Polling, like everything else, is moving online

It turns out initial Markey-Kennedy survey was no outlier

HE’S NOT EVEN officially running, but we now have two polls showing Rep. Joe Kennedy III with a sizable lead over Sen. Ed Markey in the Democratic primary for US Senate. The first, from Change Research, showed Kennedy with an eye-popping 17 point lead. That margin raised questions about how seriously to take an online poll from a newer pollster using unfamiliar methods. But over the weekend, a Boston Globe / Suffolk University poll found a similar dynamic, showing Kennedy with a 14 point lead head-to-head against Markey and a 9 point edge when all candidates were included.

The pair of polls illustrates two important lessons in polling. First, online polls are here to stay and cannot be ignored. Second, it’s impossible to tell an outlier from the start of a new trend.

The reluctance to embrace unfamiliar online polls is understandable. The track record for most online pollsters is still very short, and methods are almost as numerous as the pollsters themselves. Some, like YouGov, have been at it for a while, and have a track record of reliability for their approach. Others, like Change, are newer, and we don’t have enough to really know what to expect from them. What we have seen from Change — several presidential primary polls showing Elizabeth Warren outpolling Joe Biden — is quite different from virtually everyone else, phone or online.

Although their name and their methods are new, Change should not be easily dismissed. They employ a staff of serious people and have major investment dollars and supporters behind them. They have signed up candidates and campaigns across the country, and are beginning to form media partnerships for polling. We had Change’s chief growth officer on The Horse Race last week to discuss the Senate polls and more.

The Globe poll suggests the Change survey is at least close to the mark. The Globe crosstabs also provide a tidy explanation for Markey’s woes. His favorables are at a tepid 59 percent among likely Democratic primary voters, who should be his strongest supporters. Kennedy, on the other hand, is viewed favorably by 73 percent of the same group. Just 6 percent hold an unfavorable view. Kennedy’s appeal extends beyond the restless youth that have propelled many challengers to office. In fact, the 38-year-old’s support is highest among voters over age 65, although support varies only a little by age.

But even without a second poll showing Kennedy ahead, dismissing unexpected results as outliers is a tricky business. It’s impossible to tell at the moment a poll comes out whether a poll is just off the mark, or the start of a new trend. This issue sparked controversy in the presidential primary last month, when Joe Biden’s pollster took issue with a reputable university poll publishing a self-described outlier result showing Biden with less of a lead than most other polls. We have had our own fair share of eyebrow-raising results, and we talked about what we did with them on The Horse Race a few weeks back.

Our bottom line has been: unless there is some problem with the data, pollsters should publish their unusual looking results, even if they seem to be outliers. The alternative leads to pollsters “herding” towards a consensus view of a race. When everyone looks over their shoulders, nobody sees what’s ahead.

And very often, an unusual or unexpected result is showing something new about the electorate. In 2016, we found New Hampshire voters swinging hard toward Trump in the final days after a steady Clinton lead for months. We also were the first to show voters souring on the charter school expansion question ballot question, which went from 25 points up to 25 points down in the months leading up to voting.

Outliers and all, online polling is here to stay. Very low response rates and rising costs mean the days of live interviewer telephone polling as the clear gold standard are over. Today, a significant portion of elections and policy polling is done using a rapidly proliferating number of online polling methods. Many media outlets have current or recent partnerships with online pollsters such as YouGov, Morning Consult, Ipsos, and others.

We at The MassINC Polling Group already produced a lot of online polling, both for private clients and public consumption. We have conducted extensive side-by-side experiments and have found specific online methods to be comparable to telephone polling on a variety of measures. We are now confident enough in our approach to recommend it freely and publish the results without hesitation.

Meet the Author

Steve Koczela

President, MassINC Polling Group

About Steve Koczela

Steve Koczela is the President of The MassINC Polling Group, where he has grown the organization from its infancy to a nationally known and respected polling provider. During the 2014 election cycle, MPG conducted election polling for WBUR, the continuation of a three-year partnership. Koczela again led the endeavor, producing polls which came within one point of the margin in both the Massachusetts gubernatorial and U.S. Senate Elections. He was also lead writer for Poll Vault, WBUR’s political reporting section during the 2014 Election Cycle.

He has led survey research programs for the U.S. Department of State in Iraq, in key states for President Barack Obama’s 2008 campaign, and has conducted surveys and polls on behalf of many private corporations. Koczela brings a deep understanding of the foundations of public opinion and a wide ranging methodological expertise. He earned U.S. Department of State recognition for his leading edge work on sample evaluation in post conflict areas using geospatial systems.

Koczela is frequent guest on WBUR as well as many other news and talk programs in Massachusetts and elsewhere. His polling analysis is often cited in local, state, and national media outlets. He currently serves as President of the New England Chapter of the American Association for Public Opinion Research (NEAAPOR). Koczela holds a Master’s degree in Marketing Research from the University of Wisconsin and is a veteran of the war in Iraq.

About Steve Koczela

Steve Koczela is the President of The MassINC Polling Group, where he has grown the organization from its infancy to a nationally known and respected polling provider. During the 2014 election cycle, MPG conducted election polling for WBUR, the continuation of a three-year partnership. Koczela again led the endeavor, producing polls which came within one point of the margin in both the Massachusetts gubernatorial and U.S. Senate Elections. He was also lead writer for Poll Vault, WBUR’s political reporting section during the 2014 Election Cycle.

He has led survey research programs for the U.S. Department of State in Iraq, in key states for President Barack Obama’s 2008 campaign, and has conducted surveys and polls on behalf of many private corporations. Koczela brings a deep understanding of the foundations of public opinion and a wide ranging methodological expertise. He earned U.S. Department of State recognition for his leading edge work on sample evaluation in post conflict areas using geospatial systems.

Koczela is frequent guest on WBUR as well as many other news and talk programs in Massachusetts and elsewhere. His polling analysis is often cited in local, state, and national media outlets. He currently serves as President of the New England Chapter of the American Association for Public Opinion Research (NEAAPOR). Koczela holds a Master’s degree in Marketing Research from the University of Wisconsin and is a veteran of the war in Iraq.

Meet the Author

Richard Parr

Research Dir., MassINC Polling Group
Though it’s far too early to know how the Senate race will unfold, we are now pretty confident in its current shape. As for the race for the future of polling, both online and phone polling are viable, and should each be taken seriously.

Steve Koczela and Rich Parr are the president and research director of the MassINC Polling Group. The polling group is a for-profit subsidiary of MassINC, the publisher of CommonWealth.