PAAR in 2016: A Democratic Edge

In July of 2012, we released our state-by-state predictions of where support for the Democratic and Republican candidates for President would likely end up, using our PAAR model (Percentage Above Anticipated Result). As we discussed in the wrap-up of the predictions, the accuracy overall was something we were quite proud of. Overall, we were off by an absolute value of just 2.3 percentage points, and in the swing states the results were even closer at 0.8 points. Additionally, we predicted that Colorado would be the state that put the Democratic candidate above the 270 threshold, which was in fact the case. As a quick reminder, this model uses no polling results and ended up more accurate than most poll aggregators even though our results were release in July of 2012.

Shortly after the election, we ran our PAAR model and teased out some of the potential Republican contenders and how they might play into the most decisive areas of the 2016 election. Over two years later, Republicans seem to be recognizing our suggestion that 2016 might be decided in the upper Midwest (although we are doubtful that many Republicans read our papers), as Scott Walker continues to surge among Republicans in the latest polls.

This month, we examine the likely results for the two eventual nominees in each of the 50 states, and once again predict which state will tip the scales for our next president.

Recently, Nate Silver did a piece saying that the “Blue Wall” for Democrats is a myth. Based on our modeling, we could not disagree more. As we have been arguing for a while now, presidential elections have been, and will likely continue to be, very predictable. Only large events, such as credible 3rd party candidates (see Ross Perot), have derailed expected results. While Silver clearly would like to drive eyes to his site for the next 17 months, our model shows that the Blue Wall is real and that it will take a large change in events to derail the current path we are on when it comes to electing our President.

The biggest drivers of Democratic success in recent presidential elections have been a) demographic changes, and b) the solidifying of the Hispanic vote for Democrats since 2004. Nationally, minorities were the fastest growing population in the country over the last decade, with Hispanics accounting for over half of overall population growth, becoming the second largest ethnic group in the country. Additionally, while Democrats have traditionally been strongly supported by African Americans, they have also meaningfully increased support among Hispanics, from 53% for John Kerry in 2004 to 71% for Obama in 2012. Although this may reflect national shifts, these trends are even more significant in several electorally critical states, shoring up Democratic support in already blue states, while continually moving traditional toss-up states in the Democrats’ favor.

As we explain below, the Democratic nominee is most likely to win on November 8, 2016 and Minnesota is the most likely state to break the 270 electoral vote barrier.

The question on everyone’s mind after the 2012 election, is whether the “Obama Coalition” willcontinue to exist without Barack Obama running for President. The coalition, made up of increasing support and turnout of African-American, Latino, and young voters, allowed Obama to win easily, despite historically low support among white voters. Will Hillary Clinton or another Democrat be able to replicate that support without “being” Barack Obama?

While there was undoubtedly support driven by the magnetism of Barack Obama, it is not unreasonable to suspect that many of the gains can, and will, continue to be replicated by whomever the Democrats nominate in 2016. When we first created the PAAR model for the 2008 presidential election, it predicted that Democrats would win the election. Despite the fact that Democrats had lost the last two elections, the model recognized that the very demographics that drove the “Obama Coalition” would help whomever the Democratic nominee was. While in 2008, then Candidate Obama outperformed the model considerably (close to 3 points overall), this over performance was reduced in 2012 (+2.3).

The model, which is based heavily on demographic patterns, does not factor in specific candidates and can understate support in certain circumstances, such as a home state advantage. Thus, President Obama in 2012 did much better than expected in Hawaii. However, his results fell in line in all of the swing states, with the exception of Virginia where he over performed by 2 points. In other words, while Obama over performed expectations in 2008, in 2012, he fared as well as would be expected for any Democratic candidate.

With Republicans finding new and interesting ways to alienate Black and Hispanic voters across the country, we see little reason why support levels would change drastically for the Democratic candidate in 2016. The same theory holds for younger voters, as well as college educated white voters. As a reminder, Obama won just 39% of white voters in 2012, and given the likelihood that minority voters will make up 30% of the electorate next year (compared to 28% in 2012), the Democratic nominee could drop white support to about 37% and still win more than 270 electoral votes.

