Bernie Sanders’ loss to Hillary Clinton in Tuesday’s New York state primary served as an unofficial final nail in his campaign’s coffin. Sure, he is likely to remain in the race for a while longer, but the mathematical truth is that his current delegate deficit would require him to win impossibly large majorities in the remaining states in order to still be competitive. Sanders’ fans all along have charged that bias towards Hillary Clinton at the DNC and the DNC’s use of undemocratically appointed superdelegates, who heavily favor Hillary Clinton, have been the root of Sanders’ electoral woes. And while these concerns might be more salient if the contest was neck-and-neck, it’s important to realize that Clinton has garnered 2.7 million more votes than Sanders, giving her a 277 delegate lead, without counting superdelegates. So, instead of treating superdelegates or the DNC as scapegoats for Sanders losses, his supporters would be wise to find someone else to blame. Namely, Franklin Roosevelt.
That Franklin Roosevelt? Yep, our nation’s 32nd president actually is having a large effect on the selection of its 45th. As the only president to be elected four times, Roosevelt clearly understood how to win elections, and his best campaign tactic was assembling diverse groups of voters to support both him and his programs. Roosevelt’s coalition of urban voters, white Southerners, intellectuals, and minorities has since been dubbed the New Deal coalition. While the last 80 years have made slight adjustments to this voting bloc – white Southerners now tend to vote Republican, and younger voters now represent a stronghold of Democratic support – the New Deal coalition remains the basis for Democratic turnout.
The problem with assembling a coalition of voters, however, is that in order for the coalition to last, there has to be some unifying factor to tie the coalition together. Roosevelt’s coalition was successful because it was tightly unified in support for his administration’s New Deal programs. Unfortunately for Bernie Sanders, his failures in this primary cycle can largely be traced to his inability to unify enough support from enough facets of the Democratic coalition. Winning a primary campaign for a major party is more about breadth of support across groups than depth of support in a certain group, and Sanders’ campaign was only able to come up with the latter. For example, even in his loss to Hillary Clinton in New York, he still won the under-30 vote by a whopping 34%. Winning a group by margins like that, as Sanders has learned the hard way, is only a recipe for electoral success if the group represents a large size of the total electorate or those large margins can be repeated in other demographic groups. But in New York, under-30 voters only made up 17% of the electorate, and Sanders has been unable to translate his fervent support from young people to other classic Democratic demographics.
Take, for example, urban voters. City-dwellers have tended to lean Democratic ever since the days of the New Deal coalition, but Sanders has had trouble attracting widespread support in big cities. Just look at the county by county results in New York:
Considering that Clinton won New York readily, it’s surprising to see Sanders’ light blue dominate the state. Clinton’s key to victory was the state’s urban areas – as the only counties she actually won were in the New York metro area as well as the counties where the cities of Buffalo, Rochester and Syracuse are located. Sanders actually won both suburban and rural voters in New York, but since Democrats are most densely packed in cities, Clinton’s urban advantage was all she needed.
Sanders has also had trouble attracting minority support, especially among black voters, who have historically made up a key component of the Democratic coalition. Reasons abound for why black voters tend to prefer Clinton, from lower African-American unemployment rates during her husband’s administration to Clinton’s longer time dealing with racial issues in the public eye and Sanders’ campaign gaffes while discussing racial topics. There’s no simple underlying answer – the reason for high Clinton support among minorities likely comes from a mixture of these factors and individual voter preferences. But while it’s easy to argue about the rationale for Sanders’ struggles to win minority support, it’s hard to argue about the ultimate effects of minority voters heavily favoring Clinton. Plotting voter demographics against Sanders support shows a clear racial disparity between Sanders and Clinton supporters1:
Clinton has won fifteen of the sixteen states with more than 10% of black voters, while Sanders has won every state with less than 3% of black voters. Reciprocally, with the exception of Iowa (which voted first, when Sanders had less name recognition than he does now), Sanders has won all of the eight whitest states that have voted so far. Hawai’i is a notable exception to the general trend, as its high proportion of Asian and Pacific Islander residents makes it have low percentages of both white and black voters. However, ignoring Hawai’i, we discover that Clinton has won all thirteen of the least white states that have cast their vote. In a party that prides itself on being multiracial, that’s quite a racial divide. And unfortunately for Sanders, his campaign has cobbled together a losing demographic mix from the diverse party Roosevelt helped establish.
