--more150% the error for general elections.
If this election is more akin to a Presidential primary and a special election for the U.S. House, then Coakley's chances of winning are 56% in my methodology (which is extremely close to Nate Silver's 57% estimate), and only 12% according to Pollster.com's current trendline (Brown +7.9%). That is a clear toss-up on my end, and a strong lean toward Brown at Pollster.com.
Next, I took the comparison a step further Looking only at the 13 campaigns for which Pollster.com used its ''Loess' iterative locally weighted least squares regression" (which it only does with campaigns with eight or more polls), and where Pollster.com differed from the 15-day, simple mean by 2.1% or more, I ran some more numbers:
Primary and special election error, 13 late trending campaigns
Pollster.com
Simple 15-day
Mean error
7.46
8.38
Median error
7.36
6.46
Closest
8
5
These 13 cases compare apparent late-trending campaigns. These are important, because my 15-day simple methodology should miss the late trends (because I weight older polls more heavily) while the more sensitive, Loess regression should better pick up the trend. Although the error rate for both methods is terrible, although there are few cases, and although the simple, 15-day mean does better in terms of median error, the Loess regression still does a slightly better job overall of predicting the final results. In 8 out of 13 cases, putting more weight on later polls through a regression method was more accurate.
Now, it should be noted that there are a couple of famous cases where the simple 15-day mean accurately predicted that an apparent, massive late trend was overstated: the 2008 New Hampshire Democratic primary, and the 2009 New York 23rd special election. In both cases, which happen to be the two most prominent of the 13 in the final "study," the simple 15-day mean picked up a much smaller amount of movement toward Barack Obama and Doug Hoffman respectively. It still predicted the wrong winners (just like everyone else), but the elections were predicted to be very close and the eventual error rate was less than catastrophic.
Add it all up, and the campaign is, from a Democratic perspective, a toss-up at best. My gut, and my research, tells me that it won't be a runaway Scott Brown victory, ala the Pollster.com trendline. If I were to go out on a limb without any hard data, I would drop the Pajamas Media poll, and weight the pre-January 12th polls at only 85% of their value. That shows Coakley by 2.3%. If for no other reason, I like that as a prediction (barring more polls), because it still allows me to be contradictory with everyone else. As an election forecaster, when you are lacking a definitive conclusion, it is best to go against the herd. There is just more of an upside to being right in those cases.
However, even I can't deny that the numbers are improving for Brown seemingly everyday, and that he has a good chance of winning right now.
A 3.00% lead for Coakley gives her an 83% chance of victory, according to my numbers. To put it a different way, of the 143 closest elections from 2004-2009, in 48 cases the final polling margin differed from the final result by 3.00% or more. This means the candidate leading by 3.00% has an 83% chance of victory.
Does this conflict with other election forecasters right now? Of course! In fact, the main reason I rolled out my new election forecasting method today, along with the research backing it up, it precisely because it conflicts with other election forecasters. I had been excited about this new methodology for a while, but what better time to introduce it then when it conflicts with virtually all other election forecasters?
I could end up looking like an idiot. It wouldn't be the first time. However, I believe there is strong evidence that this is the most accurate method currently available. Right now, my numbers still put Coakley at a clear advantage, and I am kind excited that everyone disagrees with me (which, in my experience, usually means I will end up looking like an idiot).
STATE BLOGS
