Where do our political beliefs come from?
And a better answer than the usual airport book answer.
This is the third post in a series of posts about a phenomenon in American politics called (by me) the "polarization paradox". You can read about what exactly that is, some examples of it in recent politics, and its six root causes here.
The main reason we have such dysfunctional levels of polarization in the United States and other democracies is that voters have radically different beliefs about the world.
Before the “well actually” thought bubble escapes your body, I should say that of course there are other reasons for our dysfunction. But those reasons are all intermediate. In a magical Barbie-land where there is no variation in what voters believe about politics and policy, there would be no need for parties (or anything partisan, like partisan media, for that matter), and there would be little incentive for democratically elected officials to deviate from consensus policy-making to get re-elected. Basically, there would be no incentive for conflict. In this extreme case, there wouldn’t even be a need for democracy, because the whole point of democracy is to give factions of people with differing beliefs a chance to see their views represented in government.
So, what can political science say about where these Very Important beliefs come from?
The normie answer: everything matters, but we can't agree how much, why, or when
A quick aside: when I say political beliefs, I generally mean 'normative beliefs' - beliefs about how things ought to be - on specific issues of public policy. You might use this definition interchangeably with ideology, which could refer to a specific bundle of beliefs across many issues. The distinction is not super relevant here since factors that affect my beliefs may end up shifting how I ideologically present to others - let's not dig too much into semantics here.
Here's a graphic that sums up some of the best political science that demonstrates how different features of the physical world, from the very micro to the very macro, shape our political psychology:
There's a lot that we do know about political belief formation: from the role that socialization plays in ideological formation to the large effects of peer pressure and conformity on voters' party affiliation. The fact that we have empirical evidence for these findings - each of which I could write an entire newsletter about - is an amazing example of the progress made in social science over the last fifty years.
But our knowledge on political belief formation is nowhere near complete.
Firstly, social scientists fiercely disagree about exactly how "big" these boxes should be: what is the most important factor shaping political beliefs in America? Sadly we're not really doing a lot of collective research sizing these boxes relative to each other. A fundamental shortcoming of quantitative social research, in my opinion, is its piecemeal approach to building knowledge without contextualizing it (other than a lazy nod to "previous literature" in a boring lit review no one ever reads). This is often by necessity: it's hard enough to measure one damn thing! And it's hard enough to understand one damn thing too; a professor of mine, Jennifer Hochschild, once said that people can really only juggle three variables in their head when they're thinking about any social phenomenon.
Second, social scientists themselves fiercely disagree about the findings! The typology of effects I've listed under cognition, for example, is somewhat internally inconsistent. Do voters have a coherent two-dimensional ideological grid of policy preferences or are they mostly chaos monkeys lacking any stability in their beliefs (as was long believed in behavioral political science1)? The answer is: (1) it depends on the context (2) different voters may have different influences altogether. Again, more research contextualizing, or rather reconciling, these cognitive findings is much needed. See, for example, a recent seminal paper estimating that some 7 in 10 voters have views that can be well described on the standard liberal-conservative ideological spectrum. There's at least a 30% outside of this default ideological spectrum, a group we crucially need to understand better.
Finally, not all of these studies are equally good. Some are from genuine experiments that are able to identify causality rather than mere correlation, but usually in artificial or fleeting contexts; others are from "natural experiments" that still require a layer of delicate assumptions about the quasi-randomness under which the experiment was "found". Some factors — the level of industrialization experienced by people in a particular place — are so utterly confounded with others that it's nearly impossible to isolate their effects on political beliefs. Or when we can isolate them, the implied effects are miniscule even on absolute terms: one reason for the relative sparsity of recent papers in 'genopolitics' is that the presence or absence of genetic markers usually explain less than 1% variation in individuals' stated or implied political attitudes.
To further complicate all this, here are the interactions between every one of these factors:
These, too, are well-studied but suffer from similar problems of decontextualization, non-resolution, and variable credibility. Moreover, since it is so difficult to isolate each of these factors (in a real world setting, anyways), we are often not identifying their direct effects, but rather their reinforcement effects. It’s easy to claim that social media consumption reinforces existing processes of political polarization, but much harder to show that they alone alter political attitudes — as a series of recent papers in Science confirms.
But you came to this piece not to walk away juggling fifteen different answers for the titular question. So what do we think is the most important influence on our political psychology?
