For the sharpest analysis of the dust-up about Nate Silver’s “prediction” of a GOP takeover of the U.S. Senate in the November elections, you probably can’t do better than “What’s Being Overlooked About Nate Silver’s Senate Prediction” by Mark Blumenthal and Ariel Edwards-Levy at HuffPo Pollster. Here’s a taste:
‘Is Silver’s prediction an outlier?’ Jonathan Bernstein answers his own question: “Not at all. A toss up with a slight edge to Republicans is right in line with predictions by Sean Trende; the Cook Report’s Jennifer Duffy; the Rothenberg Political Report; political scientists Eric McGhee, Ben Highton and John Sides; and also political scientist Alan Abramowitz. Those forecasters come from different traditions and use different methods, but they all think Senate control is a toss-up, with slight variations about which party is the favorite.” [Bloomberg] DSCC pushes back – Guy Cecil, executive director of the Democratic Senatorial Campaign Committee: “Nate Silver and the staff at FiveThirtyEight are doing groundbreaking work, but, as they have noted, they have to base their forecasts on a scarce supply of public polls. In some cases more than half of these polls come from GOP polling outfits. This was one reason why FiveThirtyEight forecasts in North Dakota and Montana were so far off in 2012. In fact, in August of 2012 Silver forecasted a 61% likelihood that Republicans would pick up enough seats to claim the majority. Three months later Democrats went on to win 55 seats.” [HuffPost] What the reaction overlooks – First, expressing a forecast as a probability does not make it a certainty. If Silver says he is 60 percent certain something will happen, he should be wrong about it 40 percent of the time. As RealClearPolitics’ Sean Trende Tweeted on Monday, in reference to Mercer’s upset victory over Duke in the NCAA March Madness basketball tournament, “Yes, @NateSilver538 said Duke was 93% fave over Mercer, but question is really whether teams he says are 93% faves lose > 7% of the time” [@SeanTrende]
There’s more worth reading in Blumenthal’s and Edwards-Levy’s post. Read it and follow the links they provide, and you’ll have a pretty good sense of the relevance of Silver’s ‘prediction.’ Snapshot analyses may have some value to campaign strategists in suggesting tweaks in resource allocation. But no campaign for a November election should be demoralized or too encouraged by poll analysis provided in March.