Español English French Kwéyol

Haiti-Elections : Analyzing the work of the Expert Mission

By James Morrell

Submitted to AlterPresse on January 14, 2010

Many have noticed that the finding of the experts’ mission in favor of Martelly rests on a razor-thin margin of 3,225 votes out of the 1.1 million received by all candidates, or 0.3 percent. Making the conclusion even more precarious, an analysis of the 17,220 votes for Célestin that the experts disallowed shows at most 12,316 to be related to the issue of excessive votes for one candidate, leaving the other 4,904 resting on various technical issues. If, by some vicissitude, these technical exclusions were overturned for Célestin but not for Martelly at the same time, Célestin would be back in second place.

The larger question, however, that is inadvertently raised by the work of the experts’ mission is the very validity of their leaving a Célestin total in the range of 224,000, given his high negatives going into the election. The amount of votes newly disqualified by the experts, 17,220, is only 7 percent of the previous total. How valid are the remaining 224,242? The truth is that a purely statistical investigation of the doubtful material received by the Tabulation Center, after the electoral apparatus around the country had suffered a grievous assault from within and without, is dangerously limited. Or more bluntly, garbage in, garbage out.

We begin by reviewing the checkered history of the Tabulation Center since November 28, 2010 and then propose ways to cast some outside light on the vote total.

The Tabulation Center began by imposing a cutoff point of 300 votes per BV or 225 votes per winning candidate. When it became evident that turnout was light, it lowered the cutoff to 150 votes per candidate. Application of the 150 cutoff to Célestin would have reduced him from his original 279,825 by 57,933 votes, to 221,892. This last figure is very close to the 224,242 now allotted to him by the experts, and would have equally eliminated him from the race.

But in the event, the Tabulation Center did not reduce him to 150 votes per candidate but stopped at 200 votes, reducing him by 38,363 votes. This reduced his vote total from the original 279,825 to 241,462. Stopping here was just enough to leave him in second place.

What had happened? In our previous article "Send This Mission Packing" we related the still unconfirmed story that the electoral commission was pressured to remove a block of votes from quarantine to put Célestin over the top. This would have been the votes coming in between 150 and 200 votes per candidate. Two elements on both sides of this story have been confirmed: the intimidation of the commission and the existence of that block of votes.

But we do not need to depend on the story. A cutoff point of 150 takes 57,933 votes from Célestin, knocking him out of the race. A cutoff point of 200 takes 38,363, leaving him in. The difference is 19,596. Looking at this database in mid-December, the Haiti Democracy Project’s electoral mission set the cutoff point at a conservative 164, cutting off 14,408 votes. This procedure has now been replicated by the expert statisticians who, crunching the same numbers, cut off 17,220.

What the statisticians have done, in adding their 17,220 to the existing quarantine of 38,363, is essentially return the quarantine to the state it would have been in if the original 150-vote cutoff point had stood. In their report, the experts mentioned the 150-vote cutoff but don’t say what happened to it. This is an important omission because the existence of a previous standard by the Tabulation Center should have been the starting point for the statisticians’ own work. They needed to check its validity before starting anew themselves; they needed to know why it was not used.

Would it have revealed the illicit pressures on the electoral commission that were described to the Haiti Democracy Project? Was it applied and rescinded? There ought to be an e-mail record at the Tabulation Center of orders from Gailot Dorsinvil on this subject.

To summarize, there were three attempts to use the database to arrive at a slightly more realistic cutoff point for Célestin:

1. The Tabulation Center’s 150-vote threshold, cutting off 19,596

2. The Haiti Democracy Project’s 164-vote threshold, cutting off 14,408

3. The experts’ cutoff of 17,220 votes

What the three attempts confirm is the essential similarity of statistical exercises working over the same database. What they also confirm is the severe limitations of this approach.

None of the three does more than tinker with Célestin’s improbable 220,000-plus total. Theoretically, one could set the cutoff point lower. A cutoff at 100 would leave him with 191,137. But while probably valid for Célestin, such a threshold, applied to Manigat and Martelly, would undoubtedly cut into many votes that they actually got.

While the expert mission’s statisticians brought impressive qualifications, there is no statistical magic in the world that can peek behind the curtain of returns that have been falsified by the local electoral officials themselves. One can stare at them until one is blue in the face, they will give no hint.

This was brought home by the Haiti Democracy Project’s comparison of original returns in a Nord-Est deputy’s race to the results posted by the Tabulation Center. They showed that someone in the system had added 100 or 200 votes to the government candidate’s total in each BV. When it finally looked at the Haiti Democracy Project’s data, the Tabulation Center admitted that it couldn’t find anything wrong with the returns it had received. There were no visible signs of alteration. The Tabulation Center had entered the false data straight into its computers.