At the end of the day, it takes a coalition to win an election. The Democrats have had a competitive and remarkably stable coalition of voters going all the way back to Franklin Roosevelt, but Sanders’ inability to spread his appeal across the modern Democratic coalition has, more than anything else, helped hand the nomination to Hillary Clinton.
|1.||↵||2016 U.S. Census Bureau State Electorate Profiles were used to determine the racial breakdown of each state. An exception was made for Wyoming, Washington and Wisconsin, for which State Electorate Profiles were not available and 2012 Census estimates for the racial breakdown of the overall population were used instead.|
Much has been written about why Donald Trump is on track to become the Republican presidential nominee. Theories range from Trump as a response to Obama’s upending of racial hierarchy to Trump as an authoritarian figure to economic anxiety presenting itself as support for Trump. Frankly, our technical approach to punditry has very little to offer as to why Trump has appeal and so we won’t bother analyzing Trump through that lens. However, an equally interesting question is why he hasn’t been stopped. For the first several months of the Trump candidacy, pundits, politicos, and even our team were all convinced that Trump’s early support was an anomaly. It was quickly chalked up to another Santorum/Gingrich-type early surge. The thinking went that because Trump was so extreme and so despised by GOP elites, his support would drop. Man, were we wrong.
Although Trump still struggles to win majorities in most states, he is consistently winning pluralities and is well on his way to the nomination. Admittedly, it isn’t entirely clear that he will win the required majority to win on the first ballot, but he may have a sizable enough plurality to win over unbounded delegates and still pull it off. So why is it that nothing has been able to significantly stop Trump?
One of the reasons may be that, contrary to the conventional wisdom, people don’t think that differently about putting a check in a box and answering a question on the phone. Kristen Soltis Andersen at National Review summarizes the conventional wisdom as such: “Some people may say they plan to vote for Trump, but, in reality, they haven’t had to deeply consider the question and so they just say the last name they heard on the news.” (It is important to note that Andersen does not actually take this view as the truth, she merely says that it is the view held by many.) However, it is pretty clear that this is not the case. In fact, as actual elections have started and the primary calendar continues on, Trump’s support nationally continues to climb.
So why is the conventional wisdom flawed on this point? Well, the core assumption behind this theory is that when someone enters a polling place (or caucus site) and puts that check mark next to a name, they think more about the ramifications of their vote than they do when called up by a pollster. However, thinking about the voters’ incentives in the two situations reveals that there is no reason for a different thought process. When asked by a pollster1 to state their preference, the external cost of doing so is zero: polls don’t actually decide outcomes. On the other hand, the conventional wisdom is predicated on a nonzero external cost to casting a ballot2. While the cost may be nonzero, it is essentially zero, as the probability of a single voter’s vote deciding an election is so small, that any costs associated with casting a ballot can be considered zero3. This means that there is very little reason to think that people have significantly different thought processes when casting a ballot and when answering a poll.
With the idea that people would change their mind as election day approaches, the other assumption that many made was that the party would decide. This line of thinking posits that in general, candidates who are broadly acceptable to party leaders tend to win elections because they have easier access to the party’s resources. Well, kind of. There is a lot more complexity then that to the argument, which was first presented in the 2008 book The Party Decides: Presidential Nominations Before and After Reform. In this election cycle, the rise of Trump has prompted possible reevaluation of this work. Daniel Drezner, a professor at Tufts University, posits that in fact, The Party Decides is why the GOP has not stopped Trump. Drezner argues that because the broad punditsphere was so convinced that the party would stop Trump, the party forgot that it actually had to stop Trump.
