Sunday, May 29, 2016

Harry Enten Continues 538's Descent Into Political Hackery

Lying With Numbers Dept. - In the beginning, Nate Silver's 538 could have been seen as a noble experiment--a just-the-facts-ma'am approach to political contests. However one may have characterized it at its birth though, that experiment has spectacularly crashed and burned this election cycle. The 538 gang, once so committed to the numbers, have increasingly turned, for new sources, to their own orifices. Last Summer, Harry Enten boldly declared that "[Donald] Trump has a better chance of... playing in the NBA Finals... than winning the Republican nomination." Two months later, Silver himself said this about the Republican contest:
"At FiveThirtyEight, however, we’re fairly agnostic about what will happen to Trump's polling in the near term. It’s possible that he's already peaked--or that he'll hold his support all the way through Iowa and New Hampshire, possibly even winning one or two early states, as similar candidates like Pat Buchanan and Newt Gingrich have in the past. Our emphatic prediction is simply that Trump will not win the nomination."
Uh huh. On the Democratic side, 538's forecasting process has shown a rather extreme pro-Clinton bias (covered by Doug Johnson Hatlem on Counterpunch last month):
"...outside the South, 538 has a 12.2% Clinton House effect on the Democratic side. In eighteen of the twenty-one races, 538’s projection or modelling has shown a Clinton bias."
...which brings me to the Enten's particularly deplorable article the other day, "The System Isn't 'Rigged' Against Sanders":
"Sanders fans have claimed that because caucuses have lower turnout, the current national caucus and primary vote underrates how well Sanders is doing. In fact, the opposite is true. When we switch all caucuses over to primaries, Sanders actually does worse."
But of course there's no way, absent the use of magic, to "switch all caucuses over to primaries" or to work any of the other witchcraft Enten claims he's accomplished in this piece. The math Enten uses to pretend to do all this reality-altering is, like the premise of the article itself, purely Uranian. Enten pretends he can wave a wand and conjure up a reliable primary result from places where no primary contest ever took place and show, with real numbers, what would happen in closed primary states if they'd held open ones. He can't. And he knows it.

He begins his case with Hillary Clinton's "win" in the Washington state primary. This was held after Washington's delegates had already been allocated via a caucus, which Sanders had won. The subsequent "primary," which went for Clinton, was, as even Enten describes it, purely a beauty-contest. No candidate campaigned for that "primary." How much attention and enthusiasm is going to go toward a "primary" in which there's nothing at stake and no one even competing for votes? Enten sort of acknowledges the obvious answer but why even bring it up? It's a bit of sleight of hand for what comes next.

This can't be said forcefully enough: the "estimates" Enten uses here--magically turning caucuses into primaries and even caucuses and primaries into open primaries--are pure fantasy. No polling data exists to support Enten's conclusions. They're based on what he says is a "demographic model" but that model runs into a pretty serious brick wall in the form of the actual results of these contests. The Iowa caucus, for example, ended in a virtual tie. Given to Clinton, it may have even been won by Sanders--the Democratic party has refused to release the raw totals that would quantify the result. The polling leading up to the caucus tells the same tale as polling in most of the rest of the U.S.; Sanders started out as an unknown with massive 30- and 40-point gaps with Clinton, sometimes even more, and as the race proceeded, Clinton's lead progressively vanished. By the end, it was a low-single-digit race, with every new poll seeming to declare a change in the frontrunner. Polls in caucus states are, of course, typically of likely caucus participants, a much smaller group than would participate in a primary, but Enten asserts that if Iowa had held a primary instead of a caucus, Clinton would have won by a massive 24 points--a one-sided massacre of epic proportions. If Clinton supporters so decisively outnumbered Sanders supporters in the state-- by Enten's measure, nearly 3 1/2 Hillary supporters for every 2 Bernie supporters--how did we get the caucus results we did? To find Enten's conclusion at all plausible, one must believe there was some sort of catastrophic--and entirely unexplained--breakdown in getting Clinton people to the caucus. As the press delights in asserting, Sanders' strongest demographic is white Democrats (though in reality, it's actually young people of all races and ethnicities) and Enten has Clinton utterly rolling over Sanders in one of the whitest states in the U.S.. Many of Enten's results are like that. Clinton won the Nevada caucus by 5 points; Enten has her winning a theoretical primary there by 29%. Sanders dominated Clinton in Colorado, a 19% spread in the caucus; Enten's imaginary Colorado primary goes to Clinton by 6%. And so on. This is a "demographic model" that, on its face, seems to have sprung a rather serious leak. Enten pretends to provide numbers on what would happen if states with closed primaries had held open ones, which is just as baseless a fantasy.

Enten also appears to have committed some other pretty basic fallacies in assembling his presentation. When Sanders supporters suggest the nomination process is "rigged" in Clinton's favor, they aren't just referring to chicanery in the mechanics of the primary/caucus system in the many states. They're also talking about things like the nonsense with the debate schedule, which was not only changed but radically so and specifically to help Clinton, the fact that the head of the DNC, who is supposed to be neutral in the contest, is a former Clinton hand who, from the beginning, has been doing everything she can to try to help her candidate, the funneling of funds raised for downballot races into Clinton's coffers, the fact that the corporate press has been a virtual monolith in support of Clinton and opposition to Sanders and so on. Enten simply ignores all of this, focusing strictly on the question of caucuses vs. primaries vs. open primaries but pretends to be "refuting" those suggestions that the system is rigged. With regard to his "demographic model," he asserts he can create such a thing because voting this cycle has largely broken down along demographic lines, but that, of course, assumes his own conclusion within the question. Voting has broken down along those demographic lines in the election as it's being held but this is the very process Sanders' supporters have asserted is rigged--the proposition Enten is trying to refute. Ignoring their larger criticism of the process becomes even more problematic. For someone relying so heavily on demographics, Enten didn't use much demographic data in creating his model either:
"...they [the 538 gang] reduced all demographic differences between states with primaries and states with caucuses to two variables: proportion of black voters and proportion of Hispanic voters. Age? Income? Education? Geographic region? Religious differences? Economic indicators, like unemployment? Apparently none of these things matter, according to Silver and Enten... This is obviously incorrect, we know very well that at minimum age is an incredibly important variable. Young voters of all races prefer Sanders, and older voters of all races prefer Clinton... By leaving out many factors which are important for determining differences between states that could affect the vote outcomes, their analysis is subject to omitted variable bias."
Further, it's impossible via Enten's exercise to reproduce the dynamics of an ongoing campaign, where wins and losses in contests and how both are perceived by the public and presented by the press affect the next contests, the results of which affect the next and on into infinity.

Though in many cases, the system has been rigged against Sanders personally, focusing merely on that overlooks the scope of the issue. It's better to say it's rigged against anyone like Sanders, an outsider challenging the party's entrenched. Discussing the Democratic "superdelegate" system in February, DNC chief Debbie Wasserman Schultz confessed:
“Unpledged delegates [superdelegates] exist really to make sure that party leaders and elected officials don’t have to be in a position where they are running against grassroots activists."
The superdelegate system is only one aspect of that rigged process. Even as the head of the Democratic National Committee, who also happens to be an all-out Clinton supporter, can admit this is designed to protect Establishment candidates, Harry Enten continues 538's descent into hackery by playing these elaborate and absurd games that have no other purpose than to act as propaganda for one candidate in a contested race.

--j.