Presidential Campaign

How Obama will lift Clinton to victory (barely)


At the American Political Science Association (APSA) meetings last week in Philadelphia, Charles Tien and I released our presidential election forecast, as did other political science forecasters. We called a narrow win for Democratic nominee Hillary Clinton, with 51 percent of the popular vote share (the Democratic and Republican total).

{mosads}This popular vote win translates, by our calculations, into a narrow Electoral College win of 274 votes. In sum, we foresee a Clinton victory by the thinnest of margins. This essay aims to show the origins of these forecasts, and their meaning.

Since 1982, we have offered presidential election forecasts, regularly publishing them before Election Day. These forecasts come from structural statistical models, rather than from vote intention polls or prediction markets, the two leading rival approaches to scientific forecasting.

Our forecasts rest on a scholarly understanding of how American voters, in practice, choose a president. In other words, it is theory-driven, rather than data-driven. We call it the Political Economy model.

Drawing on strong theory and empirical evidence from various political science investigations, we argue that what ultimately counts are the big political and economic issues, and how the party in the White House handles them. We measure these two factors in straightforward ways. To assess the incumbent party’s handling of political issues, we employ the Gallup question on presidential job approval. To assess economic issues, we utilize national economic growth.

In words, the explanation reads as follows: Popular Vote = Presidential Approval + Economic Growth, with both factors measured in the summer before the election, in order to allow forecasting at a meaningful distance. A lead time of about three months makes the forecasts non-trivial and, at the same time, allows a level of prediction accuracy comparable to vote intention polls much closer to Election Day itself. To arrive at the precise effects of these political and economic issues, we calculate the following multiple regression equation, on the 17 presidential elections from 1948 to 2012.

Vote = 37.50 + 0.26 popularity + 1.17 growth

To forecast the 2016 presidential election, we just insert the final scores on presidential popularity and growth, as of Aug. 26, 2016.

July popularity = 51 percent and gross national product (GNP) growth (2016, first two quarters, non-annualized) = 0.20, so giving the following prediction:

Vote = 37.50 + .26 (51) + 1.17 (.20)
= 51 percent of the popular two-party vote for the Democratic candidate.

How many electoral votes would that popular vote margin yield? Across these 17 elections, this popular vote share forecasts the Electoral College vote, along an almost perfect prediction line (Electoral Vote = -198 + 4.88 Popular Vote). Applying this formula yields a prediction of 50.9 percent = 274 electoral votes, giving a bare win for Clinton.

Having laid out our forecasting procedure, we turn to how accurate these predictions are likely to be. Looking at the track record of our Popular Vote model, we see that it correctly predicts 14 of the 17 elections. In fact, its last miss was an election 40 years ago, the 1976 Carter-Ford contest. (It also missed 1960 Kennedy-Nixon and 1968 Nixon-Humphrey. For these three missed elections, the average error is just 2.7 percentage points).

Thus, the model has correctly predicted the party winner 82 percent of the time. However, that number alone may make one overconfident about a Clinton victory, since the acceptable margin of error for her prediction is only 1 percent. If the prediction of 51 percent is off by more than -1.0, she would be expected to lose, with under 50 percent of the popular vote, meaning she would also lose the Electoral College vote.

In sum, what do these numbers tell us? First, they show that a Clinton victory is more likely than a Trump victory. That is good news for Clinton supporters. However, the bad news is that the margin of victory, as predicted, is very small, which means in practice that it could tip to her Republican challenger.

Thus, while this Political Economy model puts Clinton in the lead, it also shows her to be at risk. Why is that? Substantively, the problem does not lie with weak support for President Obama. His 51 percent approval translates into considerable aid for the current Democratic nominee.

Instead, the culprit appears to be the weak economy. Growth, as measured over the first six months of the election year, has been feeble, at 1 percent or even lower. When the economy does poorly, the party in the White House suffers at the ballot box.

There has been disagreement about whether the economy has been performing well. However, considerable research has shown that, in terms of macroeconomic performance, presidential voters look closely at economic growth over the last year, especially at the last two quarters, as we have done. By that important measure of economic growth, the economy has faltered.

Further, the American people, on the whole, see this weakness. The Economic Confidence Index of Gallup, for the week ending Aug. 28, found that 58 percent of people say the economy is “worse,” with only 38 percent saying it’s better. These are not good numbers, and they are worse than from the first of year.

If Hillary Clinton takes the White House, it will be in spite of the economy, as our model suggests.

Lewis-Beck is the F. Wendell Miller Distinguished Professor in the Department of Political Science at the University of Iowa. Contact him at

The views expressed by contributors are their own and not the views of The Hill.


Tags Hillary Clinton
See all Hill.TV See all Video

Most Popular

Load more


See all Video