How the Press is Getting Things Wrong

What is the best way to get attention for one poll in the midst of this busy political season when several polls are released a day? The answer seems to be having results that are starkly different than the prevailing polling average or consensus.

During Wednesday night’s Republican Town Hall on CNN, Anderson Cooper congratulated Ted Cruz on his lead in the most recent Wall Street Journal/NBC Poll without an acknowledgment that two other polls in the field at roughly the same time found Trump with the same lead he’s had for months. Another poll this morning confirmed the findings of a big Trump lead.  On the Democratic side, the Rachel Maddow Show trumpeted the Quinnipiac poll that showed Clinton’s lead had shrunk to two points nationally even though other polls showed Clinton maintaining about an 8-point lead.

The press has a clear and consistent pattern of focusing on the outlier poll. Now as it happens outliers can sometimes be right. In New Hampshire, the WMUR Poll was the outlier on the Democratic side and it turned out to be closest to correct. But more often outliers are wrong for the simple reason that if everyone is trying their best to accomplish an identical tasking using similar methods they will get similar results. If you sent four people on a treasure hunt and three people returned and said one thing and one person said another and you had no particular reason to believe the one person, you would not treat his or her news as the only news.

Mistakes are not confined to the TV journalists whose primary objective is to put on a good show. They also plague the two most respected sources of political data journalism, The New York Times and fivethirtyeight.com

Let’s start with The Times today and move onto 538 tomorrow.

We were thrilled The New York Times caught up to our thinking on Republican delegate math, but it is clear The Times is not quite there in terms of its analysis. Here is the article.

It flags the 20% threshold that is required to get any delegates in the Alabama, Georgia, Tennessee and Texas primaries, which is incredibly important. The article suggests that Marco Rubio could get as many as 70 delegates in those states. Yet, the article neglects just how many of the delegates are awarded at the district level.

In Texas, two-thirds of the 155 delegates are awarded at the district level and only the top two finishers in each district are awarded delegates. Thus, even if Rubio charges into Texas and claims a healthy 25% of the statewide vote, which would be worth 12 at-large delegates, he’d still need second place finishes in some Congressional districts to avoid a poor showing.

In all seven threshold states, a candidate that gets 25% but never comes in second at the district level will get only 54 at-large delegates. Oklahoma gives one delegate to each of the top three finishers, thus making it somewhat less interesting. In Oklahoma’s five districts, as long as Trump, Cruz and Rubio each get at least 15%, which seems less likely, they’ll each come away with one delegate. Adding in Oklahoma that means Rubio is more likely to get 59 delegates, not the 70 the Times suggested.

At 20% (not 25%), the number of delegates drops to 48, including Texas, and 39 without. With favorite son Cruz leading and Trump coming on strong in the Lone Star state, Rubio is challenged to get the one-fifth he needs.

Still, the Times could maintain that 59 is not that far away from the 70 it predicted and that Rubio getting 25% in some of these states could lead to second place district finishes. If he came in second in 11 districts across five states (Alabama, Arkansas, Georgia, Texas and Tennessee) earning a delegate per district, he could end up with 70 delegates.

It’s hard to imagine Rubio, however, will get to second in any districts in Texas. It’s even harder to imagine given the current polling that he would actually win a district anywhere. It’s not impossible but it is a serious challenge with Kasich still biting at his tail and Bush still possibly hanging around.

The point here is that the Times implies that 20% is worth 70 delegates for Rubio, but it is much more likely that he needs at least 25% as well as reaching the threshold in Texas. Trump and Cruz taking every single delegate in Alabama, Georgia, Tennessee and Texas is as likely an outcome as Rubio getting the 70 this article implies he can get. (Reaching the 15% threshold in Arkansas looks likely for Rubio so he should be getting some at-large delegates there.)

Another problem with this article is that it refers to Colorado and Minnesota as moderate states. This might be the case if they were holding primaries, but they’re not. Both states are holding caucuses that are often not representative of the state’s leanings. While the Colorado caucus was close between the moderate Romney and conservative Santorum in 2012, the Minnesota caucus was an absolute nightmare for Romney.

The Times also hinted that with late states could make up for problems on March 1^st and that Rubio could take the delegate-rich, winner-take-all Ohio. But it doesn’t consider whether Rubio can win Ohio with John Kasich still in the race. Since Kasich is already gathering absentee ballots in Ohio, he may not even have to be in the race to keep Rubio from winning. He could have already banked as much as 10 percent of the vote in Ohio. (Old nugget: Joe Lieberman, who had quit the president’s race almost a month before, still received 5% of the vote in Connecticut with no real effort.) The conundrum posed by all of these moving pieces will be interesting to unspool after South Carolina.

For now, it is enough to note that as hard as the Times made it seem for Rubio to win, it is actually harder. Later today previews for South Carolina Republican and Nevada Democrat contest tomorrow.