|Dear Readers: This is the first of a multi-part series on the political science forecasts of the 2016 races for the White House and Congress. We’ll be featuring forecasts from nine different individuals and/or groups this year, which James E. Campbell is assembling as part of a project for PS: Political Science and Politics that we are also featuring in the Crystal Ball. These models are based on factors such as the state of the economy, polling, whether an incumbent president is running for reelection, and other indicators. They can often be a better predictor of the eventual results than polls alone, and many are finalized months before the election. We are pleased to feature Campbell’s work and the work of the many top political scientists who have built these models, both in an attempt to predict the outcome of the election and, more importantly, to identify the factors that actually affect presidential races. Following Campbell’s introductory essay are the first two of the nine models we’ll be including in this series. As we feature new models, we will update Table 1 to provide a running tally of these forecasts.
— The Editors
Normally around this time in a presidential election cycle — the “interregnum,” as it has come to be known — we would be waiting for the dust from the nomination campaigns to settle before moving on to the conventions and to considerations of the general election race. As you may have noticed, however, this is not a normal year. It has been anything but. No dust, just tons of rubble from two wildly contentious nomination fights left to clear away as we turn to a general election bout likely to be the political equivalent of a mixed martial arts cage fight.
So who will emerge from the cage next November?
Although the 2016 election promises to have more than its normal share of twists and turns in the next few months, a number of election forecasting models based on past elections and current readings of this election’s fundamentals or pre-campaign conditions may provide some foresight about what we should anticipate. They do not take on the challenge of telling us what specifically will come out of Donald Trump’s mouth or from Hillary Clinton’s focus groups, but they should give us a reading about what the electorate is likely to do when the campaign rubble is carted away and the votes are counted.
The modern era of election forecasting in political science began in the 1980s and early 1990s. Since then, the early models have been tweaked and new models added. While every model is different, they have some characteristics in common. Each is based on a statistical analysis of historical data capturing how different pre-campaign contexts have predicted the vote divisions in past elections. Among the common categories of predictors are pre-campaign public opinion measures, indicators of the economy, incumbency and the number of terms the in-party has occupied the office, and the results of past elections. The models all have different track records, but their overall performance has been impressive. In the 2012 election, seven of the forecasts made more than eight weeks before the election and, in several cases, more than three months before the election, were within 1.5 percentage points of the actual two-party popular vote.
Since 1996, a group of these forecasts have been published prior to the election, first in American Politics Quarterly and since 2004 in PS: Political Science and Politics (a publication of the American Political Science Association). This year, in addition to again publishing the forecasts in PS, we are delighted to join with the Crystal Ball in releasing each forecast as it becomes available. There are nine forecasters or teams of forecasters lined up to predict the national two-party popular vote for the major party presidential candidates. In some cases, they will offer more than one forecast based on new information as the campaign proceeds, but most forecasters offer a single forecast fixed in place for the election. In addition, several of the forecasters will be reporting their predictions of the aggregate results (net seat wins or losses) of this year’s Senate and House elections. Accompanying each forecast will be a thumbnail description of the model upon which it is based along with a link to the PS website and the article (when it becomes available) fully describing the details of the model.
There is a great deal to be said about the forecasts, their strengths and weaknesses, but readers may want to keep three points in mind. First, election forecasts are made by the forecasting models, not the forecasters who devise any of the models. Although the specification reflects the forecasters’ judgment, the measurements of the predictor variables and the estimations of their relationship to vote percentages in past elections are objective and transparent. That is, anyone with some background in statistics should be able to gather the same data and come up with the same forecast. This is simpler for some models than for others, but possible for all of them.
Second, timing is as much a part of the specification of a forecast model as the selection of predictor variables. In most cases, a model using a set of predictor variables at one point in the campaign to obtain a vote forecast may be substantially more or less accurate as the same model used at a later or earlier point in the campaign. One model may be a more accurate model in July and another better in early September.
Finally, none of the models suggests that campaigns do not matter. There most definitely will be events and statements that will make a difference to the election. The forecasts based on different ideas and measurements of the fundamentals suggest that the context in which the campaign takes place affects how the events and statements made during the campaign are likely to be received, what is to be expected — but voters and candidates have been known to sometimes do the unexpected.
