Why I Don’t Agree With Nate Silver

Number-crunching Senate prediction models are fun to follow but are not very useful.

Nathaniel 'Nate' Silver, editor-in-chief of ESPN's FiveThirtyEight blog, speaks during a panel discussion at the South By Southwest (SXSW) Interactive Festival in Austin, Texas, U.S., on Saturday, March 8, 2014. The SXSW conferences and festivals converge original music, independent films, and emerging technologies while fostering creative and professional growth. 
National Journal
Josh Kraushaar
Add to Briefcase
See more stories about...
Josh Kraushaar
April 29, 2014, 8 a.m.

I’m a num­bers guy. As a base­ball fan, I pore over box scores, reg­u­larly second-guess man­agers who use old-school tac­tics, and was prob­ably one of Nate Sil­ver’s first read­ers and an early sub­scriber to the saber­met­ric ref­er­ence book Base­ball Pro­spect­us, where he made a name for him­self pro­ject­ing play­er out­comes. In re­port­ing on and ana­lyz­ing polit­ics, I rely greatly on fun­drais­ing re­ports and polling data to in­form the tra­ject­ory of key races.

But count me un­der­whelmed by the new wave of Sen­ate pre­dic­tion mod­els as­sess­ing the prob­ab­il­ity of Re­pub­lic­ans win­ning the up­per cham­ber by one-tenth of a per­cent­age point. It’s not that the mod­els aren’t ef­fect­ive at what they’re de­signed to do. It’s that the meth­od­o­logy be­hind them is flawed. Un­like base­ball, where the sample size runs in the thou­sands of at-bats or in­nings pitched, these mod­els over­em­phas­ize a hand­ful of early polls at the ex­pense of on-the-ground in­tel­li­gence on can­did­ate qual­ity. As Sil­ver might put it, there’s a lot of noise to the sig­nal.

The mod­els also un­der­value the big-pic­ture in­dic­at­ors sug­gest­ing that 2014 is shap­ing up to be a wave elec­tion for Re­pub­lic­ans, the type of en­vir­on­ment where even seem­ingly safe in­cum­bents can be­come en­dangered. Nearly every na­tion­al poll, in­clud­ing Tues­day’s ABC News/Wash­ing­ton Post sur­vey, con­tains omin­ous news for Sen­ate Demo­crats. Pres­id­ent Obama’s job ap­prov­al is at an all-time low of 41 per­cent, and pub­lic opin­ion on his health care law hasn’t budged and re­mains a driv­ing force in turn­ing out dis­af­fected voters to the polls to re­gister their an­ger. Pub­lic opin­ion on the eco­nomy isn’t any bet­ter than it was be­fore the 2010 midterms when the un­em­ploy­ment rate hit double-di­gits. Demo­crats hold only a 1-point lead on the gen­er­ic bal­lot in the ABC/WaPo sur­vey — worse po­s­i­tion­ing than be­fore the GOP’s 2010 land­slide.

These macro-in­dic­at­ors don’t square with tar­geted Demo­crat­ic sen­at­ors — such as Jeanne Shaheen of New Hamp­shire, Al Franken of Min­nesota, Mark Warner of Vir­gin­ia, and Jeff Merkley of Ore­gon — be­ing rated heavy fa­vor­ites to near locks for reelec­tion, as the Sil­ver and Up­shot mod­els show. The mod­els are great at con­clud­ing the ob­vi­ous — red-state Demo­crats are in trouble! — but blind to an­ti­cip­at­ing fu­ture out­comes, giv­en their de­pend­ence on lim­ited pub­lic polling and quarterly fun­drais­ing fig­ures, and oth­er lag­ging in­dic­at­ors. This far out from an elec­tion, their pre­dict­ive value is lim­ited.

In­stead of trash talk­ing — I’m a big fan of the data-cent­ric sites, des­pite my cri­tique of their polit­ic­al pre­dic­tions — I figured I should put my money where my mouth is. Here are four ex­amples where my ana­lys­is runs con­trary to the mod­els’ early pro­jec­tions.