This idea may be a bit far fetched and is not necessarily true, but it starts to hit at something: Trump is succeeding because the party is failing. Specifically, the party is failing at coordinating. In economics, a coordination failure occurs when a group of firms in an industry is capable of achieving multiple equilibria. However, in order to reach the highest equilibrium, they must coordinate, which they fail to do, forcing them down to a lower equilibrium. The same concept can be applied to this election: Trump may represent an equilibrium in the GOP, but he is certainly not the highest equilibrium. However, because the party failed to coordinate, it failed to achieve a higher equilibrium, such as Marco Rubio or Jeb Bush.
So why did this coordination failure occur? Well, staying in the realm of economics, this coordination failure is likely because the costs of coordination were high and the costs of not coordinating seemed low. In order for GOP leaders to coordinate, they would have had to settle early on one particular candidate and pour resources into that campaign in order to beat Trump. This is costly because it would have required them to force other well-qualified candidates out and for much of the campaign, GOP leaders seemed unwilling to settle on a candidate. Even now, as many GOP elected officials are endorsing Ted Cruz, they are doing so rather reluctantly. On the other hand, because of a belief that Trump could never win and a low probability of support from a single person making the ultimate difference, individual GOP leaders were presented with a low cost of not coordinating. This meant that for much of the campaign, the cost of a coordination failure was seen as small compared to the costs of coordinating.
Now however, the cost calculus seems to be changing. Ted Cruz is slowly picking up more and more endorsements, indicating that many party leaders are viewing him as the last best hope to stopping Trump. But even if Cruz picks up steam, it may be too little too late: Cruz is still incredibly unlikely to win the nomination before the convention. But, the hope for many is that Cruz may be able to prevent Trump from wrapping up the nomination before the convention, leading to a contested convention and the possibility of nominating someone else.
Which leaves one last question: if it comes to a contested convention, isn’t the same coordination failure going to repeat itself? That seems unlikely. Once again, we can take a look at the incentives of delegates at the convention4. Delegates to the Republican National Convention are not normal voters; instead, they frequently have deep ties to the GOP. This may mean they are related to big donors, are elected officials or people who have a large stake in the GOP. This is important because they have a different set of incentives: Most importantly, many delegates have a selfish incentive for a strong party. Many within the party view Trump as an existential threat to the party, and therefore the cost of nominating him is astronomical. If the delegates hold that view, which is likely given the composition of delegates, then the cost of a coordination failure is equally astronomically high. Additionally, with less than 2,500 delegates at the convention, the probability of any one of them casting the deciding vote is much higher. Especially if Trump gets close to the required 1,237, the cost of any one vote becomes very high, providing further incentive to coordinate.
Of course, there is a potential cost, both to individual delegates and the party of denying Trump the nomination. Many in the base may see that action as undemocratic, putting the political careers of individual delegates and the future of the party at stake. However, it seems like many believe that denying Trump the nomination may be a necessary evil to securing the future of the party. It may be close, but a close examination of the incentives at play will continue to provide valuable insight.
|1.||↵||Or robocall or website, it makes no difference.|
|2.||↵||There are of course, other costs associated with casting a ballot. People have to physically go to the polling place, maybe take time to learn about the candidates, etc. This is not factored into this analysis as we are only concerned with the external costs of casting a ballot i.e. what effect on policy a ballot may have.|
|3.||↵||This idea and its implications are developed more in depth by economist Bryan Caplan in his book The Myth of the Rational Voter.|
|4.||↵||There is of course messiness attached to how delegates are bound and the logistics of how a contested convention may work. I will ignore this because I am more interested in determining if a contested convention where Trump loses is at all possible.|
With Super Tuesday now behind us and hundreds more delegates awarded to their respective candidates, the path to the end of the primary season seems much more clear. The big delegate winners of the night were Trump and Clinton, who won seven states each. However, the endgame of the primaries is not to win the most states but to win the most delegates, so let’s take a look at the number of delegates each candidate has won to date.
