2003 - NEW YORK AAPOR OUTSTANDING ACHIEVEMENT AWARD
THE EDUCATION OF A POLLSTER
Remarks by Michael Kagay
Thank you very much for the award. I am honored and pleased.
My remarks tonight are entitled "The Education of a Pollster." They are my way of acknowledging and thanking many of the people I have worked with over the past 35 years, and from whom I've learned so much over the years.
WISCONSIN YEARS
When my wife and I got married we went into the Peace Corps for two years, and served in India. Near the end of the two years, as we contemplated going to graduate school, we realized that we needed a university that was strong both in English literature -- her field -- and political science -- my field -- and that also answered mail from India. To our surprise there were some major universities whose names you would recognize who did not answer mail from India. In the end, the University of Wisconsin fit all the criteria, so in the fall of 1967, without Carol or I ever having set foot before in the state of Wisconsin, we rented a van and moved to Madison.
It turned out that the University of Wisconsin had a terrific department of political science -- one that allowed its graduates to leave with their love of politics still intact, rather that having had the life squeezed out of it.
It took some time for me to decide what branch of political science to specialize in. I knew I was most interested in American politics, rather than, say, international affairs. For a while I thought I might specialize in Congress; I had always enjoyed talking to politicians and I figured it would be interesting to study how they behave in groups. And Congress was a hot field just then. But partly because it was hot, the faculty members who taught the courses on Congress always seemed to be off in Washington pursuing their research and weren't always available to students.
So, quite luckily, I fell in with Professors Jack Dennis and Donald McCrone, who specialized in political behavior, public opinion, socialization and survey research. In their research they often used the Wisconsin Poll, a twice-yearly survey of Wisconsin residents the University did on behalf of the state. Half the questions were always reserved for the farmers in the sample -- their expectations for the next year's crop and so forth -- but the full sample represented all Wisconsin residents and the other half of the questions were up for grabs, so faculty members competed to get their questions on the poll. And because Dennis and McCrone used the Wisconsin Poll, they were usually on campus and available to graduate students.
I had done secondary analysis of other people's surveys before, but the first primary analysis I ever did of any survey data was from the Wisconsin Poll. And I remember thinking at the time that that was an attractive combination of quantitative methods and real people. The quantitative part of it meant that we could generalize our findings and project them to a larger population; we were not just telling biographical stories about individuals. And the emphasis on real people guaranteed that we never wandered too far from the opinions and behavior of actual human beings.
I was hooked, and have been for 35 years.
Perhaps my most notable accomplishment in graduate school was in satisfying my foreign language requirement with the FORTRAN computer language. I had been dreading doing a refresher course in German when the graduate school, in the nick of time from my point of view, decided to leave the particulars of the language requirement up to individual departments. The political science department, to its credit and my relief, approved my proposal of FORTRAN.
At that time many of you will recall, general-purpose computer programs such as SPSS and the like were not yet on the market. There were local ones like OSIRIS at Michigan, but they weren't generally available. And even with programs that did exist, it was often difficult to recode your data or manipulate it before doing a statistical test, so you often had to write your own programs to do that and FORTRAN was the language of the time. In the years after my doctorate, I certainly used FORTRAN far more than I would ever possibly have used German.
MICHIGAN SOJOURN
The second stop in this educational journey was the University of Michigan at Ann Arbor. Political scientists, who aren't political scientists for nothing, had just convinced the National Science Foundation to make them eligible for NSF post-doctoral fellowships. I won one of the first batch to be awarded, which contained a grant which would support you for a year and which you could tenure at any institution you could make a case for.
The way I figured it, if you're going into survey research and political behavior, at some point you need to see how they operated at the University of Michigan, where the Center for Political Studies at the Institute for Social Research had for 20 years been conducting the National Election Study.
There are many other political polls and surveys, but they are the frigates and PT boats of the survey research navy. I've always thought of the National Election Study as the nuclear aircraft carrier of survey research. It's huge in size, slow to change course, and can't turn on a dime with new political events, but it's loaded with every analytic weapon system the admirals in Ann Arbor can devise.