A lukewarm take: social environment (and everything else it includes) is the largest influence on political ideology.
The enduring finding in American political behavior that just won't go away is that the party you are socialized into explains your politics most of the time. So much so, that even the most sophisticated campaign models, which incorporate tons of other information about you, can't actually predict your politics that much better than simply knowing your party registration. A recent paper by political scientists Jan Zilinsky and Seo-Young Silvia Kim shows exactly this.2
The keyword here, though, is “social” — partisanship is more about social group membership than about psychological state. Cutting-edge new work from political scientist Jake Brown finds, across a massive dataset of 41 million voters, that when voters move to a new neighborhood, they actively conform to the partisanship of their neighbors and vote accordingly. It’s hard to imagine this would be the case if partisan identity was psychologically hard-wired.
But the problem is that social environment includes a whole bunch of stuff - pretty much everything else on our laundry list of factors - and we haven't really pinpointed the exact mechanism that causes the shift in beliefs. Is it talking to members of the other party? Is it switching to new media sources? Is it ... a sudden repulsion to the smell of Democrats? Or when people move to South Texas, do they begin to behave like Republicans even if they don’t think like Republicans? There is much left unanswered in this account.
The biggest problem of all is that “the social environment is important” is just a very passé 2010’s airport bookstore take. Tell me something I don’t know, please!
A hotter take: institutions deeply shape our political beliefs along every one of these pathways, much more than psychology, but it takes a long time to know.
While this take also fails to produce a clean set of belief-forming mechanisms and is ‘hotter’ than it is ‘proven true,’ it's an important re-orientation away from the over-psychologizing, One Weird Trick fashion cycle that’s dominated social science in the last few decades. Put plainly, the behavioral revolution went too far, and social science would benefit from a New New Institutionalism in its approach to studying human behavior and psychology.
The definition of ‘institution’ is extremely wide. Institutions range from markets to companies to rotary clubs to churches. One of the most thought-provoking accounts of how institutions shape political psychology, in fact, involves the church. In “The WEIRDest People in the World,” evolutionary biologist Joseph Henrich shows how the gradual increase in interpersonal trust and prosocial attitudes between groups in Western Europe can be attributed to one key policy of the Early Medieval Catholic Church: the banning of cousin marriage. This decreased “kinship intensity” or the clannishness of social groups in regions of Europe where the Medieval Church had greater influence than others. And, according to a variety of measures and the best observational causal inference we can muster, it seems to explain variation in modern Europeans’ and Americans’ group attitudes (a fundamental pillar of all political beliefs).
To take another institution, slavery was the cornerstone of America’s political and economic development in many ways, but also had long-lasting psychological effects on American politics. In “Deep Roots,” political scientists Matt Blackwell, Avi Acharya, and Maya Sen cleverly harness the quasi random-ness in cotton suitability across the U.S. South’s geography to demonstrate that “natural” shocks to the number of enslaved people in Southern counties in 1860 predict the racial resentment of whites living in those same counties 150 years later. In essence, anti-black attitudes were transmitted across generations within the same communities and are directly causally linked to slave dependence.
Finally, in one of my favorite entirely non-quantitative studies in all of political science, “Power and Powerlessness,” sociologist John Gaventa explains how a coal mining company in a small Appalachian town in the 1800s employed many different types of ‘power’ beyond political or economic, including rhetorical and cultural, in order to ruthlessly extract land resources from the region without resistance from the residents of that community. While there’s no causal relationship established to modern political attitudes in Appalachia, Gaventa’s book is a multi-faceted description of how the institutions of geography, markets, inequality, and culture interact with power to ‘crush’ rather than ‘form’ political beliefs.
As I've said from the onset in this newsletter, the default answer in social science is that everything depends on context. Institutions, whether cultural or economic, are about the most powerful things that shape this context in daily life. In the end, they probably shape our political psychology way more than our Twitter feeds or the absence of a couple of genes.
Zilinsky and Kim write:
Without information about respondents’ partisanship, even sophisticated random forests models typically only achieve out-of-sample accuracy up to 65%, and this performance metric has not increased over the last seventy years. As the predictive power of party ID grows stronger, it dominates the signal from other covariates, diminishing their marginal predictive power.
Great explanation! I’ve been interested in political beliefs for a while and have mostly learned about the topic in a piecemeal way, so I appreciate the high level overview.