‘Big Data’ threatens poll standards

Campaign polling is crossing over into dangerous territory as it meets “Big Data,” the latest craze in campaigning.

The Obama campaigns of 2008 and 2012 secretly delved into the delights of data-based voter targeting and media buying, but now the operatives that ran Obama’s cloistered “cave” operation are cashing in, working for casinos and such, disclosing and utilizing the methods they developed for voters to sway consumers.

The hitch for political pollsters is that these new methods threaten to trample on privacy and confidentiality, as traditionally practiced by  survey researchers.

The developing urban legend of the Big Data evolution is that Democrats, stung by Republicans’ great leap forward in microtargeting several cycles back, resolved to build a better mousetrap.

The Democrats denigrated microtargeting that associated voter records with consumer data like car ownership, saying that “we don’t want to sell a car.”

Instead, Democrats developed an unprecedented enormous list of voters they considered persuadable — in terms of candidate choice and turnout — and set out to develop other data that would help them communicate with those persuadable voters, through everything from Facebook friends to oddball cable TV programs.

In doing all this, the Democrats recognized some potential ethical issues. They were operating very close to the line in violating the spirit, if not the letter, of Facebook’s privacy guidelines and standards.

They supposedly took steps to cloak the identities of TV households — those with set-top boxes that record shows watched — that also turned up in their persuadable voter file. This is all fine.

But where did the original list of persuadable voters come from? My guess is that it was from an amalgam of poll and phone bank data. This is the ethical hole in Big Data’s bucket. Unless poll and phone bank respondents were explicitly told that their information would be used in such a manner, and the respondents granted their permission, use of such data is illicit.

When a pollster calls a household, there is an implicit understanding, or at least there has been and should be, that the call is confidential, that the answers an individual provides will only be reported and used in the aggregate, combined with responses of all others polled.

Pollsters who are members of the American Association for Public Opinion Research and other professional survey organizations acknowledge annually their commitment to this standard of confidentiality. This is not simply a matter of being fussy professionally. It is sound business practice.

If poll respondents believe that we are going to deposit their individual responses in a data silo to be used later to persuade them how to vote, they will stop doing interviews,  and the randomness of our samples will be compromised. Confidentiality must be preserved to sustain the broad-based cooperation of the public.

A long-standing fly in the ointment has been “phone banks” hired by campaigns and party committees that pose as pollsters to collect data like candidate preference, ideology, likelihood of voting and issue preference. These faux polls are directed not at random samples but at whole electorates, and the results are explicitly headed for databases used to target individual voters.

I don’t care for these operations, but campaigners love them. My objection is muted because the typical voter can hear the difference between a legitimate poll interview and one done by a phone bank for targeting. The latter often sounds more like telemarketing,. It’s clear that some voters are casual about protecting their privacy.

In the age of social media, sharing is in with many. But as long as others want their privacy protected, we should oblige, especially if it protects the integrity of our work product in poll results.

Pollsters must not succumb to Big Data’s entreaties to sell out our principles.

Hill is a pollster who has worked for Republican campaigns and causes since 1984.