I'm very grateful to Warren Miller and his colleagues for taking me on for a post-doctoral year. They didn't really have any obligation to me since I wasn't one of their degree students, but they were generous with their time and I made many friends there. One example is Helmut Norpoth of SUNY Stony Brook, who is here tonight. Helmut and I shared an office at Michigan in 1971 and 1972 and we've remained colleagues now for 30 years.
At Michigan I learned some of what's involved in the operation of a major research center. I had known of Warren Miller principally through his publications, and I was both surprised and impressed that at this point in his career, Warren was spending most of his time in raising money through grant applications to keep the whole operation going.
ACADEMIC POLLING AT PRINCETON
The next stop in my educational venture was Princeton University, where I spent the next ten years on the faculty and administration. I learned much about teaching from Stanley Kelley Jr. and what little I know about administration from Donald Stokes, who came from Michigan to be the dean of the Woodrow Wilson School.
Back in the 1930's and 1940's, Princeton had been a major player in public opinion and polling -- when Hadley Cantril was in the psychology department and Harwood Childs was in the politics department. Public Opinion Quarterly had been founded, edited and published there in its early decades. The archive of Gallup data that became the Roper Center was initially at Princeton.
But by the 1960's, all that had faded. Public Opinion Quarterly had gone to Columbia University, the data archives had gone to Williams College and later the University of Connecticut. Edward Tufte, one of my senior colleagues, had inherited the public opinion course, but in Harwood Childs's later years it had become a gut. Tufte once confided that the first time he taught the course, in an amphitheater where each chair had a desk in front of it, he looked up to see 50 pairs of athletic shoes in his face. Tufte recommended the course be dropped entirely from the catalog, which was done. Public Opinion at Princeton had come to a temporary end.
So by the time I arrived, much mending of fences needed to be done. I discovered, for example, that almost none of the faculty then at the University even knew any of the researchers at the public opinion firms in town, and vice versa. We took a couple steps to deal with this. The first was to create the Central New Jersey Chapter of AAPOR as a way of encouraging interchange between faculty, graduate students and researchers at commercial firms.
The second was involving local researchers in teaching. Irving Crespi, for example, who was vice president of Gallup, helped me teach one year, and Harry O'Neill of the Opinion Research Corporation, and his colleague Ken Schwartz, also of ORC, and I jointly taught a seminar for two years running in which students designed a survey and interpreted the results while ORC did the heavy lifting of sampling, interviewing and data processing. Thirteen students in the 1981 version of that seminar won the national AAPOR student award for the report they wrote interpreting the survey. All of them subsequently went into the legal profession, except for one, Keating Holland, who is now with the CNN polling unit in Washington.
And did I mention all the terrific students I had at Princeton? A teacher inevitably learns a great deal from his students. And these were some of the best.
We reintroduced the public opinion course -- I taught it nine times in ten years -- and it became popular, but I had quite jolt in the late 1970's when I watched the National Invitational Tournament basketball game on television and realized that the entire first string varsity basketball team was taking my course that semester. I thought, "My God, has it become a gut again?" So I made some discreet inquiries: were there athletes from any other sports in the course? The answer was no. The basketball team, who were very bright, traditionally took many of the same courses, so they could study together on their long road trips.
At Princeton my research projects focused mainly on voter turnout and voting behavior, how voters make up their minds and changes over the decades in the importance of issues.
HARRIS: THE COMMERCIAL SIDE OF POLLING
The next stop in the education of this pollster was at Louis Harris and Associates in New York City where I worked for five years. Harris bore the burden of converting me from an academic pollster to a commercial one, retooling me so to speak. And I want to thank Humphrey Taylor and David Krane and all their colleagues at Harris from whom I learned how to write fast (or at least faster) and how to keep projects on budget (or at least close to budget). Those were valuable lessons that I have often drawn on since.
Some of the projects we did at Harris at this time included access to health care, problems facing the elderly, public school teachers, and disabled Americans.
