A few weeks ago, on the heels of the now infamous presidential polling misses in Wisconsin, Michigan, and elsewhere, we learned that the Field Poll – perhaps the most venerable polling organization in California for the last 60+ years – was forced to shut its doors. The proximity in time of these events was not a coincidence. Telephone pollsters, particularly at the state and local level, are becoming an endangered species. Indeed, only about a third of Californians really use a landline phone anymore, so calling those numbers has become pretty pointless. And as for cellphones, the frequent mismatch between area codes and actual places of residence means that drawing samples of folks living in a specific geographical area is becoming less a science than . . . well, it may not even be an art at this point.
These and other developments have sent survey response rates plummeting — and the costs of conducting a quality poll skyrocketing. Most media, politicians, and other public-minded organizations simply cannot afford the luxury of monitoring public opinion anymore. And even when pockets are deep enough to commission a quality poll, securing demographic representativeness between the sample and the population is far from assured. Hispanics, Millennials, and people without college degrees are much less likely to take surveys, and while pollsters apply massive statistical weights to overcome this problem, the effectiveness is debatable; the young people (for example) who respond to surveys tend to differ from those who don’t in important ways (interest, personality, time), so over-counting the former to make up for the latter can sometimes make things worse. Simply put, obtaining a “random probability sample” in statewide telephone polling is about as easy these days as getting a ticket to Hamilton.
The difficulties don’t end with finding survey respondents. We have also learned that the president-elect’s level of support was understated in many traditionally “blue state” phone polls because some respondents were embarrassed to tell survey interviewers (who are disproportionately young, female, and non-white) that they planned to vote for Trump – saying they were “undecided” even though they knew what they planned to do.
There is more to say, but the bottom line is that in ten years (if not sooner), we may be talking about telephone polls the way we do VHS tapes and Netscape. And this poses real challenges to democratic governance. Elections are way too blunt a mouthpiece for citizens to really make their voices heard on important issues of the day. If we want citizens to influence policymaking on a regular basis, we need valid and reliable polling. What is more, without good survey data, researchers’ ability to understand why people think and behave the way they do will be severely compromised.
Fortunately, the industry is responding to these challenges. While there are a litany of nonscientific (and therefore useless) internet polls, two approaches to conducting surveys online provide a lot of promise as we go forward. The first approach – embraced by several prominent research organizations at the national level and by Sacramento State’s CALSPEAKS (www.csus.csus.edu/isr/calspeaks) here in California — maintain “probability-based” sampling by sending postcards to a randomized list of California addresses and inviting residents to join a panel to take multiple surveys over time (by mail for those who lack internet access). This method saves money and reduces several types of survey bias. The second approach, embodied most prominently by YouGov (today.yougov.com), employs nontraditional sampling methods but uses highly technical procedures to create survey samples that are at least as representative as those using old-fashioned telephone methods.
In the same way that near universal telephone access revolutionized the industry a half-century ago, increasingly easy access to the internet (especially on mobile devices) is doing it again. The sooner we embrace these changes, the better we will be able to understand human behavior (and to predict it on Election Day). And public policymaking just might become more democratic as well.
David C. Barker is the director of the Institute for Social Research and Professor of Government at Sacramento State University.