When people are unhappy, they look to elect people who promise to change the system.
How do people decide whom they vote for? A traditional view holds that people vote according to how the economy is doing — Bill Clinton famously said, “It’s the economy, stupid.” But, of course, the economy isn’t the only consideration as many issues come into play. Issues of identity — such as religion and race to name just two — are clearly important as well. So, in 2016, why did people vote in Trump? Was it economic uncertainty, or issues around status and identity?
Our team of researchers from Stanford, MIT, Oxford, and the University of Pennsylvania was interested in the strongest predictors of voting for Trump in 2016. Our newly published surprising finding? People’s unhappiness was a very strong predictor. Unhappiness predicted the Trump vote better than race, income levels, or unemployment, how many immigrants had moved into the county, or how old or religious the citizens were. And it did so among both whites and non-whites alike. Unhappiness also predicted the Trump election better than other subjective variables, like how people thought the economy was going or would be going in the future.
In short, the election did not seem to be about “the economy, stupid” — rather, it seemed to be about “subjective well-being” — the term psychologists use to describe the multiple dimensions of happiness. And that same effect might play out in the 2020 elections.
To measure the relative happiness and unhappiness in U.S. counties, we used data from 2 million phone surveys of U.S. households in the years before the 2016 election. Based on the questions in the phone surveys, we determined how satisfied people were, how much positive and negative emotion they experienced, and how much “purpose” they felt. To assess the Trump election, in particular “the Trump vote swing,” we also focused on how much more counties voted for Trump compared with other Republicans in previous elections.
The data suggests all parts of well-being, including emotions, purpose, and satisfaction, were equally important and did a better job of predicting the election than economic concerns.
This dataset did not determine why people were happy or unhappy. We know there are probably many causes, which include economic considerations and larger societal issues, from racial injustice to lower quality of life, compared with prior generations. For the first time, for example, the life expectancy of non-Hispanic white Americans is decreasing — some young people today may expect to lead a poorer and shorter life than their parents, in part because of the epidemics of opioid overdoses, suicides, and other “deaths of despair.” Previous work had found that counties with more deaths of despair disproportionately voted for Trump. The lines of evidence are converging between these studies and ours: There is more unhappiness in the population, and those unhappy people voted for Trump.
Our study suggests that Trump, in particular, was able to “convert” unhappiness into votes for him and “against the system,” which people may have blamed for their unhappiness and were seeking to change.
Trump used the populist advantage, in which a candidate insists on a clear distinction between ordinary people and the corrupt “elite” and says politics should represent “the will of the people” — which the populist candidate claims to represent.
The combination of unhappiness and a populist candidate can drive dramatic electoral change — people who had not voted in preceding elections voted in the Trump election, which in part explains why the prediction models got the election wrong.
This populist advantage also isn’t unique to Trump. In the primaries and the 2012 election, we found the relationship between unhappiness and voting against whoever is in power to be stronger for populists from both political parties. In the democratic primaries, we saw that unhappy counties were more likely to vote for Bernie Sanders than for Hillary Clinton.
Where does all this leave us? These findings suggest that knowing how happy or unhappy the population is matters and has real political repercussions. There has been a movement over the last two decades to measure the well-being of populations — how well our children live, the quality of our relationships, the amount of stress and joy we experience — and not just how well the economy is doing. Wrecking your car will increase the GDP (towing! repairs! a new car!) but will not increase your well-being. When parents are asked what they want for their children, they respond with the wish for a happy life, not an addition to the GDP.
Measuring the well-being is not only important to us in terms of our values and loved ones but it may also provide insights into politics and the outcome of elections.
In our work at the Institute for Human-Centered Artificial Intelligence (HAI) at Stanford, we are developing new ways to measure the well-being of large populations. We recently showed that one could collect billions of tweets to create county-level language samples and use AI to analyze this language to estimate life satisfaction and positive and negative emotions. We are laying the foundation for a new kind of 21st-century population measurement — not to sell more through better ads, but to make better policy and have a better sense where a country is heading.
This has become particularly urgent in the time of COVID — first reports indicate that up to a third of the U.S. population is currently struggling with clinical depression and anxiety in the aftermath of COVID. To anticipate mental health crises and track unhappiness in the population, we need to deploy modern measures of happiness. Artificial intelligence can help with measuring the well-being of populations.
It is hard to predict how COVID and the increasing unhappiness will affect the 2020 election — if our study is any guide, it would suggest that unhappy people vote against the powers that be. To the extent they think that Trump has become the status quo, they may vote against him. To the extent they think that he fights against an underlying larger status quo (e.g., the “deep state”), they may again vote for him.
Johannes Eichstaedt is a computational social scientist and is jointly appointed as the Ram and Vijay Shriram HAI Faculty Fellow and assistant professor in psychology.
Stanford HAI's mission is to advance AI research, education, policy and practice to improve the human condition. Learn more.