When talking about an election that is still well over a year away, most political pundits and experts add an asterisk to any and all predictions and polls, stating a lot can happen in 17 months and it’s just simply way too early to be making these predictions (although clearly that doesn’t stop them). The logic goes something like this: they believe that candidates themselves can make or break an election, or that their home state can give them an extra boost that a generic candidate wouldn’t receive. Perhaps they focus on the money each candidate (or Super PAC) is able to raise that will play a decisive role. Or maybe, in addition to many other factors, pundits simply believe that presidential campaigns are a long process, with lots of time for candidates to make gaffes, get a little too close to their opponent in a debate, or say something they shouldn’t. All of these factors that can turn an election in the favor of a candidate cannot possibly be presumed 17 months out.

In theory this line of reasoning is correct, in practice, however, it is a different story. If you look at the2014 Senate PAAR results, there were very few elections where an over or under performing candidate had a meaningful impact on the result. In 2014, 15 Senate Democratic candidates were more than three points off their anticipated result, as were 20 Senate Republicans. While this may seem to be a significant number, the fact that Nels Mitchell in Idaho over performed expectations by 3.7 percentage points is impressive but not significant since it only brought him up to 34.7%. Likewise, Senator Lindsey Graham underperforming his expected result by 5.7 percentage points did not change the fact that he won his seat by a safe 18 point margin.

In the few elections that the candidate and campaign did matter, such as the poor performances of Bruce Braley in Iowa and Mark Udall in Colorado, it cost both of them the election. While on the Republican side, Joni Ernst and Cory Gardner used Braley and Udall’s poor performances, respectively, to come out on top. Additionally, both Bill Cassidy and Tom Cotton’s campaigns turned what would have been a nail biters into landslides, both performing at least 6 points above anticipated. As a result, there were only 4 elections, out of 32, where the campaigns and performance of the candidates actually had any meaningful impact on the result.

Additionally, both presidential candidates will raise such exorbitant amounts of money that, at the end of the day, it doesn’t make that much of a difference. Overall in the 2012 election, $1,107,114,702 was spent in support of Obama’s reelection and $1,238,097,161 was spent in support of Romney[1]. While $130,982,459 (the difference in their spending) is a lot of money that would make a huge difference to most people, when we’re talking about billions of dollars spent and TV ad time being bought out months prior, that extra hundred million doesn’t mean a whole lot, and obviously didn’t help Romney on Election Day. As we argued after the 2012 election, once spending gets to this level it reaches the point of diminishing returns. This cycle, like all recent cycles, both nominees will raise copious amounts of money, but the candidate who happens to have the extra hundred million dollars to spend on their campaign in the final weeks will not change the results in any dramatic way.

Given the changing demographic trends, it should be no surprise that our model once again predicts a Democratic victory next November. What is even more interesting, in our opinion, is the shift in where the states are likely to land compared to other states, especially in swing states.

In 2012, we correctly projected that Colorado would be the decisive state to push the winning candidate over 270 electoral votes, with Iowa and Ohio being the next two closest states. In 2016, the model projects that Minnesota will become the new deciding state. Colorado has moved up in the rankings and while the expected vote percentage has risen in Minnesota for the Democratic candidate over the past 4 years, the increase is relatively small (0.2 percent) compared to Colorado which increased by 1.9 percent.

As the top ten deciding state rankings shows (along with the full list in the Appendix), the status of the states continue to trend in favor of the Democratic candidate on the presidential level. In only seven states has the expected result moved in a negative direction for the Democratic candidate while it has increased in 43 states, plus the District of Columbia. In four of the states expected to decrease, the movement will have little effect on the overall results. While Louisiana and Arkansas are expected to increase their Republican vote by over a point, that does not affect either party’s game plan. Neither does Massachusetts or New Mexico’s expectation to vote under a point less Democratic this time around. The only three critical states where Democrats have lost momentum is Ohio, an important swing state that decreases by 0.5 this cycle, Arizona, an up and coming swing state that decreased by nearly a point, and Missouri which is expected to vote 3.6 points less Democratic than it did in 2012. However, across the country, save a few states, demographic trends will continue to help the Democratic candidate while hindering the Republican candidates.

This is part of a broader trend for Democrats. States with large minority populations, such as New Mexico and Nevada, which were once difficult pick-ups for Democrats have now moved firmly into Democratic territory. New Mexico was barely won by Gore in 2000, lost by Kerry in 2004, and now the model predicts Democrats will earn a higher share of votes in New Mexico in 2016 than New Jersey.

