POLL-ution

Eleven candidates took the stage at last Wednesday's GOP debate, but three unorthodox, outside-the-beltway candidates stole most of the headlines. In a field heavy on governors and senators, the CEO, neurosurgeon and casino magnate-turned-reality TV star are gaining attention and took the top three slots in CNN's latest poll, confounding the party establishment. But before you assume any of them are destined to win the Republican nomination-much less the White House-a word of caution: Early polling rarely resembles the final vote. This is likely even truer today than in years past, as polling has become increasingly challenging and unreliable. 2016's contest remains very much up in the air, making it way too early for investors to fear or cheer any one candidate's potential market impact.

Before Rick Perry dropped out of the race last week, 17 candidates were vying for the Republican nomination. While 24 million people tuned into the first GOP debate in August, we reckon the percentage of voters who can rattle off all 17 names and each candidate's last two jobs, general world view and top-three policy proposals rounds roughly to zero. It's summer, four months before the first primaries. This is still the season of road trips and vacations, not a time for buckling down on inside politics. Hence, early polling is based primarily on name recognition. We reckon a pollster's September 2015 interview with Sally Voter would go something like this:

Poll Person: Which of the following candidates do you support for the Republican nomination? In alphabetical order: Jeb Bush, Ben Carson, Chris Christie, Ted Cruz, Carly Fiorina, Jim Gilmore, Lindsey Graham, Mike Huckabee, Bobby Jindal, John Kasich, George Pataki, Rand Paul, Rick Perry, ... (takes deep breath) ... Marco Rubio, Rick Santorum, Donald Trump, Scott Walker.

Sally Voter: Wait what was the middle one?

Poll Person: Uhhhhh.

Sally Voter: Oh this is going to take way too long. Ummmmm, Trump. At least I know who he is. Are we done now?

This, of course, is a tongue-in-cheek hypothetical, but 2008's early polls illustrate the point. In summer 2007, polls showed Hillary Clinton and Rudy Giuliani leading their primary races. The former First Lady and the world-famous former New York mayor. Also high ranking then? Former Tennessee Sen. Fred Thompson, who also happened to spend five years serving as District Attorney on a little show called Law and Order. ^[i] Lest you think the polls are more accurate this time, the last year has been a colossal embarrassment for pollsters. Even surveys on an election's eve have been wildly off the mark. In Denmark, former Prime Minister Helle Thorning-Schmidt called a snap election in May, aiming to consolidate power with polls showing her center-left Social Democratic party gaining momentum over the center-right opposition Liberal Party. The day before the vote, polls showed her coalition slightly ahead, but on election day the opposition won, with the anti-EU, anti-immigration Danish People's Party winning the second-most seats. In the run-up to May's UK general election, polls showed a dead heat between Labour and the Conservatives, with neither poised to win a majority, yet the Conservatives won a 12-seat majority. In Israel, polls showed incumbent Prime Minister Benjamin Netanyahu potentially losing to challenger Isaac Herzog, yet he ran away with it. Before last year's US midterms, polls predicted the Republicans would win the Senate by the slimmest of margins and likely lose a few House seats-but they trounced expectations. But perhaps the biggest shock of all came in Greece's July referendum. Days before the vote on bailout terms, some polls showed "yes" with a slight edge. But "no" won by a whopping 22 percentage points.

Political scientists have done a lot of soul searching on these woeful polls, and theories/explanations abound. Perhaps the biggest hurdle to accuracy is technology. More and more people have ditched land lines and use their cell phones exclusively, making it difficult for pollsters to reach a sufficient number of people. Pollsters have historically used auto-dialers to reach the most people in the shortest time, but per the FCC's interpretation of the 1991 Telephone Consumer Protection Act, they can't auto-dial cell phones. Humans must manually dial each number, usually going through several before reaching another human who's willing to talk (and making polling more costly and time-consuming). Caller ID and visual voicemail add another wrinkle-people are less apt to answer calls from 800 or unknown numbers.

As a result, it is far more difficult for pollsters to gain a sufficient data sample that represents the overall voting population. Or even one that zeroes in on folks likeliest to vote. After all, if a majority of respondents say they lean towards candidate X, yet the ones who prefer candidate Y end up voting disproportionately more, the poll will whiff. To overcome this, pollsters ask a variety of questions (beyond the obvious "how likely are you to vote?") to sniff out likely voters, but this is more art than science. Different organizations use different questions to assess respondents' likelihood to vote, and it isn't clear any have found the magic set of questions. The line between likely and unlikely voters has become even more fuzzy of late, as fewer people respond to calls.

Online polling might seem like the obvious solution, but it presents problems, too. The people most likely to vote (older folks) use the Internet the least, and the people most likely to respond to online polls (younger folks) are generally the least likely to vote. There is also a greater potential for dirty data, as anyone in the world can theoretically vote in an online poll. Polling online also robs pollsters of the chance to target households in certain demographics, which political scientists regularly do to ensure the makeup of their polling sample is a rough proxy for the broader electorate. Some outfits try to solve this problem by polling voluntary pre-screened panels, but that removes the "random" sample element. People who sign up for polls and expect to be surveyed regularly probably don't represent the average voter. One day, e-polling might be a viable solution, but for now it is a work in progress.

The fact outsider candidates are polling so high is another reason to question the accuracy. Traditionally, people most likely to favor anti-establishment candidates are highly unlikely to vote in primary races, which by definition are establishment affairs, especially in caucus states. Support for out of the mainstream candidates tends to come not from party stalwarts and activists-those most likely to caucus or vote in primaries-but from the party's fringe or independents (or even those registered with other parties) who are tired of the old boys' network. In some states, closed primaries prevent the latter from participating, which could massively skew the results. On the other hand, those who trek out in the bitter cold to vote in early primaries and caucuses much more often represent the party's base who are less inclined to favor outsiders. Note, Saturday's Michigan straw poll looks very different from other recent polls, showing Kentucky Senator Rand Paul leading with 22%. (That probably isn't any more accurate, but a counterpoint is a counterpoint).

Sometimes polls are startlingly accurate but never see the light of day, due to a creeping trend known as "pollster herding." Polling agencies have an increasing tendency to discard results that differ too much from the consensus, fearing outlying results will make them look bad. After the UK election, Damian Lyons Lowe, founder and CEO of polling firm Survation, admitted he'd withheld a poll conducted the day before the election that showed the Conservatives with a six point lead over Labour: "The results seemed so 'out of line' with all the polling conducted by ourselves and our peers-what poll commentators would term an 'outlier'-that I 'chickened out' of publishing the figures-something I'm sure I'll always regret." Days earlier, Lord Ashcroft-a member of the House of Lords as well as a highly regarded UK pollster-published a poll that also showed the Tories ahead by six points. The commentariat quickly wrote it off as an outlier and potentially biased (Lord Ashchroft is a Conservative). Another agency, ICM, caveated its own mid-April poll showing the Conservatives at 39% vs. Labour's 33% by saying it might have rung up too many Conservatives. The more pundits disregard outlying results, the less incentive there is to publish them.

So whether you're thrilled or scared to bits by what the polls show, don't get too excited. It's way too early. Not all of the 16 Republicans still in the race^[ii] will be in the race come primary season, and there is no telling how supporters of soon-to-be-dropouts^[iii] will shift. Or whether the outsider leaders will keep their momentum. Or whether (and whom) gaffes will strike. The time will come to weigh the election's outcome and its potential market impact, but that time is not here yet.