In RealClearPolitics, a provocative thesis, from Vikram Maheshri (U. Houston) and Cliff Winston^* (Brookings):

The United States experienced a Great Depression during the 1930s causing one-quarter of its workforce to be unemployed. Although not formally recognized, a growing body of survey evidence indicates that the US has been experiencing a Second Great Depression for decades, worsened by events such as 9/11, the Great Recession, the growth of social media, and the COVID pandemic. However, the causes and consequences of this depression have been largely psychological, not economic, with a notable fraction of the population becoming socially disengaged and depressed.

…

…Trump has made few efforts to address the Second Great Depression. Instead, he has exploited its malaise to win two presidential elections by convincing an important share of the public to vote for him because he gives voice to their fears and anxieties and encourages them to join a movement of like-minded people. Indeed, a closer examination of the links provided above show that the Second Great Depression disproportionately afflicts men and younger and rural Americans—that is, people who form the bedrock of Trump’s political support.

…

The entire article is here, and does not provide statistical data as this is a difficult thesis to quantitatively and rigorously assess. Clearly, self-reported depression is up, as noted in the article.

Source: Gallup, May 2023.

(*full disclosure: I was Dr. Winston’s RA 40 years ago).

Was Donald Trump’s candidacy more attractive to those who suffered from mental depression? That’s much more difficult to assess, and would require micro data to evaluate.

I can evaluate at the state level the following correlation between the prevalence of depression (2020) and voting for Trump in the last election.

Figure 1: Trump vote share (vertical axis) and prevalence of depression (horizontal axis), both in %. LOESS (locally weighted regression fit, 60% window) (red line). Source: NBC, HHS.

OLS regression results:

Robust regression results:

The point estimates indicate that each 1 percentage point of increasing prevalence of depression is associated with a between 1.3 to 1.7 percentage point increase in Trump voting share.

Of course, correlation is not causation. And I expect that these are “fragile” regression results in the Leamer sense. However, I was surprised at how much variance was explained by a simple bivariate regression.

One could try to account for the endogeneity of depression prevalence by using 2SLS, but I’ll leave that to others to try. Better yet would be to correlate voting behavior with diagnosis with depression at an individual level. For now, this is an interesting correlation that (I think) buttresses the Maheshri-Winston thesis.