The US Presidential Election is decided, even though Trump and his dwindling number of allies as predicted will continue to contest it, even if for example Fox News has long called it for Biden.
Both the popular vote and some state-level results are different from what many expected, including some social scientists and journalists who rely on various forms of social science-style work (polling etc.).
As a professional social scientist, I wanted to write a short note with some preliminary thoughts on what the outcome might mean for social science.
I wrote a similar note after the 2016 US election, and – perhaps a sign of my advanced age and calcified ways of thinking – was tempted to just post it again. The observation that “we don’t know how most people feel about politics and how it ties in with other aspects of their lives and identities” and that “a desk is a dangerous place from which to watch the world” I’d be happy to just reiterate.
But I’ll add a few more observations, both around social science, and its interface with journalism, and will look forward to others’ thoughts. (Read responses to this thread with lots of smart people engaging – much there that I won’t try to cover here, on forms of citizenship, popular forms of political communication and political expression, varieties of journalism, the more hands-on role of platforms this time, and much more.)
All politics is to some extent identity politics
First, on social science, I don’t think we can understand the 2020 election outcome without putting identity in center place, especially if we are to understand why many white Americans have responded favorably to Trump’s explicit racial appeals. Identity was central in 2016, and this year again I think.
Much pioneering work on voting, especially coming out of sociology (the “Columbia School”) did this already a half-century ago, with e.g. Bernard Berelson et al’s observation that voting has its origins “in ethnic, sectional, class, and family traditions” and is a matter of sentiment and disposition rather than “reasoned preferences”, an approach further develop in the “Michigan model” that put party identification at the center.
For decades, more rationalist models of “median voters”, assumptions about “informed citizens” and the like were in vogue in some quarters, models that have greater affinity with many journalists’ tendency to assume (or pretend) that voting is some form of political arithmetic where people collect information (or are exposed to misinformation), analyze it, arrive at a conclusion, and then act on it. (American political scientists, and perhaps American journalists, may have been more enthralled by this view – at least in my experience, it never seemed as prominent in other countries.)
But in recent years, identity has been back front and center of much impressive social science on American politics, though less frequently centered in the particular subset of social science-style work journalism and pundits often rely most on – polling done by polling companies and various kinds of independent analysts.
Take for example Identity Crisis, where John Sides, Michael Tesler, and Lynn Vavreck came to the conclusion that Trump’s victory in 2016 was foreshadowed by changes in the Democratic and Republican coalitions that were driven by people’s racial and ethnic identities, and by the Trump campaign actively exacerbating these divisions by hammering away on race, immigration, and religion. Similarly, Liliana Mason in Uncivil Agreement argues that political partisanship is increasingly mapped onto race and other social divisions and Ashley Jardina in White Identity Politics draws attention to, well, what the title suggests.
None of these analyses would leave one surprised that 2020 was not a Democratic landslide or that attempts to combat demonstrably factually wrong misinformation may not have done much to influence the result, as important as they are in other respects. All of these analyses suggest that tens of millions of predominantly white Americans voted for Trump because they support him and what he stands for, not because they were misinformed, uninformed, or intellectually incapable of processing the latest long investigative piece, scorching op-ed, or pithy tweet denouncing the President for his failings.
Other work underlines that we should not accept the (for some) comfortable illusion that white identity politics is narrowly tied to Trump and thus perhaps could be written off as an outlier. Carol Anderson in White Rage documents a century and a half of how social progress for African Americans has been countered time and again by deliberate and organized white opposition, just as Doug McAdam and Karina Kloos in Deeply Divided draws a line from contemporary America back to the Civil Rights struggle and the white backlash against it.
As my friend Daniel Kreiss, who has introduced me to much research on these issues, has pointed out, we have decades of important work by social scientists and historians documenting how a central theme in American politics both now and historically is whites’ power to determine who a citizen is and should be. That, not Trump, is at the root of rage-tweets like “STOP THE COUNT!” He is arguably a symptom of white identity politics, and, however aggressively and effectively he has pursued it, not its cause.
Social science and political journalism
Some journalists I think share this line of thinking. As Matthew Yglesias wrote in 2015, “All politics is, on some level, identity politics.” But more broadly, I think there is a big gap between how a significant number of social scientists analyze and understand politics as based in large part on identity, and how many journalists analyze, understand, and present politics.
Sometimes that gap is fine, even necessary. While they have some shared commitments and overlap, journalists are not social scientists, and social scientists not journalists.
But if journalists turn to social scientists and people like pollsters who do social science-style work as sources they rely on to provide evidence and insight into how politics works, and in turn present this to their readers as the best obtainable version of the truth, or at least as strong, credible predictions… then they might be better off turning to social scientists who do it well, just as presumably journalists would prefer using competent doctors and public health experts as sources for stories rather than people who demonstrably and repeatedly get things wrong, or who offer self-confident opinions dressed up as evidence?
One problem, of course, is that social scientists can’t always offer analysis and insight with the precision, timeliness, and certainty that we might like, and that journalists may seek.
I get it, that’s inconvenient and irritating. We all abhor information vacuums, often feel uneasy with uncertainty, especially faced with important events, and tend to, when no credible sources are available, improvise, rely on heuristics, and work with other sources.
But sometimes we don’t know exactly what will happen, where, when, and to the last decimal. That’s true for social science too – uncertainty is a defining feature of research science, as is organized skepticism (we are essentially a tribe united in challenging one another by constantly asking “how do you know?”). We know things. But we don’t know everything instantaneously.
