On Tuesday, September 18, I watched the webcast of the Americas Society / Council of the Americas event called “The Road to Venezuela’s Elections” (you can watch the whole thing here). Of course a lot was said, but one statement especially caught my attention (start at 32:00). Pollster Luis Christiansen of Consultores 21 ended his presentation addressing the wide discrepancy in results among Venezuelan pollsters. As he put it there, even if you only look at those pollsters with a history and reputation, there is a range of results from Capriles up by 2 (which is what Consultores 21 has with Capriles at 48 and Chavez at 46) to Chávez up by 13 (referring to a pollster that he did not name but which is probably Datanálisis which in August had Chavez up by 13.5% with Capriles at 34.3 and Chavez at 46.8). He argues that both numbers cannot be right. He suggested to the audience that they use the following rule of thumb for judging the poll numbers that different firms release.

Take the intended vote percentage for Capriles and convert it into absolute numbers–estimating that around 14 million votes will be cast. And ask yourself if that absolute number is reasonable or not. How can one judge if it is reasonable? In each electoral process we have had in Venezuela the opposition has achieved a higher vote than in the previous process. In the latest electoral process of 2010 the opposition achieved 5.9 million votes, since then the electoral registry has grown by a million and a half voters. We are in a polarized scenario, so let as suppose that those votes will be split in half [between Capriles and Chávez], this means that an absolute number useful for reference is 6.6 million votes [For Capriles]. If you take the percentage of votes for Capriles and turn that percentage into absolute numbers, thinking that 14 million will vote, and that absolute number is very far below 6.6, then that is not a reasonable number. I invite you to do this exercise because one sometimes loses sight of the fact that what makes percentages meaningful is the real basis on which they are calculated.

Christiansen is effectively giving us a rule of thumb with which we can move beyond percentages and think about the polls in terms of real people. His argument can be broken down into the following statements.

  1. The opposition has obtained a higher vote in each electoral processes.
  2. [Unstated premise: it should therefore get as many or more votes in this electoral process]
  3. In 2010 the opposition got 5.9 million votes.
  4. The Electoral Registry (RE) has grown by a million and a half.
  5. Those votes will be divided in half by the two candidates, i.e. each will get 750k new votes.
  6. 14 million people will vote.

The conclusion is that any number less than 6.6 million votes should generate some skepticism. Brought back to percentages, any poll putting Capriles at less than 47% should generate skepticism.

Let’s take a look at each of these premises.

1. The opposition has obtained a higher vote in each of the recent electoral processes.

This is actually only true of the last two elections since in the 2008 regional elections the opposition got fewer votes than it did in the 2007 elections. Below you can see the numbers for the opposition’s vote for each of the last four electoral processes.

  • 2007 4.5 million (51.1%)
  • 2008 4.1 million (40.8%)
  • 2009 5.2 million (45.14%)
  • 2010 5.32 million (47.2%)

Before this the opposition vote tally actually went up and down even more. Arias Cardenas got fewer votes in the 2000 presidential election than Salas Romer did in 1998. The number of votes went up with the 2004 recall referendum but then back down as opposition denunciations of fraud generated abstention in the 2004 regional elections, and opposition candidates pulled out of the 2005 parliamentary elections.

2. [Unstated premise: the opposition should therefore get as many or more votes in this electoral process]

The implicit idea in Christiansen’s assertion here seems to be that a person that votes for the opposition will not go back to Chávez. There are, of course, good reasons to doubt this. First, can we really assume that relative support for Chávez versus the opposition will be the same in a presidential election as it was in a legislative election? Within the PSUV Chavez is a battleship among rowboats. PSUV politicians at all levels are less popular than Chávez and it should be expected that they would get less support in a legislative election than he would in a presidential election. Of course, perhaps the same thing could be said about Capriles. But I don’t think it is true to the same degree. There is not a disparity between Capriles and other MUD politicians like there is between Chávez and other PSUV politicians. Furthermore, voting for Capriles is a vote for a radical change in government whereas voting for an opposition legislator was not.

