Dur­ing the last pres­id­en­tial elec­tion, Demo­crats mocked Re­pub­lic­an com­ment­at­ors for sug­gest­ing that pub­lic polling show­ing Pres­id­ent Obama lead­ing Mitt Rom­ney was flawed be­cause more Demo­crats than Re­pub­lic­ans were in­ter­viewed. Con­ser­vat­ives ques­tioned the motives of poll­sters and the me­dia out­lets that com­mis­sioned the sur­veys for pub­lish­ing (they said) such ob­vi­ously biased res­ults, and web­sites sprang up that “un­skewed” the num­bers to re­flect what Re­pub­lic­ans thought was closer to real­ity.

Ul­ti­mately, their cri­ti­cism was un­foun­ded. The polls cor­rectly pre­dicted Obama’s vic­tory.

But it turns out that GOP cri­tique may have come two years too early. Demo­crat­ic and Re­pub­lic­an poll­sters alike agree that most of the pub­lic sur­veys on the big 2014 con­gres­sion­al races are un­der­es­tim­at­ing the level of Re­pub­lic­an sup­port in a midterm elec­tion year, which tends to be more con­ser­vat­ive than the rest of the vot­ing-age pop­u­la­tion.

That’s be­cause most pub­lic polls con­duc­ted for me­dia out­lets or by aca­dem­ics are sur­vey­ing the en­tire uni­verse of re­gistered voters, with little re­gard for wheth­er those voters will ac­tu­ally cast a bal­lot on Elec­tion Day. And in the past few midterm elec­tions, Demo­crat­ic-lean­ing voters haven’t turned out at close to the same rates as those who typ­ic­ally back GOP can­did­ates.

Cam­paign poll­sters, on the oth­er hand, con­tact only those who are likely to vote.

“We know a lot about the dif­fer­ence between” the likely midterm elect­or­ate and the over­all pool of voters, said Demo­crat­ic poll­ster John An­za­lone, whose firm works for scores of Demo­crat­ic can­did­ates run­ning this year. “It’s gonna be older, it’s gonna be whiter, it’s gonna be more Re­pub­lic­an.”

Re­pub­lic­an poll­ster Glen Bol­ger offered vir­tu­ally the same as­sess­ment. “Likely voters tend to be a little older, a little bit more Re­pub­lic­an, a little more white,” he said. “And that’s the nature of the elect­or­ate, par­tic­u­larly in non­pres­id­en­tial elec­tions. Re­gistered voters are a little more likely to match a pres­id­en­tial-year look.”

Exit polls aren’t in­fal­lible meas­ures of the com­pos­i­tion of the elect­or­ate, but they are in­struct­ive in show­ing the dif­fer­ence between a pres­id­en­tial- and midterm-level turnout. In 2006, 79 per­cent of voters were white. That dropped to 74 per­cent in 2008, but jumped back up to 77 per­cent in the 2010 midterms. In 2012, just 72 per­cent of voters were white, a re­cord low.

Young­er voters, in par­tic­u­lar, drop off in midterm years. Voters un­der 30 made up 12 per­cent of the 2006 and 2010 elect­or­ates, com­pared with 18 and 19 per­cent of the elect­or­ate in the last two pres­id­en­tial elec­tions, re­spect­ively.

But pub­lic poll­sters ar­gue that their sur­veys at this stage of the cam­paign aren’t meant to be pre­dict­ive. They are a snap­shot of where the elect­or­ate stands now, and just be­cause the midterm elect­or­ate has his­tor­ic­ally been older and whiter than in pres­id­en­tial years doesn’t mean it’s cor­rect to as­sume it will hap­pen again this year.

“His­tor­ic­ally, the elect­or­ate is more Re­pub­lic­an in off-year elec­tions, it’s whiter, it’s older,” said Doug Schwartz, dir­ect­or of the Quin­nipi­ac Uni­versity Polling In­sti­tute in Con­necti­c­ut. “First of all, you don’t know for sure that those his­tor­ic­al pat­terns are go­ing to hold up, and even if they do hold up, you don’t know how much more Re­pub­lic­an, how older, how much more white” the elect­or­ate will be.

Schwartz says Quin­nipi­ac, which will be sur­vey­ing the com­pet­it­ive gubernat­ori­al races in Con­necti­c­ut, Flor­ida, and Pennsylvania, among oth­ers, won’t start screen­ing for likely voters un­til after the sum­mer.

“The main reas­on is that voters are not really tun­ing in closely to the cam­paigns un­til after Labor Day,” he said. “In a lot of races, you don’t have the two can­did­ates set yet…. You don’t know who’s go­ing to be a likely voter un­til that time. It’s too early to as­sess who’s go­ing to be likely to vote.”

While that’s true, it also means that plenty of voters who won’t cast bal­lots are be­ing in­cluded in these sur­veys — and that Re­pub­lic­an can­did­ates might be in stronger po­s­i­tions than pub­lic polls in­dic­ate. In Oc­to­ber 2010, the re­spec­ted NBC/Wall Street Journ­al poll showed more re­gistered voters favored a Con­gress con­trolled by Demo­crats, by a 2-point mar­gin. In Novem­ber, Re­pub­lic­ans romped to a his­tor­ic midterm land­slide that the sur­vey didn’t an­ti­cip­ate. Like­wise, the same sur­vey this month showed Demo­crats with the same 2-point edge.

Demo­crats aren’t auto­mat­ic­ally doomed be­cause polls of re­gistered voters overrep­res­ent voters who won’t turn out on Elec­tion Day. The demo­graph­ics of the 2010 elect­or­ate looked just like 2006, but the res­ults were vastly dif­fer­ent. Some Demo­crats also point to the Obama cam­paign’s soph­ist­ic­ated turnout op­er­a­tion, and the chances that ef­fort could com­pensate — at least in part — for the drop-off that usu­ally oc­curs in midterms.

So how can in­ter­ested ob­serv­ers ac­count for this dis­crep­ancy? For one, polit­ic­al pro­fes­sion­als pay less at­ten­tion to the pub­lic sur­veys than most would be­lieve, es­pe­cially judging by the volume of press re­leases gen­er­ated by the can­did­ate lead­ing in such a sur­vey.

“I nev­er get a false sense of hope from some pub­lic poll,” said An­za­lone, the Demo­crat­ic poll­ster. “As a pro­fes­sion­al, the only thing I look at is Pew,” re­fer­ring to the Pew Re­search Cen­ter’s na­tion­al polling on the gen­er­ic House bal­lot and oth­er is­sues.

And us­ing a crude in­stru­ment to add points to the GOP can­did­ate’s vote share to re­flect an ar­bit­rary turnout tar­get — see 2012’s “Un­skewed Polls” move­ment — isn’t the an­swer, either.

Still, if re­cent his­tory holds, when pub­lic poll­sters start screen­ing for likely voters, those track­ing these races should ex­pect a shift to­ward Re­pub­lic­ans. Un­til then, pub­lic sur­veys are best viewed as snap­shots of the over­all elect­or­ate, with­in which par­tis­ans on each side can de­term­ine at what turnout levels their can­did­ates might pre­vail.

Un­skew away.