Warren Mitofsky’s departure last September was the leading story in the nightly news in Mexico, and it was widely commented upon in the media. Warren’s last exit poll was the work he did for the Mexican presidential election of 2006. My intention in this brief account is to tell the story of Warren in Mexico, and in some way assess his outstanding contribution to Mexican polling, to the Mexican polling industry, and to the construction of a reliable electoral system in Mexico.
I first met Warren in the winter of 1990. I was at the time pollster to the president of Mexico, Carlos Salinas de Gortari, who had just been elected under widespread accusations of fraud. From our first meeting I did my best to make my case: The pollster of the Mexican president wanted him to do independent exit polls in Mexico—ideally, to bring the American networks to do it. Warren just couldn’t understand it. Or perhaps he just didn’t believe me. He made absolutely clear that he didn’t see any possibility of the networks doing it, and there was no way he would ever work for a government, let alone a political party. The only possibility he could see was if one or some of the Mexican media hired him.
I kept frequent contact with Warren, but I realized he was in the middle of important changes in his life. In those years he left CBS, and he began to think about forming his own company, Mitofsky International. In the midterm election of 1991, Televisa, the largest Mexican network, hired one of the most renowned and prestigious American polling firms to do an exit poll. The PRI, the official party, won every single district and more than 60 percent of the popular vote. An easy election to call, but the large and famous American firm could not give its estimate of the popular vote until three o’clock in the morning, and only because the client pressured them to do it. It was time to bring Warren Mitofsky to Mexico.
After our first meeting, Warren returned to Mexico on various occasions to see my office do exit polling for some local races. The election for governor in the state of Hidalgo in the spring of 1992 was a turning point. We were watching the work of the interviewers in a very rural precinct when a very angry woman walked by complaining because, as she put it, she had had to vote twice. I asked her why she had participated in the interview, and she answered, because it would be impolite—“de muy mala educación”—not to answer those poor guys who had been there all day long. Warren decided instantly to work in Mexico. He knew the value of high response rates.
In his visits to Mexico, Warren could see up close the efforts Mexicans were making to build a reliable electoral system. He met several times with various officials of the electoral commission, and he knew firsthand all the changes taking place. He maintained occasional contacts with them and accompanied these changes all along the way.
Finally, in 1994, the Chamber of Radio and TV hired Warren to do an exit poll for the presidential election. He worked with two Mexican polling firms. The announcement of the winner was made as the polls closed, and it was precise to the decimal point.
The following year, Roy Campos, who had been providing technical support for the exit polls we had been conducting in local races since 1989, founded Consulta, an independent polling firm. Roy knew Warren, and two years later Consulta became Consulta-Mitofsky, the Mexican partner of Mitofsky International. They put together a plan for Televisa, and, beginning with the midterm election of July 1997, Consulta-Mitofsky announced their results in every single local and national election on Televisa as the polls closed. On no single occasion did Mitofsky’s estimates differ from the official results announced later in the night. TVAzteca, the other network, followed his lead.
Electoral polls during the campaigns and exit polls and quick counts on election nights became a common feature of Mexican elections, and Consulta-Mitofsky’s polls became an indispensable reference. After many years of distrust, the attentive public could predict the winner with some degree of certainty after a good and fair competition; and as the polls closed, the media confirmed the expected winner, and the defeated candidate conceded. Elections became citizens’ events. This common feature of elections in any advanced democracy was of enormous importance for an electoral system under construction and would never have happened without a robust polling industry working for independent media. One election at a time, the electoral system gained the trust of the people.
Before the presidential election of 1988, only five polls were published, and no exit poll or quick count was done. In the 2006 presidential election, between January and June, just over one hundred polls were published, and eight exit polls and quick counts were released on election night by nine different media, conducted by a dozen different polling firms. Half of these firms had, in some way or another links with Warren Mitofsky, and all of them knew him at least somewhat. Warren became an indispensable point of reference for good, honest, independent polling in Mexico.
I think the best testimony of Warren’s legacy in Mexico is the overall quality of the Mexican polls. In Table 1, three estimates are shown of the differences between the last polls taken prior to elections and the actual vote in Mexico. M1 is the “error” in the winner, M2 is the average error in the three major candidates, and M3 is the error in the margin between first and second. Pollsters are identified with letters only, because in all cases the means for first and second place are within the margin of error, and hence all estimates are basically the same.
Table 2 shows the average of M2 for different elections in Mexico and the United States. As can be noticed, the average error seems to be larger than that in the United States. This larger error does not stem from any difference in methodology. It is the result of the forced silence imposed on Mexican pollsters by the electoral laws. In Mexico it is forbidden to publish poll results eight days before the election, and the last poll has to be taken almost two weeks before, since, with telephone coverage limited to middle- and upper-class Mexicans, all polls have to be conducted face to face. The important point, however, is the improvement in accuracy of the Mexican polls over time.
Warren was generous and open, and many of us interacted with him to different degrees about every aspect of polling. I will never forget the hundred times he took a pencil and explained to me or my collaborators this or that methodological point. But the most important advice I received from him was not about sampling or estimation. Many times he mentioned how much it surprised him that people asked about those things, while never mentioning the most important aspect of good exit polling: quality control of the information. I do not know what Warren taught other pollsters. What I do know is that he did not teach us how to do exit polls; he did more, much more than that. He trusted us, and set very high the ethical standards for the Mexican polling industry he helped develop. Above all, Warren was an integral participant in the construction of a reliable electoral system, and, therefore, a better democracy.
It would be unfair to Warren to finish this account on such a solemn note. At Warren’s funeral, Kathy Frankovic, director of polling at CBS, read a beautiful piece in which she described a WAPOR seminar in Mexico she and Warren had participated in. She described how she could hear “Mitofsky” whispered like a kind of prayer. I have to make a small correction. Mindful of the very many problems he had helped us solve in Mexico, we were only trying to figure out the most difficult one Warren had put to us: Where should we take him for lunch?
Ulises Beltrán is general director of BGC, Beltrán y Asociados, S.C.