By Lambert Strether of Corrente
Strong data about an election always trails the election results by many months, and interpretation of that data trails the data release. This month, we have the Cooperative Congressional Election Study (CCES) data release:
This is the main release of the 2016 CCES Common Content Dataset. The data includes a nationally representative sample of 64,600 American adults. This release includes the data (including vote validation for respondents) and a full guide and codebook for the dataset.
Laudably, the data is freely available. Table I lists the names of the principal investigators, and describes the effort:
This project is the collaborative effort of 60 research teams and organizations. Individual teams had their own principal investigators and research groups and designed their own team surveys.
(I find it reassuring that I don’t recognize any of the investigators’ names. Also, the PDF is produced in the typesetting language TEX, a reassuring sign of academic and not media provenance.)
In this post, I’ll briefly present the initial framing of the CCES study, discuss today’s reframing — since that framing could be used tendentiously in current Democrat factional infighting — and finally present some alternative frames.
Initial Framing of the CCES Study
Initial coverage of the CCES study focused on voters who flipped from Obama to Trump, as in this story from Nate Cohn in the New York Times from August 15:
The Obama-Trump Voters Are Real. Here’s What They Think.
(I briefly discussed that story here.) After discussing several other stories, Cohn focuses on CCES, writing:
Here’s what one survey, the C.C.E.S., says about these voters…
Taken together, the data indicates that Mr. Trump had considerable and possibly unique appeal to an important slice of Democratic-leaning voters. Mr. Trump adopted a platform tailored to white working-class Democrats. In doing so, he neutralized many traditional Democratic lines of attack against typical Republicans like Mitt Romney. Many of these voters backed him in the primary and seemed to prefer his brand of populism, suggesting they probably would have backed Mr. Trump no matter which Democrat he faced….
And Cohn concludes:
All considered, it does seem likely that at least a portion of the Obama-Trump vote can be lured back to the Democrats — especially against traditional Republican candidates who emphasize small government, free markets and social conservatism.
Whether that means it should be the crux of the Democrats’ path to power is another question. But it will most likely be a part of it, and will probably need to be for Democrats to secure parts of the Rust Belt that continue to play an outsize role in American elections.
(Here Cohn, as nearly as it is possible for a reporter to do, takes a stand in the “Democrat factional infighting” mentioned above.)
Today’s Reframing of the CCES Study
So, last night I saw this come across my Twitter feed, also based on the CCES study. August 23:
— John Fund (@johnfund) August 23, 2017
The chart comes from Taggen Goddard’s Political Wire in the article to which Fund links. Goddard qualifies the headline:
That so many Sanders primary voters chose Trump over Clinton in the general election is an extraordinary finding based on the tremendous policy differences between the two men.
But in the end, it seems the 2016 presidential election was less about public policy than it was about rejecting the status quo.
From there, Newsweek and WaPo picked up the (reframed) story, reframing it further. Newsweek’s headline, August 24:
BERNIE SANDERS VOTERS HELPED TRUMP WIN AND HERE’S PROOF
ZOMG, “PROOF”!!!!! The body, which in essence appropriates the Tweet storms of the previous evening, qualifies the headline:
While the self-confessed socialist Sanders and the “America first” embracing Trump represent polar opposites on the ideological spectrum, they both carried an outsider message and appealed to voters frustrated with the staid political establishment. Although Sanders railed against the policies and rhetoric offered up by Trump, the Republican at times sought to reach out to Sanders voters, emphasizing the similarities of the two candidates’ positions on issues like trade.
(Why is it that there’s no such thing as a “self-confessed” neoliberal?) And WaPo‘s “Monkey Cage” coverage, August 24. The headline:
Did enough Bernie Sanders supporters vote for Trump to cost Clinton the election?
Once again, the body qualifies the headline, and is in fact a useful survey of all the surveys, not just CCES. The conclusion:
In short, it may be hard to know exactly how many Sanders-Trump voters there were, or whether they really cost Clinton the election. But it doesn’t appear that many of them were predisposed to support Clinton in the first place.
Now let me circle back to the Tweetstorm of the previous evening. Brian Schaffner, co-principal investigator of the CCES study, looked at the data to further characterize the 12%-ers, as we might call them, and what Trump’s appeal to them might have been. He rules out trade:
For those suggesting this is about trade, note that opposition to TPP is not dramatically different among defectors. 8/n pic.twitter.com/Fu0GhL58bU
— Brian Schaffner (@b_schaffner) August 23, 2017
(I’m sorry about Twitter’s miserably inadequate embedding function.) And rules in race:
But there does appear to be a racial component to this, as defectors are much more likely to disagree that whites are advantaged in US 9/n pic.twitter.com/zcoHm9APNf
— Brian Schaffner (@b_schaffner) August 23, 2017
Let me remark here that other factors that correlate with voting for Trump include poor credit scores, battlefield casualties, employment situation, and deaths of despair. These do not seem to have been considered by CCES, or if considered were not brought forward by Schaffner, and it is not clear to me that views on trade or race are adequate proxies for them.
