In the counterfactual world of no pandemic, at least four election prediction models based partially or entirely on economic factors suggest Trump could have been a formidable candidate, using economic data available early this year. The new reports add data not previously available that would be exceptionally favorable to an incumbent running for reelection. Without the pandemic overwhelming all other factors, the new data would have enabled Trump to appeal to a wide spectrum of voters and to strengthen his case in the campaign’s closing weeks.

Specifically, the Census Bureau’s latest release of “Income and Poverty in America” reports that median household income grew more in 2019 than in any previous year on record: 6.8%. Measures of inequality declined: Income grew much more among Black and Hispanic people than among white people, and median earnings increased three times faster for women than for men. Poverty fell to the lowest level since 1959, and among Black and Hispanic people it fell to the lowest levels ever.

Separately, the Fed’s triennial “Survey of Consumer Finances” shows further evidence of decreasing inequality. Using data collected between mid-2016 and early 2020, it found that incomes fell during that period among families at the top of the wealth distribution and rose among those in the middle and bottom. Incomes also declined among families with a college degree and rose among those with a high school diploma. Average incomes fell for white non-Hispanic families and rose slightly for Black non-Hispanic families.

Early in the year, before that data was known and before the pandemic arrived, Trump already looked strong in prediction models. In February, he was the clear favorite in the model used by the Oxford Economics consulting firm, based on the pioneering work of Yale professor Ray Fair. As the incumbent—his opponent was then unknown—he would get 55% of the vote, said the model, which has correctly called 16 of the past 18 elections. Just three months later, with some 40 million Americans suddenly unemployed, Trump had been transformed into an underdog of historic proportions. The model now recently said he’d lose the popular vote to Biden 65% to 35% for the worst defeat of an incumbent party since Republican Warren Harding demolished Democrat James Cox in 1920.

The Economist’s model, recalculated every day since early March, showed Trump and Biden equally matched during that month, trading the lead every few days. Their paths diverged decisively on April 1 and have never re-crossed. As of October 2, the model showed Biden winning 340 electoral votes to Trump’s 198.

A model developed at Chapman University in Orange, California, has predicted the popular-vote winner in 15 of the past 16 elections and shows Biden winning by 10 points. Changes in GDP and jobs in the election year are major factors, so it’s likely that if last year’s economic trends had been uninterrupted by the pandemic, Trump could have been tough to beat. FiveThirtyEight’s prediction model, which includes economic data, began on June 1 and has shown Biden strongly ahead every day. As of October 2, it predicted Biden winning 335 electoral votes to Trump’s 203.

None of this means that if Trump loses, he and his supporters can blame a stroke of terrible luck. Luck is what you make of it, and some argue that he could have made the pandemic a political advantage, calling on the nation to unite against the coronavirus and casting himself as a Republican FDR fighting the Great Depression. In addition, non-economic factors, notably the thousands of protests over racial inequities and police behavior, may play a larger than usual role this year.

But Trump’s opponents should not believe, as many did eight months ago, that he couldn’t be reelected. He could have been, and the latest data suggest his prospects might have been even better than they appeared pre-pandemic. Now, how will having COVID-19 affect his chances? The experience of the year so far says: Assume nothing.