Mellman: Parsing the polls

Mellman: Parsing the polls
© Greg Nash

In recent days, two sets of very smart people have disputed the nature and meaning of current polls.

Rather than get myself in trouble by wading directly into the conflict, I’ll confine myself to offering a few thoughts on some of the issues that have been raised.

ADVERTISEMENT
First, there is little doubt the presidential race has tightened considerably. Ignore the individual polls for a moment and consult the averages.

Huffpost Pollster’s model gives Democratic presidential front-runner Hillary ClintonHillary Diane Rodham ClintonChris Murphy’s profile rises with gun tragedies DNC, RNC step up cyber protections Gun proposal picks up GOP support MORE a lead of 1.6 percentage points as of Tuesday. That compares to her lead of 10 points at the beginning of April. RealClearPolitics used a straightforward average and gave presumptive GOP nominee Donald TrumpDonald John TrumpIvanka Trump pens op-ed on kindergartners learning tech Bharara, Yates tamp down expectations Mueller will bring criminal charges Overnight Cybersecurity: Equifax security employee left after breach | Lawmakers float bill to reform warrantless surveillance | Intel leaders keeping collusion probe open MORE a 0.2 percentage point advantage. In mid-April, its method gave Clinton an 11-point margin.

But today’s tight race does not necessarily imply a tight race in November. As we get of tired of saying and hearing, polls are a snapshot in time, not a prediction of what will happen in six months. 

Indeed, we are now in an unusual period in which we have come to expect a candidate still in the midst of the primary process to be experiencing difficulty relative to someone who has “won” his nomination.

In 2008, Sen. John McCainJohn Sidney McCainRubio asks Army to kick out West Point grad with pro-communist posts The VA's woes cannot be pinned on any singular administration Overnight Defense: Mattis offers support for Iran deal | McCain blocks nominees over Afghanistan strategy | Trump, Tillerson spilt raises new questions about N. Korea policy MORE (Ariz.) clinched the GOP nomination in March and was leading Barack ObamaBarack Hussein ObamaAll five living former presidents to attend hurricane relief concert Overnight Health Care: Schumer calls for tying ObamaCare fix to children's health insurance | Puerto Rico's water woes worsen | Dems plead for nursing home residents' right to sue Interior moves to delay Obama’s methane leak rule MORE, who was still battling Clinton for the Democratic nod. Obviously, despite his March lead, McCain didn’t win in November.

One of the reasons for this phenomenon is that candidates consolidate their parties after becoming the presumptive nominee. We see that among elites, as Speaker Paul RyanPaul RyanThe Hill Interview: Budget Chair Black sticks around for now Gun proposal picks up GOP support GOP lawmaker Tim Murphy to retire at end of term MORE and even Sen. Lindsay Graham (R-S.C.) curtail their hostility toward Trump.

That process is reflected in the data as well. I’m looking at states where Trump has picked up a net of 14 points among Republicans. The increase was a net of 16 among Republicans nationally in the ABC/Washington Post poll. 

While it’s been suggested the recent polls offer radically different results, they really don’t. 

In the last five major national surveys, Clinton has garnered between 42 percent and 48 percent of the vote, close to the margin of error for comparing these polls. On average, these five put her at about 45.5 percent, well within the appropriate margin of error.

Meanwhile, these same polls put Trump at between 41 percent and 46 percent of the vote, for an average of 44 percent. Again, it’s a number well within the margin of error. 

There’s no huge variation, especially after focusing on each candidate’s relationship to the averages.

Some disputants have also veered dangerously close to the famously foolish “unskewers” of cycles past, noting what seems to them like errors within individual subgroups.

In giving Trump 28 percent of the Hispanic vote, they exclaimed, the NBC/SurveyMonkey poll was wildly off. SurveyMonkey’s methodology is still experimental, in my view, but it put the critics’ charge in context — think about its real-world impact.

Pew Hispanic Center analyses of exit poll data indicate that Republican candidates have garnered between 21 percent and 40 percent of the Hispanic vote since 1980. Mitt Romney, with his focus on self-deportation, got the backing of 27 percent in as the GOP nominee 2012, and McCain received 31 percent.

In that context, 28 percent seems high, but not bizarrely out of sync.

But what difference does it make? Precious little. 

Say the critics are correct, and Trump is really only getting 13 percent of the Hispanic vote. That would reduce his national total vote by just 1.5 percentage points — far less than most surveys’ margins of error.

Battling over 1 or 2 points, or even 5, in the May polls is not a particularly informative exercise. 

The race looks close now, but let’s check back again after Clinton clinches her nomination, and then after the conventions, before we start making fine-grained predictions about the outcome. 

Mellman is president of The Mellman Group and has worked for Democratic candidates and causes since 1982. Current clients include the minority leader of the Senate and the Democratic whip in the House.