CEPR on Venezuela's recall: Mark Weisbrot et al spinning the truth away
By Aleksander Boyd
London 24 Sep. 04 - Last week I was visiting the renewed website of the Venezuela Information Office (VIO). The original domain veninfo.org that was registered by former Center for Economic and Policy Research (CEPR) employee David Levy redirects users now to another one called rethinkvenezuela.org, which is administered by a new recruit of the VIO, namely Eric S. Wingerter. However Wingerter is not the only rookie in the Venezuela's inanity propagating machine in Washington: Jo Ellen Chernow, Andres Mateo, Jarrin Cuvi and Robert Naiman have also been recently recruited according to the Foreign Agents Registration Act (FARA), whilst Isaura Gilmond, Deborah S. James and Stacy Elizabeth McDougall renewed their contracts. Curious as ever, I googled the new spinmeisters and yet again CEPR pops up for Robert Naiman was a senior policy analyst of the think-septic-tank. But it gets better.
The Center for Public Integrity is a Washington based nonprofit, nonpartisan, tax-exempt organization that conducts investigative research and reporting on public policy issues in the United States and around the world. One of its researchers, Kevin Bogardus, seems to have spent some time looking at the VIO files in the FARA unit and decided to write a report on it. Delighted, as I was upon learning that someone else besides me deemed appropriate to expose the propagandistic endeavours of Chavez' regime in Capitol Hill, I took up the phone and rang Bogardus. He explained that his aim is to investigate how the oil monies from OPEC countries are spent in lobbying firms in Washington, hence he came across the VIO. "The relationship between the VIO and CEPR is bizarre" he said for he did not quite understand how "the same day the VIO re-registered with the Justice Department, a letter authored by the CEPR [to draw attention upon Venezuela] was posted on a Web site frequented by large donors to progressive causes. Seems very odd" he concluded. His article is basically a summary of what I have denounced here previously, i.e. the double discourse of the Chavez regime and the obvious partisan stance of the spinmeisters of CEPR. In that sense, Weisbrot comment in regards to his relationship with the VIO " Sure, I talk to them all the time. They are a great source of information" wraps it all up quite neatly for we hear, from the horse's mouth, where does he shop for information when it comes to Venezuelan matters, which brings me to the point. As everyone knows there is a huge fraud cloud above Chavez' recent electoral win. The reason is quite simple, the international observers -namely OAS and Carter Center- did not do the job properly nor were they allowed to conduct and control the audits that would have cloaked with legitimacy the end result.
At the request of Sumate, the civic electoral NGO, Harvard's Ricardo Hausmann and MIT's Roberto Rigobon came up with a hypothesis on the electoral fraud that has become the benchmark against all posterior studies have been compared to. The Carter Center (CC) on the other hand, trying desperately to save face, recourse to Stanford's Jonathan Taylor, whose initial findings were readily commented upon by Jennifer McCoy in The Economist. As it happens, Taylor's backtracked on it, conceding that he had an error in his calculations. We then see yet another publication produced by the CC where an attempt is made to ensure the randomness of the sample used for the second audit. Let me make a halt here for there seems to be a confusion between the first audit and the second.
The first audit, a.k.a. "hot audit" could not be conducted in fair and speedy manner according to Jennifer McCoy due to the following reasons:
"the sites were restricted by the CNE for logistical reasons -- since the sample was only drawn in the afternoon of the August 15 (by previous agreement with the political parties in order to avoid the possibility of the machines being tampered with before August 15), it would not be possible to send the CNE auditors to far-flung or rural places in time for the scheduled closing of the polls and the "hot audit". Our reports all refer to the second audit, August 18."
Ergo all reports produced so far by the CC, as admitted by Dr McCoy, refer to the second audit, that is that of August 18. Nevertheless a clumsy first audit attempt was made.
