The Carter Center version
Jennifer McCoy, who led the Carter Center delegation (well, apart from Carter himself) also has a piece in the Economist this week, which details their findings. I find it pretty convincing. Here are three key paragraphs:
So, were the audited mesas selected at random? And whose statistical model are we supposed to believe?
The only way the boxes could have been altered would be for the military—historically the custodians of election material in Venezuela—to have reprogrammed 19,200 voting machines to print out new paper receipts with the proper date, time and serial code and in the proper number of Yes and No votes to match the electronic result, and to have reinserted these into the proper ballot boxes. All of this in garrisons spread across 22 states, between Monday and Wednesday, with nobody revealing the fraud. We considered this to be supremely implausible.
This second audit showed that the machines were very accurate. We found a variation of only 0.1% between the paper receipts and the electronic results. This could be explained by voters putting the slips in the wrong ballot box. An additional piece of corroborating evidence was the result from the 15% of polling stations that used the old-fashioned manual ballot. These stations (in mostly rural areas without telephones) were even more favourable to the president, voting 70:30 against recall.
If the machines were accurate, how do we explain the three suspicious factors noted by the opposition? First, the mysterious “tied” results or “caps” on the machines. We found that 402 of 8,100 mesas (each with one to three machines) had two or three machines with the same result for the Yes vote; and 311 mesas had the same results for the No vote. So the phenomenon affected both sides. We consulted Jonathan Taylor, a statistician from Stanford University. Using various mathematical models, he predicted that 379 mesas would have ties (of two or three machines) in the Yes votes, and 336 mesas would have ties in the No votes. The error range would be plus or minus 36 mesas. So the actual results fell within the range of probability, and do not provide evidence of fraud.