Democrats are bullish about the election. Prominent Republicans like Colin Powell have endorsed Joe Biden. Recent national polls have established a Biden lead by a margin of between 8 and 13 percentage points over Trump. The Economist unveiled a sophisticated forecasting model that as of June 12 gave Biden an 86 percent probability of winning the electoral vote and becoming president, as well as a 97 percent probability of winning the irrelevant national popular vote.
But Democrats should not be deluded into believing these polls guarantee anything, or thinking that they can sit on a lead. Polls today mean little as signals of how the electorate will feel come election day. And as the history of elections and the configuration of the electoral college both dictate, the 2020 contest is likely to depend on a specific number: the average unemployment rate at the close of the third quarter — the months of July, August and September — in the 10 states where the presidential vote margin was the closest in 2016.
To see this we built an election model to predict the probability of winning the 270 electors that produces a president using a forecast of the unemployment rate in the third quarter. Based on that, if the swing state unemployment rate in the third quarter is 7 percent, Trump has about a two-thirds chance of winning re-election. Biden has about a two-thirds chance of winning if that rate is 13 percent. If the rate is 10 percent, the election is a toss-up. In that event, a higher turn-out helps Biden, but the unemployment rate remains a more important factor.
The Federal Reserve Board Chairman Jerome Powell is predicting unemployment will be 9.3 percent at the end of the year. If he is right, the election will likely be too close to call.
Our model ties four variables together: economic conditions, turn-out, a state’s percentage of white voters, and the partisan lean of a state’s voters. With a state-by-state model we forecast a range of possible democratic vote shares in each of the 50 states and Washington, D.C. that have an electoral vote. In 40 states, the result is virtually preordained by partisan lean and white-voter percentages, two largely unchanged factors since the last election. Contestable by both parties are the 10 states where the plurality was the closest in 2016: Michigan, New Hampshire, Wisconsin, Pennsylvania, Florida, Minnesota, Nevada, Maine, North Carolina, Arizona.
To make an assumption about economic conditions we compared three different third quarter unemployment rates for 2020 with the rates in the same quarter of 2019. In that way our model reflects whether voters feel the economy is much worse, somewhat worse, or only modestly worse than a year before.
We took as a starting point the average unemployment rate across the ten key swing states in May of 14%, and then forecasted three different scenarios as averages for the 10 key swing states based on how we think unemployment will evolve in each state: 13 percent, 10 percent, and 7 percent. To unpack this, here is a grid showing what the three scenarios represent in each state.
To determine the impact of turnout, we used three scenarios: the low turnout of 1996 (52 percent percent), high turnout of 2008 (62 percent percent), and the average across elections from 1980 to the present (57 percent percent). The risk of contagion may discourage turnout at polling places while widespread vote by mail policies could lead to increased participation, making the wide range used here appropriate given the level of uncertainty in how things will settle.
In the following grid we show the results of our model. The percentages are the likelihood of Joe Biden winning at least 270 electoral college votes, and therefore the presidency.
As is shown in the upper left of the grid, if unemployment is below 7 percent in the average swing state, and voter turnout falls below the historical average, then Biden has only a one-third chance of winning. Even if turn-out is high by historical standards, Trump has a 59 percent chance of being reelected.
However, if the average unemployment rate is 13 percent in the swing states, Biden has about a two-thirds chance of winning. Turn-out plays a relatively minor role in this scenario.
The election could go either way if swing state unemployment averages 10%. High turnout gives Biden a modestly higher likelihood of prevailing, but the unemployment rate is a more important factor.
The President has more capability to influence the unemployment rate than does an out-of-power rival. That may be why polls show Americans trust Trump more Biden to cut the unemployment rate. Our model suggests that Trump will pay little attention to the blue states while promoting economic recovery in the swing states.
Moreover, some swing states matter more than others. Holding the Republican National Convention in a state with 29 electoral votes (Florida) makes more sense than conducting it in a state with only 14 electoral votes (North Carolina). In this connection the Treasury Secretary may use his substantial discretion to allocate hundreds of billions of dollars to swing states. The model also explains why the president, who claimed that he would have won the national popular vote in 2016 if all votes were fairly counted, now would discourage voting methods that increase turnout.
In response, the Democratic Congress needs to set forth a blueprint for getting more than 20 million people back to work. In the early months of responding to the pandemic, it went along with the Republicans in spending more than three trillion dollars, but the money was dedicated to disaster relief as opposed to new job creation.
In presenting its message to the voters, Democrats must note, for example, that four out of five Americans support government action to create clean energy jobs. The Biden campaign should describe the stimulus plan he would pass at the beginning of 2021, in an uncanny but critical repetition of the Obama-Biden stimulus of 2009. Under no circumstances should Democrats compromise on important points to get a half-baked stimulus through a Senate led by Mitch McConnell. For Democrats, the election must turn on who has the best plan for a full economic recovery in 2021 and beyond.
In their current tactics, the Republicans demonstrate that if they are not using a model like ours, they clearly have intuited its results in their strategic decision making. The Democrats should take only cold and temporary comfort from poll results that will be as ephemeral as the currently high unemployment rate is short-lived.
Vinod Bakthavachalam (@vinod__b) is a data scientist who writes about economics, politics, and policy. He has previously written for the Harvard Business Review, the World Economic Forum, and the New York Times.