# Has Ouattara won? Here’s some strong evidence he has

February 6, 2011 32 Comments

### Much has been written about the late 2010 elections in Ivory Coast and although there has initially been near unanimous support for and acceptance of Ouattara’s win by the international community, South Africa’s Zuma and Uganda’s Museveni have recently decided Gbagbo’s claim of fraud in the North deserves further scrutiny and would thus like to see an investigation. Because of these requests I have decided to look at the publicly available data myself with a view to answering the question: Was it fraud in the North that caused Ouattara’s win?

In order to do so I have eliminated the contentious second round vote in the North from my considerations. I have found a way of doing so by taking the first round election results which are considered credible as the basis of my analysis. As the first round had twelve additional presidential candidates running who together achieved 30% of the vote, the question that needs to be answered is who have their voters supported in the second round when they had to chose between only Gbagbo and Ouattara? To determine this trend I have completely ignored the second-round Northern vote and fully concentrated on the vote in the Gbagbo-controlled South which most certainly is free of electoral fraud and intimidation benefiting his opponent. It turns out that in the South 65% of these “new” voters have on average voted Ouattara, leaving 35% to Gbagbo. I think it fair to assume that Ouattara would have got at least the same percentage in free and fair elections in the North, and so it’s fair to apportion a minimum of 65% of these voters to him country-wide. Remembering that Gbagbo achieved 38% in the first round and Ouattara 32% that leaves us with a couple of simple mathematical formulas to work through:

Ouattara: 32% + 0.65 * 30% = 51.5%

Gbagbo: 38% + 0.35 * 30% = 48.5%

We see that even when we take only the Southern part of Ivory Coast as an indicator of who the twelve eliminated presidental candidate’s voters have chosen in the second round, Ouattara is still ahead with a tidy 3% margin. That’s more than six times the 0.46% margin Mills achieved in the Ghanaian election 2008. To my knowledge this is the best evidence yet – based on only publicly available data – that Ouattara really did win the election. The argument derives its strength from the fact that it is based solely on data approved by the country’s Constitutional Council, headed by Gbagbo’s election campaign manager Yao N’Dre. Feel free to verify the numbers and assumptions that went into the calculation yourself.

Even if there were reasonable doubts that Ouattara has won the election – and I don’t see any – there are much, much graver doubts that Gbagbo has won it. It is nothing but preposterous for him to stay in power claiming a mandate on the basis of these election results. It is even more disappointing to see that African democracies South Africa and Ghana still refuse to fully support President-elect Ouattara.

Brilliant!!!!!

Pingback: Côte d’Ivoire: Did Alassane Ouatarra Really Win? @ Current Affairs

Your “analysis” is severely flawed. Not only does it NOT use any actual political statistics methodology, but it is also based upon completely faulty assumptions. It’s great that you want to use this simple mathematical formula to “prove” your point, but unfortunately—it’s incredibly arbitrary and incorrect. I could use a number of simple mathematical formulae to “prove” the exact opposite based on a number of other possibilities, and they would all be just as faulty.

You cannot arbitrarily decide that the first round would result in the same voting patterns as the second. This is why we have a second round. You can not also arbitrarily move back and forth from one vote to the other to grab your numbers when it suits your fancy. Your “formula” for figuring out the 65% to Ouattara in the second round and 35% to Gbagbo is completely incorrect. Let’s just look for a moment at each area.

During the second round in Agneby, Gbagbo got roughly 84% of the vote and ADO got 16%. In Dix-Huit Montagnes, Gbagbo got 51%, while ADO got 49%. In Fromager, Gbagbo got 67%, while ADO got 33%. Haut-Sassandra, Gbagbo got 57%, ADO got 43%. Lagunes, Gbagbo got 52%, AGO got 48%. In Lagunes outside Abidjan, Gbagbo got 71%, ADO got 29%.In Marahoue, Gbagbo got 55% and ADO got 45%. Moyen-Cavally, Gbagbo got 64% and ADO got 36%. Moyen-Comoe, Gbagbo got 54%, ADO got 46%. Sud-Bandama, Gbagbo got 62%, and ADO got 38%. Sud-Comoe, Gbagbo got 63%, ADO got 37%.

So please tell me how it is possible, by looking at these numbers (where ADO has not gotten more than 50% of the vote in ANY of these regions), that Ouattara has obtained 65% of the Bedie /other candidate vote? Your math here is very, very wrong. Not that that even really matters, because it is based upon assumptions that are faulty to start with.

The best way to “prove” that Ouattara won—have the UN release the tallies from each and every polling station in the country and let us actually run the numbers properly. The fact that they are unwilling to do this is only compounding the crisis. Instead of calling a winner, they should have called for transparency of numbers that would allow proper scrutiny by the public. Now, we are left with numbers that don’t even add up properly in their current form (Incrits do not match the UN number of registered voters, percentages don’t up in a bunch of places, participation doesn’t add up, etc.) and are left thinking that something is fishy, when it could easily be proven with the release of the proper numbers.

I suggest you leave the mathematics of politics to those who have actually studied methodologies, because it is clearly not your strength.

I was in a rush, and when I tried to copy out the numbers in my above comment it was missing a chunk there that now appears to be lost. No matter.

So, after work today when I had the actual time, I took a look at the numbers and began to crunch them— and I found quite a different result than you would have us believe in your post.

