Almost as soon as the 2016 election was over people noticed similarities between maps of presidential candidate support and where the opioid crisis is greatest. Now the correlation has been confirmed, and partially explained.

To test whether the relationship is more than a trick of the eye Dr James Goodwin of the University of Texas used the proportion of a county's population who were prescribed at least 90 days worth of opioids in 2015. Of course prescriptions are just part of the picture, with about half opioid deaths coming from illegal drugs. However, since heroin dealers seldom publish databases of where they are selling, and criminal enforcement rates vary, the Medicare data Goodwin drew on is probably the best indication we will get of the geographic distribution of the opioid crisis. It's already been established, there is a strong geographical connection between high opioid prescription and deaths from overdoses.

In JAMA Network Open, Goodwin confirms the correlation is real, at least for a sample mostly aged over 65. In the 693 counties with opioid prescriptions well above the national average, Trump scored almost 60 percent of the vote, compared to less than 39 percent in the 638 counties with rates significantly below average. A weaker, but still substantial, correlation was found between opioid use and voting for Republican congressional candidates. There was virtually no correlation, however, between presidential vote and insulin prescription, indicating the relationship is not about medical service.

The chance that this effect is random is less than one in a thousand. Some people may jump to the conclusion being on drugs inspired Trump-voting, but correlation does not equal causation. Instead, it is much more likely common factors influenced both opioid consumption and voting behavior.

Goodwin and co-authors went searching for these underlying causes. Although they were unable to find all of them, they did find factors that explain around two-thirds of the measured association.

High unemployment, rural status, and education were all predictive of both opioid use and Trump voting. As to the other third of the relationship, Goodwin speculated to NPR the destruction of community opioid over-prescription induces may have driven a desperation for change. "That can lead to a sense of despair. You want something different. You want radical change."

Just last week, a study found the opioid crisis is only one facet of what has been dubbed “deaths of despair” including suicide and alcohol-related deaths. Perhaps voting for Trump is another symptom of this lack of hope, a theory supported by previous studies showing swings from Romney to Trump were highest where life expectancy has fallen. However, where taking drugs to numb the pain is self-destructive, the consequences of a presidential vote has the capacity to cause far wider pain.