Two Approaches to Predicting Which Party Will Win in 2016

By Jonathan Ladd

The 2016 presidential election is still more than a year away. It is highly likely that Hillary Clinton will be the Democratic nominee, but it is difficult to tell who the Republican will be, with 18 currently declared candidates and Republican elites slow to generate a front-runner through endorsements. Given that we don't know the comparative personal strengths of the two tickets, and considering that political scientists think the effect of the candidates is small anyway, what can we say about the prospects of the two parties? There are two different perspectives on evaluating which party is likely to win the presidency in 2016. But I think one is much more useful than the other.

The National Conditions Perspective

Douglas Hibbs "Bread and Peace" model, a prominent model based on
the National Conditions Perspective. For more, see
http://www.douglas-hibbs.com/

Political scientists know that we can explain and predict a large portion of the variation in presidential election results by knowing three things: economic growth in the first three quarters of the election year (more is better for the incumbent party), the number of terms that the incumbent party has held the presidency (more terms is worse for the incumbent party), and in the post-World War II era, the cumulative number of casualties in any ongoing overseas wars started by the incumbent party (more casualties hurt the incumbent party). Some models also include presidential approval (higher is better for the incumbent party), but I would argue that approval, while predictive, is largely a consequence of these first three variables, not a separate cause.

These models are how political scientists most often explain aggregate presidential election results in academic journals and the popular media. This perspective is often called predicting using "the fundamentals." There is almost always a panel at the Labor Day weekend Annual Meeting of the American Political Science Association in presidential election years where political scientists make presidential predictions based largely on different ways of measuring and modeling these types of variables. A symposium of this style of predictions usually appears in the APSA journal PS: Political Science and Politics before the election, and a follow-up symposium in the same journal after the election allows the modelers to assess their predictive accuracy this time.

An important part of this type of modeling is that it doesn't just ignore the candidates who are nominated. It also treats the parties themselves as interchangeable. All you need to know is which party currently holds the presidency. If the opposite party held the presidency under similar circumstances, these models would predict the exact opposite result. They assume no underlying difference in electoral strength between the parties. Parties would perform identically under identical circumstances.

The Strength of Party Coalitions Perspective

An alternative way to think about how the parties will fare in 2016 is to look at the demographic coalitions of the Democrats and Republicans. The percentage of different geographic, racial, age and other types of groups supporting different parties changes over time. Sometimes groups that are strongly supportive of one party (and associated with that party in the public's mind) gradually become more evenly divided between the parties. And sometimes  a demographic group that was evenly divided gradually begins to overwhelmingly support one party over the other. On top of this, even if the partisanship of each group stays the same, changes in the demographic make-up of the country might matter for election outcomes. If the groups that support one party grow as a percentage of the electorate, that party is in a much stronger position.

It is very tempting to use the changing demographic support bases of the parties to try to predict the outcomes of presidential elections. And given that voters' party identifications and the demographic composition of the electorate both change only gradually over time, it is logical to suspect that, if one party gains a demographic advantage, it might persist for many years.

Kevin P. Phillips engaged in this type of analysis in his 1969 book The Emerging Republican Majority. It might appear that his prediction was right. After its publication, Republicans won 4 of the next 5 presidential elections (1972 through 1988). Viewed from the  National Conditions Perspective, these outcomes could be explained by very strong economic performance in 1972 and 1984, very week conditions in 1980, moderate growth in 1988, and middling economic performance in 1976 producing close to a popular vote tie. But from the Strength of Coalitions perspective, this pattern could be explained by Republicans holding a larger coalition supporting it at the national level, interrupted only by the Watergate-influenced 1976 loss.

In addition to National Conditions providing a strong rival explanation, another complication with interpreting these elections this way is that there wasn't a Republican advantage in party identification. More people identified as Democrats than Republicans in the American National Election Studies throughout this entire period. But perhaps there was an advantage in people willing to support the national Republican Party. There was heavy split ticket voting in these years, with many voters telling pollsters they saw themselves as Democrats, while voting for conservative Democrats for Congress and for Republicans for president.