The Democratic side only has two candidates competing for the nomination, so it should be the easier side to analyze. However, giving delegate totals for the Democratic side is actually a little tricky due to the large amount of superdelegates whose votes are not tied to any primary or caucus result. Many, but not all, of these superdelegates have already expressed preference for one candidate or the other, allowing us to include them in each candidate’s delegate total. However, it is important to recognize that these superdelegates can change their support at any time leading up to the nomination.
Including superdelegates, Hillary Clinton has 1,052 delegates to Bernie Sanders’ 4271. 2,383 delegates are required to win the nomination. This means that Clinton still requires 1,331 delegates to clinch the nomination. Considering that there are 3,284 Democratic delegates not yet allocated, a little bit of division tells us Clinton needs to win 40.5% of the remaining delegates to secure her position as the Democratic nominee. On the other hand, Sanders would require 59.6% of the remaining delegates to win the nomination.
However, there is a critique to counting superdelegates in the totals at this state of the nomination process. Since superdelegates are not bound to candidates, they can change their preferences and affect our previous calculations. Many Sanders supporters argue that if Sanders starts outperforming Clinton later in the primary process, superdelegates that currently support Clinton will be pressured to support Sanders. Since Clinton currently has an overwhelming lead in superdelegate support, this would make the nomination easier for Sanders. We can mathematically adjust for this assumption by ignoring the existence of superdelegates and calculating how many regular delegates each candidate would need to clinch the nomination. This model works under the assumption that superdelegates will ultimately cave in to popular support. Ignoring superdelegates, Clinton still has a delegate lead of 594 to 405. In this hypothetical system, winning 2,026 regular delegates ensures a majority, and Clinton would need to win 46.9% of the remaining delegates while Sanders would require 53.1%. So yes, Sanders supporters are correct in arguing that a late superdelegate shift would help their candidate, but it is important to note the effects of such a shift are minimal.
The Republican side is a little more complex, as there are more than two candidates to consider. It takes 1,237 delegates to win the Republican nomination. Of the 695 delegates already awarded, Donald Trump has 319, Ted Cruz has 226, Marco Rubio has 110 and John Kasich has 252. Interestingly, even though Trump has been billed by the media as the presumptive Republican nominee, he has only won a plurality of the delegates so far and not the majority he would need to win the nomination without a brokered convention. Trump needs to win 51.7% of the remaining delegates to win the nomination, while Cruz needs 56.9%, Rubio needs 63.4% and Kasich needs 68.2%. Of course, due to the possibility of a contested convention for the Republicans, the candidates know that they do not necessarily need to win a majority of the delegates to have a shot at the nomination. Instead, they simply need to prevent other candidates from winning a majority.
So, if Hillary only has a slim delegate lead over Sanders and Trump hasn’t even won a majority of delegates so far, why is the media portraying them to be the presumptive nominees? The answer is less to do with math and more with scheduling. The early Republican primaries and caucuses tend to allocate delegates more proportionally, while the later ones allocate using a winner take all system. In most states, Trump has rather safe, steady leads, and instead of having to split delegates proportionally with the losers, he will soon win all delegates from each state. This is crucial for Trump, who leads many states with a comfortable plurality of voters but not a majority. That situation has cost him delegates up until now, but starting March 15th when every contest is winner take all, he will be racking up all the available delegates without having to worry about his margins of victory.
Meanwhile, the early primary and caucus states have actually been rather generous to Sanders demographically. Nationally, he is losing to Clinton 52% to 38%, and his 14 point deficit puts him in an especially dire situation because, unlike the Republicans, Democrats award their delegates more proportionally. This means that he not only needs to catch up to Clinton in a more diverse slew of states than he has already won, but he also must win by large margins in order to overcome his current delegate deficit.
