Hanging Chads two
March 30, 2011 § 1 Comment
And here is part two which Think Africa Were good enough to publish in its entirety. Here is a version I did earlier which had links to the google map which I thought was a nice touch in that you could see where the places discussed were (but please also check out the Think Africa version as well, they’ve done the graphics really nicely and have some fantastic coverage of other issues in African Politics).
Issues in the Chad elections
To understand what is happening, it is necessary to understand type of electoral system in operation in Chad. Chad has a system whereby if any party wins over 50% of the votes they win all the seats. It they do not then seats are allocated by the highest remainder formula (Hare quota). For simplicity I am treating the MPS and allies as one for the purposes of this analysis.
The first issues arise on the question of turnout. Turnout was given as 56.6% and the figures relating to the numbers of votes cast do give this turnout figure if one takes the Chad Electoral Commission’s (CNEI) statement of figures for registered voters which the CNEI’s statement appears to do.
However this then raises several problems. For a start the figures don’t add up: adding the totals for each constituency gives a figure precisely 15,000 short of the published total. Moreover the statement of election results gives totally different figures for registered voters – listing some 570,881 registered voters fewer than the official figure. Moreover this is a far from blanket disparity: Dababa and Dagana have lost over 51,000 (55%) and over 49,000 (68%) of voters respectively between the statement of voters registered and statement of results, whilst Lac Lere has gained over 71,000 voters (77%).
It is not clear what the reason for the discrepancy is, or which is the correct figure. The fact that the former figure is the basis for turnout suggests the former whilst the fact that it is the latter that appears on the statement of election results suggests the latter. Yet whichever figure is correct, and whatever the reason for this disparity, there are further questions to be asked about the turnout.
If we accept the statement of electoral results as being at least internally consistent, and take their figures for registered voters, then that suggests improbably high turnout in a number of constituencies. In Bahr Sara and Dar-Tama more votes were cast than there were voters in the constituency, whilst in Kobe 99.8% of people voted (implying only 54 people in the entire constituency did not). This also gives turnout figures in the 90s for a further 8 seats. Dar-Tama and Kobe were uncontested seats, so high turnout is unlikely but an error would not benefit any side directly. However Bahr Sara is a seat in which the MPS won after the URD missed out on a seat by 2.5% of the vote on an apparent 105% turnout.
Conversely, if we were to accept the statement of registered voters as the correct figure then different questions arise in different constituencies. Votes in most constituencies now occupy a more normal distribution (although turnout in three safe MPS seats drops to sub 30%) but at the high end there are some puzzling results. Two seats (both of them uncontested MPS) post turnout in the 90s and one (also uncontested MPS) posts turnout of slightly over 100%. But the real problems arise in La Nya Pende and Lac Lere. In La Nya Pende the MPS picked up two seats by just 4.5% of the vote on what was seemingly a 113% turnout. In Lace Lere (where the UNDR won all three seats by just 3% of the vote) turnout was a credibility shattering 161% of registered voters.
We then move on to look at spoiled ballots. Spoiled ballots are a fact of life, and in a country with a literacy rate of 25% one would expect there to be many. Indeed there are – 13.9% of votes cast. However what is interesting is that the number of spoiled ballots is far less even than one would expect. In three constituencies less than 2% of votes were spoilt and in 15 spoiled ballot rates were below 5% whilst in Tandjile est, where the MPS topped the poll with 38% of the vote a full 38% of ballots were spoilt, and there were 10 seats where the number of spoilt ballots was over 20%.
This extreme level of variation is unusual but possibly partly explicable: the seats with low rates of spoilt ballots are all either safe MPS constituencies or uncontested seats; this makes a certain amount sense, it is much less probable that a ballot will be spoilt in an uncontested election, and in a safe seat ballots are less likely to be challenged. In contrast Tanjile est is a marginal seat with fifteen different parties vying for the vote, and in the ten seats where spoilt ballot rates were unusually high there were considerably more parties standing for election (an average of 12.8) than was usual (average of 6.8). Nevertheless one does have to question what sort of a mandate one can claim in a seat where more ballots are spoilt than are cast for any party.
Then we look at the actual results themselves. The problem is clear from the very first result on the CNEI’s website where it is announced that the MPS have won the seat of Bahr El Gazal Nord with 105.67% of the vote.
The number of votes cast hardly ever adds up to 100% (indeed another surprising thing about the results in Chad is that in six constituencies they actually do add up to exactly 100%, as these are all in either very close or totally uncontested constituencies it is possible that this is due to either exhaustive recounts or to the official just writing the convenient figure in). Voters can walk out of the booth without lasting their ballots, the occasional vote can get lost and the occasional vote can get counted twice. A handful of miscounted ballots is not the end of the world, but when the number gets into hundreds, or thousands, then there is cause for concern. Indeed in the UK 66 extra votes in a safe seat once caused quite a stir.
