Forecasting the 2020 election based on the results of one or even a few past elections is a risky business for professors and political pundits. Recent history is littered with a series of busted political predictions. In 2004, for instance, after the Republicans’ third straight election given them control of the House, Senate, and Presidency, there was talk that Karl Rove had created an 1896-like realignment; the 2006 and 2008 elections quickly shattered that illusion. Those two major Democratic wins caused, among others, James Carville in 2009 to claim that the Democrats would govern for a generation, but the 2010, 2014, and 2016 elections showed that generations are shorter than they used to be. Similarly, in 1994 and 2010, the Republican congressional landslides did not lead, as some predicted, to Clinton and Obama being one-term Presidents. In short, one election doesn’t necessarily determine the results of the next. But does it tell us anything? Are there any takeaways from the 2018 midterms that will likely influence the 2020 Presidential elections? Our answer is yes: We believe that 2018 offers several specific insights for looking ahead to 2020.

The first lesson is that Donald Trump generates intense feelings, which in turn drives unusually high turnout. In the 2014 midterm elections, some 83 million Americans voted; by contrast, 118 million Americans cast a ballot in 2018, a 40 percent increase. Indeed, it has been more than 100 years since an off-year election produced a better turnout. Presidential elections, of course, attract an even greater share of voters, and the 2020 electorate will be younger and less white, both of which are factors that normally favor Democrats. If the extraordinary 2018 turnout is any indication, the 2020 Presidential election vote totals will be enormous, exceeding 2016.  This development would benefit Democrats.

But the second takeaway, a variation of the first, is even more important: The midterms were a referendum on Trump. Figures 1 and 2 show the generic congressional vote, broken down by presidential approval numbers. From the very beginning of YouGov’s election polling in September, 80 percent of those who approved of Trump planned on voting Republican; the opposite held true for those who disapproved. Those proportions rose over the course of the campaign, however, and by Election Day, 90 percent both approvers and non-approvers intended to vote for congressional candidates consistent with their views of the Trump presidency. Since more voters disapproved of Trump than approved of him, the Democrats had an advantage going into November. If the President and his allies are worried about these numbers, they’re right to be. Given a host of ongoing scandals and negative publicity—the Paul Manafort and Michael Cohen revelations, the National Enquirer scandal, the fallout from the Mueller investigation, and the problems with his charitable foundations—it seems unlikely that Trump will reach 50 percent approval before the election.

What else might carry over from 2018 to 2020? Possibly the relative size of four major electoral gaps in American politics: race, education, gender, and age. Table 1 shows these gaps for the 2014, 2016, and 2018 elections. The racial gap is generally the largest, usually topping 35 percent, but it actually decreased in 2018 as compared to previous years. This change is entirely attributable to the increase in the proportion of whites voting for Democrats. In 2018, 76 percent of Blacks, Latinos, and Asian Americans chose Democratic candidates, almost exactly in line with the proportion that did so in 2014 and 2016. By contrast, white voters temporarily reversed their trend toward the GOP, with 44 percent voting Democrat (as compared to 37 percent in 2016).

Unlike the racial gap, the gender gap did not shift significantly in 2018. Indeed, the difference in the preferences of men and women was actually slightly smaller than it was in 2016. The lack of a net change masks an important trend though—both men and women voted Democratic at significantly higher rates than they did in 2016 (5 percent higher for women and 6 percent higher for men). These statistics indicate that the scores of suburban women who turned against Trump were joined by a slightly larger number of men.

In contrast to race and gender, educational attainment was more divisive in 2018 than it was in the previous four years. College educated citizens voted 59 percent for Democrats, versus 49 percent of those without a degree. As such, the education gap was 10 percent, up from 8 percent in 2016. With Trump on the ballot in the next election, we expect this gap to persist in 2020.

Finally, the most surprising result was the dramatic increase in the age gap between those 18-29 years old and those age 65 and over. In 2016, the gap was only 10 percent, but in 2018, it grew to 19 points. While 55 percent of younger voters supported Democrats in 2016, that number rose to 67 in 2018. Now, we are mindful of the fact that, as people age, their voting habits change, so we do not conclude that Democrats have captured the under-30 generation. Nevertheless we are confident that nearly everyone in that demographic who voted Democratic in 2018 won’t shift and vote for Trump for President in 2020.

Since non-white voters consistently support Democrats at high rates, it is also useful to examine these same demographic splits among white voters. The percentage of white voters voting Democratic varies substantially from year to year, and understanding how they vote by gender, educational attainment, and age might provide some insight into this unstable group. We present these figures in Table 2.

Generally speaking, the gender, education, and age gaps among white voters resemble the magnitude of the gaps among the electorate as a whole. Nevertheless, some differences are worth noting. In particular, the education gap is substantially greater among white voters than it is among the broader electorate—among the former, the gap is 16 percent, as opposed to 10 percent for the latter. By contrast, the age gap is smaller among white voters as compared to the entire electorate; however, there was still a double-digit increase between 2016 and 2018 in the percentage of under-30 white voters who preferred a Democrat. Given that young voters are typically more numerous in presidential years, this trend stands to benefit Democrats.

In addition to extraordinary turnout and subtle demographic shifts, the 2018 election was notable for its reversal of the famous 1992 Clinton slogan, “It’s the Economy, Stupid.” In late October 2016, 94 percent of Republicans thought the economy was either stagnant or getting worse; by January 2018, 83 percent of Republicans thought that the economy was either holding steady or improving, with a full 50 percent thinking it was getting better. The opposite shift occurred in the economic perceptions of Democrats. In late October, 2016, 73 percent of Democrats thought the economy was holding steady or improving; on January 17, 2018, 75 percent thought the economy was stagnating or getting worse, with 42 percent saying the latter.