At the end of the day, can we say for sure that Trump and Clinton have their respective nominations locked up? The numbers say no. But perhaps more importantly, the calendar says that unless something drastic happens soon, the numbers won’t be so tenuous for long.
|1.||↵||All delegate totals are current as of the time of publication on March 2, 2016.|
|2.||↵||Other candidates who are no longer in the race acquired delegates, so the totals of the individual candidates detailed here do not add up to 695. The calculus of where those candidates’ delegates go is a bit complicated and dependent largely on state rules. For example, some are not able to vote for other candidates (even if they have been “released”) until the second ballot, while others can vote for whoever the want on the first ballot. Additionally, in states like Iowa, where the delegates are ultimately decided at a state convention, these delegates may be reallocated before the convention. For the sake of simplicity, we simply ignore these delegates.|
Yesterday, 735,000 Republicans in South Carolina and about 80,000 Democrats in Nevada1 headed to the polls and voted or caucused in the third day of voting in this year’s primary cycle. Here, we break down what exactly happened yesterday and what it means going forward.
South Carolina Republican Primary
Coming out of New Hampshire, Donald Trump had won big, John Kasich pulled out a surprising second place finish, Jeb Bush was riding high on beating Marco Rubio, who was suffering after an embarrassing debate performance, and Ted Cruz still confident after his win in Iowa. New Hampshire and Iowa had served to narrow the field a bit, forcing Mike Huckabee, Rick Santorum, Carly Fiorina and Chris Christie out. In South Carolina, Trump looked poised to pull off another large victory, leaving Cruz and Rubio vying for third and Bush just not wanting to finish too far behind them. Kasich had absolutely no expectations going in and Ben Carson is somehow still running. And, as expected, Donald Trump ended up the big winner of the night.
Trump won all of the delegates from yesterday’s contest, giving him a big lead heading into the Nevada Republican Caucuses and Super Tuesday. Rubio and Cruz ended up pretty much tied, but more importantly, Rubio showed that he has mostly recovered from his poor debate performance and is once again the conventional wisdom establishment-backed candidate. More unfortunately for Cruz, South Carolina is an example of why the delegate math is going to be challenging for him; Ted Cruz is expected to do best in many Southern states going forward, a large chunk of which are winner-take-all in each congressional district, meaning that a close second may not mean much in terms of actually winning him delegates.
But the biggest news coming out of South Carolina on the Republican side is that Jeb Bush has announced he is suspending his campaign. This is significant for two reasons. Although there are myriad think pieces dissecting why the Bush campaign failed, I will a chart which contains just one more reason.
This chart adds together all of the advertising minutes from Iowa, New Hampshire and South Carolina of TV ads from both the Bush campaign and Right To Rise, Bush’s primarily supporting SuperPAC. The biggest take away is that the campaign failed to really advertise until Bush had largely slipped in the polls and had little chance of recovery. It poured money into the first three states, but only right before they were to vote, meaning that they could only sway late-breaking voters. While there certainly are many late-breaking voters, they tend to vote strategically, meaning that if their primary focus is say, stopping Trump, they are more likely to vote for someone like Rubio than Bush. Bush’s advertising campaign simply came too late to substantially change public perception of him as a viable candidate.
Bush’s dropping out is significant for both electoral and monetary reasons. On the electoral side, it frees up the “establishment lane” of the GOP. With Bush and Christie now both out, more moderate, establishment-minded Republican voters really only have a choice between Kasich and Rubio. Taking a look at polling2, we can see specifically that this helps Rubio the most.
We can then take the 6.3% of support that Bush had in the latest HuffPost Pollster average and split it up to the different candidates.
Clearly, Rubio is the biggest beneficiary to Bush dropping out, but that 6.3% being split multiple ways means that it honestly does not have a huge difference on pure polling numbers.
The second effect of Bush dropping out, freeing up big donors, is much more significant. Although Bush never got very far off the ground enticing voters, he was very successful in enticing donors. However, with him out of the race, many of these donors will be able to move their money around freely. And the general thinking is that many will move their money over to Rubio who can then use it to execute the same kind of attack-ad heavy campaign that allowed Romney to knock out his rivals in 2012.