POLLING FOR NEWS WITH THE NEW YORK TIMES
The next chapter takes place at the New York Times. My first contact with the Times was early in 1977 when they hired me as their academic polling consultant while I was still at Princeton University. I was hired by Henry Lieberman, who looked like a newsman of the old school. His most recognizable characteristic was the stubby cigar he always carried in his left hand. When I met him he was in his 60's and his doctor no longer allowed him to smoke the cigar but he would chew on it occasionally and I never saw him without it. But Henry was a new breed of newsman. He was trained as an engineer, most unusual for a newsman, had been a foreign correspondent in China, had been the science editor, and oversaw all the Times's coverage of the first manned landing on the moon. He had a home-brewed computer in his house, and now he was in charge of increasing the amount of quantitative material in the newspaper.
One way he did this was to put The Times Best-Seller List on a scientific basis with a national sample of bookstores, weekly interviews with an employee at each store, and electronic data processing of the sales figures. For the past 25 years Deborah Hofmann, our staff member who oversees the Best-Seller List, has carried on Henry's tradition in this area.
Henry was also a co-founder of the New York Times/CBS News Poll, whose first full year had been the presidential year of 1976. It was originally conceived as an election year project and at the end of 1976 the contract between the Times and CBS News had expired and their original academic consultant, Gary Orren, had gone back to Harvard. Early in 1977, someone posed the question of whether the poll could possibly generate good news articles just based on public opinion about issues, unconnected with any election. If we had failed to prove that, the poll might have disappeared until the next election year. But it was a favorable time. Jimmy Carter's administration was in its early months, the energy crisis was upon the country, and there were lots of good issues to poll about. Now, I had some confidence we could succeed in this mission. After all, the Gallup Poll had been doing it successfully since 1938 and the Harris Poll for a decade or more. But the Times hadn't yet demonstrated it in its own articles with its own poll and that was what we set out to accomplish that year.
EXIT POLLING
Also at the Times I first learned about exit polls. Based on questionnaires filled out by voters leaving their voting stations, and analyzed in real time on election night, exit polls were invented by television news. There's simply no academic counterpart and I had never seen one before.
Before exit polls, a news story written on election night usually went something like this: "Under gray skies, voters trooped to the polls today after a hard-fought campaign at the end of which both parties claimed victory." In other words, it told you nothing at all. After exit polls were invented, an election night story could say voters trooped to the polls with issues X, Y and Z foremost on their minds and (if the margin was big enough) which party seemed in the process of winning or (if the vote was close) whether labor turned out in strength and how the Hispanic vote was splitting. In other words, a real story could be told. The exit poll solved the perennial problem of what to say in the early editions of the paper the morning after an election, a story that almost always has to be written before any ballots are counted.
The first exit poll I worked on was in June of 1977 in the New Jersey primary for governor. That's the year that Brendan Byrne eventually won the govenorship and it went fairly smoothly and I was impressed with how accurate it was. I looked forward to my second exit poll, which was in September in the New York City mayoral primary. That's when Ed Koch became mayor. But that poll turned out to be one of the worst disasters I've ever been a part of. Bud Lewis was polling for Channel 2, our partners in exit polling that year. He had done exit polls for years at NBC News, but in this instance he made an unusual arrangement with a women's club. As a fund-raising project the members of this club would volunteer as interviewers on Election Day and would spread out around the city at randomly chosen precincts, hand out the questionnaires, and phone in the results every few hours. In return, Bud Lewis would pay into the treasury of the club the money that would have gone to the interviewing firm that normally handled such work.
Well, on Election Day it rained, and I mean really rained, and by 2 p.m. when we first started looking at data it was obvious that the majority of precincts were not reporting in (and they never would). E.J. Dionne, who was a young metro reporter, and I had set up our computer terminal in Henry Lieberman's office and as soon as we saw what was happening, we started bailing water and told the reporters to go back to the "Under gray skies" lead. At that very moment, Henry Lieberman threw open the office door and brought in a guest who turned out to be Theodore H. White, author of the "Making of the President" series of books which had revolutionized political journalism. Henry raved on about how White was witnessing here the new technology and the wave of the future and how all elections would eventually be covered in this way. E.J. and I glanced at each other, nodded, and just kept bailing water. It was a new technology all right; maybe the wave of the future, and most elections might eventually be covered with exit polls, but just not on that particular day.
