A Year-End Look At The 2014 Senate Map And Why Others Are Wrong

I received a number of questions over the weekend regarding the PPD 2014 Senate Map, which currently predicts a far more favorable 2014 election outcome for Republicans than other maps, particularly Crystal Ball and to a lesser extent, predictions by Rothenberg Political Report, and so on.

Thus far, I have released expanded analysis for the following Senate races:

Alaska, Arkansas, Iowa, Kentucky, Louisiana, Michigan, Montana, and North Carolina, with West Virginia — which is rated “Safe Republican” on the 2014 Senate Map — soon to be released as the ninth article.

So, while I have the utmost respect for Larry Sabato and others at the University of Virginia Center for Politics, I would like to expand on my rationale for disagreeing with many of their ratings and offer some insight in to the model used at PeoplesPunditDaily.com.

First and foremost, far too many pundits do not take their role as a predictor of elections seriously, violating the public trust. What political analysts or talking heads say about the status of a race, as well as the findings of pollsters, can have far more sway than I think should be the case. Together, they can impose a sentiment of inevitability or otherwise damaging narratives to one candidate or another, and unfortunately, they know it.

However, that being said, let me provide a few examples of different variables that go in to my election model.

Polling

When the electorate and the media want to know who is leading in an election, we all look to the polls. While polls certainly help to gauge a race, “polls do not vote, people vote.” In order to account for what is often their gruesome inaccuracy, I rate pollsters by accuracy, which I weigh when factoring an average on PeoplesPunditDaily.com.

Most of my liberal friends may just be surprised to find that Rasmussen Reports, for instance, is actually one of the lower rated pollsters. In 2010, though other pollsters and pundits mocked Rasmussen for predicting a sweeping GOP victory, which obviously came to fruition, individual race margins were way off from the actual outcome.

Perhaps it was the exit of Scott Rasmussen or their desire to regain credibility after their 2012 disaster, but post-election 2012 has gotten even worse. They were once far more Republican, but now they have swung to the other side too much, with the Virginia governor race — which had McAuliffe winning by double digits a week before the election — being the last straw for me.

Though I will not get into a back-and-forth or provide a list at this point, I will say that SurveyUSA and Gravis Marketing, on the other hand, are among the top-rated pollsters for their accuracy. Therefore, according to the PPD model, they are more influential in averages and final ratings.

The model of rating pollsters is similar to FiveThirtyEight, but there are significant problems with FiveThirtyEight’s forecasting models. One such problem, in their own words:

These ratings are designed to form an objective assessment of pollster quality with respect to one particular function: their aptitude in accurately forecasting election outcomes, when they release polls into the public domain in the period immediately prior to an election.

Except, as is the case with PPP (Public Policy Polling) and Washington Post polling, for example, FiveThirtyEight does not hold pollsters accountable for their dubious tendency to release surveys aimed at creating or reinforcing a narrative that benefits those they ideologically favor, and often even provide the funding for their surveys. Polling conducted far before or in the weeks leading up to elections too often aim to do just that, rather than accurately predict election outcomes.

The Washington Post, just as Rasmussen, had a ridiculous double-digit lead a few days before the Virginia gubernatorial election, with even more ridiculous margins for Obama in Virginia, Ohio, and so on. Pat Caddell, a Democratic pollster for Jimmy Carter, blasted media-funded public polling during the 2012 presidential election cycle for this, and he was right.

Therefore, their influence on ratings assessed in the months – not weeks – leading up to an election, cannot be as significant as it may be in the weeks leading up to the election, if it turns out that they are even accurate, which many aren’t. Otherwise, the pundit is falling into the very trap the bias pollsters have set for them, and at that point the pundit is just being used to create a false “air of inevitability.”

It is clear, particularly because they don’t offer up much but flawed pollster findings, that the guys at Crystal Ball fall in to this trap, with PPP being the single-most cited pollster at the University of Virginia Center for Politics, from which the Crystal Ball hails.

Bottom line: Polling should corroborate your model, your model shouldn’t rely upon polling. All of this must be weighed, along with state-by-state demographic trends and so on.

Partisan Voting Index (PVI), Demographics, Presidential Approval Rating

Cook PVI, or Partisan Voting Index, carries a good deal of weight in my model at PeoplesPunditDaily.com.

In 2010, for comparison, in states where the Cook PVI was more Republican than D+2, Republican candidates won just under 85 percent of the races. On the flip side, in general, the GOP lost races in states where the electorate was more Democratic than D+2.

This is probably a good time to address some of the more ridiculous ratings at Crystal Ball, and others.

First, Iowa, is trending Republican, with the state disapproving not only of President Obama by larger than average margins (38 – 59 percent), but the entire Democratic agenda. By a 2 to 1 margin, Iowa voters say they want a senator who opposes ObamaCare and a path to citizenship for illegal immigrants, with a plurality saying they want someone who is opposed to stricter gun control laws.

Worth noting, presidential approval rating is still the most telling variable in midterm elections, historically and empirically speaking.

The PVI has held steady in Iowa at D+1 — well within the threshold that carries a 83 percent chance of Republican victory — but also two polls have found a very real anti-liberal sentiment. Braley, who is fairly well known, is trailing a generic Republican in Quinnipiac (46 – 41) and Harper Polling (42 – 38), with Iowans favoring a Senate controlled by the GOP.

The same is true for Louisiana, where Mary Landrieu is favored in Crystal Ball. That is just a ridiculous call, “period.” Data inside the state demographics tell us that the growing black voter population is not even close to compensating for the state’s Republican trend.

When Landrieu was first elected in 1996, 40 percent of Louisiana’s voters were white Democrats. Today, white Democrats account for only 22 percent of Louisiana’s electorate. Yet, it is even worse than it sounds when we look at ideological leanings.

Voter registration is trending against Democratic candidates, as well, but 48 percent of Louisiana voters say they agree with Republicans despite which party they say they chose when they registered to vote. Only 36 percent say they agree with Democrats.

Also, when we look at the Cook PVI (Partisan Voting Index), we see the state is R+12, well above the D+2 threshold for an 85 percent chance of Democratic loss. No poll, particularly a survey conducted by PPP, should outweigh that reality without real corroboration.

Both Iowa and Louisiana are rated “Leans Democrat” on Crystal Ball, while both are justifiably rated “Toss-Up” on the PPD 2014 Senate Map.

Candidate Recruitment And Organization

I placed Cook PVI just prior to candidate recruitment and organization to underscore just how important candidate recruitment and organization can be. In fact, when we look at recent election cycles with state-by-state PVI in mind, Todd Akin and Richard Murdock represent the tip of the iceberg when it comes to blown elections that Republicans should have won.

It is a long list that includes more than I care to cite, but the point is moot and evident now. But that doesn’t really matter to the model, at least at this point, because until the Republican Party makes a clear mistake during the nomination process we cannot simply give the Democratic incumbent the benefit of the doubt, as they seem to do at Crystal Ball.

Bottom Line

It is true that Republicans unseated just 3 Democratic incumbent senators in the last 10 years, but data is data, and we have never had an election in which ObamaCare was put on trial, save for 2010.

Whether the GOP candidates can effectively make the case against their opponent is yet to be seen, but the pundits cannot assume that the 2014 candidates will behave as the 2012 candidates did, and would do better to focus on the fundamentals of the midterm election. The environment clearly favors the Republican Party, and if you thought 2010 was a defeat for the Democratic Party, you ain’t seen nothing yet.

View the PPD 2014 Senate Map