Firstly, you don’t take the simple difference in first and second voting amounts and calculate the percentage increase. It’s more complex than that– that’s why people take years of stats classes before being able to properly calculate and interpret this type of data. Secondly, you have to actually work it out region by region, instead of averaging it out— which is dangerous to do considering that some regions are significantly differently populated than others (for example Lagunes region which has nearly ten times the population of some of the other regions). It’s actually much closer to a split percentage and even that would be dangerous speculation based on the assumption that the first and second rounds had consistent voting patterns AND it also ignores the problem of the changes in null votes as well, as there was a major difference in the number from first to second round which could significantly alter the results. Not to mention that the numbers on the tally online don’t add up themselves in the first place, making the numbers there suspect.

Yes, simple calculations taking the difference in voting numbers for each candidate from one round to the other, then figuring the percent increase can show you 65/35; but it’s not proof of anything actually statistical. Sorry.

“prove” the exact opposite based on a number of other possibilities

I was talking of “strong evidence”, not “proof”. I agree that you can set different assumptions and get to a different result. I didn’t mean to suggest you can’t. You could e.g. say that Bedie (and other) voters in the North have a strong dislike for Ouattara and are therefore much less likely to vote for him than those in the South. You can then assume only 40% of them voted for Ouattara and you have a generous win for Gbagbo. It does seem to me however – without any evidence to the contrary – to be a fair assumption that Ouattara got the same or more from Bedie voters in his Northern strongholds than in Gbagbo’s Southern strongholds. I’m saying “or more” because I have observed myself that when you live in an area where people predominantly hold political views different from your own and you befriend some of them, and you have conversations with colleagues, shop-keepers, people at the market and fellow travellers, you watch their television programs and you read their local newspapers, then you invariably become more open to their political views. So I expect all else being equal Bedie and other voters who live in the South predominantly with Gbagbo supporters to be more likely to side with Gbagbo and those living in the North to be more likely to side with Ouattara. It would be interesting to see if we could glean something from the data to support or refute that view.

You could also say that Gbagbo had such a fantastic election campaign in November that he must have won 10% of Ouattara voters over, or the opposite. I happen to think that assuming the Ouattara and Gbagbo voters did not change their view is a fair assumption unless we have evidence to the contrary. You attack my assumptions in general a bit later but you do not explain sufficiently what you see as wrong with them for me to be able to respond.

I haven’t picked the data “as it suits my fancy”, instead I have responded to the allegations of fraud in the North in the second round and used the full set of remaining data to piece a full result together. I haven’t picked a few regions here, a few regions there and then arbitrarily calculated my favourite result, instead I have taken the full set of data that is considered credible by both parties remaining in the race for the second round.

Your “formula” for figuring out the 65% to Ouattara in the second round and 35% to Gbagbo is completely incorrect.

Please explain how it is wrong. I do not understand how the data you enumerate proves me wrong. The fact that Ouattara did not get more than 50% of the vote in total in any of the Southern regions doesn’t say anything about how Bedie’s (and other) voters have chosen to vote after most of them have been urged to vote for Ouattara. Bedie and Ouattara are in coalition and a vote for Ouattara was at this point the only way for these voters to get Bedie into a position of at least some power. I would expect them to consider this a strong incentive to vote for Ouattara and I’m not surprised that the data shows the majority in the South did do so.

I understand your gripe with regard to the unavailability of data but please do note that what I am doing here is use the available, credible data for analysis. I would be grateful if you could point me to the information you have on: “Incrits do not match the UN number of registered voters, percentages don’t up in a bunch of places, participation doesn’t add up, etc.” While I have read about something along those lines, I haven’t found any quantification or localisation (just in the North or everywhere?).

I took a look at the numbers and began to crunch them— and I found quite a different result than you would have us believe in your post.

The reason I published all the gory detail of my analysis in that Google Spreadsheet including data, formulae and assumptions is so that it’s up for discussion and critique. Could you please do the same with yours?

you don’t take the simple difference in first and second voting amounts and calculate the percentage increase

Why not, and how would you do it correctly?

You say I “have to do it region by region”. What I have done is that I have checked how much Ouattara needed of the Bedie + others vote in the North in order to win and found that number to be just over 60%. I then looked at the twelve areas in the South and found that Ouattara has been over 60% in two thirds of these (all except Agneby, Fromager, Haut-Sassandra, Sud-Bandama). It does show that different social cleavages can lead to different results: If I had based my calculation on just Agneby and projected that onto the North it would have shown a win for Gbagbo. But I do believe basing the projection on the maximal amount of data that I have is the most likely approach to yield a correct result.

It’s actually much closer to a split percentage

Can you please show us the “region by region” calculation with which you arrived at this?

it also ignores the problem of the changes in null votes

Can you please elaborate on the issue you see with this and propose how to deal with it. I suspect we have a similar situation with the changed number of Registered Voters. I think it might well be relevant but I do not yet see how.

Not to mention that the numbers on the tally online don’t add up themselves in the first place, making the numbers there suspect

I have seen some instances of the numbers as shown by abidjan.net not entirely adding up but no huge differences which would affect the result. As to whether the numbers are suspicious or not I’ve taken the view that they have been confirmed by the Independent Electoral Commission, the UN and the Constitutional Council, so they are the best set I have available. I have tried to be clear, honest and exhaustive in my approach and I have certainly done a better job than the Constitutional Council that brute-forced seven districts, a total of roughly 600,000 Northern voters out of the election, thereby swinging the result. Your help to make it even better would be more than welcome.