I took from that experience the lesson: never poll with volunteers or amateurs on the team. Always call in the paid professionals. Bud Lewis went on to found and direct The Los Angeles Times Poll and I doubt if he ever again used volunteer interviewers either.
It was at this time I made perhaps my longest lasting contribution to polling -- the Kagay Arrow. Exit polling questionnaires consist of the front and backside of one piece of paper, which the respondents fill out themselves. At the bottom right corner of the front side, there was usually an instruction to "Please Turn Over" and sometimes a wimpy arrow directing the same thing. But there was always a fall off from the front side to the back side in the number of respondents filling it out, such that the back side usually had 5 to 10 percent fewer respondents than the front, because of people failing to turn it over. So I suggested the Kagay Arrow, which was a massive two-inch long and half-inch wide solid black arrow, boldly instructing the respondent to turn over the damn page. I'm proud to report that the fall-off in respondents from front side to back side typically went down to something like 4 or 5 percent.
THE CLYMER YEARS
At this time I also first met Adam Clymer. I had worked with several reporters during the spring and the summer, but for the October 1977 national poll I was told there would be a new guy, Adam Clymer, who had just joined the Times as a political reporter. It only took one poll for me to realize that this guy Clymer was different: when there was a political poll story to be done, he came early, stayed late, took home computer printouts between polls, called me for extra tabulations, and suggested new questions for the next poll. No one else had ever shown that much involvement and persistence. Poll stories became Adam's specialty over the next few years, and no one was surprised when around 1983 he was appointed head of the polling department.
In 1987, Adam asked me to be his deputy, and I was delighted. I moved from Harris back to the Times and in the next several years we did many polls together, both on elections and issues. Much of what I know about news I initially learned from Adam Clymer. When he went to Washington as the Chief Congressional Correspondent (and later became Washington Editor), I succeeded him as head of the polling department for the next 13 years.
CBS NEWS
I also want to thank our longtime partners in polling, CBS News. Warren Mitofsky, who was a co-founder of the Times/CBS News Poll, and Murray Edelman and Kathy Frankovic, have been totally professional, cooperative, and good humored over the past 28 years. The Times couldn't have asked for more honorable partners than CBS News.
I don't know if it's widely known, but every Times/CBS Poll gets two sets of eyes cast upon it. Every new question proposed by one of the organizations has to survive the scrutiny of the other organization, which, lacking pride of authorship, is free to say, "That's the dumbest thing I've seen this month." And every set of poll results is analyzed simultaneously by two independent teams located in different parts of the city. Once the results are available, the Times and CBS never really talk to each other, except to point out a data error if one is spotted. Each organization's conclusions and interpretations are its own business and each is eager to find some subtlety that the other group missed. I've always thought that keeps everyone on his or her toes and adds extra zest to the whole operation.
NEWS SURVEYS STAFF
I also want to thank the staff of the News Surveys Department at the Times: Janet Elder, Marjorie Connelly and Deborah Hofmann. For the last 15 years they have made me look good, and I'm very grateful. So I'm happy to report that in the last five years, the Times has enabled them to look good themselves under their own byline. Previously, poll stories were allowed only one byline, the name of the reporter who wrote the story, and it was always left for the reader to wonder just who else was involved in designing, conducting and analyzing that poll. Now the paper allows a joint byline that includes the News Surveys staff member who did the most work on that poll. Readers now see the bylines of the reporter and Janet Elder or the reporter and Marjorie Connelly. It was well deserved and long overdue.