"On Saturday, August 14, 1 (one) day before the recall referendum, the Electoral Council decided that this audit, in which the totals of ballots contained in the boxes would be contrasted with the printed results of the electronic tally, would only be done in 20 (twenty) of the 336 (three hundred and thirty-six) municipalities, in 14 (fourteen) of the 24 (twenty-four) states of the country. On Sunday, August 15, between 6:00 PM and 7:00 PM, a sample of 192 (one hundred and ninety-two) machines was selected, which had to be audited immediately upon completion of the voting event in the same voting table, reviewing the content of the boxes and checking its exact coincidence with the tally of the corresponding machine and with the number of registered voters in the corresponding electoral roll (or cuaderno). It is worth noting that on the day of the recall only 76 (seventy-six) of the 192 (one hundred and ninety-two) stipulated boxes were audited. The promoters for the SI were present in only 27 (twenty-seven) of these audits and the international observers in approximately 10 (ten) tables;" [Sumate report page 39] nonetheless in those audits witnessed by the opposition the SI option (Chavez to be recalled) won with 63% of the votes It is also extremely important to note that none of the international observers were allowed into the tallying room of the National Electoral Council (CNE) -as it was agreed- during the tallying process. Former OAS Secretary Gaviria is on the record correcting Jimmy Carter with respect to their absence in the said room. They did not witness the process that led to the announcement of the preliminary results on August 16 at 4AM by CNE's director Carrasquero. In light of the aforementioned, how on earth could they have endorsed those preliminary results is something that truly escapes me.
But it gets better still in the second audit. The opposition, feeling quite betrayed by the international electoral observers, requested for a second audit whose aim was to dissipate any doubts vis-a-vis the results by taking away the control that the CNE had over the first audit and passing it to the international entities, OAS and CC that is. A conciliatory Gaviria sort of force fed the initiative to a Jimmy Carter who had itchy feet to abandon the country. Sufficient assurances were given to the opposition in relation to the manner in which the second audit was to be conducted. Key element here was the software to be used to randomly generate a statistically representative sample of voting centres. Again and again it was assured by OAS and CC representatives, that solely their software programme were to be used. However the latter screwed it all up the second time too. Allegedly their random generator was "faulty" and as a result a very diligent Jorge Rodriguez, CNE director, considered fitting to use the same software programme that was used on the first audit. Hence the CNE had absolute control of the second audit for the data fed to a CNE's computer was introduced by CNE's employee Tibisay Lucena using a CNE devised software and not the one proposed by the international observers as agreed. Moreover, the second audit was done on August 18, by then the 'CNE randomly selected' boxes had been out of the international observer's watch for 60 hours. Alas the second time round they did not see anything out of the ordinary either. Never mind their lies and breach of duty with the opposition.
The icing on the cake has to be of course the 'report' of that beacon of truth and honour known as the Center for Economic and Policy Research. Its intellectual prostitutes led by Weisbrot come to Chavez' rescue and pretend to strike off, as wild conspiracy theories, all of the above. But as with everything in life, there is always someone that has the time to call their bluffs off. Navigating through Nueva Venezuela's forum I came across this little gem penned by I. Torres, that simply destroys the arguments put forth by the CEPR. I shall add none to it, you be the judge.
Beginning
There are certain distinctions that need to be made regarding Hausmann and Rigobon's (H&R) work -or anybody else's- to reply to any claims of flaws.
H&R accepted to work on the RR data for personal reasons when Sumate requested this of them (first distinction is their personal 'reasons'). They then made certain assumptions in *choosing* the statistical methodology for analyzing the data (second distinction is 'assumptions'). They then performed the statistical work very rigorously (third distinction is the statistical 'work' itself). Finally, they drew some conclusions from the results of the statistical analysis (fourth distinction is 'conclusions').
Given these distinctions (Reasons, Assumptions, Work, Conclusions), I'll stress, again, so far no flaws have been pointed out in their *work*.