Hi,

Unfortunately there is fraud which invalidates the results of the trend.

According to this site ( http://resultatelectionpresidentielle2010-cotedivoire.ivoire-blog.com/archive/2010/12/14/des-anomalies-troublantes-dans-les-chiffres-de-la-participat.html ), election results were proclaimed by the President of the IEC and the media with a 70% of voters on November 29.

Three days later, this same president proclaimed a participation rate of 81% (Hôtel du Golf).

Why the participation rate is not close to the one announced on November 29 ?

How to calculate the true trend without the real numbers ?

Palop

Yes, I have read the reports of an estimated 70% participation. What’s news to me is that this was officially pronounced by the IEC or in fact that anything regarding the results was announced by them the day after the election apart from perhaps the vote abroad. When I considered this I came to the conclusion that this type of fraud would have to have been nearly countrywide because the only regions apart from Etranger (Abroad) that are below, say, 75% are N’zi Comoe, Bas-Sassandra, Zanzan. (All btw being areas in which Bedie (and/or potentially other eliminated candidates) had taken a significant part of the vote if my cursory look is anything to go by, the significance of which I’d have to think about.) And the turnout overall was higher in the South than in the North. I think that’s pretty hard even for a pro-Ouattara IEC to pull off in the South. And this fraudulent behaviour would have to have been done such that the UN doesn’t notice or the UN must have been complicit in it. Well. I think it is a fair point and you could be right that this has occurred but it does seem unlikely to me and that is why I upheld my assumption that it’s possible to glean something from this data.

Look at the news from the 28th until the 1st of December or so.

For example, an English site for you: http://www.english.rfi.fr/africa/20101129-cote-divoire-announces-first-partial-vote-results. One of MANY news sites that says this. We heard this all around within the country and numerous foreign media outlets also printed this– that is up until December 2nd, when suddenly it changed. The CC results have listed a 71,28% turnout for the second round, which I think is quite closer to the actual turnout.

Voter turnout was higher in the south than the north? Um. No. It wasn’t. I suggest you actually calculate this. There were more voters in the south than the north, but the turnout was higher in the north. If you make a trend of the change in voter turnout from the first to second rounds using the CEI results, you see that in the southern districts, the turnout decreased, whereas, in the six more northern areas, voter turnout actually increased in the second round. Voter turnout in Denguele in the first round was 84.35, up to 91.59% in the second. In Bafing, it was 81.57, up to 83.66% in the second. In Savanes it was 84.26% in the first and 88.12% in the second. In Worodougou, it was 83.68% in the first, and 90.54% in the second. The only two more northern regions to have a decrease in voter turnout were Vallee du Bandama and Zanzan, who both went from around 81% to 77%. In the southern regions, only ONE region experienced higher turnout in the second round than in the first– Agneby. All the rest supposedly had lower turnout rates. The average turnout rate in the northern regions is also higher than that of the southern regions.

“And this fraudulent behaviour would have to have been done such that the UN doesn’t notice or the UN must have been complicit in it.” The UN had how many observers for more than 20,000 polling stations, and how many of those were in the north and how many in the south? I think it’s quite possible to say that they could have missed quite a lot while monitoring less than 5% of the polling stations.

I think it is still quite reasonable to suggest that Ouattara won, but as I said before, it makes little difference or not at this point what the actual results say if Gbagbo is unwilling to accept them.

For some reason wordpress doesn’t allow me to respond to your post directly. There’s no Reply button on it, so I’m responding to my own, hoping it’ll end up directly underneath yours anyway.

Many thanks for the RFI link, it does confirm that even the IEC announced a 70% turnout and I agree that does bring up some very valid questions as to what has caused the difference between the first and the second announcement and whether something’s fishy. At the same time I wonder why the Constitutional Council hasn’t picked up on this and worked on credibly discrediting the vote results on the basis of that evidence. If they had seen a chance to do so, surely they would have done it?

The CC results have listed a 71,28% turnout for the second round, which I think is quite closer to the actual turnout.

But what’s your point here? Surely you don’t agree with the view of the CC, as expressed by the invalidation of 600,000 votes in seven districts in the North (and confirmation of all the other data), that voter turnout was zero in these seven districts but in Abidjan and elsewhere exactly as announced by the IEC — going straight against your own anecdotal evidence?

Voter turnout was higher in the south than the north? Um. No. It wasn’t. I suggest you actually calculate this.

I have calculated it and published my calculation on a Google Spreadsheet linked to in the originial article, but there I’m including all areas won by Ouattara (incl. Bas-Sassandra and Etranger) as “North” even though they aren’t really of course. I’ve done this to avoid having them in the trendsetting calculation and have people then complain that because they were won by Ouattara the results there were potentially fraudulently obtained (it weakened the trend to 65% when I moved them North, so that was to the benefit of Gbagbo). When I delete the Etranger line and copy Bas-Sassandra back over to the South I get 2nd round voter turnout of 81.87% (South) > 81.18% (North). Is the abidjan.net data different from the data you are basing your calculation on or have I made a mistake?

I think it’s quite possible to say that they [the UN] could have missed quite a lot while monitoring less than 5% of the polling stations.

Yes, but what I mean is that as far as I remember the UN got copies of the same tally sheets that the IEC got straight after the election. If the IEC found a first rough estimation of 70% turnout on the basis of these sheets a day later and then fraudulently fiddled the numbers by adding tally sheets or changing the numbers on the tally sheets then that would have to have been noticed by the UN. But if the tally sheets were already fraudulently changed before the IEC and the UN got them, then it’s hard to understand how the IEC concluded 70% voter turnout on the basis of the fraudulent sheets that later showed >80%.