Ultimately, the general thinking is that Bush’s dropping out leaves more room for Rubio to win over establishment support and use that support to win the nomination. However, with Trump and Cruz still putting up strong performance, even if Cruz cannot win many winner-take-all states, it seems increasingly likely that this race will drag on well past Super Tuesday.
Nevada Democratic Caucuses
On the Democratic side, yesterday brought more of a return to status quo than any substantial change. Although Bernie Sanders did better than expected in Iowa against Hillary Clinton and followed it up with a demanding victory in New Hampshire, Hillary Clinton has steadily been a strong favorite to win the nomination. Going into Nevada, it seemed as if some nebulous “momentum” might be on the side of Sanders, although little was certain because of a dearth of polling in the state. However, Hillary Clinton pulled off a convincing win over Sanders.
Although her victory was not the 20+ point lead that was seen in the polling last fall, reflecting the overall growth of Sanders as a competitor to Clinton, it likely effectively stopped any momentum that Sanders had coming out of New Hampshire and Iowa. And heading into South Carolina, it will likely prevent the downward trend in Clinton’s polling there that she has experienced since December. Nevada also showed that Clinton’s lead among Latino voters is not as commanding as it was eight years ago3 and South Carolina will let us know if her lead over Sanders among Black voters is as strong as it is assumed to be.
All in all, the Nevada caucuses did very little to change the Democratic race drastically; rather, it restored it back to the conventional wisdom that Clinton will win the nomination. South Carolina will be another test of this hypothesis. If Clinton is able to come away there with a commanding lead, she will have the momentum, and demographics, behind her.
|1.||↵||Because of the nature of the caucuses, it is almost impossible to know exactly how many people turned out.|
|2.||↵||It is worth mentioning that there is a high margin of error on this kind of a crosstab. Because Bush supporters make up such a small subset of those polled, breaking it down further to who their second choice is creates a high level of error. However, because Rubio’s lead among Bush supporters is so commanding, we feel confident that the general thrust of this poll result is correct.|
|3.||↵||Although it would is hard to argue that Sanders is definitively leading among Latinos, as the entrance poll in Nevada would suggest. For a more thorough discussion of why the entrance poll proves very little, check out .|
Before the Iowa primary, we showed the data behind the conventional wisdom that outperforming expectations in Iowa generally leads to a polling bump in New Hampshire. The logic behind this argument makes perfect sense, as positive media coverage and the bandwagon effect attracts voters who are still making up their minds. And plotting the data from the last few elections, we were able to see a clear positive trend in support of this theory. However, the recent New Hampshire results demonstrate an important but often overlooked truth about poll-based predictions – they assume the status quo remains unchanged. The poll averages that many media predictions are based on do a decent job of measuring current support, but they simply cannot account for future events that have nothing to do with the numbers. So when a surging candidate like Marco Rubio trips up in a debate with heavy media attention, all our predictions go out the window.
For all the uncertainty on the Republican side of things, our simplistic model of converting Iowa support to New Hampshire success actually worked quite well for the Democrats. We first calculated the percent that Bernie Sanders overperformed and Hillary Clinton underperformed against expectations in Iowa. To calculate the expectations, we modified the HuffPost Pollster average with corrections for both the viability rule that hurt Martin O’Malley’s final vote tally as well as for undecided voters1. Using this calculation as our baseline, we found that Bernie Sanders overperformed by 1.2% and Hillary underperformed by 0.6%. In our article on Iowa, we explained that in the last few elections, each percentage point above expectations in Iowa tends to raise New Hampshire performance by 0.55%. This allowed us to calculate, using our historical model2, a prediction for the New Hampshire vote and compare it to the actual results:
|HuffPost Pollster Average||Average Adjusted for O’Malley + Undecideds||Actual Result||Result against Expectations|
|HuffPost Pollster Average Feb. 1st||Average Adjusted for O’Malley + Undecideds||Actual Result||Model’s Predicted Result|
All in all, the historical model was fairly accurate at predicting the actual New Hampshire results, missing by just 1.3% for Sanders and 3.4% for Clinton. While the magnitude of the shift was slightly larger than predicted by our model, the model was successful in adjusting the expectations in the correct direction for both candidates. The relative success of the model for the Democrats was likely caused by the Democratic race progressing more or less how the model’s assumptions predicted them to. Sanders’ campaign beat expectations in Iowa, received energetic press, and saw a small bump in New Hampshire. The Clinton campaign faced media questions about the rising threat of Sanders, and saw a small drop in support in New Hampshire.
