Political Journalism and Political Science: Still a First Date

At Poynter, James Warren writes about last weekend’s meeting of the Midwest Political Science Association, focusing on a panel titled, “The Media and the 2016 Election: A View from the Campaign Trail.”  While I appreciate the journalists who would show up to such a thing, if Warren’s report is any indication, even the journalists interested in political science still have a lot to learn from it.

Steve Peoples of the Associated Press suggested the 2016 election was leading him to question all of his assumptions, which is probably a good practice for most people in general.  But Warren reports that Peoples wondered what journalists would do if you cant trust the polling.

If this was Twitter, I’d be hashtagging that sentiment #facepalm and #headdesk for several reasons.

First, it is usually the case that post-election seminars feature journalists confessing that too much of election coverage is focused on the horse race.  Political scientists would tell you there’s good reason to be concerned about it:

“Patterson (1993; 2005) and others fear that the focus on the game over substance undermines the ability of citizens to learn from coverage and to reach informed decisions in elections or about policy debates. Capella and Jamieson (1997) argue that the strategy frame portrays candidates and elected officials as self-interested and poll driven opportunists, a portrayal that they show promotes cynicism and distrust among audiences. Farnsworth and Licther (2006) go so far as to suggest that horse race coverage in the primary elections results in a self-reinforcing bandwagon effect with positive horse race coverage improving a candidate’s standing in subsequent polls and negative horse-race coverage hurting a candidate’s poll standings.”

The 2008 and 2012 elections had much the same problem.  And 2016 was no different, with horse race coverage accounting for most of the reason a candidate like Donald Trump got mostly positive coverage.  Indeed, while Nate Silver is a data journalist rather than a political scientist, his analysis supports the bandwagon thesis: the media covered Trump well in excess of his poll standings, ultimately driving those standings higher despite bad favorability numbers.

In contrast, you can check Jack Shafer‘s 2008 hot take defending horse race coverage to see how much worse it sounds now than then (and it sounded bad then).

Second, while there was a small systematic error in the 2016 polling, Nate Silver explained before the election why his model showed a 28.6% chance of Trump winning and the reasons he gave pretty much explained in advance what happened.  And even if you don’t buy the precision of a model like Silver’s (and you probably should not), it was Sean Trende (who holds a poli sci degree) noting that a 25% chance was like flipping a coin and having it come up heads twice in a row — hardly shocking.

Instead, journalists and more conventional pundits tended to see 25% — or even 14% — as 0%, when in fact, sometimes unlikely results occur.  That does not wipe out the laws of probability.  The chances of rolling a six on one die are only 16.67%, but it still happens and when it does, it doesn’t mean the die is loaded or defective.

Third, polling isn’t the only thing political science has to offer journalism.  Political science could also offer a number of fundamental reasons — 2016 being an open seat election in a mediocre economy involving two poor candidates and a Democratic Party that had been losing white working class voters for decades — that helped account for Trump’s victory, all of which could have been considered and incorporated into journalists’ thinking well in advance of election day.

Molly Ball and Nia-Malika Henderson apparently commented on the sorry state of the Democratic Party.  Ball thought it was “hard to underestimate how screwed the Democrats are,” but noting their situation wasn’t hopeless, recalled that Barack Obama was a little-known state senator before the 2008 election.

I’m hoping Warren mischaracterized Ball, as this is almost entirely incorrect, and any good political scientist would have been able to correct her.

First, by the time of the 2008 cycle, Obama had been elected to the U.S. Senate and had been the highly-publicized and highly-lauded keynote speaker at the 2004 Democratic National Convention.  Political scientists would identify such a person as a rising star, well positioned to compete in the “invisible primary” of party officials, donors and influencers that occurs before a single vote is cast.

And in fact, Obama proved to be a prodigious fundraiser from both Wall Streeters and small donors alike.  While it was certainly possible that he could have fizzled had he lost the Iowa caucuses, political scientists would have predicted he could mount a strong challenge to Hillary Clinton.

[Aside: The fundraising is usually crucial because of the cost of paid media.  In 2016, Donald Trump entered the race with high name-ID and a press willing to provide free media well in excess of his poll numbers.]

Second, as for the Democrats being screwed, Jay Cost (another political scientist by education, iirc) has observed that “[i]f the Republican party were a publicly traded company, January 20 would be the day to sell, sell, sell.  This may sound counterintuitive, but the verdict of history is clear, if not quite unanimous: The moment a party achieves total control of the government is the moment just before power begins to slip through its fingers.”

Finally, Ball apparently wants to know if there has been a lasting realignment of the parties, or whether 2016 was an anomaly.  Trende’s book, The Lost Majority, would tell you no such thing truly exists.  See also Jay Cost:

In addition, while the panel apparently noted that Hillary did well with college-educated whites, I have noted previously that Trump was outpolled by down-ticket GOPers in many races, often by appealing less to working-class whites and more to college-educated whites.  John Judis — a progenitor of the Emerging Democratic Majority theory — noted the GOP’s overall improvement with white voters, but particularly college-educated white voters, back in 2015.

The GOP having Trump as its public face might change those trends in time, even if it did not occur in 2016.  But a political scientist would tell you that’s where the analysis starts.

PS: Consider subscribing to WHRPT in the sidebar (the posts come straight to your inbox; no muss no fuss). And following WHRPT on Twitter.  Thanks for reading and sharing!

Upsets Happen. No, Really.

Before we get too far away from the SuperBowl, let’s revisit ESPN’s win probability chart:

We all laughed. We all made jokes comparing the Biggest SuperBowl Comeback to the 2016 election.

What we didn’t do was conclude that Tom Brady repealed the laws of probability.  People who have watched pro football over the course of their lives didn’t need a chart to tell them that when a team is ahead by 28-3 (as the Falcons were at one point), the odds of the opponent winning are slim indeed.  We also didn’t need a chart to remember that sometimes big comebacks do happen.

Yet there are a lot of people who seem to believe that the 2016 election proved that polls are worthless and polling models doubly so.  Before the election, Nate Silver wrote about why FiveThirtyEight’s model gave Trump better odds than others and why Hillary Clinton was in a weaker position than Barack Obama had been.  But people just wanted to treat the topline numbers as Gospel.

Nate Cohn, despite the NYT giving Trump worse odds, wrote just before Election Day that he was within striking distance of winning because of his huge lead with white voters without a college degree.  The NYT concluded that Clinton’s chance of losing was about the same as the probability that an NFL kicker misses a 37-yard field goal.

You don’t have to have been a longtime NFL fan to at least vaguely recall that the Vikings’ Blair Walsh missed a 37-yard FG attempt in 2016.  Or that the Bears’ Connor Barth missed a 31-yarder.  Or that the Bucs’ Roberto Aguayo missed a 32-yard attempt in 2015.

Of course, if a kicker is consistently bad, he’ll get cut; just ask the Mighty Bengals.  Then again, if you never campaign in Wisconsin, maybe you’ll lose to Donald Trump.

When we see unlikely things happen in football, we seem to have more rational reactions than when we see them happen in politics.  After all, if you’re not a fan of data journalism (and to be fair, it’s far from perfect), it’s an easy slam.  And if you’re invested in pushing a narrative of Trump as the Colossus who remakes the GOP and American politics generally, it’s a useful slam and a way to dismiss unfavorable data as “fake news.”

But the laws of probability have not been repealed.  And while the polling industry faces big challenges, it’s not dead.  People will ignore data at their peril.

PS: Consider subscribing to WHRPT in the sidebar. And following WHRPT on Twitter.  Thanks for reading!