Also, I can imagine tally sheets to have been fiddled with for the benefit of Ouattara in the rebel-controlled North prior to them reaching IEC+UN but it’s hard to see how they would have been able to do so on a grand scale in the Gbagbo-controlled South in the middle of a curfew. And I’ll say it again: the overall turnout is higher or at least about the same there.

As to the UN covering less than 5%: I suspect that what they had to do is strike a balance between perfection (observing all) and affordability. Could it not be that their presumably well-trained statisticians that you suggest I leave the mathematics of politics to, have decided that the less than 5% they’ve chosen to look at represent a statistically relevant subset likely to help them confirm or deny the full results they see?

“At the same time I wonder why the Constitutional Council hasn’t picked up on this” They have. And no, I certainly don’t agree with the invalidation of votes in the north. And the turnout isn’t the same in all the other districts between the CEI and CC versions… it’s quite different actually. Thousands apart in each region.

I’m using the data based on the abidjan.net published results—and that is not the result I have found… The north had a higher voter turnout overall, as well as witnessed an increase in almost all regions from the first to the second rounds as I pointed out in my last comment. The south, on the other hand, showed a decrease in voter turnout in all but one region from the first to the second.

What Kind of rubbish is it that because some news channel, RFI, publishes an estimate that was actually announced on the night of Sunday the 28 Nov to Monday 29th Nov that we now have fraud?

The 70% estimation was given by the CIE and UN and everyone for that matter

even as voting was in progress..An estimation is a is an educated and calculated guess based on a small random sample of voting stations. This seems to have been wrong and was corrected as more results came in. Simple, very simple… I really don’t see any basis for fraud allegations here!There was fraud in the other side too,

The CC overturned the results of the 3 most important ADO areas, based on nothing.

They should have restarted this election, they should have done this, we would not be talking about pretty – not so obvious – trend.

It was the best thing to do.

Palop again, the french guy

Actually, you have picked and chose your data. I’m not meaning to suggest you were doing this with any malicious intentions, but rather assume that you are not a political scientist or statistician and so did what most uninformed persons do when trying to “analyze” this type of data. Unfortunately, it doesn’t work out to actual statistics, but rather layman interpretation—and should be taken as just that.

“I do not understand how the data you enumerate proves me wrong.” As I said in my second comment, the full numbers I put there did not get copied—but I’ll reiterate again, if you do not know stats, it’s next to impossible to explain how to run a “proper” analysis in a tiny comment. Most statisticians are in school for years to understand how to run the numbers properly. If it was as simple as taking the difference in voting numbers and figuring the percent increase, nearly everyone could be a political statistician. Basically, you’d have to run data looking at the percentage point difference across the board, and run it district by district factoring in several different variables that were not included in your analysis; but unfortunately, with the weak data that is provided, the margin of error is too high for this to even be considered realistic in the slightest. The lack of data is not your fault, but unfortunately, makes this type of analysis essentially useless.

“Incrits do not match the UN number of registered voters, percentages don’t up in a bunch of places, participation doesn’t add up, etc.” While I have read about something along those lines, I haven’t found any quantification or localisation (just in the North or everywhere?).” It’s simple. Tally up all the incrits, district by district from the CEI version available online (the CC version doesn’t have incrits available I believe—so we can’t calculate using their version). What’s the number? Now, look to the SRSG’s certified number of registered voters from before the election. You will find they don’t match. There are tens of thousands of extra voters from across the country. This has nothing to do with the extra voters in the North— as this information has never been made available publicly for scrutiny. It could easily be dispelled if, as I said above the UN released the numbers polling station by polling station. We could easily see if the numbers were off then. This however, is a problem. Where did these extra voters come from—who are they—and why did they get to vote if they weren’t on the registered voting list?

Again, if you go across the numbers in the CEI or the CC versions provided and manually calculate percentages, and several other numbers—there are quite a few that have been added up incorrectly. This is clearly visible across the board. Either, they were entered incorrectly, or there has been some fudging of the numbers—I don’t know the reason—but it does make a difference.

I don’t have a website, and I haven’t the foggiest of how to add a spreadsheet into a comment, if that’s even possible. As I said above however, it makes little difference as the margin of error is so high that it would make it useless anyway. It wouldn’t “prove” anything, other than the fact that this type of calculation can’t be realistically done with the information provided because the margin of error is too high.

As for the null votes; if you calculate the percentage of null votes during the first round and compare them to the second round, you will notice that the difference between the two rounds is significant. This will also affect the margin of error for the calculations, as there were more than a hundred thousand null votes in the first round than in the second. These people could have voted for anyone in the first round, and obviously, many more of their votes were counted in the second round… that is not factored into your equation in the slightest. In some regions, there are nearly 15 times less null votes in the second round than in the first. When you start adding up all these problems and factoring them in—you begin to realize the margin of error is significant.

I’d have to agree with Palop as well. I currently live in Abidjan, and nearly everyone I knew voted in the first round, but skipped the second round because of fear from intimidation and other factors. I also saw significantly shorter lines in the second round from the first round at every polling station I passed. When I heard the announcement originally of around 70% participation—I thought even that was high… but days later when I heard the CEI say there was nearly 81% participation—I thought, there is NO way. Something is fishy here. The local monitors cited around 70% participation (http://www.societecivile-csci.org/images/stories/pdf_MOE/Declaration%20du%2030%20novembre%20%202%20tour%20election%202010.pdf), as did every media outlet until several days later when suddenly it was nearly 10 points higher.

