There is a dormant Trump vote. The best place to look probably isn’t Facebook memes.

2

Politics – washingtonpost

The reason Fox so often wins in the ratings, though, is largely disconnected from the idea that Americans crave pro-Trump coverage. It’s connected, instead, to the fact that Republicans seek pro-Trump coverage. So Fox News gets a much higher density of Republican viewers than CNN or MSNBC do of Democratic viewers, since people looking for less sycophantic coverage have more options from which to choose. In other words, Fox News wins in the ratings the way Trump won in the 2016 Republican primary: It has a loyal core of support while its various competitors are fighting over those not enamored of the president.

This dynamic is important to consider when reading Kevin Roose’s fascinating assessment of the role Facebook plays in the conservative media ecosystem. Roose has become transfixed by the fact that the posts that generate the most traffic on the site each day tend to be ones published by conservative commentators or news sites.

“What sticks out, when you dig in to the data, is just how dominant the Facebook right truly is,” Roose writes. “Pro-Trump political influencers have spent years building a well-oiled media machine that swarms around every major news story, creating a torrent of viral commentary that reliably drowns out both the mainstream media and the liberal opposition.”

He created a Twitter account that lists the top 10 link-sharing posts each day. Using that account and categorizing each of the top-performing outlets into broad categories, we can get a visual sense for what Roose is observing: The resulting chart is heavily red. (Data for Aug. 1 and 2 were not included.)

Roose extrapolates from the existence of a conservative, right-wing Facebook ecosystem to a common refrain of the president’s: There’s a hidden group of supporters he can rely on in November’s election. Perhaps, Roose speculates, this group of “silent” supporters (who inexplicably don’t show up in polling) is manifested instead in its behavior on Facebook.

“Looking at Facebook’s lopsided political media ecosystem might be a useful reality check for Democrats who think Mr. Biden will coast to victory in November,” he writes. “After all, Mr. Trump’s surging popularity showed up online before it showed up in any polls in 2016.”

(I’m skeptical of that last claim, given how Trump seized online attention essentially from the outset, but that’s a different article.)

Before we get too far into this, there are a number of caveats worth applying.

The first is that conservative commentator Ben Shapiro, who frequently turns up in the top 10 along with his site Daily Wire, was found earlier this year to have been participating in an arrangement that manipulated Facebook’s algorithms to boost viewership of his articles. This doesn’t make those views somehow invalid, but it does suggest that Shapiro and Daily Wire have a leg up in the metric under consideration.

Another consideration is scale. Shapiro, Roose writes, has gotten 56 million interactions on his Facebook page in the past month, far more than, say, The Washington Post.

And then there’s the third point, the one with which we started: If Shapiro has a loyal, engaged audience that fervently tracks and shares his content, that doesn’t necessarily mean the audience is a politically significant audience. They might not be of voting age, in the United States or registered to vote.

Again, scale is important. Fox News averaged 1.9 million viewers a day in the first quarter of this year, a figure that corresponds more closely to individual people than the 56 million interactions number, given single people could be responsible for multiple interactions.

Also important: 1.9 million people is 1.3 percent of the total turnout in the 2016 presidential election. Two million people engaging with your news stories or video clips is not insignificant, particularly in a close election, but it’s not as though Fox News is being viewed regularly by a huge swath of the electorate.

The Trump campaign talks about its silent support for two reasons. The first is that it is trailing in the polls, leading Trump and his team to speculate that people simply aren’t telling pollsters they support him. (Notably, this was never raised during the 2016 primary, when Trump was consistently leading.) The second is that there actually are a lot of voters who would probably vote for Trump if they went out to vote — but they don’t go out to vote.

This is a minefield for campaigns and pollsters alike. How do you capture support from infrequent voters who are suddenly spurred to go vote? How do you even assess if that’s a likely outcome? How do you determine if that would actually make a significant difference. As elections loom, pollsters focus on likely voters because, well, they’re likely to vote. Belaboring the point, infrequent voters don’t vote very often, by definition, so assuming they’ll all turn out this time is a risky proposition. (This problem arises perennially with young voters, who always seem like they’re on the brink of rushing to vote and then always seem not to. Yes, there are exceptions; no need to email me.)

