Latin America – September 13th 2004
Global Economic & Strategy

UBS Investment Research

(Extract on the Venezuelan Recall)

Venezuelan referendum (yet) again Latin American Economic Perspectives One constant in Latin America is a complicated political scenario. In that context, Venezuela continues to stand out, and we wanted to make a comment or two on recent developments in the standoff between the government and the opposition, and in particular, on the likelihood of a post-referendum .reconciliation. between President Chávez and the opposition.

An important obstacle to even a limited reconciliation is the opposition.s belief.whether right or wrong, a quite genuine belief, in our view.that the referendum was fraudulently .stolen. from them. This belief recently received analytical support from a study by noted Venezuelan academics Ricardo Hausmann (Harvard University) and Roberto Rigobón (MIT). It is no secret that these individuals are more sympathetic to the opposition than to the government, but they wrote a careful piece of statistical analysis that we believe deserves to be read on its own merits.

They are forced by data limitations into a somewhat indirect approach, and they investigate the data in two main ways. First, they argue that errors from two separate ways of predicting the referendum vote (by precinct) should be uncorrelated, if there was no manipulation of the count. But some kinds of manipulation (notably those that affect a subset of voting precincts differently than others) would generate a correlation. And they find a strong correlation, which they cite as evidence that the vote was manipulated. There are a lot of steps in this argument, and some clever person might be able to come up with an alternative explanation for the correlation, though to our knowledge none has done so yet.

There is a separate section of the paper that we found even more convincing, because the statistical argument is less indirect. The authors asked the question, were the election results in the electoral precincts that were audited by the Carter Center .typical. of those in the country as a whole, or were they different? The question is a natural and important one to ask, because the electoral council supposedly delivered a random sample of precincts for the Carter Center to audit (though they were unwilling to allow the Carter Center to provide the computer program designed to choose the centers, demanding that their own computer program be used instead). However, if you think (as some in the opposition do) that the electoral council.s program .steered. the audit toward precincts that were undoctored, or less heavily doctored than the precincts that the Carter Center didn.t get to audit, then there would likely be differences between the audited precincts and the unaudited ones.

What Hausmann and Rigobón find (simplifying a bit) is that the correlation (across voting precincts) between the referendum vote (as announced by the electoral council) and the (much earlier) vote in favor of the petition to hold a referendum was higher in those electoral precincts that were audited by the Carter Center than in the centers that were not. As Charlie Chan would have said, this is very interesting. In principle, any significant difference between the statistical properties of the centers that were audited and those that were not is suspicious, because a truly random process should have picked a sample that is representative of the entire population of centers. There will, of course, be statistical variation in the statistical properties of any subsample, but the magnitude of this variation is quantifiable by normal statistical analysis and the authors find it highly unlikely (in the jargon, .at a 1% confidence interval.) that normal statistical variation explains the difference they found. Moreover, the particular correlation that they identified seems like one that would be found if the electoral authorities .adjusted. the results of a number of machines (thus plausibly, though not necessarily, attenuating the correlation between the announced vote and .true. preferences, as indicated by exit polls or the earlier collection of signatures for the referendum) while leaving a subset of the machines unaltered for later auditing by the international community.

Now what?

Did any of this happen? We have no idea, and are supremely poorly positioned to make a judgment. Moreover, we don.t think a judgment is required of us in our .day job,. because unless you think a potential fraud can be proven in the months to come, which we consider unlikely, the existence (or not) of fraud is unlikely to change the political context, or (therefore) have investment implications. But the political context is importantly affected by the strong belief on the part of the opposition in a fraud, and we want to finish up with some thoughts on the lasting implications of this belief.

First, while the Hausmann-Rigobón analysis may or may not be right, it highlights the genuine doubts that the opposition has about the validity of the referendum results, and that these doubts are based upon more than self-delusion, wishful thinking, or statistical illiteracy. As noted, we think the prospects for a reconciliation between the government and the opposition, and an easing of the political polarization, are severely reduced by the persistence of these doubts. While some in the private sector will probably feel forced to cut deals with the government to ensure their survival, the odds of a genuine dialogue between the government and the opposition seem very low to us.

Second, the episode has done potentially lasting damage to yet another institution in Venezuela, this time one that is critical for the functioning of the country.s democracy. This is the first time in modern history that a Venezuelan election result has not been accepted by the loser. Unfortunately, it is unlikely to be the last.

Third, we highlight the fact that the government could easily have erased the opposition.s doubts, and their very damaging consequences, by agreeing to a count of the paper ballots that were emitted by the voting machines during the voting process. It.s hard to understand why the electoral council and the government have been unwilling to permit this recount, if they are genuinely convinced that the vote was clean and sincerely interested in promoting some kind of national reconciliation. The Carter Center may have absolved the government of any external pressure to conduct a more thorough verification of the results, but its blessing will not undo the damaging long-term consequences of permitting a substantial portion of the population, rightly or wrongly but certainly unnecessarily, to believe the voting process was subverted by fraud.

regreso a documentos