As much as I may wish Ouattara had won over Gbagbo (though frankly I hate both and think both are going to run this country into the ground, but Ouattara is the lesser of two evils)—the data available is not all that credible and so no one can say with any degree of certainty. The CC acted inappropriately and should have taken time to investigate the numbers and if necessary call for a re-election. They didn’t do that, and so most likely, they realized Gbagbo was finished. The UN also behaved atrociously during these proceedings and not at all in the spirit of diplomacy. I was thoroughly disappointed in them. The EU’s recent report is also thoroughly disappointing and full of clear mistruths, which frankly I can’t understand. It’s not as if one needs to make up fodder to complain about Gbagbo—there is PLENTY there for that, but their report is just pure bullshit.

I wish people outside the country would stop trying to be the “voice” for the Ivorian people. The Ivorian people are speaking their mind, and are capable of handling their own situations. They are not helpless. If the population wanted to be rid of Gbagbo so thoroughly, they could easily take to the streets as the Egyptians and Tunisians have. People just want to get on with their lives here, whoever is in charge. All these international bodies are only escalating problems here, when they could have been a voice of reason. Their outside sanctions and other moves are hurting the people, not Gbagbo. Cocoa farmers are suffering severely. Workers are being laid off and having difficulty getting their pay. Had the UN, when faced with the initial crisis, not taken sides, but rather said—look—we have the results from each polling station—here it is—there wouldn’t be soo much doubt. The public could scrutinize it to an extreme. If there were irregularities that couldn’t be resolved—then they could urge a re-election. Now, recounting and scrutiny are off the table. Re-election is off the table. Any real negotiation is off the table. There are fewer options each and every day. Frankly, I don’t see the point in trying to “prove” Ouattara won the elections at this point. It will not make one iota of difference. Whether he won or not is irrelevant to the situation on the ground and will not make the situation any more peaceful. We’re beyond that now. Even if one could prove beyond a shadow of a doubt that Ouattara won, would Gbagbo suddenly step aside? Nope. So now it’s time to work at the crisis from other angles and leave the results in the past.

I have also read I think on the apeaceofconflict blog about the mismatches and extra voters and if this has occurred in a significant way it does of course call the data into question. If you could provide a link I’d be grateful as my Google mojo seems to elude me at the moment.

The anecdotal evidence that I have is that people were galvanised for the second round and all of the ones I personally know and have talked to, returned to vote (although one’s parents couldn’t because of intimidation in the West).

I don’t know why you keep characterising what I’ve written as an attempted “proof”, with double-quotes for increased hilarity. As I said, it was never intended as such. I’m not proposing this as a new method for the UN where they are certifying elections, I would hope they have more complete data and indeed fully trained staff for the purpose.

null votes… that is not factored into your equation in the slightest.

Yes, that’s a fair point. Thank you.

In some regions, there are nearly 15 times less null votes in the second round than in the first.

I did notice that the difference is significant being reduced by more than half in the second round. In a country with 18% illiteracy (CIA Factbook) I find it credible that the percentage of null votes we have seen occurs and that when they try again a month later with only two instead of 14 options they will improve. Perhaps there were a few eager people in some areas going out of their way to help other voters get it right the second time, explaining the major improvement in some of the places but not in others. Do you think the outliers are an indication of fraud?

To publish a spreadsheet you only need a Google account and go to their Docs & Spreadsheets, then upload your spreadsheet (or start a new one from scratch), open it, set Sharing at the top right and copy the link you see at the top of your browser in a comment or email.

If the population wanted to be rid of Gbagbo so thoroughly, they could easily take to the streets as the Egyptians and Tunisians have.

Easily? That seems like a cynical comment to me. Gbagbo has his way of dealing with demonstrators. Even allowing for the fact that the idea of “peacefully” taking the TV station was bordering on the naive on Ouattara’s part and has contributed to the result of this, the 16th of December and after have shown the Liberian or Angolan mercenaries have been hired for a reason: to kill when the need arises. And with 1000s of Young Patriots reportedly being armed and trained it doesn’t seem like a great idea to go for a demo. I’m told that there are specific measures to stop people from gathering in groups at least in some areas and that it’s not uncommon to be stopped and beaten up for nothing but the hint of an attitude. People are routinely killed. This is anything but “easy” to pull off and while I applaud the Egyptians and Tunisians for having developed the courage I cannot fault the Ivorians for not having done so (yet?). It also seems far more likely to me to lead the country back into civil war than result in a peaceful overthrow of the government. Also Egyptians seemed like perhaps 80% against Mubarak, the split is far more narrow in CIV. The outside involvement (minus Zuma et al) is welcome to the Ouattara supporters I have spoken to and they’d like more not less.

Whether he won or not is irrelevant to the situation on the ground and will not make the situation any more peaceful.

Well, I didn’t exactly expect Gbagbo to step down after publishing this blog entry. I don’t think I can make much of a difference to the situation except perhaps for the people I am close to.

You have to literally express your formula in order for me to understand what you want to do.

For instance :

q is the trend for the south,

so we get,

32% + q * 30% for ADO,

38% + q * 30% for LKG.

Now, you can explain how you computed your trend.

I was stuck with your initial formula because it didn’t mean anything to me.