In Chad, by the election commission’s own figures, there were 11 seats where more than 2% of the vote went missing, and in the worst cases, Arrondisment N’djamena 5eme and Dar Tama, 5.6% (1145 votes) and 4% (1716 votes) of the vote went missing. Worse still there were 37 seats where there was more than 2% more of the vote than there should have been, and in seven constituencies it was more than 10%. In the worst case, Arrondisment N’djamena 10eme, there was a full 15.2% (3329 votes) more votes counted than were cast.
In ten cases this could have influenced the result:
|Constituency||Winning party and how||Margin of victory||Disparity votes cast to votes counted|
|Gueni||MPS over 50%||0.8%||2.1%|
|Batha Ouest||MPS over 50%||2.3%||4%|
|Haraze al Biar||MPS over 50%||3%||4.2%|
|Abtouyour||MPS over 50%||5%||5.9%|
|Barh Azoum||MPS over 50%||6.2%||7.8%|
|Baguirmi||MPS over 50%||6.8%||9.3%|
|Mayo – Leimi||URD (MPS short on 2nd seat as under 50%)||-6.6%||10.8%|
|Arrondisment N’djamena 1ier||MPS over 50%||1.1%||12%|
|Arrondisment N’djamena 4eme||MPS over 50%||1.37%||12.3%|
|Arrondisment N’djamena 10eme||MPS over 50%||10.8%||15.2%|
Even more concerning is the fact that the given percentages for votes cast in each seat do not actually match the percentages one arrives at when calculating the percentages from the data itself. Checking the number of votes cast for each political party does give the same answers as those provided by the electoral commission, but the totals in each seat add up to radically different percentages to those given. An exhaustive survey of potential methodological differences does not throw up any potential explanations for where CENI’s figures came from.
Calculating one’s own percentages directly from the given figures gives a greater number of seats (37 as opposed to 23) where the numbers of votes cast is within broadly acceptable (2%) limits of the number counted but also greater extremes of missing votes.
There is now one constituency, Arrondisment N’djamena 9eme, where nearly 20% (18.5% or 9238 votes) of votes are missing and another, Lac Lere (once again), where over 10% (10.5% or an incredible 25,170 votes) are missing. There are, admittedly only now three constituencies where there was more than 10% more votes counted than were cast. Arrondisment N’djamena 3eme is now the most overvoting constituency with 16.4% (641 votes, it’s a small constituency with low turnout) more votes counted than were cast.
In eleven cases this could have influenced the result, changing thirteen seats:
|Constituency||Benefiting party and how||Margin of victory||Disparity votes cast to votes counted|
|Arrondisment N’djamena 9eme||UNDR (MPS short on 2nd seat as under 50%)||12.3%||18.5%|
|Lac Lere||UNDR over 50%||3%||10.5%|
|Baguirmi||MPS over 50%||6.8%||9.3%|
|Batha Ouest||MPS over 50%||2.3%||6%|
|Lac Iro||MPS over 50%||3.2%||4.1%|
|Mando Oriental||MPS over 50%||1.1%||4.1%|
|Guera||MPS (PDSA short on 2nd seat as under 50%)||1.5%||1.9%|
|Kimiti||MPSx3 over 50% (with five seats allocated, MPS would have only won 2 if under 50%)||0.8%||3%|
|Gueni||MPS over 50%||0.8%||3.9%|
|Mayo-Lemie||URD (MPS short on 2nd seat as under 50%)||6.6%||9.8%|
|Arrondisment N’djamena 9eme||MPS over 50%||1.4%||13.3%|
In conclusion therefore we find serious anomalies both with the turnout in certain constituencies, and with the published results of the election. The scale of some of these anomalies is large enough that, if there is not a reasonable explanation for them, they call into question the validity of the poll. In addition it is possible that these anomalies could have resulted in the wrong result being declared in 19 constituencies, with the potential effect of giving the MPS and allies up to 19 more seats than they were due, and the opposition up to three.
These last figures are interesting. If these anomalies were down to random chance we would expect them to affect all parties equally. As the MPS and allies won 71% of the seats, we would expect 71% of the seats this throws into question to be MPS seats. Yet 19 of the 22 seats in question are MPS. Whilst this does not prove electoral fraud it does suggest a worrying trend of puzzling anomalies which systematically benefit the MPS.