for center-left, big data more useful for finding racists than understanding elections

Vox is center-left opinion with well-documented conflicts of interest masquerading as objective analysis, but every now and again they have something resembling reporting. Such was actually the case — despite the explosive headline “Persuasive proof that America is full of racist and selfish people” — when they interviewed Seth Stephens-Davidowitz, a former Google data scientist and occasional New York Times contributor. In addition to backing up the assertion that anonymity fosters both nastiness and honesty, that article established an apparently strong correlation between the frequency of certain internet searches and broader trends in our polarized voter population around election day. I found the interesting big data analysis lurking behind the volatile headline to be most fascinating for supporting what I believe to be a general dysfunction of the mainstream progressive dialogue today and a major liability for Democrats.

While the article focused mainly on racist google searches and sad things like evidence of the lack of access to safe abortions in many quarters of the nation, one interesting correlation seemed to slip through without much focus or attention. Speaking about “find my polling place” and related queries, Davidowitz asserts:

In [the 2016] election, you saw very, very clearly in the data that there was a huge decrease in these searches in cities with enormous African-American populations, for example. It was very clear in the Google search data that black turnout was going to be way down in 2016, and that was one of the reasons Clinton did so much worse than the polls predicted.

That’s of particular interest to me since I was so puzzled by the high popularity of Clinton among Southern blacks in the 2016 primaries. This blog attempted to make sense of those trends in early March of that year and basically landed on ageism and lack of access to information technology; but that latter argument is now certainly called into question by Davidowitz’s data and warrants some revisiting.

2016 turnout trends

In April 2016, Nate Silver of FiveThirtyEight made the argument that it wasn’t just blacks but minorities generally that tended to favour Clinton. In particular, the Democrats of the diverse and populous states of Texas and Florida had a pronounced preference for Clinton, so the argument turned to a correlation between the racial makeup of the states Clinton was winning and the overall mix of the party nationally. On the very same day in April, the New Yorker ran a piece about the “myth of the monolithic black vote” that ultimately concluded, citing FiveThirtyEight analyst Harry Enten, that Clinton “just did better in the South” across all races partly because Southern party fealty better rewarded a longtime partisan than an insurgent socialist. But in June when the Wall Street Journal analyzed Clinton’s primary victory through various demographic lenses, the very first explanatory data point was that she “won African-Americans by more than 50 percentage points.” Maybe not a monolith, but that’s a clear bias that mostly explained the size of Clinton’s delegate rout.

In any event, Davidowitz was surely right about low turnout in November. According to analysis performed late last year by this site’s sister blog, turnout in the general election for Democrats was down in 2016 for virtually every state as compared with every presidential election since 2004 [1]. It’s worth noting that the Republicans fared only marginally better by that same measure, but overall turnout was more volatile over the same period. In short, Democratic turnout was uniquely depressed in 2016. Clinton fared worse than Obama in both elections he won in every Southern state, with the lone exception of Texas (which slightly favoured Clinton as compared with Obama in 2012 but still slightly preferred Obama in 2008 to her).

Clinton herself acknowledged that disparity late last month, saying “President Obama broke that racial barrier, but you know, he’s a very attractive, good-looking man.” This blunt sexist angle wasn’t new to Clinton’s retrospective: a senior campaign advisor less than a week after her loss attributed it in part to the sexism of “Bernie Bros”. But the refusal of the Democratic machine to accept responsibility for its failures motivated even Vox to cover the absurd levels of dodging that these lines of argument represent.

coalition shifts

In truth, the lost votes of thousands of Southern blacks — many of whom had surely been voting along explicitly racial lines when they showed up for Obama — don’t alone explain Clinton’s general election defeat. The 2012 Obama coalition actually slipped away largely in the Rust Belt, and Florida was an important loss too. There was a lot of post-election handwringing about the apparent losses or flips of white voters that resulted in these defeats [2]. But more than 10% of the population in four of the six lost states is black, and there are some big cities in each of the type Davidowitz’s data suggests probably had substantially lowered black turnout. But was it actually markedly lower? If so, did it matter?

Overall population data from the Census Bureau.

In March, the New York Times claimed through a lot of analysis but without much sourcing that the answer appeared to be “yes but no”: “Turnout Wasn’t The Driver of Clinton’s Defeat.” While noting that “to the extent Democratic turnout was weak, it was mainly among black voters,” the main argument is that white turnout was higher than expected and slightly favoured Trump. In other words, their analysis supports flip claims. But it was also noncommittal on that point (emphasis mine): “The turnout was slightly better than expected for Donald Trump, but maybe not quite by enough to decide the election.” The analysis seems dubious for other reasons too; for example, it declares 2016 to be “such a high-turnout election” by comparing it with 2014, a midterm election that was also one of the lowest turnout elections in recent times.

