Take the 'Macher' Approach to Polling

A recent blog item in the Washington Post about opinion polling brings to mind someone my Jewish mother would call a "macher," which, by the way, can be a Janus-faced label of both praise and ridicule. A macher can describe someone who has lots of connections, many of whom are in high places, or when accompanied by the modifier "big," as in "big macher," may describe someone who has delusions of grandeur. Either way, according to the math the Post persuasively provides, the macher can influence public opinion disproportionally because the opinions of those who are heavily networked carry more weight than those held by someone like me whose Facebook friends list is in the double digits. The Q score of a macher probably earns him free stuff all the time. This blog item (and a recent column in the New York Times Sunday Styles section) suggests the rapid change in attitudes toward gay marriage that galvanized the Supreme Court decision was, in part, a function of massively networked young people, whose attitudes toward sexuality are more relaxed than their older counterparts and who are more likely to have gay friends. Because people's opinions are actually not formed in a vacuum, but rather as a function of what they read and who they talk to, in this highly networked world, opinions can spread like the flu in mid-winter in Chicago.

The macher model of opinion formation is a violation of just about every principle opinion pollsters hold sacred. It essentially says you can predict the opinions of people in the future by knowing the opinions of the machers in the past. The more friends and family the machers have, the more influence their opinions will have. Most polling firms aim for the idealized "random" sample or stratified "random" sample by using some form of random digit dial (be it landline or cell) sampling frame. The basic statistical principle, called statistical independence, is at the root of random sampling and requires that the people you sample cannot share known characteristics (family, friends, neighborhood) because it will inflate the margin of error and make your estimates unstable. Pollsters would prefer you lived on Pluto -- by yourself. In fact, they assume you do.

What does that mean then for the accuracy of opinion polls? Not a great deal. Opinion polls take a stable photo of a single point in time and capture what people think on that day. If laws were made on that day or politicians were elected on that day, polls would be decent predictors of what happens. People are not so fickle as to change their minds in a day or two, but they do tend to change their minds in several weeks or months. Polls as predictors of the near future may, therefore, suffer the fate of Pluto -- they will soon be demoted to an inferior status. The opinion macher now works faster and with better tools as the digital world has put schmoozing on warp speed. Thus, it is hard to take a still photo of opinions because they are moving while you are focusing. A better mechanism for predicting opinions in the future may be to identify those folks with millions of Twitter followers and see what they have to say. Social media is actually the tool for collecting the data that mimics the way in which the data are created. As Marshall McLuhan said, lo these many years ago, the medium is the message.

The trick now for those engaged in opinion polling is to understand how to knit the machers together into a representative framework and to sort the "big macher" from the true macher. As we know from our own Facebook feeds, our networks are substantially segmented so that often we exclude people who don't agree with us. The question of how we fit all of these networks together and follow the representative opinion makers, who are now regular folks like you and I but with more "friends," is the interesting challenge that lies ahead.