For many, Georgia’s appearance on the top ten list above is surprising. While public polling in the 2014 Senate and gubernatorial elections were more favorable to the Democratic candidates than the end result, when we look at the Senate PAAR, Georgia behaved as was generally expected. Indeed, Michelle Nunn over performed expectations by 2.8 percentage points. Therefore, we have no reason at this time to believe that the presidential PAAR numbers are overly optimistic or rosy for the eventual Democratic nominee and Georgia will be the next state to move into the swing state range.

In 2012, the model predicted seven swing states (Colorado, Ohio, Iowa, Missouri, Florida, Virginia, and Nevada). However, this cycle, the model does not even put Nevada in the top ten, as it is predicted to vote more Democratic than both Pennsylvania and New Hampshire. This cycle, the model predicts 8 swing states (Minnesota, Ohio, Virginia, Florida, North Carolina, Georgia, Colorado, and Iowa). Two of the top three swing states in 2012, Colorado and Iowa, are at the bottom of the swing state list for 2016, as they are both predicted to favor the Democratic candidate by at least 3 points, an approximate 2 point increase from 2012. Additionally, North Carolina and Georgia’s 3.2 and 4 point increases, respectively, has moved them into swing state contention. Of the swing states in 2012, Ohio is the only one that has not moved positions, even becoming slightly friendlier to Republicans as previously mentioned. For Republican presidential candidates to have a chance in the Electoral College in the future, the GOP first needs to come to terms with the demographic changes that are solidifying Democrats’ place in the White House, and then find a way to address their deficiencies in this area.

For argument’s sake, let’s say there is truth behind all of the talk about no Democratic candidate being able to follow Obama’s impressive performances the past two cycles and that the Obama coalition will not hold as strongly as we predict. Our model allows us to account for the effect of his candidacy in 2008, which was an increase of 2.9 percentage points over the expected result.

In that scenario, Democrats still win Minnesota with 49.9% of the vote to squeak by with 272 electoral votes. One could look at these numbers and say that at that point, it’s anyone’s race, however, this is the absolute worst case scenario for a Democrat. If the Democratic nominee runs a terrible race and the Republican nominee is infallible, the Democrat would still win. Unless some major event happens that changes the face of the election, the Democratic candidate will be the one getting sworn in in January 2017.

As we’ve discussed, this model has performed quite well over the last couple presidential elections as well as in the Senate elections. Clearly there have been campaigns that have exceeded expected results, such as Heidi Heitkamp and Scott Brown in 2012, Greg Orman and Susan Collins in 2014, and President Obama exceeding the expected result in Virginia by two points in 2012. With the predictability of presidential races, the PAAR model is designed to also assess the effectiveness of campaigns. As was the case in 2012, Obama won all but one of the “swing states,” however, that does not mean to say that we should make all of those victories equal. Obama’s over performance in Virginia should be viewed as much more impressive than the fact that he won Florida, at least from a statistical standpoint. We still have 17 months to go until the presidential election, but the numbers show that the Democratic nominee will be victorious come November 2016.

All of this is not to say that Democrats can afford to be complacent about the 2016 election. They are not winning back the House in 2016 or likely any other time this decade. Additionally, the Senate, which we will project results for later this year, looks likely to stay in Republican hands (although by a smaller majority). Democrats losing the presidency would put Republicans firmly in control of the executive and legislative branches of government for the first time since 2004. While Democrats hold a strong advantage in the race for the White House, the party should not just sit back and relax. Democrats must continue to build upon and learn from the lessons of their past campaign failures and successes to solidify a 2016 victory.

On the other side of the aisle, Republicans need to address their lack of support among minority voters. A large undertaking is apparently underway, however we have yet to see a return on their investment. Any debate over moderate versus conservative candidates is just masking the systemic problems the party faces when it comes to the Presidential elections. In the end, Republicans do not need to win a majority of support among Hispanic voters, but they must keep the level of support above 40% if they want to have a chance of winning the magic 270 electoral votes.

There is about a year and a half to go before the votes are counted on November 8, 2016, but our model points to a very predictable end result. The Blue Wall is a real phenomenon and anyone saying otherwise is just trying to generate attention for the next year and a half.

Appendix 1

***Tipping point state for both parties.

***Tipping point state for both parties.

For a downloadable pdf of the white paper, click here.