Sunday, August 22, 2004

A mechanism for an electoral electronic fraud in Venezuela

Considering the number of mail I have received and the observations left in the comments section, I think it would be good to give a more detailed example on how I think that the electoral fraud could have been committed. A little bit as a follow up of the diagram that I posted a few days ago and that seems to have helped a few folks.

The first thing, and I must stress that, even if a fraud has been made by the chavista side, this does not mean that it did not win the vote. And that is the saddest part of all! They might have won anyway but no matter what happens in the future chavismo electoral victory is tainted, with all the consequences for political stability that this represents.

This presentation is done in three parts, plus conclusions and what can be done.

1) how the fraud parameters were set

2) how would have this affected a favorable result for chavismo

3) how would have this affected a victory of the opposition

1) An example of how the fraud parameters could have been set

The first thing to keep in mind is that we are talking of alterations done to the electronic components of the voting machines. Only counting ALL the ballots printed by the machines could clear the air (first hint that this is a possibility is the CNE fighting real, real hard to minimize as much as possible any ballot counting: if this one is so certain of the results while not just go ahead and count all the ballots and shut up the opposition once and for all with a crushing victory?).

Let's imagine the district of Tucusiapon where the glorious revolutionary leader Pepe Bimba was born. He wants to make sure that the district remains in chavismo hands for personal pride. Unfortunately the polls indicate that it will be a tight race. What to do?

First, examine the vote distribution. The electoral system works as follows. There are three polling stations with a variable number of machines depending on the population. One polling station, A, is in a pro Chavez section of the town. This one has 4 voting machines. Polling station B with three machines is an electorally ambiguous part of town. And the opponents are concentrated in the district of polling station C with 3 machines.

Second, predict the possible outcome at these 3 districts. For that one has access to polls, local informants and any signatures that were collected for the Recall Election petition.

Third, set a top for the SI at several machines. Thus we obtain the table below. We can see the total voters per machine. For the sake of simplicity I have omitted abstention and assumed that since it is a small community everybody will vote (remember, this is a simplified imaginary situation set to illustrate how the fraud mechanism could have worked; including abstention would make the table more complex while strengthening even more my point). The next two columns show the expected SI and NO votes.

Next column is the difference between predicted SI and NO, the NO advantage,used to set the "top" SI. The programmers selects two machines in the A district because it is politically important that all the popular districts show a clear victory for the NO. Then two machines in B to avoid "surprises". Finally, to make sure the SI are happy, as they will the expected a victory in the C station, it will be granted to them but the total will be limited by a 76 top.

total votersExpected SIExpected NONO advantageset topsSINO
Polling station A 12359645625964
151619029 6190
131557621 55 76
Polling Station B 11757603535364
1015150-1 5150
10552 531535253
Polling Station C 1428062-18767666
Results 129263365926 617675
49.0%51.0% 47.8%52.2%

The reader can observe that only based on the expected results the fraud would be undetectable. Polls are predicting a narrow victory for the NO and this one gets a little bit more than a narrow victory. You can observe that at no point there are two identical results in any station. There is only two 64 NO but in different stations, something quite probable. The 26 vote advantage of the NO is magnified to 58, a believable advantage and a solid protection in case the SI get a little bit better. If the NO wins by a better margin, the fraud would be diluted and undetectable even if ballots are counted.

So, what happens when the results are not as expected? Namely when the SI get more than what was predicted.

2) Chavismo wins by a slightly lesser margin than expected

This is where problems starts. The next table show the results with a narrower than expected NO victory.

total votersReal SIReal NOtopsReported SIReported NO
Polling station A 1236063626063
1516388 6388
1315378 5378
Polling Station B 1175760535364
1015348 5348
Polling Station C 1428062767666
135795676 7659
Results 1292642650 622670
49.7%50.3% 48.1%51.9%

The "real" columns are the results that are in the ballot boxes, if counted. The "reported" results are the ones reported from the "ACTAS" printed by each voting machine when voting is over.