What the Tabulation Center and OAS mission didn’t address is how many more of these innocuous-looking but false returns had flooded into the counting room without their noticing. One can be sure that the fraud was not conveniently confined to one race in Ouanaminthe. The government’s own presidential candidate now has 55,583 votes from all over the country marked for quarantine. The government party had effective control of the electoral apparatus in many places. To arrive at a realistic vote total, one must abandon the narrow confines of the Tabulation Center and reconstruct what actually happened at some polling places, interpolating from there.

The Haiti Democracy Project has proposed to the electoral commission and U.S. government the systematic collection of the carbon-copy returns of party poll watchers, to compare against the posted results. Since the poll watchers kept their own handwritten tally sheets and followed the count closely, their returns (or tally sheets) should be good. They are the last independent record of what happened.

The responses of the electoral commission and U.S. government were revealing. Laurette Croyance said that the candidates could use their copies of the returns for their challenges. Asked about the CEP’s responsibility to verify the returns before publishing them, she replied that there were eleven thousand polling places. Verification was the task of the candidates. As for the U.S. government, it simply didn’t know the situation.

The collection of this independent record faces many obstacles. It would take a systematic effort. Their production is at the discretion of the candidates. Many candidates turned in their only copies when they filed their challenges.

Nevertheless, this independent record exists and holds the key to verifying the November 28, 2010 election.

In its absence, a final recourse is comparison with a known good standard.

The Haiti Democracy Project deployed fifty electoral observers in Ouanaminthe. Because of the number of observers, the vigilance of opposition candidates, and the extent of popular demonstrations, the ruling party was unable to make the fraud stick there. They certainly tried. Their rent-a-mobs wielding rocks and bottles ran through the voting-center courtyards. They fired shots in the street. Yet because of the opposition’s vigilance, the original presidential vote was preserved. This gives us a valuable base for analysis.

Célestin gets only 14 percent of the 14,663 votes from this town. Yet Ouanaminthe is in no way exceptional in its political leanings. Célestin and the ruling party have no more following anywhere else in the Nord’Est or Nord.

Ouanaminthe then with its 113 polling places, although only 1 percent of the country, is a starting point to understand what happened in the Nord’Est province, and from there the rest of the country.

Just down the road from Ouanaminthe, a Unité deputy candidate was busy on election day: 58 percent for Célestin in Capotille and 68 percent in Mont-Organisé. The candidate made sure to pad his own totals by 200-plus votes as well. The fine-toothed comb of the experts eliminated not a single vote for Célestin in either Capotille or Mont-Organisé. Much the same story around the province: Unité candidates raised the Célestin vote to 38 percent in Sainte-Suzanne and 47 percent in Carice, 28 percent in Ferrier, 31 percent in Vallières, 24 percent in Fort-Liberté: altogether 35 percent around the rest of the province, as against 14 percent in our known good sample of Ouanaminthe. Applying this control, Célestin’s total from the province would shrink from 17,816 to somewhere near 8,463, a broad-brush approach but at least one proceeding from real information. The statistical exercise of the experts did not reduce his percentage from this province at all, in fact raised it.

As with the Nord’Est, so with the Nord. Célestin gets 7 percent in Cap-Haïtien and l’Acul du Nord and 6 percent in Plaine du Nord. But even after the passing of the experts’ fine-toothed comb he gets 42 percent in Milot, 64 percent in Grande Riviere du Nord, 47 percent in Saint-Raphael, and 45 percent in Pignon. The difference here is not the sentiment of the people but, as in Capotille, Mont-Organisé, and Sainte-Suzanne the hyper-activity of actors like palace favorite Moise Jean Charles and Nawoon Marcellus and the unobstructed field in which they could work. One fondly recalls Jean-Charles’s capture of Milot in the 2009 senatorial elections when he claimed 12,158 of 12,784 votes cast in this town, reflecting a nearly 100-percent turnout as against 8 percent for the rest of the Nord Department. The Tabulation Center accepted these risible figures and it took weeks of vigorous protest by the opposition candidate and civil society to get them disallowed.

Certainly, the Haitian people, who knew whom they had voted for, heaved a sigh of relief when the experts released their result. And surely, the experts did a better job than the politically-motivated OAS mission on which they were superimposed. But we can only judge a recount by the degree to which it captures the voting reality. A work of experts of that caliber should have done more than tinker; it should have torn through the curtain of suspect returns, salvaged the independent documentation, and assembled a base of real numbers on which to reconstruct the count. Haiti would be on far safer ground if it had.