What does that have to do with the relation between social science and journalism? I think it speaks to how journalists do and do not rely on social scientists, among other sources, as they work to make sense of the world in real-time and meet people’s demand for information.
Take this two-part observation from Margaret Sullivan, one of the most insightful observers on US journalism and media: “Polling seems to be irrevocably broken, or at least our understanding of how seriously to take it is.” ‘Polling is irrevocably broken’ and ‘our [journalists’] understanding of it is broken’ are two very different claims! Both could be true. And there is no question that we, again, need to examine much more closely the performance of different polls and how they were used. (Is polling broken? It absolutely has taken some big knocks, again, but still – predicting how more than 140 million people will vote within a few percentage points still beats tea leaves? And that some polling is broken does not mean that all polling is broken.) Sullivan’s second observation, that part of the problem here is the gap between social scientists and journalists (“our understanding”) is also important. As my colleague Ben Toff has shown, the way in which US journalists and news media use polls is at least in part driven by a growing interest in and reliance on polling aggregator websites fueled by many journalists’ demand for precise predictions which, when combined with sometimes limited expertise within newsrooms to adjudicate between surveys and a more competitive space with more and more polls published, can lead to news media offering readers what is fundamentally at best spurious precision and at worst outright misleading.
We see this side of the issue, many journalists’ demand for something akin to certainty most clearly with polls, but it is not limited to polls. Various news media have published electoral predictions based on everything from betting markets, Wall Street trading, to sentiment analysis of social media posts (as well as of course just self-confident men paid to have opinions). Social scientists, on the other hand, deal in uncertainty. As Andrew Gelman wrote after 2016, “There’s a theory that academics [are] petrified of making a mistake, hence we are overcautious in our predictions; in contrast, the media … reward boldness and are forgiving of failure.” Forgiveness I suppose is a virtue, only perhaps not without complications if the price for failure is paid by the public. Is it sometimes worth trying some new sources, even if they sometimes come across as less dazzlingly self-confident, acknowledge more uncertainty, and offer less precise predictions?
What does journalism treat as social science, and who do journalists recognize as social scientists?
That leads to questions around how journalism and social science intersect.
Number one, can more journalists and editors accept that social science and social science-style work cannot always provide certainty and precision? I know this is inconvenient for reporters with stories to file, and for data visualization teams with maps to draw and needles to move, but sometimes the best obtainable version of the truth is that we don’t know, or that we only now this, or roughly that, but not these other things. At its best, journalism accepts this when it comes to e.g. medical research and natural science. Not so much social science. At a personal level, it is an almost daily experience for me as a source to have to disappoint journalists by responding “I don’t know” to questions. Many of them then later turn to someone else with the same question, often of course people who actually know – but also just sources who simply seem willing to give a straight, certain, seemingly precise answer (especially consultants, lobbyists, think tanks, campaigners and the like, sometimes drawing in part on social science-like work). As Chris Anderson (another friend) suggest in his Apostles of Certainty, sometimes journalism perhaps need to convey more provisionality and uncertainty rather than risk spurious precision?
Number two, what gets to count as social science, and who gets to count as social scientists? Social science is quantitative, but it is also qualitative, and social science can be focused on the present, but also historical. I am originally a qualitative researcher myself, though with colleagues in recent years increasingly doing quantitative work focused on the present, but I continue to be concerned that the fuller, more nuanced, more robust – fundamentally more credible, convincing, more complete, can I even say, more true? – understanding of politics require qualitative and historical perspectives. I don’t have data at hand to back this up, but my impression is these types of social science are far less featured in political journalism. If it was recognized, perhaps it would be more clearly recognized in daily coverage how central identity and racial divisions are to US politics? Perhaps. (A further complication here is that analysis focused on identity politics, racial appeals, and racisms necessarily relies on a vocabulary that is both analytical and moral – calling Trump a racist on the basis of his campaign and his statements is an analytical conclusion, but also a moral judgement, and the latter may strike some journalists as partisan, no matter how well founded the former.) Similarly, in terms of who gets to count as social scientists, it is hard not to notice how many of the pollsters cited and sources used are white men (like me). There are many other outstanding social scientists, and initiatives like Women Also Know Stuff and People of Color Also Know Stuff are aimed at addressing precisely this problem by providing journalists with easy access to a wider range of expert sources.
We do social science, not because it is easy, but because it is hard
Finally, this stuff is not easy, neither the journalism nor the social science.
I am deeply conscious of how little we know, and that we often disagree on how to interpret, or even approach things. As I wrote after the 2016 election: “Science is hard. We are in the dark, poking at the world with different sticks.”
But I am also tremendously proud of social science at its best, including the work of the many colleagues I’ve cited above. In my view, we collectively produce far more robust and reliable knowledge about social life than most other professions do, and thought we do it slowly, often in obscure and in nearly unintelligible ways, and with many, many gaps, we do so from a more disinterested vantage point. That’s why I think we have something very distinct and valuable to bring to the rough process of public discussion, something I think it is important journalists recognize, and that scholars, especially those in privileged positions, should prioritize making available to the public.
Finally, did social science do a terrible job this year? Some of it, sure. We may not sound like it, but we are human too. Overall, I’m not sure. The unweighted average predicted Electoral College vote for Trump across seven forecasts published in advance of the election in PS: Political Science and Politics was 237. The final result, if Trump as projected wins North Carolina? 232.