The Republican results in New Hampshire, however, defied the expectations set by our model as well as most political pundits. Most pundits declared that Marco Rubio and Ted Cruz beat expectations in the Iowa caucus, while Donald Trump underperformed. Following conventional punditry as well as our quantitative model, we would thus expect Rubio and Cruz to see a bump in support in New Hampshire while Donald Trump’s numbers fell.
In actuality, the exact opposite happened. Trump’s final results actually surpassed his New Hampshire HuffPost polling average from the day of the Iowa caucus, even though his campaign had to fight off negative media attention arguing many of his supporters were unlikely to turn up at the polls. Meanwhile, Rubio’s and Cruz’s final vote tallies were lower than their polling averages a week earlier. What happened? Well, to put it simply, events happened that broke the assumptions made by both our model and the political pundits. Our model predicted that momentum from Iowa for someone like Rubio would carry on to positively influence his level of support in New Hampshire. And this prediction seemed to be coming true, as Rubio’s HuffPost polling average in New Hampshire was climbing the week before the primary. However, an unforeseen and unpredictable debate gaffe likely cost him supporters in the crucial few days before votes were actually cast. Cruz faced favorable demographics in Iowa but unfavorable demographics in New Hampshire, likely playing a role in his disappointing finish. And Trump’s surprisingly high turnout proved to the media once again that it is notoriously difficult to compare Trump’s supporters with historical examples.
So, in conclusion, New Hampshire’s results show that while our models and punditry occasionally get it right, both attempts at political prognostication work best when certain assumptions hold. The Democratics had no last-minute surprises, and estimates were pretty close to the final result. However, the last-second poll swings of the Republicans demonstrate that even the so-called experts base their predictions on imperfect assumptions, and it’s sometimes best to take both polls and pundits with a grain of salt.
|1.||↵||These adjustments are necessary because polls often come in with lower raw percentages than final results because voters can declare themselves as undecided. The problem is exacerbated in Iowa where, due to a viability rule that hurt Martin O’Malley on the day of the caucuses, O’Malley’s final support was much lower than his polling numbers. For the adjustments, the lost O’Malley support was divided among Clinton and Sanders according to the second choice candidates of O’Malley supporters via a pre-Iowa PPP poll. Then, undecided voters were given to Sanders and Clinton in proportion with their polling percentages. This is not a perfect system of modeling O’Malley support or undecided voters, but it is the best way possible to reconcile the differences between a poll and actual election results.|
|2.||↵||We do not claim that this historical model is supposed to be 100% accurate. It is based off a small data set and there is still a great deal of variability. It is simply used as a rough quantitative way to show the post-Iowa bump that pundits often reference.|
Tonight, starting at 9pm ET/8pm CT (or earlier if results come in earlier), Tyler and Matthew will be kicking off our Iowa Caucuses Liveblog, providing analysis and commentary as results come in. You can find the liveblog here!
After months of speculation about who will run, many debates, and discussion of candidates’ chances, tomorrow, February 1st, we get to learn the first results of the election cycle as Iowa heads to the caucuses tomorrow evening. Although we will be providing rolling analysis as results come in tomorrow, we have also put together this article to help you know what to watch for tomorrow.
Arguably the most important thing to watch for tomorrow is how the candidates do relative to expectations. The first way to track expectations is by looking at win probabilities; if a candidate who is given a small chance of winning the caucuses wins, that win will mean much more than an expected win. FiveThirtyEight has put together what they call a “polls-plus” model1 to assign a probability of winning Iowa to each candidate. PredictWise uses betting market data to derive a similar probability for each candidate.