The Post’s Lenny Bronner dug into voter registration and election survey data to try to determine how many voters there are in the country whom Trump might be able to rely on if he could get them to the polls. The model Bronner developed, explained in all of its complexity at the bottom of this piece, moves currently registered voters into four buckets: those who likely voted for Trump in 2016, those who likely would have if they had voted, those who likely voted for Hillary Clinton and those who likely would have.

The upshot? There may be 8 million potential Trump voters in states he could conceivably have flipped in 2016 but who didn’t turn out to vote. He might have won states like Colorado, Delaware, Maine, Nevada, Virginia if — and this is a massive if — every one of his possible voters came out to vote and Clinton’s didn’t.

Our point here isn’t that this is likely to happen, but it does serve as a reminder that there is a big pool of voters who mostly aren’t captured in polling and who, if the campaign can get them to vote, could make a significant difference in the outcome of the race.

Again, though, some important caveats. The first is that campaigns often don’t spend a lot of energy and money trying to turn out infrequent voters because, again, they aren’t as likely to vote. The Republican convention throwing out a barrage of arguments for Trump and against Democratic nominee Joe Biden was, in part, an effort to spur some of those infrequent voters to show up in November. It’s unlikely they’ll similarly focus a lot of time on this group as the election approaches, given the urgency of turning out people who are more likely to cast a ballot.

The second caveat is that Democrats, too, will be trying to turn out the people who didn’t show up for Clinton in 2016. There were 4.4 million voters who supported Barack Obama in 2012 who stayed home four years later, a group that is heavily non-White and heavily young. According to Bronner’s model, there are about 12.8 million voters who would probably have supported Clinton who didn’t vote, enough to win 16 states that Trump did if (big if) all of those voters had turned out and Trump’s hadn’t.

If every voter turned out in every state and voted the way Bronner’s model expects, Clinton would have won Michigan, as well as Florida, Georgia, North Carolina and Pennsylvania, the latter four of which were essentially Clinton-leaning ties once the model of actual turnout is applied. (Why didn’t Wisconsin, one of the closest states, flip? It’s less racially diverse.)

Here’s what the model shows, if you’re curious.

It is absolutely the case that there will be voters spurred to vote this year by things they read on Facebook from conservative and right-wing commentary sites. But you probably already did the math: The 8 million voters mentioned above is four times the number of people who watched Fox on any given day in the first quarter of the year that, itself, is a bigger audience than what Shapiro’s likely getting.

And that’s just in theoretically flippable states. Overall, Bronner’s model figures there are about 30 million probable Trump-supporting voters who didn’t vote in 2016. That’s the “silent” group that Trump’s team is trying to figure out how to mobilize, and it’s a group that almost certainly swamps Shapiro’s regular audience.

It’s also about 10 million voters smaller than the group that would likely have supported Clinton in 2016.

Lenny Bronner contributed to this report.

How the model works: To estimate the potential number of additional voters for either candidate, we need to estimate the probability that an individual voter would have voted for Trump or Clinton. We do this using a multilevel regression and post-stratification model (MRP), which we feed data from the Cooperative Congressional Election Study (CCES). This study, conducted after every major election, asks more than 50,000 people who they voted for and records their demographic attributes. Using those responses, we estimate the probability of having voted for Trump/Clinton for every registered voter in the United States, using their gender, race, age and location (down to a county level).

The specific type of model lets us infer behavior of similar groups, even if we don’t have enough data. Say we have a lot of responses from African American women in Jefferson County, Ala., aged 30-45, so we have a pretty good estimate of the probability that an individual in that group voted for a candidate. At the same time, not enough Black women in that age group answered the survey in Pickens County, Ala. Using the probability for African American women in that age group, generally, the probability of people in Alabama, generally, and even the probability of people in the South, generally, to have voted for Trump/Clinton, we can infer the probability of the group we are missing data for in particular. After estimating this probability for each subgroup (over 3,000 population subgroups in Alabama alone), we apply our probability estimates to the number of registered voters in each subgroup, using a national voter file, provided by L2.

For example, we estimate that White men between the ages of 18-29 in Autauga County, Ala., had a 20 percent chance of voting for Clinton, and we also know there are 415 such registered voters. So if every registered voter in the United States had turned out, Clinton would have received 83 voters and Trump would have received the other 332.



Source link

Comments are closed.