Now, if you go beyond the headlines in the Newsweek and “Monkey Cage” articles, and follow the Tweetstorms, you’ll see that there were more Clinton defectors from Obama in 2008 than there were Sanders defectors from Obama, if indeed the Sanders voters can even be characterized as “defectors” (as Schaffner does), given that they seem not to identify strongly as Democrats in the first place. They also seem to be more “conservative” (whatever that means these days), and not necessarily supporters of Sanders’ platform, as opposed to being supporters of Sanders, the authentic candidate.
However, not everybody goes beyond the headlines, in good faith or not, so no doubt the blame cannons will be deployed. There will be a barrage of Tweets that “Bernie Bros are racists who handed the country to a fascist,” or, personalizing, “Sanders is a racist…”, or, factionalizing “brocialists are racist Nazis,” or “Why will not Sanders apologize?” and so forth. This is normal, and it’s not my purpose here to debunk what liberal Democrats consider best practices on Twitter. (I just realized I forgot to include something about the Russians. My bad.) Rather, I want to look at an alternative frame.
Further Reframing the CCES Study
Let’s start with this tweet on voters in swing states.
Sanders -> Trump voters…
Trump win margin…
— G Elliott Morris?? (@gelliottmorris) August 23, 2017
First, consider the campaign at the tactical level. Let’s assume that Sanders had been nominated and that no Clinton voters would have stayed home or voted for Trump. (If we take the attitudes of the Democrat establishment as a proxy for all Clinton voters, that must be a reasonable assumption, albeit a pleasing inversion of “they have no place to go.”) Now shift those voters WI, MI, and PA voters into the Sanders column from Trump. That means Sanders wins those swing states, and the election. So “Bernie would have won.” Sanders would have picked up Rust Belt voters in the swing states, instead of racking up useless votes in California, as Clinton did.
Second, consider the campaign from the strategic level. I’ve always remembered this passage of realpolitik from Ian Welsh in 2010:
If the left does not do this, left wing politics and policies will be discredited with Obama….
The first step to fixing America is fixing the Democratic party, and the first step in fixing the Democratic party is fixing Barack Obama and destroying, forever, publicly and in the most high profile way possible, the idea that Democrats can ignore and abuse their own base.
That didn’t happen in 2010; if the above interpretations are correct, it did happen in 2016; Sanders voters (no matter how one qualifies them) “took Clinton down” and — critically — “were seen to.” If I were the General Grant of the left, which I most definitely am not, I wouldn’t necessarily see Welsh’s realpolitik coming to fruition as a strategic disaster, and might even see it as an opportunity. As Frank Herbert wrote in Dune: “He who can destroy a thing has the real control of it.” So, if the first step to saving America really is fixing the Democrat Party, 2016’s outcome was A Good Thing.
One thing I learned from election 2016 is how vast this country is, how various voters are, and how many voters take their votes much more seriously than pundits imagine they do. In the words of one of the “defectors” quoted in the Washington Monthly:
But as one retired long-time Democrat who voted for [Trump] told me: “You can’t give the guy 100 days, or five or six months. You’ve got to give him at least a good year, or two years. We need a change. Then we’ll ditch him, if he doesn’t do anything.”
 I have to take the sampling methodology on authority. Page 12:
The 2016 CCES survey was conducted over the Internet by YouGov. The Common Content was asked of 64,600 adults interviewed in October 2016 (for pre-election data), and in November 2016 (for post-election data). The sampling method uses YouGov’s matched random sample methodology.
(I would link to the CCES PDF, since there’s a discussion of the methodology starting on page 12, but it’s not online as of this writing.) Pew Research says that YouGov “consistently outperforms” its competitors (actually, they anonymized the pollsters, but later confirmed the top rating to YouGov).
 It’s possible, of course, to view both the Establishment “Better Deal” and the Sanders “People’s Platform” as “lures” to the Rust Belt. Much will turn on whether the voters being so lured — phished for? — regard the offers as made by candidates acting in good faith.
 The article is paywalled, so I actually subscribed to Political Wire to read it. The key paragraph that accompanies the chart:
In their newly-released Guide to the 2016 Cooperative Congressional Election Survey, three university researchers find that 12% of voters who backed Sen. Bernie Sanders in the Democratic presidential primaries actually voted for Donald Trump in the general election.
Since this paywalled chart is now all over the Twitter, let me take a moment to be cranky about provenance. Goddard writes “three researchers,” but as we have seen, there are 60 research teams, so it would have been nice to have their names. Second, I downloaded the PDF for the “Guide to the 2016 Cooperative Congressional Election Survey,” scanned every page, and didn’t see the chart. Third, I googled the chart headline, “What Sanders voters did in the general election,” and I got “no hits” (as of Thursday, August 24, 11:26AM). So, I’m assuming the researchers generated the chart from the dataset with their technical toolkit and sent it to Goddard. Fourth, if that scenario is correct, it would have been nice if Goddard had followed CCES’s best practices for citation, instead of simply linking to a download page. Now, nobody, including the academics, challenges the chart, so I accept its validity on authority. And it’s entirely possible I’m being stupid about this. However, Goddard’s article isn’t helping me not to be stupid.
 I’m assuming that Sanders would have outcampaigned Trump. That seems reasonable. I’m also assuming that the Clintonites and the political class, especially the press, wouldn’t have sabotaged Sanders. That seems less reasonable; see the Iron Law of Institutions.