Miller Center / Flickr

I met Andy Kohut nearly 30 years ago, when I was enlisted to join a small group of experts constructing a new way to look at the American electorate. Times Mirror, owners of The Los Angeles Times, Newsday, the Baltimore Sun, and other papers, had decided to invest handsomely to change the way polling and analysis of Americans had been done. This was the brainchild of a brilliant Times Mirror executive, Don Kellermann, who had once worked for Senator Jacob Javits of New York, along with his colleague Pat Butler, who had been a top assistant to Howard Baker of Tennessee.

They in turn had brought in Andy, who had been running Gallup, to shape the ideas and implement the surveys. And a small group of academics (me and my best friend Michael Robinson) and others from political and policy worlds worked alongside Andy, Don, and Pat for several years designing the program and its rollout. The initial result, in 1987, was a new typology of voters, using values along with party identification. It was imaginative and path breaking, and successive generations of typologies have followed, the most recent coming out last year. Soon after that initial success, Times Mirror created an ongoing operation, the Times Mirror Center for the People and the Press, headed by Kellermann, and Andy assumed the helm in the early 1990s.

The years spent devising the study, crafting questions, creating and describing voter types, analyzing the implications, and writing up the findings—not to mention presenting them all over the country, including at party-nominating conventions in 1988, were exciting, challenging, satisfying, and more fun for me than anything else I have ever done professionally. Our group has gotten together periodically to reminisce and celebrate our friendship and our successes.

I like to think I played a part in it all. But the frank reality is that Andy Kohut was the driver here. Andy brought imagination, discipline, integrity, and indefatigability to the project. Don Kellermann and Times Mirror got the study and the Center going, and provided incredibly generous funding and backing—enabling us to do thousands of face-to-face interviews, unheard of in this day and age. But the typology and design of the survey came from Andy, and the execution, making it a model survey among surveys, was a tribute to him.

The Times Mirror Center became a casualty of Times Mirror’s struggles in the new age of newspapers and media. It was not clear it would survive, until Rebecca Rimel and the Pew Charitable Trusts picked it up and transformed it in 1996 into the Pew Research Center on the People and the Press, now simply the Pew Research Center, with Andy at its helm. Along the way, Andy positioned the Center at the leading edge of research on global public opinion, as well as that of the United States.

It should come as no surprise to anyone sentient who follows American politics that polling itself is in crisis mode. Any fly-by-night hustler, or any ambitious consultant or institution, can hang up a shingle that says “Acme Polling,” release results, and have a good chance of having the results picked up by hungry news organizations that don’t care whether Acme Polling is any better than the Acme Company that outfitted Wile E. Coyote, much less equivalent to reputable surveys. A lot of surveys are done by partisans, and—surprise—release results that are stunningly favorable to their candidates or the companies or industries that pay them. For telephone surveys, response rates continue to drop (they are now below 9 percent). Most surveys can’t afford full and robust samples of cell phones to complement randomly dialed landlines. But the huge swaths of the population that no longer use landlines, the widespread revulsion against unsolicited phone calls, and the lack of regular availability of people at set hours have combined to create major problems for pollsters trying to conduct a survey overnight, or over a couple of days. As costs have gone up, many pollsters have turned to automated poll calls, adding to the unreliability and instability of the results they produce. Internet surveys may be the wave of the future, but have their own challenges. Other “pollsters” dispense with surveys and do “focus groups,” as if twenty or thirty people in a room are enough to divine public opinion. Then there is another reality, pointed out decades ago in a brilliant paper by my late mentor Phil Converse, that we often measure “non-opinions,” assuming that there is some knowledge base when respondents answer questions.

Doing polls under these conditions requires more art than science. It means weighting the surveys so that they can approximate proportions of people, by gender, age, income, race, partisanship, in the population being surveyed. Many good models are off by enough to skew the results—but that nuance does not affect the headlines or analysis. For example, a bunch of recent Quinnipiac polls had samples that were in many cases older and whiter than the likely 2016 general election voting population—meaning poorer numbers for Hillary Clinton than a more accurate count would have given. But one had to dig very deeply to discover that, and no analysis of the results in states like Pennsylvania, Ohio, and Florida reflected that.

That is just sampling. We also know that question wording, question placement, and many other factors can have a huge impact on survey results. Approval and disapproval of the recent Iran deal, for example, varied widely depending on the how the questions were worded.

I raise all of this because Andy Kohut, through the company he created and ran with his wife, Diane Colasanto, and through his work for Times Mirror and Pew, was not just the gold standard for public-opinion polling—he was the platinum standard. Andy knew everything about the realities and problems with polling. He insisted on rigid standards, and honed the questions and survey formats to be as accurate and as utterly objective as possible. He used some of the resources at Pew to examine what kinds of instruments work and don’t work in surveys, and to keep the industry as honest and competent as possible. There are good pollsters and good polling organizations out there, and a lot of smart and conscientious people in the profession. But no one, frankly, came close to Andy—and no organization even begins to rival the Pew Research Center he built.

Andy died from leukemia on Tuesday morning, after a long battle. For his friends, who treasured his humor, decency, integrity, and family, his loss is a terrible blow. But for the broader American society, trying to make sense of where the public is and what it really stands for at a time of enormous political turmoil, his loss cuts even deeper.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.