1. Re­pub­lic­ans are in bet­ter shape than you think in Iowa, and Demo­crats are in bet­ter shape than you think in Michigan.

The cur­rent con­ven­tion­al wis­dom says that former Michigan Sec­ret­ary of State Terri Lynn Land holds close to even odds in re­tak­ing Michigan’s Sen­ate seat for Re­pub­lic­ans. Sil­ver pegs the GOP’s odds at 45 per­cent, while the Up­shot team at The New York Times puts the GOP’s chances at 50/50. The mere fact that a Michigan Sen­ate race is com­pet­it­ive says a lot about the na­tion­al en­vir­on­ment, giv­en that the last time Re­pub­lic­ans won a Sen­ate elec­tion in the state was two dec­ades ago — Spen­cer Ab­ra­ham’s vic­tory in 1994.

Yet the bullish pre­dic­tions for Re­pub­lic­ans in Michigan are largely in re­sponse to polls show­ing Land run­ning com­pet­it­ively against ex­pec­ted Demo­crat­ic op­pon­ent Gary Peters and her sol­id fun­drais­ing early on. But at this point, the early polls are akin to a crude gen­er­ic-bal­lot test than any­thing re­flect­ing a cam­paign between two vet­er­an politi­cians. Most voters are un­fa­mil­i­ar with both can­did­ates des­pite their deep re­sumes. And the com­pet­it­ive polling is af­fected by the early mil­lions spent in anti-Obama­care ads by Amer­ic­ans for Prosper­ity against Peters, whose al­lies haven’t yet matched their spend­ing on air.

In Michigan, Re­pub­lic­ans have a long his­tory of see­ing highly touted statewide re­cruits (Dick De­Vos, Mike Bouchard, Pete Hoek­stra) fade in the wake of or­gan­ized labor’s ef­forts, which has a sol­id re­cord of turn­ing out Demo­crat­ic voters. If the highly touted in­vest­ment in turnout tech­niques suc­ceeds, it will have to work in Michigan — a pres­id­en­tial battle­ground with a siz­able share of Afric­an-Amer­ic­an votes and where Demo­crat­ic or­gan­izers have plenty of ex­per­i­ence.

And I’m not as sold on Land’s strength as a can­did­ate, simply based on her sol­id fun­drais­ing num­bers. It took a while for Sen­ate strategists in Wash­ing­ton to em­brace her cam­paign, and she has kept a low pro­file in the race. Peters may not be a par­tic­u­larly strong can­did­ate, either — the con­gress­man already re­placed his cam­paign man­ager — but has shown some savvy in win­ning a tough primary in 2012 and by vot­ing with House Re­pub­lic­ans to roll back ele­ments of Obama’s health care law. All told, Peters is more than a slight fa­vor­ite in the race, bar­ring a large Re­pub­lic­an wave.

By con­trast, Re­pub­lic­an state Sen. Joni Ernst is look­ing in­creas­ingly for­mid­able in the Iowa Sen­ate race, thanks to her com­pel­ling pro­file and rare abil­ity to unite the party’s es­tab­lish­ment with the grass­roots. A fe­male Ir­aq War vet­er­an, Ernst has struggled with fun­drais­ing but caught late mo­mentum thanks to a catchy ad ar­guing her hog-cas­trat­ing back­ground will lead her to cut pork in Wash­ing­ton. Mitt Rom­ney and Sarah Pal­in both en­dorsed her, and she has ta­cit sup­port from Gov. Terry Bran­stad. Ernst is run­ning neck and neck with GOP busi­ness­man Mark Jac­obs, who has out­spent her in the primary.