More recently, in 2004, in a deliberate homage to Phillips' book, John Judis and Ruy Teixeira published The Emerging Democratic Majority, claiming that the demographic trends favor a growing Democratic advantage in presidential elections. While George W. Bush's reelection seemed to contradict Judis and Teixeira's thesis, Obama's wins in 2008 and 2012 with a younger and more racially diverse coalition has led some to proclaim that  Judis and Teixeira were right.

During the 2012 campaign, Ronald Brownstein reported,

Republican strategists clearly feel the weight of trying to assemble a national majority with so little support among minorities that they must win three in five whites. “This is the last time anyone will try to do this,” one said. A GOP coalition that relies almost entirely on whites could squeeze out one more narrow victory in November. But if Republicans can’t find more effective ways to bridge the priorities of their conservative core and the diversifying Next America, that weight will grow more daunting every year. 

Charlie Cook and David Wasserman recently examined demographic trends and found them looking fairly ominous for the GOP:

In 1980, when nonwhite voters were just 12 percent of the electorate, Ronald Reagan won 56 percent of white voters and was elected in a landslide. But in 2012, when nonwhite voters accounted for 28 percent of the electorate, Mitt Romney took 59 percent of white voters—and lost the presidential race by 4 percentage points. Without a total brand makeover, how can Republicans expect to prevail with an even more diverse electorate in 2016?... 

If the electorate evolves in sync with the Census Bureau's estimates of the adult citizen population (admittedly, a big if), the white share of the electorate would drop from 72 percent in 2012 to 70 percent in 2016; the African-American share would remain stable at 13 percent; the Latino portion would grow from 10 percent to 11 percent; and the Asian/other segment would increase from 5 percent to 6 percent. If the 2012 election had been held with that breakdown (keeping all other variables stable), President Obama would have won by 5.4 percentage points rather than by his actual 3.85-point margin.

In addition, the group with which the GOP does best—whites without college degrees—is the only one poised to shrink in 2016. President Obama won just 36 percent of these voters in 2012, while 42 percent of white voters with college degrees pulled the lever for him. But if the electorate changes in line with census estimates, the slice of college-educated whites will grow by 1 point, to 37 percent of all voters, while the portion of whites without degrees will shrink 3 points, to just 33 percent of the total.

Even Supreme Court Justice Ruth Bader Ginsburg appears to support the Democratic demographic advantage thesis. One of the reasons that she will not strategically retire during Obama's presidency appears to be that she thinks that Democrats can reliably win presidential elections for the foreseeable future: Ginsburg told the Washington Post in 2013: "I think it’s going to be another Democratic president [after 2016]. The Democrats do fine in presidential elections; their problem is they can’t get out the vote in the midterm elections.”

Evaluating the Perspectives

Which perspective better helps us understand what will happen next year? I tend to favor the National Conditions Perspective, as I think do the majority of political scientists who study elections. The reason is that one can explain all the post World War II presidential election results pretty well just looking at national conditions, without considering the parties' demographic support bases at all. In contrast, an analysis based on the demographic support bases of the parties has a tougher time explaining swings from a party winning convincingly in one year, to the other party winning convincingly 4 or 8 years later, as often happens.

However, earlier eras of U.S. history and other democracies show that it is possible for a party to hold a long-term electoral advantage. Prior to the Great Depression, there was a string of eras where one party dominated U.S. presidential elections. The Democratic-Republican Party choose every president in the 1800 through 1820 elections. After Andrew Jackson and Martin Van Buren created the modern Democratic Party in 1828, the Democrats defeated the Whig Party in 7 of 8 presidential election from 1828 through 1856. The election of Abraham Lincoln and the Civil War ushered in an era of Republican dominance, in which the GOP won 14 of 18 presidential elections from 1860 through 1928, with one of those 4 losses the three way contest in 1912 in which two Republicans split the GOP vote. And if you look internationally, examples like Japan's Liberal Democratic Party after World War II, South Africa's African National Congress after apartheid, and the Israel's Mapai/Labor Party from independence through the 1970s indicate that a party can hold a dominant position in a country for decades. But I believe that there has been no similar era partisan dominance in the post-World War II U.S., and I would need to see much more evidence to convince me that one was on the horizon.

