ON Aug. 21 Mexicans will cast their ballots in a national election that carries enormous implications for the country's political development. Will the ruling Institutional Revolutionary Party (PRI) have its 65-year hold on power broken? And, especially important for Mexico's stability, will the election be fair?
The presidential contest six years ago, which saw Carlos Salinas de Gortari come to power, was widely criticized for alleged vote fraud. Mr. Salinas took office under a cloud, although he later recovered and has enjoyed broad public approval. The Salinas administration has invested in changing voting arrangements to make fraud difficult.
Even if the election is conducted fairly, however, much is riding on whether it will be seen as open and honest. The uprising in Chiapas and the assassination of the first nominee of the ruling party has left social stability fragile.
Opinion surveys have become a factor in the electoral drama. They offer another means - besides the essential one of the ballot box - of determining voter sentiment. A string of preelection survey results shows what Mexican voters say about their intentions. In addition, a distinguished US researcher, Warren Mitofsky, the founder of ``exit polling,'' is working with Mexican colleagues, preparing to conduct a definitive poll of voters as they leave voting stations on election day.
Perhaps not surprisingly, the election surveys themselves have come into question. Those following the Mexican polls have seen a stream of massive contradictions.
There is some agreement in the polls. Virtually all of them show, for example, that Diego Fernandez de Cevallos, the candidate of the more conservative and market-oriented National Action Party (PAN), greatly improved his standing through a strong performance in the nationally televised presidential debate held on May 12.
But contradictions abound. On May 16, for example, Market Opinion Research International (MORI) of Mexico released a poll showing the PAN candidate leading with 25 percent of the declared preferences, followed by the PRI candidate at 20 percent, the Democratic Revolutionary Party's (PRD) nominee with 13 percent, and an extraordinary 42 percent not indicating a choice. On May 21, though, GEOP (Gabinete de Estudios de Opinion Publica) put the PRI comfortably ahead with 39 percent, followed by the PAN at 25 percent and the PRD way back at 12 percent, with a far more modest 19 percent undecided or otherwise not responding. On June 12, MORI released a poll that again put the PAN candidate ahead, this time with 31 percent, followed by the PRI at 23 percent, the PRD at 14 percent, and 30 percent not responding. On June 6, though, the newspaper Reforma had released a survey showing the PRI at 41 percent; PAN, 29; PRD, 9; and 18 percent not responding.
The MORI polls have been highly publicized, running in the magazine Este Pais and frequently reprinted in US publications. They have contributed mightily to the notion that, if the election really is fairly conducted, the ruling PRI is likely to be toppled. That conclusion may be valid, but there are strong reasons to doubt it.
The MORI studies are not opinion surveys in the US meaning of the term. In the US we now insist, rightly, that opinion surveys be based on systematically drawn probability samples. The MORI polls are based on street-corner interviews in just five large cities - leaving out most cities and rural areas. Each MORI poll includes just 320 interviews.
There is a chance that a street-corner poll like that will get results that match what one would have gained from a truly national, probability-based study - but the chance is very remote.
It must be noted that reliable election polling is difficult to achieve in Mexico, no matter how careful one is. Many are reluctant to tell interviewers their real preferences in a country where one party has ruled since 1929. But things can be done - like letting respondents fill out a ``secret ballot'' and drop it in a sealed box - to minimize the problem. And national probability sampling must be carefully employed.
At least two surveys have been conducted since the May 12 debate using national probability samples. The Reforma survey cited above is one of them. A second was done May 17 by Covarrubias y Asociados for the magazine Voz y Voto. The Covarrubias study put the PRI well ahead at 48 percent, followed by the PAN at 26 percent, the PRD at 9 percent, with 14 percent not responding. These two studies thus showed the PRI candidate ahead by an average of 17 percentage points over the PAN, and showed the PRD nominee trailing badly.
All polls are not created equal. Interested readers need to be aware of critical matters involving sampling. Thirty percent of the Mexican population lives in the countryside, and to be reliable surveys must take them into account.
Polling bears the burden in the Mexican election of providing entirely unbiased findings as to the preferences of the Mexican electorate, and findings derived according to the highest professional standards. Mexican voters may change their minds, of course, but their current preferences need to be ``read'' in a way that all careful students of polling find reliable. That's not easy, but it's achievable. All friends of Mexican democracy have a stake in having polls perform their essential democratic mission.