If she wins the nom­in­a­tion, Re­pub­lic­ans could boast a com­pel­ling fe­male con­tender — in a pre­dom­in­antly-white state where dis­ap­prov­al of Obama runs high. Mean­while, Bra­ley’s com­ments in­sult­ing farm­ers at a Texas fun­draiser are dam­aging and bound to hurt him when Re­pub­lic­an groups use the foot­age in cam­paign ads. The Up­shot pegs the GOP chances at a mere 14 per­cent, with Sil­ver at 25 per­cent. I’d put the GOP’s chances much closer to 50-50. (Even if the deep-pock­eted Jac­obs wins the nom­in­a­tion, Re­pub­lic­ans should be able to run com­pet­it­ively here, al­though Demo­crats would have more op­pos­i­tion re­search to util­ize against him.)

2. Sens. Jeff Merkley and Mark Warner are more vul­ner­able than the mod­els sug­gest.

As in Iowa, the mod­els struggle with as­sess­ments fea­tur­ing chal­lengers with low name iden­ti­fic­a­tion. Merkley and Warner are both favored to win, but by pre­dict­ing them as near locks for reelec­tion — the Up­shot puts their odds at 99 per­cent or more — the mod­els are un­der­es­tim­at­ing the strength of their less­er-known op­pon­ents.

Warner is a per­son­ally pop­u­lar sen­at­or and former gov­ernor, but his ap­prov­al rat­ings have been soft in re­cent polling. He leads Ed Gillespie, a former Re­pub­lic­an Na­tion­al Com­mit­tee chair­man who’s well-known in Wash­ing­ton but is vir­tu­ally an­onym­ous to most Vir­gin­ia voters. Yet Gillespie will have the re­sources to get his mes­sage out — he raised $2.2 mil­lion in the last quarter — in a battle­ground state that isn’t im­mune to the na­tion­al en­vir­on­ment. At this time in 2006, former Sen. George Al­len was be­ing touted as a fu­ture pres­id­en­tial can­did­ate; he lost in a wave elec­tion to Jim Webb. With signs that 2014 could be an­oth­er land­slide year, this time for Re­pub­lic­ans, it’s a race worth watch­ing.

An­oth­er sleep­er race is in Ore­gon, where Re­pub­lic­ans are likely to nom­in­ate phys­i­cian Mon­ica We­hby to chal­lenge fresh­man Sen. Merkley. Just watch We­hby’s in­tro­duct­ory ad and you can see why GOP strategists think she could pose a ser­i­ous chal­lenge to the sen­at­or. The prob­lems with Ore­gon’s health care ex­changes — state of­fi­cials just aban­doned the troubled sys­tem for the fed­er­al ex­changes — also makes the health care is­sue a loc­al one that a Re­pub­lic­an doc­tor could ef­fect­ively ex­ploit.

Merkley’s cam­paign re­leased their in­tern­al num­bers, show­ing the in­cum­bent at 52 per­cent against We­hby be­fore the primary. That’s a de­cent place to be, but far from safe ter­rit­ory — es­pe­cially if out­side con­ser­vat­ive groups de­cide to spend ag­gress­ively to tar­get the in­cum­bent.

3. Scott Brown gets no re­spect.

For a former sen­at­or whose entry in­to the race bolstered Re­pub­lic­an hopes of re­tak­ing the Sen­ate, the mod­els don’t like his chances. Like Sil­ver, I share his skep­ti­cism that Brown is as for­mid­able a can­did­ate as the early hype says he is. He’s run­ning against a well-liked sen­at­or and needs to over­come skep­ti­cism that he’s a car­pet­bag­ger. But his chances are bet­ter than the 12 per­cent that Up­shot gives him and the 25 per­cent chance that Sil­ver does.

New Hamp­shire is a state that fre­quently fol­lows the na­tion­al en­vir­on­ment — just look at the un­stable two-seat House del­eg­a­tion that flips with many re­cent wave elec­tions. The Sen­ate dy­nam­ic is sim­il­ar: John Sununu de­feated Jeanne Shaheen in 2002 — a good Re­pub­lic­an year — and Shaheen re­turned the fa­vor in 2008, a great Demo­crat­ic year. In the GOP wave elec­tion of 2010, GOP Sen. Kelly Ayotte crushed Demo­crat­ic Rep. Paul Hodes, who rep­res­en­ted half the state, by 23 points.

