[It] seems to have been reserved to the people of this country, by their conduct and example, to decide the important question, whether societies of men are really capable or not of establishing good government from reflection and choice, or whether they are forever destined to depend for their political constitutions on accident and force.

Alexander Hamilton, Federalist 1

The answer may be “accident and force.” This graph, derived from Christopher Achen, shows an almost perfect correlation between presidential election results and economic growth during two quarters before the election, adjusted for how many years the incumbent party has held office. It implies that who wins the White House in November will depend almost entirely on what the COVID-19 virus does to GDP during this quarter and next. If the US economy manages 1.5% positive growth despite COVID-19, Trump should win. If it declines, he will probably be done, regardless of the Democratic nominee.

One should always be careful about correlation graphs with relatively small numbers of data points and carefully contrived axes. You can fish for combinations of variables that generate uncannily neat pictures. However, this graph shows the result that you would predict if you hold a theory of electoral politics that goes back at least to Joseph Schumpeter in 1942:

• People have better things to do than follow politics closely, let alone form and test careful hypotheses about the impact of political positions on outcomes (holding other factors constant). Voters are not going to be rigorous social scientists.
• Instead, voters will assess the most prominent political leaders of the moment by evaluating their own circumstances. They won’t only think about economics, and certainly not only about the nation’s GDP. However, GDP growth is a decent proxy for how well a whole population is faring in their everyday lives compared to last year.
• People will judge politicians using other heuristics, such as partisanship, demographics, and ideological labels. These factors also determine what we hear or read about the world beyond our doors. (Watching the Fox News homepage during the COVID-19 crisis, as I have done, is an object lesson in ideological framing.) However, in a system like ours, two closely equal voting blocs with their own media networks will emerge as an equilibrium. Who actually wins any given election depends on the main variable that changes from month to month: GDP growth.

This is bad news for any theory of democracy that envisions millions of people deliberating about ideas and making choices. With Schumpeter, we might at least hope that a national election functions as a test of performance, rotating failures out of office. We would expect Trump to lose in November because things will go badly between now and then, and that will serve as a vivid test of his fitness for office, his allies in the media and Congress, and his ideology (nationalism), which has guided his response to the epidemic. The voters would demonstrate collective rationality by throwing him out.

The problem with that theory is the massive element of randomness. The signal can easily be lost in noise. If COVID-19 hadn’t hit now, Trump wouldn’t be seriously tested. If Hillary Clinton or Bernie Sanders were president now, the government’s public health policy would be better than Trump’s, but we would still face a global pandemic and probable recession. Then the signal would convey that neoliberalism or democratic socialism–not nationalism–was the failed ideology.

To make matters worse, politicians are systematically rewarded for the wrong things. Andrew Healy and Neil Malhorta show that spending (or not spending) money on prevention has no effect on electoral outcomes. However, relief spending is a big boost to an incumbent. This means that Trump may benefit from COVID-19 if things play out fortuitously for him. If we are in recovery by November and Trump is handing out stimulus relief, the crisis may carry him to reelection. In that case, not only would the public receive a false signal about his competence and ideology, but his policy of doing nothing to prevent a crisis would have been rewarded.

Should we therefore give up on democracy? I definitely think not, for these reasons (and I leave aside the tired argument that it is better than the alternatives):

First, all the models discussed above are based on political leaders who are plausibly competent and whose stances and worldviews put them in sync with close to 50% of voters. Trump has always risked not quite meeting those criteria. Up to now, his approval ratings have been well below what you would expect for an incumbent presiding over historically low unemployment. If he were to lose because his statements and policy choices have alienated a significant minority of voters who would have voted for him otherwise, then we’ll learn that national elections at least serve to weed out true losers. Four years too late, but better now than never.

Second, it matters which groups coalesce into the two large blocs of active voters. Those groups are better served when their side wins. Therefore, it matters which citizens we engage and motivate to vote.

Third, a national election, although important, is an outlier among all forms of politics. It is episodic and short-term. Millions of people participate, each having a microscopic impact. It is entirely mediated, since only a tiny proportion of us actually know the candidates. Given our electoral system, it is filtered through a party duopoly.

Local politics can be much better. So can national politics, over a longer time-span. Consider the improvement in mainstream attitudes toward sexual minorities in the US, which affects the stances of presidential candidates as well as many other aspects of our society. That, too, is politics: the result of advocacy, organizing, discussion, and learning. Our expectations for self-governance should be much higher than our expectations for presidential elections, where we must hope that Alexander Hamilton’s “accident” is benign.