Now we got this trend, what can we do with it ?

Probably not what you did. I explain :

32% + 0.65 *30 = 51.5 means ADO has won in the very North, but also has won in the very South, and also in the North + South, and so on,.

Trust me. I don’t know how to explain it, but that means nothing.

I’ll try to explain here, a simple way to use the trend.

So let’s take this trend for what it is and try to understand its behave. We apply it on the final results in the South (57% LKG – 43% ADO)

• There were approximately 3 millions voters in the South

• There were approximately 1.5 million voters in the North

time t1: No one has voted in the North. Gbagbo has won with 57%

time t2: ADO needs a positive trend which is the case. But he also needs a number of voters to win this election

• 10 000 voters in the north is too little, he loses

• 100,000 voters in the north is too little, he loses

• 200,000 voters in the north is too little, he loses

• etc …

time t3: There are enough voters. ADO has won.

Do you see what I mean ?

Your trend has to evolved and described a real situation. Your trend has to be use with an amount of voters.

If you don’t do that, you just give me a final result for a particular value.

you viciously just told me that 35% weren’t enough for Gbagbo to win. so what ?

If there were only 1000 people who were living in the north, gbagbo would win anyway.

But your formula won’t stop telling me that Gbagbo lost. Do you see what I mean ?

You can’t use trend the way you’ve used it.

I’m editing this comment for the 3rd or so time, deleting the rubbish I’ve previously written. You had me confused here. But I think it all does make sense.

From our previous conversations on the French blog I think we agree on the calculation that 65% of the Bedie (and other eliminated candidates) vote went to Ouattara in the second round in the South — under the stated assumptions.

So we have these 65% in the South and we have an assumption stating that the Bedie (and other) voters in the North have voted the same (or more). So we’re assuming the same 65% apply in the North to all Bedie (a.o.e.c.) voters. But if 65% have voted for Ouattara in the South and 65% have voted for him in the North, then 65% have voted for him

in the whole country.Then we are using the 1st round result. Everything from the first round: The number of votes for Gbagbo (38% = 1756504), the number of votes for Ouattara (32% = 1481091), the total turnout (100% = 4616817), everything. Then we share the vote that went to Bedie and other eliminated candidates (30% = 1379222) between Ouattara (65%) and Gbagbo (35%) according to the trend we have found from the Southern second round vote.

In percentage terms this is:

ADO: 32% + 0.65 * 30% = 51.5%

LKG: 38% + 0.35 * 30% = 48.5%

In actual vote terms this is:

ADO: 1481091 + 0.65 * 1379222 ~= 2,377,585

LKG: 1756504 + 0.35 * 1379222 ~= 2,239,232

Again, this is valid because of the assumption that Bedie (a.o.e.c.) have voted

across the whole countryas they have according to the second round trend from the South.So, you’re calculating the result from the 1st round and you’re applying the trend of the 2nd round on it.

That was my first thought on the french blog.

Why do you think there is a second round ?

In order to avoid these kind of operations.

And the null votes, what did you do with them ? The people who stayed at home ? I dunno, the whatever things that could have happened those days ?

So that’s it.

Are you aware that if there is just one election in this world, where a candidate with a positive trend is given winner in the first round, he loses in the second round anyway, your formula (which by the way is not – I cannot see one variable) will be completely false.

You got lucky with this one. But what you did was nothing more than a survey after a second round based on your own choices.

I hope you’ll see my point

My apologies for my reaction

Here’s I’m going to calculate the formula for the trend in order for people to understand what they’re really dealing with :

VC1, VC2 are respectively the amount of voters of the first and the second round, for the first candidate.

VD1, VD2 are respectively the amount of voters of the first and the second round, for the second candidate.

I’m going to calculate the trend (T1) for only the first candidate :

T1 = ABS(VC1-VC2) \ [ ABS(VC1-VC2) + ABS(VD1-VD2) ]

Now,

P1 is the percentage for the first candidate in the first round,

P2 is the percentage for the second candidate in the first round,

P3 = 100 – (P1 + P2), is the percentage for the others all together in the first round,

So we had got this formula :

P1 + T * P3 -> (32% + 0.35 * 30% ) which was the formula you tried to show us your way.

There is the full formulas :

FakeP1 = P1 + { ABS(VC1-VC2) \ [ ABS(VC1-VC2) + ABS(VD1-VD2) ] } * P3

or

FakeP1 = P1 + { ABS(VC1-VC2) \ [ ABS(VC1-VC2) + ABS(VD1-VD2) ] } * ( 100 – (P1 + P2) )

FakeP2 = 100 – FakeP1

which is far more complicated.

The trend is so important, we cannot just hide it.

OK, you’re making fair points, let me try and respond to them one by one:

Yes, it’s absolutely correct that I could be wrong and Gbagbo really won the second round election. What I am saying is that what I have observed and shown here is making it look very likely that he lost. I’ve stated the assumptions under which this is true and everyone is free to agree or disagree. I have

not proventhat Ouattara won.The trend is so important, we cannot just hide it.

I have not hidden it. I have published all the detail of my calculation and pointed to it twice in my article. I object to you saying I’ve hidden it. I have not detailed the calculation in the article itself because I wanted to keep it short and readable even to people who hate maths. (I once read that book authors can expect every formula in their books to halve their potential readership!!)

It is true that I did not properly separate the null votes out from the other additional voters and I apologise for my mistake. My brief look at it suggests that doing so wouldn’t really change much about the result. I will try and look into that soon and update my document accordingly.