These two methods produce fairly similar results, but both are worth noting. FiveThirtyEight’s probabilities represent what the polls are showing may happen, so those who follow polling closely will build their expectations to be similar to that model’s. On the other hand, PredictWise uses betting markets, so those probabilities more closely represent the “common wisdom” of what will happen. Because the “common wisdom” is affected so much by media portrayals and how the media spins Iowa will be crucial, it is especially worth paying attention to.
So, what do these probabilities actually tell us? Well, on the Republican side they tell us mainly that Donald Trump is a favorite to win the caucuses. That means that Trump needs to win Iowa or else his campaign will be seen as falling apart. Similarly, despite close polling between Hillary Clinton and Bernie Sanders, Clinton is the overwhelming favorite to win Iowa. However, because she has only a small lead in the polls, a loss by Clinton would probably not affect her as much as a loss by Trump would affect him.
In addition to looking at the probabilities of winning, another key part of playing the expectations game is how a candidate does relative to their polling. To understand this effect, we looked at all candidates in the last three cycles (2004, 2008, 2012) and averaged the three Iowa polls released closest to the Iowa caucuses. We then compared these to actual results of Iowa result to get an “Iowa Result against Expectation.”2 We did a similar three poll average of New Hampshire polls prior to Iowa held their caucuses and a three poll average of New Hampshire for the days after Iowa held their caucuses. This allows us to detect a change in New Hampshire’s polling which can, in part, be attributed to the results in Iowa3. Using this data4, we created the following chart. In this chart, the dot size indicates the actual share of Iowa vote that each candidate won.
While not incredibly strong, there is clearly a positive correlation between Iowa results against expectation and changes in New Hampshire polls. By looking in the upper left quadrant, it is also very clear that candidates who beat expectations significantly and win a fairly large share of the vote tend to get the strongest boost in New Hampshire. However, to quantify this relationship a bit more, we can run a linear regression through this data.
By running this linear regression, we find that a candidate who beats expectations by 1 percentage point in Iowa can expect, on average, to improve their standing in New Hampshire by 0.583 percentage points5. So, the effect is not huge, but it is distinct. Certainly, Trump’s and Sanders’ leads in New Hampshire polling are large enough that it seems unlikely for this boost from Iowa to propel anyone ahead of them. However, in the Republican race, Ted Cruz, John Kasich, Marco Rubio, Jeb Bush, and Chris Christie are all so tightly packed that a stronger than expected showing in Iowa (or a weaker than expected one) could change the race significantly.
So, tomorrow when the results start to come in, while everyone will be watching mainly to see who wins, make sure to pay attention to how the candidates perform relative to current polling in order to see how New Hampshire and the rest of the race may shake up. And, of course, join us for our liveblog!
|1.||↵||Their model incorporates state polling, endorsements and national polling. For their full methodology, check here.|
|2.||↵||For example, an average of the last three polls before Iowa showed Rick Santorum sitting at 17% in 2012. Santorum ended up winning 24.6%, so his result against expectation is 24.6-17=+7.6%. On the other hand, Michelle Bachmann was polling at a 7% average before Iowa and ended up winning 5%, so her result against expectation is 5-7=-2%.|
|3.||↵||Certainly, there are many potential confounding factors, but because we are limiting the number of days between the first NH average and the second, we hope to cut down on as many of these as possible.|
|4.||↵||Which is available in whole in this spreadsheet.|
|5.||↵||We also find that by comparing pre-Iowa polling in New Hampshire to the New Hampshire results, beating expectations by 1 percentage point improves standing in New Hampshire on average by 0.55 percentage points, meaning the effect possibly wears off a bit. However, we decided that the difference in time over the three cycles between Iowa and New Hampshire made the methodological inconsistency too great and so we stick to pre-Iowa and post-Iowa polling in New Hampshire.|
With the House of Representatives entering their third week of this year’s session, the team at Electoral Statistics looks back and launches an interactive reviewing the House in 2015. You can check it out here and please provide us with any feedback or fun things that you notice!