Second, there are several reasons to think that the government is in a stronger position than it was in September 2010. Chavez’s approval ratings and trust ratings are higher now, as is personal optimism. And party identification—which favored the opposition in September 2010— now favors the government.

But in a highly polarized environment like Venezuela, one could imagine that perhaps there is some validity to this idea. Perhaps through the act of voting for the opposition, you strengthen your own opposition to the Chavez government, and will continue to vote against it. I personally don’t find this convincing; but neither do I find it completely absurd either.

3. In 2010 the opposition got 5.9 million votes.

What Christiansen counts as the “opposition” vote here is misleading for it includes not only the MUD’s vote but also the vote for the PPT and all other parties. This is logically correct since the MUD and all of the other small parties “opposing” the PSUV got 51.7% of the vote. However, it would be just as logically correct to say that the parties “opposing” the MUD got 52.8% if you add the PSUV and all other parties. In fact, in 2010 the PPT openly ran as a third option, clearly distinguishing itself not only from the PSUV but also from the MUD. This elision is common in opposition discourse and is the cornerstone of any number of specious analyses currently circulating on the internet guaranteeing a Capriles victory. However, the opposition properly considered as the MUD got 5.3 million votes in 2010.

4. The Electoral Registry has grown by a million and a half.

It has actually grown by 1.3 million. The 2012 RE is 19 million, up from 17.7 million in 2010.

5. Those votes will be divided in half by the two candidates, i.e. each will get 750k new votes.

Note that this assertion breaks with the logic of looking at actual numbers of voters rather than percentages. Furthermore, not all of those new people on the RE will turnout. If we assume a turnout rate of 75%, 975,000 of these newly enrolled voters will vote. If they split evenly between the candidates, each would get 487,500.

6. 14 million people will vote.

This would be a turnout rate of 75% which is what it was in 2006. Seems like a safe assumption.

Conclusion: Any number under 6.6 million votes for Capriles does not seem reasonable.

Let’s assume that Christiansen is correct in his analysis that “once you vote opposition you never go back,” and let’s pardon his sliding back into percentage thinking when talking about new voters. However, let’s use the actual correct numbers for his calculation.

If in 2010 the MUD got 5.3 million votes, and if they will get 487,500 of the newly registered voters, then Capriles should get at least 5.8 million votes. If 14 million people vote in the 2012 election that would mean that Capriles can safely count on 40% of the vote. Put differently, and in keeping with Christiansen’s original intention, the suggestion would be that any poll giving Capriles less than 40% of the vote should generate skepticism.

This leads to one more issue. So far most pollsters have not given final predictions but include in their presentations a large number of undecided voters. For example, Datanálisis’ numbers include around 19% of all respondents saying they are undecided. The actual election, in contrast will, for all practical purposes, be zero-sum with a very small number of null votes and votes for the other candidates. Datanálisis’ numbers suggest that if the undecided were to split evenly between the two candidates, that would put Capriles at 43.8% and Chavez at 56.2%

Thus Christiansen’s attempt to contextualize percentages by going to absolute numbers and assuming the opposition will continue its upward trajectory, does not help us distinguish between the results of the most recognized pollsters. Most specifically it does not help us decide between the poll that has Capriles up by 2 (Consultores 21), and the poll that has him down by 13 (Datanálisis).

My own personal tendency would be to turn Christiansen’s suggestion around. One of the unique characteristics of Venezuelan politics in recent years has been the high number of people who identify neither with the Chavez government nor with the opposition—at most times “Ni Ni’s” (Neither Nor’s) are the largest block in questions on political identification with around 40%. That being the case, polling firms that have the undecided vote in the low single digits—Consultores 30.11, Consultores 21, and Varianzas—raise my suspicions. My guess is that these companies are not providing the actual percentages from their data, but are reassigning most of the undecided to one candidate (Capriles in the case of Consultores 21 and Varianzas, and Chavez in the case of Consultores 30.11), or simply dropping them from the sample.

On Tuesday Datanálisis will be releasing new numbers and will probably make some projections. I think that will give us a good sense if there has been significant movement in voter intentions.