The total NO result indeed is lower than what the programmer had prepared but still more than one percentile point than the real result. What went wrong in the prediction here is that machine 1 on station B had more votes than expected, as well as all the machines in station C. The consequences is that some results are identical! The two 53 of station 2 are coincidental, but the three 76 vote results of station C become suspicious!

There is also a NO repetition BUT again in different stations!

In this case the "fraud" could be considered acceptable and the opposition would probably accept it since it also had exit polls that told it that the election was very close in Tucusiapon. No ballot counting might be asked in spite of the strange coincidence. But what happens when the opposition wins more decisively?

3) The opposition wins

In this next table the opposition wins. In popular districts chavismo did not do as well as planned and in the opposition districts this victory was wider. Again, the first two columns indicate the real SI and NO votes and the "reported" SI and NO come from each machine acta.

total votersReal SIReal NOtopsReported SIReported NO
Polling station A 1236558626261
1516586 6586
1315576 5576
Polling Station B 1176156535364
1015645 5645
Polling Station C 1428062767666
Results 1292663629 631661
51.3%48.7% 48.8%51.2%

Now the possibility of fraud becomes more obvious. Exit polls indicated a victory for the opposition and yet it loses the vote. But now coincidences appear in station A (two 62) and in station C (three 76). Note that curiously this time the accidental coincidence in station B disappears. And note that the coincidence in the NO result is the same as before, two 64, BUT in two different stations!

If it were not for the three 76, a casual observer would have found all fine, even though the rigging resulted in an outright switching of the results! And if the result had been a little bit more favorable, the coincidence in station B might have appeared as the third machine would have reach the 53 set top!

4) Conclusions

Proving fraud is not going to be an easy task. However it can be done beyond any doubt if fraud indeed has taken place, without even the need to count the ballots! "How could that be possible?" the reader would ask.

Well, the reader may start by observing that the "coincidences"
in the simple case presented here are for the SI votes IN SAME POLLING STATIONS!
This is very different from the similarities observed for the NO across
polling stations.

Second, it should be obvious from this example that the frequency of similarities between SI results should be comparable with the similarities between the NO results. That is, if 20% of voting stations show "strange" SI results, 20% of voting stations should also show "strange" NO results. If the "strange" result between SI and NO is too elevated then there is a problem that should be investigated. And if it is way off any statistical scale, then the election should just be voided. In the speculative case here presented the SI coincidence are more elevated than the NO coincidence AND have a different distribution pattern.

Finally, it bears repetition that EVEN if fraud is proved beyond all doubt that does not mean that Chavez lost the Recall Election! But it will be a total discredit to his rule, creating considerable difficulties for Venezuela as other countries would stop trusting a government that is able of electoral fraud. And the chavista section of the CNE should go to jail.

5) What can be done?

If the CNE is absolutely certain of its results, then count all the ballots! Chavez international stature could only benefit from such a move.

While the CNE stalls, the opposition must do a thorough analysis of "coincidental" SI and NO results, and of abstention results (not shown here but to which the same statistical rules should apply). If the numbers that will eventually be presented indicate statistical improbability, then the OAS and the Carter Center will have to swallow hard (have they have done elsewhere in the past)and declare that the result of the election is questionable. Then maybe a political compromise could be reached to make sure that the regional elections are fair and we will have to wait until 2006 for the final and decisive show down.

If the statistics are not convincing, well, the result will still be tainted and the opposition will have to accept it nevertheless. What can it do after is everyone's guess.

Or, no matter what, Chavez will say "see if I care!" and then all is possible.

PS: a piece of unsollicited advice for the readers of this blog: if you live in a system with electronic voting with no paper trail, I would suggest that you get seriously involved in your community groups checking on it.

No comments:

Post a Comment

Comments policy:

1) Comments are moderated after the sixth day of publication. It may take up to a day or two for your note to appear then.

2) Your post will appear if you follow the basic polite rules of discourse. I will be ruthless in erasing, as well as those who replied to any off rule comment.