In the upcoming months every single politician will care about what you think, what you fear and what you expect, and follow closely your ideas with all the interest of their hearts. It is, of course, time for elections! You may probably be thinking about the forthcoming challenge to Barack Obama's administration, or the French elections or indeed the London Mayoral Elections, but we also have an interesting battle in Mexico. On July 1st, Mexicans will decide who their next president will be for the next six years. And I honestly believe this process will be more interesting than the American one. First of all, Mexican legislation does not allow any form of re-election, meaning that no matter what, we get a brand new president. Meanwhile, in the US, we get to see the same old Obama campaigning around the country, all over again, which leaves Americans with only a new competitor. Boring. Here in Mexico, on the other side, we have four candidates, three of those with a real aspiration of winning the big day.
Ok, let's get serious for a moment. Although the US elections are certainly more closely watched by the rest of the world, there are some really good reasons why the upcoming presidential elections in Mexico are interesting to statisticians, reasons closely related to a story which started in 2006.
Once upon a time, in a beautiful and magical kingdom called Mexico, the 2006 presidential election turned out to be the closest match that has ever occurred in the country. The official results granted the presidency to the official party, Partido de Acción Nacional (Party of National Action), with a 35.89% of voting preference, just a bit above the Left opposition party, Partido de la Revolución Demócrata (Party of the Democratic Revolution), which got 35.31% of votes. The difference was about the same that was seen in the 2000 presidential election in the US (Dubya v Gore) which led to some harsh months of lawsuits, trials and lots of uncertainty about the outcome. The exact same thing happened in Mexico and, during that period, one unlikely victim was blamed for the political chaos: surveys.
Usually, you would think that surveys help to bring certainty to elections. By letting the voters know the tendencies, changes and sudden reactions in electoral preference, the people may easily accept official results when they make sense with previous surveys. The problem was that, with such a close race, not every survey presented the same result, but particularly, not every survey used the same methodology. During 2006, not a single survey company had any obligation to talk about its methodology which, sometimes, not even its client could know. This fact was certainly misused, producing hundreds of surveys that were somehow presented in the media and often confusing voters. And I don't talk about illiteracy when interpreting statistics, no, this was much more serious. Just imagine seeing the results of one phone survey today and tomorrow one with actual face-to-face interviews. Different winners each day. Both surveys may be methodologically flawless (ok, more about that later), but a phone survey has a complete different population of measurement since you have excluded half of the Mexican population who do not have a phone at home. You can't compare results or tendencies based on two different procedures. What's worse, you never knew whether it was a telephone-based survey or not, so even if I had a PhD in statistics I couldn't judge their results.
It got really bad. Newspapers and the media didn't know which results to publish. In the end I discovered newspapers presenting results of about a dozen surveys at once. I even heard of “meta-analysis” of surveys producing an agglomerate result (just a mean if I should guess) of different survey tendencies. That's just how desperate the media became.
Of course, these procedures weren't exactly scientific, but the same could be said about some surveys. I'll talk about a beautiful example I recall of a particular company that produced surveys by asking drivers in the streets. They usually waited next to a traffic light until every car stopped. Then, they would block the street with a huge banner that said: “We are conducting a survey: turn on your lights if you are going to vote Blue.” Selection units? Sampling frame? Weighting? Estimation procedures? Who cares when I have the chance to interview a Ferrari! Seriously, I'm even tempted to link you to their website but I think that would be rude.
Certainly with all these numbers popping up everywhere, the candidates didn't worry too much about methodology and just accused every single company of producing self-induced surveys, devoted to aid a particular party. Forget about trash campaigns, we surveyors had a full six months of being trashed by every single politician. Something had to be done.
After the presidential election and the not-so-boring chaos after results were declared (you know, the losing candidate demanding justice, etc.), it was time to rethink the role of surveys in elections. A very interesting debate started which, thankfully, included expert surveyors. One of the main points was deciding whether surveys could drastically modify the tendencies which some suggested could be done by publishing them or just by doing them. This resulted in the first formal regulation for publishing surveys in periods of elections, which passed in 2007 and was first applied in the mid-term elections of 2009 in Mexico. The official organisations concluded that surveys have the right to be published but that these publications should at least be somehow regulated. The country is currently celebrating its first presidential election with these new rules.
But what has really changed? First of all, it is important to point out the fact that the new legislation in no way limited the publication of survey results. The Institute of Federal Elections, the official organism that organizes and regulates the federal elections in Mexico, demanded that every published survey must be accompanied with a full extensive discussion of some specific details of the methodology used in it. This methodology appendix must be delivered to the Institute after publication of the results. With this information, the Institute is ordered by law to produce and publish a report every month with the main results and methodological details of every survey published regarding the elections. For those interested, these reports can be found online in the official web page of the Institute (in Spanish).
But what is the Institute asking for? What are the basic methodology issues that must be described? First of all it is important to clarify that the Mexican authorities consulted with surveyors, statisticians and other related professionals the recommended procedures to design surveys in order to determine the international standards required for these exercises. After reuniting in 2008 with these groups, an agreement was produced in which there was specified the minimal methodological requirements that a survey that is to be published must divulge. These “general scientific criteria” that every survey must declare about its methodology are the following:
1) Objective of the study
2) Sample frame
3) Sample Design, which should at least include:
• Definition of the target population
• Procedure followed to select units in the sample
• Procedure followed for the estimation of the parameters
• Sample size and how it was obtained
• Quality of estimation: Confidence and maximum error accepted in estimation
• Frequency and treatment of non-response
• Rejection rate of the interview
4) Method of gathering the data, either by phone, mail or face to face surveys
5) Copy of the questionnaire applied
6) Processing for obtaining estimations and confidence intervals
7) Software used for the analysis.
Personally, I think some details are rather vague and could be left open for interpretation, but I hope these requirements will at least scare some people without full knowledge in statistical procedures for surveys and prevent them from publishing results where the specific target of observation is a vehicle. It is also important to mention that the information shall only be presented to the Institute and in no way will it be verified by professionals, so even if the estimation methods, sample design and everything else is awfully wrong, it won't stop your survey from being published. Also, I've noticed that some published results have not been yet been reported to the Institute. I'm not aware of any fines or procedures against these companies, which may imply some loopholes in the legislation. Still, this counts as step in the right direction as it will help promote statistical knowledge and at the same time save the citizens from tons of useless information.
Let's see how everything turns out in the next election. Anyway, it appears this year the results will not be so close so surveys won't be blamed this time around. Yet there are two months of campaigning to come so maybe the spotlight may move to statisticians once again.