1) REASONS. Some people would have me believe (I'll use 'me' to avoid making generalities) that I should ignore H&R's work because their reasons for having accepted to do the work may be partisan. Given that those saying it have reasons of their own to be criticizing H&R, I should --to be consistent-- ignore the criticisms of these persons. But I don't. I do not cease to listen to the analytical work of those who criticize H&R just because I know they are Chavez supporters, as I don't listen to Chavez opponents just because I know that they are Chavez opponents. I listen to everyone's work and analyze it for myself. Oddly enough, some people have implied that this is wrong or inconsistent, but they won't explain how they reach that conclusion.
2) ASSUMPTIONS. Some people would have me believe that I should ignore H&R's work because H&R's assumptions may have been baseless, or, worse, altogether wrong. H&R's work included some very creative statistical choices because of their assumptions. H&R had two kinds of assumptions.
Of the first kind, H&R assumed that if there had been a fraud of this magnitude, then it would have to be a fraud of certain characteristics. Namely, they assumed that the fraud would have foreseen that standard statistical tests would be performed and that, therefore, would be designed to *pass* standard statistical tests. They didn't have to show that this was true; they only had to assume this to explain the choice of statistical methodology for their work. Whether this assumption was correct or incorrect does not diminish the quality of the results of their work. If someone believes that incorrect assumptions of this type diminish the quality of the work, they should have to point to where the work is dependent on the assumptions. On this kind of assumption, they were only used to choose a non-standard statistical method that would detect a fraud despite it being designed to pass standard tests.
Interestingly, CEPR’s Mark Weisbrot, David Rosnick, and Todd Tucker ( WR&T) also volunteer the possibility that perhaps H&R were wrong in their assumption of such kind. So they offered yet another scenario of fraud that, of course, would fail to explain all observations. Perhaps they should look into the claim that the opposition observers got to see 27 of the audit centres, and that in those 27 the 'Yes' won by the percentage shown by the exit poll conducted by Penn & Schoen et al. They might also wish to contrast the Chavez 2000 election numbers in the centre of Petare where the 'Yes' won in the RR, by the amount measured by the exit poll. I'm not in a position to know if these claims are true, but I am in a position to realize how easy it should be for them to prove these claims wrong, but they don't and I can’t help to wonder why. The point is there are many possible assumptions of this kind that have not been considered, which do not need to be considered because they do not affect the quality of their work.
Of the second kind, H&R assumed that the signature data and exit poll data were independent variables measuring, to some degree, the intent of electors. WR&T criticized this assumption that the two independent variables were, in fact, independent. WR&T are correct in pointing out that *if* the variables are not independent measures of electors’ intent, *then* the errors might be correlated by virtue of the lack of independence. Big *if*! Do they provide any proof of the lack of independence? NO. Do they provide any logistics of how such independence might have been jeopardized? NO. They merely point to the reasons the pollsters may have had for providing fraudulent data, and leave it at that. They even admit, "We have no idea how they [the data] were gathered or if there was fraud involved in their collection." They did not find a flaw in the H&R work; they just made an accusation of fraud drawn from a presumption of reasons.
3) WORK. Some people would have me believe that I should ignore H&R's work because there are flaws in the work itself. But they have never pointed to a single flaw in the work done. The Weisbrot, Rosnick, and Tucker work centres on concluding that H&R's report *must* be flawed because *if* the assumption that the audit sample might have been a random sample of a random sample is true, *then* the audit sample is representative of the universe, which showed the 'No' winning. (Note in the last paragraph of page 7 to the second paragraph of page 8, WR&T goes from describing the work to analyzing the assumptions). They also conclude that H&R's work *must* be flawed because *if* the independent variables used are not independent, *then* the correlation of errors *can have* a different interpretation than "fraud".
Firstly, what if the assumption of random sample from a tamper free random subset was wrong? That does not diminish the bullet-proof-ness of the work, just because the assumption in choosing which statistical method can be wrong. What if, for example, the audit sample was not random but carefully chosen? H&R's work still detects the fraud *because* they chose a method that would detect regardless of whether their assumption is correct or not. You will note that Chavez supporters sidestep the crux of the matter:
Relation of factors with respect to audit sample is different than relation of the same factors with respect to universe points to non randomness, giving a probability of 1% of this occurring merely by chance.
