The current conventional wisdom, expressed for example in this NYT Upshot piece, as well as by Bret Stephens on MSNBC yesterday (June 5) is that the vocal “left” of the Democratic party has lost touch with the authentic base of the party, and therefore risks re-electing Trump by veering too far left.
I believe this analysis suffers from a theoretical mistake that enables its pundit providers to ignore certain empirical evidence while trumpeting other such evidence. Here’s why.
In general, the evidence for these claims comes from public opinion polling that asks people relatively general questions about their political identity (as in the Hidden Tribes project) or about their stances on particular questions (impeachment, Medicare for All, free college, etc.). When the polling results show a less “woke” electorate, or one apparently less attracted to state intervention in health care, inequality, etc., the conclusion is that the coastal (or Twitter) elites are endangering Democratic fortunes by dragging the party too far to the left of the “real” electorate.
The results from even these polls are more mixed than the naysayers suggest — single-payer health care, for example, polls well, depending on how and when the question is asked. But the bigger problem with the cautionary tale is how it understands the political subject and the public opinion poll that represents her.
At least since Zaller, public opinion scholars have at least paid lip service to rejecting the “file drawer” model of public opinion: the idea that most people are walking around with pre-formed ideas on every conceivable issue, and that the task of answering a poll is essentially the same as reaching into a file drawer and pulling out the answer to the question the pollster asked. That’s not actually how public opinion works; although academics and pundits may walk around with such ideas in their heads–and organize the world in terms of what they see as coherent political worldviews–most people do not. (I say “lip service” because many actual practitioners continue to treat polling data as if if it worked that way, even though they ought to know better.)
In fact, many respondents don’t demonstrate what’s called opinion constraint, which is the requirement that opinions on different questions be compatible with one another. A person might, for example, oppose higher deficits, oppose tax raises, and support higher government spending. And a person might change opinions on a given question over a relatively short period of time, particularly if the political environment changes. Indeed, in the article cited above, Sullivan et al. show that constraint increased in the context of the politically charged 1964 election campaign: a sign that the structure of public opinion is malleable and contextual.
In my view, the strongest theoretical understanding of public opinion is that poll responses (like other forms of action) are low-stakes, context-dependent, identity-reinforcing, responses to partially-contolled stimuli. In the aggregate, they measure some combination of durable and contingent responses in a population, and that combination is certainly of substantive interest. But it cannot be assumed to be mostly durable, mostly portable (i.e., applicable to unasked questions that seem to the analyst to be similar), or mostly deeply authentic.
There is ample evidence for this position: evidence the caution-preachers ignore. The Obama-Trump voters constitute one important piece of evidence; there is no pundit-approved political philosophy that, remaining constant, would result in a vote for Obama in 2012 and for Trump in 2016. Conservatives’ newfound love for border walls–not to mention liberals’ for free trade–is evidence for opinion structure responding, sometimes dramatically, to changes in the political environment. I think the pundits ignore this evidence because they continue to assume a file-drawer version of the political subject, even though the theory and evidence tends toward rejecting that image.
The early opinion-constraint literature considered constraint a sign of a smart, capable electorate; the alternative was a sort of chaos of opinions, disorganized and therefore not worth considering. I don’t think that’s right either. Rather, it’s reasonable to understand aggregated private opinion (polls, that is) as organized and coherent for a particular time, place, and population relative to a particular stimulus. Thus the polls are at least as much about the environment and the stimulus as they are about the durable dispositions of the respondents. The principle is illustrated in the memorable “Green Beans” episode of West Wing. CJ: “Everyone’s stupid in a campaign year, Charlie.” Charlie: “No, everybody gets treated stupid in a campaign year, CJ.”
What this means for the 2020 candidates is that they should pay attention to the politics they can make, not just the opinions that predate them. Voters are not (all) stupid or chaotic. But they’re also not packages of opinions waiting to be represented as a whole. Campaigns are prime learning opportunities for voters. The best — and, I hope, the winning — candidate will be the one that raises the voters’ best hopes and ambitions, not the one that just follows their existing views.