But the Times piece is worth highlighting for being one of the few articles I found explicitly acknowledging lowered black turnout at all. It’s also worth noting that the turnout dip didn’t just come from people staying home: there was a flurry of articles within a few days after the election reporting the surprise that Trump did better with blacks and Hispanics than Romney did in 2012! But while argument about how to blame whites for Clinton’s loss went on well into 2017, November appears to have been both the beginning and the end of the post-election discussion about the racial coalition that nominated her.

easy targets

It’s clear that there’s a lot of data and a lot of disagreement about what it means, but the common denominator in Clinton loss retrospectives was a focus on white voting patterns. It’s emblematic that the motivating Vox article headlined a nuanced interview about big data with a reductive accusation of racism. This is a dangerous road since, for example, we’ve seen there’s some basis for an argument that ‘blacks and Hispanics elected Trump’. After all, if the “angry working-class whites” explanation is the mostly right one for the “Rust Belt Rebellion” then it might not have even happened if Sanders had won the primary with his working-class populism. Such a speculative argument is obviously perilous. The difference is that no one attempted to make the argument that blacks swung the nomination to Clinton only to not show up for her in November, but much was said about Sanders’s mostly young, whitenaive supporters hurting Clinton’s chances. Placing her defeat in the general election mostly at the feet of whites was an inevitable next step.

Who’s right?

This blame game seems like a natural consequence of  “white privilege” and “white fragility” dialogues, which are serving as self-righteous litmus tests within lefty echo chambers. Seeking to explain massive racial disparities in modern society, blanket attacks on whites as entitled, selfish, complicit, and ignorant is now progressive lingua franca. And much like the patient who proves she’s in denial by arguing the diagnosis, criticism of white culture is unassailable doctrine for proponents of the ‘white guilt’ line of argument while any other racially charged criticism is oppressor taboo proving cultural insensitivity if not outright racism. No doubt reality corroborates most narratives of minority oppression, but this development seems an abrupt pendulum swing for which clickbait factions within Vox and Huffington Post eagerly supply rhetorical ammunition. As a political tool, though, emphasizing the fact of oppression without much in the way of a path forward that addresses diverse interests is a losing strategy that ignores important realities and doubtless alienates would-be sympathizers.

where’s the autopsy?

Consider: Clinton’s 2016 campaign platform was essentially “not Trump”, and that was expressed through decrying racism, xenophobia, misogyny, cronyism, Russian collusion, and so on. It didn’t win. The Democrats continue the anti-Trump campaigns, and they’re still losing. Democrats across the country poured money into a single special election in Georgia recognized even internationally as a “referendum on Trump”, making it the most expensive congressional election in history, but still lost by four points last week. An outsider in precisely the wrong way, the losing Democratic candidate John Ossoff didn’t even live in the district he was running for. The whole mess is an outsized paragon of the inevitable consequences of the insular, self-reinforcing information bubbles that are increasingly the norm in online debate.

Clearly, the Democratic party is in shambles and its messaging is ineffective. It’s losing elections and, seemingly in denial itself, is refusing to engage in the soul-searching “autopsy” the Republicans famously and publicly embarked on in 2013. Liberal media meanwhile seem to gleefully focus on criticizing white people en masse, with the motivating Vox article forming just a drop in the ocean. The left is doubling down on its message of blame and despair. I’m worried that this is a doomed strategy which seems to be deliberately hiding the widespread conclusion about how the election was lost — that the Democrats ran a crushingly unconvincing campaign — as well as other interesting factors which explain it, like the drop in black turnout. I suggest that Democrats and the liberals who love them would be better served reforming the party with a truly progressive platform that can rebuild its coalition rather than obsessively playing “hunt the racist.”


[1] Kerry 2004 and Clinton 2016 were about evenly matched, though Kerry generally fared better except in the South and a few Western states. One must turn to Al Gore before finding a candidate about whom we can say Clinton performed significantly better across many states; but Virginia and Maryland were the only two Southern states she actually carried, and Gore carried one of them (Maryland) too.

[2Salon claimed Trump flipped working class whites, implying economic distresses; the dominant narrative, held among others by Jimmy Carter, was that Clinton didn’t address their concerns. The Atlantic explicitly disagreed with that last December and doubled down in May to add that it was “cultural anxiety” rather than economic concerns that drove those whites into Trump’s arms. Slate disagreed that working-class whites were flipped, but were simply lost. New Republic says it wasn’t even the working class; “blame college-educated whites” instead, they say. Contra that, Rolling Stone pleaded that we stop obsessing about white working-class voters because Clinton actually lost poor and uneducated voters across all races. This is obviously a confused mess, but the takeaway from everyone does appear to be that Clinton lost votes, which makes the repeated reminders on her irrelevant win of the popular vote all the more toxic.

Leave a Reply