With that, here are descriptions of the first two presidential election models and their predictions:
The Primary Model
Forecaster/Team: Helmut Norpoth of Stony Brook University
Description: The Primary Model relies on presidential primaries and an election cycle as predictors of the vote in the general election. For elections from 1912 to 1948, all primaries were included. Beginning in 1952, only the New Hampshire Primary has been used, as a rule; South Carolina has been added to gauge primary performance this year. For the record, the Primary Model, with slight modifications, has correctly predicted the winner of the popular vote in all five presidential elections since it was introduced in 1996. In recent elections the forecast has been issued as early as January of the election year, although this year’s came in early March.
Predictors: Primary Score of Democratic Candidate, Primary Score of Republican Candidate, Democratic vote in last election (2012), and the Democratic vote in the next-to-last election (2008).
The Forecast: Donald Trump 52.5%, Hillary Clinton 47.5%. Posted on March 7, 2016.
Likelihood that Trump will win: 87%
Leading Economic Indicators and the Polls
Forecaster/Team: Robert S. Erikson of Columbia University and Christopher Wlezien of the University of Texas at Austin
Description: On the eve of the election, the impending result of the presidential vote can be seen fairly clearly from trial-heat polls. Earlier in the election year, the polls offer much less information about what will happen on Election Day.1 The polls capture preferences to the moment and do not — because they cannot — anticipate how preferences will evolve in the future, as the campaign unfolds. Various things ultimately impact the final vote. The standing of the sitting president is important. The economy is too. Both can change as the election cycle evolves. To make matters worse, late-arriving economic shocks have a bigger impact on the electoral verdict than those that arrive earlier. This complicates accurately forecasting the vote well in advance.
Our solution to the problem of early forecasting has been to turn to The Conference Board’s index of leading economic indicators (LEI). In previous articles2 we have shown that the weighted growth in these indicators through the spring of the election year — quarter 13 of the election cycle — is a strong predictor of the vote. This is for two reasons: (1) It provides a summary of the state of the economy leading up to the election year; and (2) it gives advance indication of changes in the economy (and presidential approval) during the election year. We use the quarter 13 measure of LEI growth in conjunction with current trial-heat polls to predict the vote at different points during the election year.
The forecast can be updated as the election campaign unfolds using new poll data, keeping in mind that it is the only variable in our model that changes — the value of leading economic indicator growth remains the same. Our intent is to produce new forecasts at two points prior to the fall, general election campaign: (1) the week before the Republican Party convention and (2) the week before Labor Day. Of course, forecasts also can be provided at any other point between now and Election Day.
Predictors: Growth in leading economic indicators (cumulated for quarters 1-13 of the election cycle); incumbent party candidate’s current share of trial-heat poll averages.
The Forecast: Democratic candidate 52.0%, Republican candidate 48.0%. Forecast is as of June 13, 2016.
Uncertainty Estimate (percentage likelihood that the forecast has predicted the popular vote winner — that the estimate is above 50 percent): 75%
Table 1: Forecasts of the 2016 two-party presidential vote
Notes: Table will be updated as model results are released or updated with new information.
James E. Campbell is a UB Distinguished Professor of Political Science at the University at Buffalo, SUNY. His previous books include The American Campaign and The Presidential Pulse of Congressional Elections. His latest book, Polarized: Making Sense of a Divided America, is forthcoming from Princeton University Press in July.
1. See James E. Campbell, The American Campaign: U.S. Presidential Campaigns and the National Vote (College Station, TX: Texas A&M University Press, 2008); Robert S. Erikson and Christopher Wlezien, The Timeline of Presidential Elections: How Campaigns Do (and Do Not) Matter (Chicago: University of Chicago Press, 2012).
2. Christopher Wlezien and Robert S. Erikson, “Temporal Horizons and Presidential Election Forecasts,” American Politics Quarterly 24 (1996): 492–505; Christopher Wlezien and Robert S. Erikson, “The Timeline of Presidential Election Campaigns,” Journal of Politics 64 (2002): 969–93; Christopher Wlezien and Robert S. Erikson, “The Fundamentals, the Polls, and the Presidential Vote,” PS: Political Science and Politics 37 (2004): 747–51; Robert S. Erikson and Christopher Wlezien, “Leading Economic Indicators, the Polls, and the Presidential Vote,” PS: Political Science and Politics 41 (2008): 703–07; Erikson, Robert S., and Christopher Wlezien, “The Objective and Subjective Economy and the Presidential Vote,” PS: Political Science and Politics 45 (2012): 620–624.