I also want to thank Rich Meislin for his kind remarks here tonight. Rich is our new department head and he will be leading us in an expanded mission. The thing to know about Rich is he was one of Hank's Boys -- one of the group Henry Lieberman collected around himself in the mid 1970's. They were all a new breed of newsmen: comfortable with numbers and familiar with computers. Rich was there at the beginning of the poll, and although he went on to do many other things -- foreign correspondent, head graphics editor, and editor-in-chief of the Times website -- he's now returned to where we think he belongs.
The Times has enabled us to do many good projects over the years: issues facing women, race relations, economic downsizing, the new world order, Gulf Wars I and II, health care, taxes and spending, impeachment, and the presidential elections of 1988, 1992, 1996, and 2000.
These projects have enabled us to work with many fine reporters over the years. A great benefit of polling for the Times is that the paper already has on its staff a world-class expert on almost any subject you would need to poll about: the reporter who covers that particular beat. That's why we normally have the beat reporter write our poll stories. I feel I have learned from all of them.
AAPOR
I also want to acknowledge AAPOR itself as a learning experience. In the mid 1970's in the Princeton area, Bob Bezilla of the Benson and Benson polling firm was the first to propose a central New Jersey chapter. He observed that many senior people at the local firms saw each other once a year at the national AAPOR conference but seldom in the months in between, and many junior people knew only their colleagues inside their own firm and had no opportunity to meet people in other companies. And, as I noted, University faculty and graduate students hardly knew anybody at all in the polling community. For more than 25 years now, the local chapter has met many of those needs. A few years ago the name was changed from Central Jersey to the New Jersey Chapter and I've now heard rumors they are considering renaming it the Mid-Atlantic Chapter.
For years my only AAPOR experience was with the chapter. The principal national activity, the annual Conference in May, always occurred during Princeton's final exam week, and if a faculty member left campus for several days then, he could be shot for desertion under fire, so it was only after I moved to Harris that I regularly was able to attend the annual Conference.
My first experience with the national Executive Council was in 1988 and 1989 when I was the associate conference chair under Bob Groves from Michigan. A conference chair, like almost no other officer, has to learn all the constituencies who make up AAPOR and understand their interests. Bob's conference in 1989 was in St. Petersburg and was a big success.
My conference in 1990 was in Lancaster, Pennsylvania, where, the program went well but not the food. I will never forget the evening before everyone arrived the hotel's chef walked out without giving any notice at all. The food service operation just collapsed. I can remember at one of the lunches sitting at a table for ten with each of us staring at the gray hockey puck on our plates.
One person said: "Might be chicken."
Another person said: "Could be veal."
And there were other theories as well.
Someone suggested that we taste it.
One person said: "Might be bad chicken."
Another person said: "Could be bad veal."
And there were further opinions.
The lesson I drew from that was: in a meeting or conference that depends on food, never leave it to chance if you can help it. It's just too important.
My presidential year was 1998 and '99. That was also the year of Impeachment for Bill Clinton. It boggled my mind and astounded much of the nation that, when the Monica Lewinsky scandal broke, Bill Clinton's job approval rating went... Up.
Trying to understand the public's reaction to that scandal, and how it might develop, occupied us for much of the year. Initially it seemed that much of the country was rallying around the president as if in a time of crisis, but there was no telling how long that would last. We had to keep polling in case the public eventually turned on Clinton. But as the year went on the attacks on him seemed increasingly partisan to the public, and the public never did turn. That was a year when public opinion really mattered, and therefore polling really mattered. I don't think there has been any other year in my professional lifetime when public opinion and polling made such a difference.
CONCLUSION
I want to thank my wife Carol, my wife of 38 years. From our Peace Corps days until the present one of her many endearing features has been what Ernest Hemingway called, "A built-in, shock-proof crap detector." She has saved me from many errors over the years.
Finally, after 35 years in the profession, polling still seems to me an attractive combination of quantitative methods and real people. It allows us to generalize, and to project, but never to stray too far from the opinions and behavior of real human beings.
I am still hooked.
Thank you again for the award, and for your patience.
NYAAPOR 152 Madison Avenue, Suite 801
New York, NY 10016
Tel: (212) 684-0542
Fax: (212) 481-3071 info@nyaapor.org