Let me show you something.

Bear in mind I used the exact formula I described previously for you (The non reduced one)

and obtain the very same result as you had for Gbagbo and Ouattara.

ADO: 32% + 0.65 * 30% = 51.5%

LKG: 38% + 0.35 * 30% = 48.5%

Now look at this election :

1st round :

C1 has 46%

C2 has 18%

We have 36% left for the rest.

2nd round :

C1 has 49%

C2 has 51%

Hence C2 is the winner.

We have these trends :

T1 : 0.1538461538

T2 : 0.8461538462

Guess what ?

FakeP1 = 51.5384615385%

FakeP2 = 48.4615384615%

Therefore C2 is the loser.

which means – This formulas is not error prone. We just have an estimation based on the first round and estimations could be wrong. Couldn’t they ?

We must not use this formula for a two rounds election.

Apologies, but I don’t understand the example. You’re quoting percentages, but then you’re saying you’ve applied a formula that asks for actual vote counts without providing that data (T1 = ABS(VC1-VC2) \ [ ABS(VC1-VC2) + ABS(VD1-VD2) ]). Could you please spell it out for me? Also, if any of the values in the ABS(…) expressions is negative I think you are describing a case different from the one I’ve looked at and I think you’ve extended the theory to these cases, which may or may not work.

This example is not valid. I made a mistake when I was copying the datas from the spreadsheet, but there are three examples of faulty evaluations in the pdf file I sent you.

“I have also read I think on the apeaceofconflict blog about the mismatches and extra voters and if this has occurred in a significant way it does of course call the data into question.” Yes, that is where I originally was alerted to this problem, and how I found your blog as Rebecca is a close friend of mine and said to me “you should run those numbers”. It’s pretty easy to verify the difference. What link are you looking for there—I’m unsure? It’s a matter of calculating the data regarding inscrits from the CEI version. Or do you mean the SRSG certified number? That is available here: http://www.un.org/en/peacekeeping/missions/unoci/documents/cote_d%27ivoire_elections_round2_%20factsheet24112010.pdf

“I don’t know why you keep characterising what I’ve written as an attempted “proof”, with double-quotes for increased hilarity. As I said, it was never intended as such. I’m not proposing this as a new method for the UN where they are certifying elections, I would hope they have more complete data and indeed fully trained staff for the purpose.” No, but you suggest that it is the “best evidence yet… that Ouattara really did win the election”, and almost characterize your data set as nearly flawless. Calculate your margin of error on your calculations and you’ll see that it is extremely high, enough to invalidate the point of the exercise even.

“Do you think the outliers are an indication of fraud?”. No, not my point in the slightest. Merely that this will be a factor that must be considered as it will significantly affect any calculation.

As to the comment about the population taking to streets. People in Egypt and Tunisia also faced thugs hired to kill and abuse them, and there were specific measures taken to stop people from gathering in groups as well. In both these places, it was also extremely likely to lead the country to more violence, and not to result in an overthrow of government… that’s why so many people were so shocked when it happened. “Easily” is the wrong choice of words, but it is definitely possible for the population to be far more pressuring than they are.

It’s a matter of calculating the data regarding inscrits from the CEI version.

As shown on the abidjan.net site or are you refering to a different source?

No, but you suggest that it is the “best evidence yet… that Ouattara really did win the election”, …

Yes, I still think it is when based on only publicly available data. (*) Where have you seen better? Here, http://news.abidjan.net/h/390674.html? Or here, http://www.mediapart.fr/node/106759? The drawback of both of these is that they assume that results in the North were correct except perhaps the cancelled regions while I think the implicit allegation by N’Dre is that fraud was widespread in the North and cancelling these seven districts accounted for that. In a way I’m responding to this perceived “fraud was all over the North in the second round” allegation with my analysis.

(*) Without the public data constraint I’d say that Choi saying Ouattara won is the best evidence.

… and almost characterize your data set as nearly flawless.

Yes, I’ve called it credible and based this on the data having been approved by the IEC, the UN and the CC, three organisations not overly likely to agree on anything these days. You have given some evidence that the data might not be that trustworthy (the question of 70% vs. 80% voter turnout + the fact that in some areas there’s higher turnout than number of registered voters) but you wisely haven’t claimed to have proven me wrong, just as I haven’t claimed to have proven anything.

null votes… a factor that must be considered as it will significantly affect any calculation.

I will look into this and update my calculation but it might take a few days until I get to it.

Here are some examples of trends calculated for a second round and applied to a first round of an election. The document is in pdf format. It contains four examples. The first one concerns this very election. I have put it there just to show you that the results I get were the same as yours. You’ll see with the other examples that, even though you have the real trend, the estimations may still be wrong.

http://www.2shared.com/fadmin/18449082/5f2221b5/examples_elections.pdf.html

Here’s the spreadsheet I created with openoffice calc (Direct competitor of excel) from the previous different formulas based on your own point of view. It can calculate any kind of estimations, and in addition It may be interesting for those who would like to understand the mechanism involved. It’s a complete program therefore it contains macros, you’ll be warned of when you open the document. These macros are not paramount, so you can ignore them if you want.

http://www.2shared.com/fadmin/18449079/4f68e4f2/election_estimation.ods.html

You can download openoffice from :

http://download.openoffice.org/

This second document is a small and complete program (or complicated spreadsheet) for those who would like to know more about their elections and numbers in general. For instance, based on the first round, and if we consider the results of the second round as valid, the trend Gbagbo needed to win was a negative trend of 40.03%.