Brown won his first Sen­ate race rid­ing the na­tion­al wave against the pres­id­ent’s health care law, and he’s po­si­tioned to cap­it­al­ize on dis­sat­is­fac­tion with the law’s im­ple­ment­a­tion this time around. That doesn’t make him a fa­vor­ite against Shaheen, but it gives him more than a fight­ing chance.

4. Don’t bet against Sen. Mark Pry­or, but don’t over­es­tim­ate his chances, either.

The two mod­els hold wildly dif­fer­ent views on the state of play in the Arkan­sas Sen­ate race. Sil­ver is bullish on a Re­pub­lic­an pickup, giv­ing GOP Rep. Tom Cot­ton 70 per­cent odds to oust the Demo­crat­ic in­cum­bent, Mark Pry­or. But thanks to a re­cent New York Times/Kais­er Fam­ily Found­a­tion poll show­ing Pry­or with a double-di­git lead, the Up­shot rat­ings now view the Demo­crat­ic sen­at­or as the fa­vor­ite, with just 42 per­cent odds of los­ing. Part of the dif­fer­ence between the two is the mod­els’ re­li­ance on lim­ited pub­lic polling — Sil­ver’s mod­el was re­leased be­fore the fa­vor­able Pry­or poll came out.

But in real­ity, there’s not a whole lot of dif­fer­ence between the polls con­duc­ted in the race. All show Pry­or in trouble — stuck in the mid-40s, an un­com­fort­able spot for an in­cum­bent — but hold­ing a re­serve of per­son­al fa­vor­ab­il­ity that can sus­tain him through a tough en­vir­on­ment. And all show the na­tion­al en­vir­on­ment in Arkan­sas is treach­er­ous for any in­cum­bent Demo­crat — par­tic­u­larly one who voted for the pres­id­ent’s health care law.

In­deed, the New York Times poll that drove Re­pub­lic­ans batty — the RNC ac­tu­ally is­sued a state­ment con­demning the sur­vey meth­od­o­logy — con­tained lots of good news for Cot­ton. His “double-di­git” de­fi­cit was at­trib­ut­able to his lower name re­cog­ni­tion. As my col­league Ron­ald Brown­stein poin­ted out, Cot­ton is win­ning only about half of Obama­care dis­ap­provers in Arkan­sas, thanks to many people not know­ing who he is. That num­ber should rise con­sid­er­ably as his pro­file grows, and would close the gap.

At this point, this race is a toss-up, though Demo­crats will need to dis­qual­i­fy Cot­ton as an ac­cept­able chal­lenger to main­tain their lead.

What We're Following See More »
WH Never Released NIger Statement
42 minutes ago

"A staffer at the National Security Council drafted a statement of condolence for President Donald Trump to make almost immediately after a deadly ambush of U.S. soldiers in Niger earlier this month. But Trump never issued the statement, and, some two weeks later, is now in hot water over his initial silence on the soldiers’ deaths and alleged controversial comments he made to a widow of one of the dead."

2 US Navy Sailors Die of Drug Overdose Same Week
1 hours ago

"Two US sailors based at Naval Submarine Base Kings Bay in Georgia have died of apparent drug overdoses in the last week, according to a US Navy public affairs officer." The submarine force and Kings Bay leadership have ordered more drug tests and are taking the events very seriously.

U.S. Immigrant Population Hit 43.7 Million in 2016
2 hours ago
Second Lady Pence Rolls Out Initiative
3 hours ago
Five Shot at Maryland Office Park
5 hours ago

Welcome to National Journal!

You are currently accessing National Journal from IP access. Please login to access this feature. If you have any questions, please contact your Dedicated Advisor.