Also, what if the assumption of variable independence is wrong? Did WR&T demonstrate how dependent they would have to be to have correlations in the errors with respect to the vote? NO. Did they provide the same exercise with different data of elector’s intent that would be certainly independent (by their own example, pollster information that indicated Chavez would win)? NO. Did they do any work to find a flaw in the assumption on which H&R's work was based? NO. You will note that Chavez supporters sidestep the crux of the matter:
Correlation of *errors* between independent variables measuring the intent of the electorate and actual votes, giving a probability of 1% of this occurring merely by chance.
Secondly, and so as not to be accused of the same thing (i.e., sidestepping the crux of someone's work), they are also statistically mistaken. WR&T seem not to be aware that H&R did not perform and analysis of the Audit Sample, directly; they performed an analysis on the relationship between the certain factors and the Audit Sample and its universe. The reason this distinction is important is because their result cannot be used as a descriptor of the Audit Sample to be extrapolated to the universe. They calculated a descriptor of the relationship, not of the factor itself. For this reason, Hausmann actually stated in an interview that he really did not know *from his work* if one side or the other won, only that there was a 1% probability that the Audit Sample could have been a random sample of the universe, by chance. The method he used is accurate in stating that there is a certain percent variation in the value of one versus the other, but the method is not the appropriate one for making any claims regarding the values themselves, which is what WR&T are doing.
They, perhaps unwittingly, acknowledge the importance of finding a flaw in the work and not doing it when they state that they do not *need* to find a flaw in H&R's work because, logistically, they had already provided sufficient information to conclude that it was *impossible* for there to have been a fraud of the type H&R assumed there had been. Quote:
"But it must be emphasized that there is no need to determine exactly how the model in Hausmann and Rigobon's analysis may have been mis-specified. The fact remains that their theory of how the fraud could have taken place is untenable, as the Carter Centre has demonstrated. And the audited sample, which has been shown clearly to be a random sample of the entire universe of voting machines, matched the electronic results almost exactly. On this basis we can safely reject Hausmann and Rigobon's econometric evidence."
All speculation
4) CONCLUSIONS. Some people would have me believe that I should ignore H&R's work because conclusions they drew from their work may fail to be true. As with the assumptions, conclusions that are drawn independently from the work do not affect the quality of the report. An example immediately comes to mind: Taylor's work. His report produces the result that the probability that the number of 'Yes' repeats to occur by chance is 0.3%. I conclude from such a number that one should investigate further for there is a high enough probability of non-chance going on. Carter Centre, however, concludes that 0.3% is within the parameters expected (or something like that). Two opposite conclusions, from the same, undisputed work of Taylor , do not affect, in any way, the now flawless nature of the report.
To wrap up: Someone needs to point to a flaw in the H&R and Taylor reports, or come up with a non-chance, non fraud explanation for the statistically significant and valid results of both H&R’s and Taylor’s work. The onus is on them, because the probabilities of observations not being by chance of only the three results of the two pieces of work are 0.00003%. If no other explanation comes up, fraud probability still stands at 99.99997%.
WR&T's analysis of the logistic impossibility is also very limited. They did not address, in their summary the possible loopholes in the observers' perspective in detecting fraudulent activity. To name one of the loopholes, the OAS’ observers were not allowed in a totalization centre. Interestingly, this very loophole was one of the stated complaints in Peru , justifying their not wanting to take part in the second round of the election.
WR&T go off on an all too familiar rant as to how it is "best" to move on. To WR&T Co. I ask, did you ever watch Columbo, or more recently, Monk?
Find a flaw in the WORK! Stop giving me criticisms on people's personal REASONS, irrelevant ASSUMPTIONS, or deduced CONCLUSIONS. Give me FACTS.
End
send this article to a friend >>