This trend is nothing new. As Figure 3 shows, the percentage of Democrats and Republicans thinking the economy is getting better is closely tied to whether the respondent’s party holds the White House. In 2018, though, partisan loyalty truly dominated economic perceptions when it came time to vote. Among Republicans who believed the economy was improving, 98 percent supported a Republican candidate for the House. However, 61 percent of the Republicans who believed the economy was getting worse still voted Republican. The trend was the same among Democrats, with 78 percent of those who believed the economy was improving still supporting a Democratic candidate for Congress. In short, the true state of the economy seems to have almost no influence outside the margins. Democrats who think the economy has been strong under Trump will still vote Democrat, while Republicans who believe the economy has been weak will still favor the GOP.

Beyond the economy—and with the midterm elections becoming largely a referendum on Trump’s presidency—the Republican Party is now clearly the Party of Trump. A simple plot of Trump’s approval ratings among Republicans confirms this diagnosis. Figure 4 shows us that as time has gone on ever more Republicans have come to “strongly” approve of Trump, rather than “somewhat” approving of the President. In March of 2017, less than half of Republicans “strongly” approved of the President, while nearly 40 percent “somewhat” approved. The “somewhat” approve number remained above 30 percent through April of 2018, while the “strongly” approve number hovered around 50 percent. Moreover, those who were not strong supporters of Trump were less likely to approve of his trade policy, share his views on immigration, and were much more supportive of gay rights, among other issue differences. By the time of the 2018 election, however, internal resistance to Trump had largely crumpled among Republicans. The fraction that “strongly” approved of his performance rose to nearly 70 percent, with nearly 90 percent of Republicans at least “somewhat” approving.

At the same time that Trump’s popularity was rising among Republicans, the proportion of the GOP who wanted him to run in 2020 increased as well. In September 2017, only 60 percent of Republicans wanted the President to seek reelection—by November 2018, that number was almost 80 percent. In sum, for now the 2020 nomination is Trump’s for the asking.

Thus far, our analysis has focused on the general electorate and the Republican Party. As we know, however, Trump’s victory in 2016 was the result of narrow wins in a few key states. Turning then to the Electoral College perspective, is there anything in the 2018 elections that provides insight into 2020?

Again, the answer is yes. In 2016, Trump won five states by a margin of four or fewer points. In Arizona and Florida, Trump won by 3.4 and 1.2 percent respectively, while in Michigan, Pennsylvania, and Wisconsin, he won by less than 1 percent. In addition to these Trump states, there were seven states where Clinton won by five points or less: Colorado, Maine, Minnesota, New Hampshire, Nevada, New Mexico and Virginia. Collectively, these states are the most likely to change hands in an election where Trump is a candidate. Tables 3 and 4 examine these states, comparing 2016 results to 2018 across races for House, Senate, and Governor. In the five highly competitive Trump states, there were 47 Republican House members compared to 29 Democrats after the 2016 election; the 2018 midterms narrowed that margin to 39-37. While the Senate remained the same (six Democrats, four Republicans), the Democrats also picked up two governorships in those five states. Turnout in these areas was also up by more than six million votes in 2018 as compared to 2014, an increase of 37 percent; moreover, the Democratic vote share was up by more than 7 percent. In short, Democrats did well in these states in 2018, and that bodes well for them in the 2020 election.

In the seven competitive states won by Clinton, the results were similar: The Democrats picked up five House seats, a Senate seat, and three governorships; overall turnout in these states was up; and the Democratic vote share was up compared to 2014. Taken together with the five Trump states, it seems that the most competitive swing states in 2016 moved decisively toward the Democrats in 2018, which, like our earlier results, favors them in 2020.

Indeed, all of the 2018 takeaways we have discussed thus far could be regarded as good news for Democrats. Turnout will be higher in 2020 than it was in 2016, and the electorate will be younger and less white than the strongly Democratic 2018 electorate. Trump has not increased his base over his first two years, and, compared to 2016, educated voters, young people, and white voters have moved toward the Democratic Party. Finally, the key Electoral College states moved back toward Democrats in 2018, bolstering the party’s hopes for 2020.

Right now 2020 is shaping up to be the Democrats’ to lose, all else equal. Of course, all else may not be equal. The party seems headed for a contentious, drawn-out primary and could potentially end up with an aggressively hyperliberal nominee. Such a choice would cut into the Democrats’ advantage among independents. During the course of YouGov’s election polling, it asked Democrats and independents what they thought about the Party’s ideology, and if they preferred Democratic congressional candidates who resembled Barack Obama or nominees who resembled Bernie Sanders. The responses are informative; only 27 percent of Democrats preferred candidates resembling Obama, while only 31 percent of Independents preferred candidates resembling Sanders. Likewise, a majority of independents (who were crucial to the Democratic victory in 2018) thought the Democratic Party was too liberal, as compared to only 12 percent of Democrats. These incongruous responses suggest that the presidential nominee Democrats are likely to choose might not have much cache among independents. Accordingly, it is possible Democrats choose an unviable general election candidate and lose again in 2020.

Of course, another alternative is that President Trump decides not to run in 2020. Perhaps reeling from the Mueller Report and from whatever additional political blows House Democrats manage to land via congressional investigations, Trump might declare that he has already achieved more than any other President in just one term, pack his bags, and ascend back up the escalator at Trump Tower. If, however, Trump runs and the Democrats don’t shoot themselves in some significant part of their anatomy, the GOP is in for a rough ride in 2020.