Outside of an era where one party dominates, the Strength of Party Coalitions Perspective doesn't add that much. Say I want to predict the 2016 result based on national conditions. I already know that the Democrats have held the presidency for 8 year and that Obama has not started a new costly overseas war with ground troops. So I would say that the result will depend on the economy in 2016, with the Democrats needing a stronger economy than they needed in 2012 because it is their second attempt to retain power. I can simply wait until 2016 and make a prediction.

How could I make a prediction with the Strength of Party Coalitions Perspective? I know that, if party voting within each demographic group stays the same, population change will increasingly favor the Democrats, but will it be enough this year? Hillary is taking positions on immigration, LGBT rights, criminal justice reform, etc. suggesting that she is trying to mostly maintain Obama's demographic coalition, not recreate her husband's 1990s coalition. Yet her race and gender may make small changes at the margins, but how much? It is plausible that African American voters will still overwhelmingly support the Democrats, but will both turnout and vote Democratically at slightly lower rates than in 2008 and 2012. It is also plausible that swing state whites will vote Democratically at a slightly higher rate than in 2008 and 2012 due to racial biases. But how will the small increase in white support and drop in African American support balance out? Which will have a bigger affect? It is very hard to tell. Not only can the National Conditions Perspective better explain variation in past election outcomes since 1948, it also provides clearer guidance for predicting the next election.

Integrating the Perspectives

One could treat these ways of understanding elections as mutually exclusive. But they can be integrated. The authors of the 1960 book The American Voter separate short term and long term forces affecting the vote. They observe that major long term shifts in party loyalty toward the Democrats during the Great Depression meant that, in the 1950s, the Democrats had a much larger portion of the electorate predisposed to support them. This gap in party loyalty constituted the long term forces affecting 1950s elections. But elections are also affected by short term forces, such as the popularity of the candidates, the state of the economy, and foreign wars like Korea.

The American Voter is often cited for its claim that party identification is stable and highly influential. But in the two presidential elections it examined, it explains that short term forces overcame the Democrats' long term party identification advantage to elect Eisenhower. (These were "deviating" elections, in that they deviated from the long term partisan balance.) The Democrats' advantage in party brand loyalty just meant that that Republicans needed a larger advantage in short term forces to overcome it and win. If short term conditions favored neither party or gave Republicans only a slight advantage, Democrats would prevail.

Another possible way to integrate the National Conditions and Strength of Party Coalitions perspectives comes in James Stimson, Michael MacKuen and Robert Erikson's 2002 book, The Macro Polity. In their main model of the two-party presidential vote, they show that it is strongly correlated with "macropartisanship" in October of the election year (see Table 7.9 on page 273). This is a very useful finding. But I don't think it tells us too much about potential long term party attachment advantages lasting for several decades. Macropartisanship is measured using Gallup's question, "In politics, as of today, do you consider yourself a Republican, Democrat or Independent?" This question moves around a lot more over time than the standard party identification question used by the American National Election Studies, the General Social Survey, and other commercial polls, which asks: "Generally speaking, do you think of yourself as a Republican, a Democrat, an Independent, or what?" The Gallup macropartisanship question combines some long-term party identification with a lot of one's current feelings about which party one wants to vote for right now.

In contrast, the National Election Studies question is stable over time, even when a particular election circumstance leads a voter to temporarily defect to the other party. It is these long term party attachments that people must mean when they talk about a partisan advantage that lasts over a series of elections. But I don't know of a modern model that uses the national balance of long term party attachments as a way to improve predictions of specific presidential election outcomes.

The Bottom Line

My bottom line is that, if you want to know which party is going to win the 2016 presidential election, I would use a standard national conditions-based model, with economic performance, number of terms the incumbent party has held the presidency and war casualties. This is the best way to predict presidential elections that I know of. The balance of long term party loyalty matters, but it may only matter when one party has a larger advantage than either party has held since World War II or will hold in the near future. I wouldn't count on it to guarantee the Democrats victory in 2016.