It may be interesting if you’re in a middle of an election and you’d like to have some predictions or if you want to control some elements about two rounds elections.

As “someone who actually studies stats” said (which I completely agree with most of it), this isn’t as simple as we might think. I agree that if people study those kind of things, there must be some good reasons.

In a mathematical (arithmetical actually) point of view, due to the fact that those formulas could be wrong, we can’t use them to prove things like that. Statistically, with lots of probability calculations, we may have something interesting but again, because of accusations launched by the opposite side (fraud), we cannot be 100% or a strong percentage sure that these formulas are the absolute evidence. There is a sociological side to take in consideration, I don’t know, maybe there are studies that prove that many people may not be gone mad between a first and a second round. I mean decided not to vote in the first round and vote in the second for the opposition (They kept airing on TV strong evidence of crimes between the two rounds that psychologically could have influenced people choices, …), etc.

A second round mustn’t have something to do with a first round. First round we have our choice, second, we forgot. I found myself voting a completely different way that I could never have imagined between a two rounds election more often than I thought. Maybe that could happened in CI, I don’t know, but I really don’t think these formulas are enough to prove that one of the candidate has win.

I have followed the first link and had to click through some stuff to get to the actual Download link. Then it downloaded “FDMSetup.exe” instead of your html file and my virus killer went bonkers. Is it possible that 2shared is dodgy? Could you upload your document to a Google Docs account, would the formulas still work? Alternatively could you email it to me (address on About page of this blog)?

… Ahh, I’ve now seen your next post with the other links. That works much better.

I’m afraid I made a mistake.

http://www.2shared.com/document/U2nYB9Qw/examples_elections.html

http://www.2shared.com/file/30UuRLJA/election_estimation.html

OK, I think the spreadsheet works much better for me than any HTML or PDF so I’ve taken the election_estimation.ods document and uploaded it to my Google Docs account. It’s public so you can see it here: https://spreadsheets.google.com/ccc?key=0An9LqYI8mJVpdFdyRnNpYWIyei1RV0JOV3FZcXdRUUE&hl=en

I’m also right now sending you an email via Google Docs so you’ve got permission to edit it. What happens in the first example is indeed that one of the ABS(…) expressions has got a negative value which gets turned positive and I think that’s why it fails to predict the trend properly.

I have based the trend calculation on a few assumptions, one of which is that previous Gbagbo/Ouattara voters returned and voted as they did in the first round, so within the bounds of my theory this cannot happen. Or I guess it can happen, but then my maths is not necessarily applicable. Or am I misunderstanding something?

Thank you for your response. Feel free to share that spreadsheet, I don’t mind. Maybe I will add some functionalities as a “fraud discovery button” or some charts (I want users to see the percentage of probability that a candidate may have in order to win an election).

The pdf document contains all the 3 examples. It does not have the same purpose as the .ods. This .odf contains some values by default, users will have to enter their owns (it does not contain a real example – that may explain this negative value you get).

The ods will allow you to enter your own values for any sort of elections you would like to investigate based on your point of view. However, it’s a really good way to see how things evolve in an election. I would really like you to understand that there is some error percentage that may be possible in any kind of election. It’s true that when we get the real trend, we may wipe out a lot of possibilities of errors, but not all of them (look at the .pdf).

– IMPORTANT POINT –

This .ods file is a spreadsheet I created with a free but powerful software [openoffice calc : http://download.openoffice.org/ ] which is slightly different from Microsoft Excel. The spreadsheet won’t be displayed correctly with Microsoft office Excel, so it’s really important than you use a version of openoffice (sorry about that but I am more used to openoffice than Microsoft Excel). There are some controls form in this spreadsheet that are not exported in Excel. Some fields are not even displayed, so really test it again with the original software if you want to see how powerful it is.

This .ods file also contains 2 macros, the first one displays a message when you click on the Help button, the second one, is a button that just displayed a red box to inform you than the estimation and the real results are different.

I might translate this program for Excel but I will only be able to do it next week end .

PS : I’ll send you the files directly to your mail. I added a chart which give the percentage for each candidate to win based on a the threshold.

null votes… that is not factored into your equation in the slightest.

I’ve now looked into how to deal with null votes and found that it doesn’t change much overall. The thing is that when you separate the group of votes that were null votes in the first round but became successful votes in the second round from the trend-determining group for Bedie and other eliminated candidates (“Bedie+”), a reasonable assumption such as “these voters have voted as their Southern/Northern counterparts” will still lead to a change of the Bedie+ trend which works in the opposite direction.

Even when you apply the most extreme and unrealistic assumptions (“all those newly non-null voters voted Gbabgo, or all of them Ouattara”) and when you only do so for the South favouring Gbagbo even more, the best result I get for him is just under 49%.

I’ve now detailed my thoughts and calculations on this at the bottom of the spreadsheet. Would be grateful for any feedback.

All this is very interesting but i don’t think statistics is going to prove who won in Ivory Coast, because I suspect it is just like it was in Kenya – both sides cheated. This is a recurring trend in Africa based on Ethic lines.

What Ivory Coast now needs is the two to come together in fresh talks. Instead of singing ‘peace’ which neither of them mean if they don’t win, they need to tackle the issues at their root i.e what is the real cause of disunity the country. Otherwise i might as well hope to see North Ivory Coast and South Ivory Coast in future world maps..