A Networked Theory of Trump Support

Visualization of a Typical Social Network (Image by Martin Grandjean)

Those studying the 2016 presidential election have been baffled by the rise of Donald Trump, looking for demographic factors that seem to best correlate with support for that candidate. And in general, they have been confounded: the phenomenon cannot be reduced to a simple correlation with blue-collar white males who are uneducated or recently disrupted by economic forces they cannot control. Indeed, this article suggests that the best correlation is geographic, identifying places where the next generation is faced with poorer prospects than the last. Geography does offer some hints, but the still-emerging field of network science offers perhaps our best shot at understanding this phenomenon.

First, let’s consider this idea: people’s political beliefs are shaped primarily by the people they are closest to.

Our popularly accepted narrative is that humans are rational actors who process inputs like lived experience, identity, education, and self-interest to render a coherent worldview that then determines their political orientation — left or right, fiscally conservative or liberal, religious or secular. The corollary to this theory of political identity is that people’s views can be changed by exposure to new facts or to argument.

However we have seen over the last two decades in particular that facts are mostly irrelevant. People seem to believe what they believe, and then cherry-pick the facts that support their beliefs.

So the evidence seems to suggest a different model for how political identity is established and then maintained: I will call this the networked model. In this model, a person believes in a candidate or platform because their social network encourages or at a minimum tolerates support for that platform or candidate.

People who support Donald Trump and the current Republican platform are people whose social network encourages or tolerates support for Donald Trump. People who support Donald Trump are people who effectively risk ostracism or financial deprivation by doing otherwise.

It is has been widely noted that this election has been marked by its sense of “otherness” — you will hear people on the left say that “they don’t know a single person that supports Trump.” Likewise, on the right, you hear people constantly calling for Hillary Clinton to be “locked up” and claiming that she is a criminal, and that “those people” are going to ruin America once and for all.

So it seems that the fundamental characteristic that defines this situation is isolation of two networks of people, who each demonize the other. What can lead to that kind of isolation?

Geography is a pretty good candidate. After the 2012 election, I did some analysis that discovered that population density was a key predictive metric in determining whether a county would vote Democratic or Republican. Above about 800 people per square mile, there was a 66% chance that a county would vote Democratic. Below 800 people per square mile, there was a 66% chance the county would vote Republican. So it is clear that geography presents some sort of divide.

If we return to a networked theory of political belief formation, it is not hard to imagine a theoretical network of United States citizens as follows:

• urban areas, which consist of multiple relatively densely networked communities, separated somewhat by forces like segregation, language, and income level

• rural areas, which consist of fewer, moderately densely networked communities

• suburban areas, which link to both urban networks and rural networks, and are moderately densely networked

Given that the left has purchase on the urban networks and that people in rural areas often feel that their interests are opposed to those in urban areas, it is reasonable to see how a networked theory might work — people in rural areas adopt platforms and candidates that run counter to those primarily favored by urban populations. In this case, that tendency landed rural dwellers in bed with Donald Trump, and seeded rural and some suburban networks with his supporters.

Once this occurred, and Trump effectively became the only choice on the right, the remainder is an inevitable product of social network structure. Peer pressure kicks in, and people’s support becomes a matter of who they must support based on who they are surrounded by.

This is then further reinforced by the rapid change in social networks made possible by social media tools. Political brawls online over Trump/Clinton can only go on so long before they end with the use of the Block button. This only reinforces the phenomenon: each network becomes more steeped in its own self-righteousness and determined to undermine the other, all the while severing social ties with those in their network that might demand a more nuanced view.

Some readers will counter with arguments about third parties. But they are niche versions of the same phenomenon. People who vocally support third parties are people whose social networks demand, encourage, or tolerate support of third party candidates. In practice these are peculiar regions of the social network — odd protuberances and outgrowths on either the left or the right.

Is anything actually new here, or has it always been thus? I don’t think this is a new phenomenon. Indeed the Republican Southern Strategy was an early recognition of some of these network properties, and successfully moved many voters from the Democratic Party.

But until now, we have attempted to maintain the illusion that our political process was a fair and open marketplace based on facts and ideas. What we see now, and what is different now, is the increasingly rapid drumbeat of evidence that demonstrates that model is simply incorrect. There are other factors at work, and they are clearly dominant over previous models.

The genius of Donald Trump is that he had an idiot savant’s sense that the networked model was true, and exploited that fact to win the nomination. My sense is that he developed that instinct through his work in television and with decades of exposure to the media. What he does not seem to have is the good sense or will to execute a successful campaign.

So what, if anything, can we do about this? I propose three next steps:

• first, create a mathematical version of “the network model” that could support the evidence we have observed: this would consist of a synthetic network map that would simulate the friend networks that seem to exist in cities, suburban areas, and rural areas. That model could then be seeded with the assumptions we posit above and we can see how it responds.

• second, we should aim to collect data that would verify whether the social network we observe in the wild comports with the simulated model. Social networking companies like Facebook could be especially useful in this step.

• third, we can begin to propose changes to social networks that may lead to more balanced views. If we want people to have moderate, balanced views, it is clear we cannot just try to sell them on facts and reason: we have to influence who their friends are.

This set of solutions may sound dystopian. But the alternatives are even worse.

References for further reading:

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.