Egypt’s potential district boundaries
The new districts for Egypt’s parliamentary election were leaked the other day. You can find the links for the the nominal and list tier respectively, here and here. It appears that the districts will be based around Egypt’s 27 governorates, which makes some sense, but will also reduce proportionality. The allocation of PR districts is strange; district size is either four or six seats, regardless of governorate size. So Cairo, with 28 seats, is divided into seven districts of four seats. Not surprisingly, most parties are upset about the new boundaries, although the Muslim Brotherhood – to my knowledge – hasn’t really commented on them.
While there are limitations on what we can analyze (and I welcome comments from those more knowledgeable) we can examine many aspects of the new boundaries. The first, and perhaps most important, question to ask is if the boundaries are relativity fair, or malapportioned. This is easy to examine as population statistics are available. (One important note, for population I added the populations of Helwan to Cairo and 6th of October to Giza, as those two governorates were recently reintegrated into their old territories.)
Figure one shows that the population of a governorate and the number of PR seats it was awarded are largely related. The regression has an R squared score of .9178.
Figure two (click to enlarge) does a better job of showing which governorates received more or less seats than they would have if seats were perfectly distributed. Giza comes out pretty poorly, being underrepresented by almost six seats, followed by Qalyubia and Shariqia. At first glance this looks like a wide disparity, but in reality I think this isn’t too bad. The fact that the chart is compressed (so it was easier to read) gives the impression that there is a lot of range, but most governorates fall between one and two seats outside of their expected level. It is notable, however, that with few exceptions, higher population areas do worse than low ones. The governorates in the chart are ordered from highest population to lowest, so we can see a pretty obvious bias towards the right side.
Next, we can look at how proportional each governorate is. I’m aware of some proportionality indices, but they only work if we have party votes. (If anybody knows a good way to measure without this information, please let me know). So absent this information, I tried another idea. I took the the percentage difference between district magnitude if the 252 PR seats were perfectly distributed, and the actual average district magnitude in each governorate. So, for example, if seats were distributed perfectly by population, Cairo would receive 29 seats. Then the most proportional distribution would be for all of it’s 8,968,694 residents to vote in one district with 29 seats. In reality, however, the average district magnitude is 4, with each one having 2,242,181 residents in it. Comparing the percentage difference of these numbers across governorates should allow us to view how proportional each region is in comparison to each other. Figure three shows there is once again a pattern; large population governorates are the least proportional while small ones are more so. Lines in red have an actual district magnitude lower than the expected value, while blue lines have a DM that is actually higher than we would expect. This pattern isn’t terribly surprising and should be expected when you hold the district magnitude constant across all regions, regardless of population.
Cairo stands out as the least proportional governorate; its average DM of four means there are 1,933,549 more people per district than there would be if the governorate had a district magnitude of 29. (A 152% difference in actual district magnitude). Figure four shows the relationship between population and this measure of disproportionality.
So the largest governorates enjoy a small, but real benefit as far as overall representation, but suffer with regards to proportionality. Overall, however, nothing seems too off with these boundaries.
So what are people saying about the boundaries? Well, mostly bad things. Some of the complaints, however, seem unfair:
Mohamed Farag, assistant secretary-general of Al-Tagammu Party, said that the distribution of constituencies was “illogical”.
“Giza governorate for example is divided into three constituencies … north, middle and south,” Farag explained, adding that the north combines villages with more developed districts.
“The needs and types of people in each area are different … and impossible to include in one constituency,” he added.
Heaven forbid an MP would have to represent more than one type of person. However, others are complaining that the large sizes of the territories will favor candidates with more money, as it costs more to run a campaign in larger areas. This probably has some merit with regards to the nominal tier of seats, but shouldn’t matter for the list tier. It is somewhat ironic, however, that this is a result of complaints by these same actors. The allocation of seats was originally planned with the nominal tier accounting for 2/3 of all seats. Parties, however, argued for the PR tier to count for more. They got their wish (the tiers are now split 50/50) but the decrease in nominal seats invariably means they will need to be larger.
Update: A recent discussion with an Egyptian friend of mine shed light on some of the complaints people are making with the district boundaries, especially those in the nominal tier. Districts, like they were in the past, are based around the presence of police stations. Yes, you read that correctly. Needles to say, this means that many districts aren’t exactly drawn in a manner most optimal to creating fair representation.
Update 2: Changes made in second graph after reading comments.
Posted on September 13, 2011, in Electoral Systems, Middle East and tagged boundary delimitation, Egypt, electoral systems. Bookmark the permalink. 5 Comments.
The indices of (dis)proportionality can be applied to any case where we are interested in how closely one quantity corresponds with another. So use each district’s voters as a percentage of all voters and each district’s allocated seats as a percentage of all seats. Same quantitative problem as party votes and seats.
I assume you do not have actual district voter (or population) totals? Obviously, using aggregated districts at the governorate level, where a governoarate is divided into two or more districts, is less than ideal. But what you have done here is still a useful analysis.
But you seem to imply that more populated areas are over-represented. That would be really unusual, and I am not sure that it would hold up in an actual district-by-district analysis.
Is the horizontal axis in Figure 2 meaningful? I could not tell if observations are arrayed by population (as the text implies) or not, because the axis is not labelled.
It would make more sense to me to plot (and regress) the logarithm of population vs. logarithm of seats (or disproportionality).
Thank you. You’re correct about the districts; the information knowledgeable people have told me recently indicate that the districts are pretty off due to the fact that they are based on police stations, and nothing else.
I will review the other suggestions/comments, and potentially update when I get a rare, spare minute!
Strange principles lead to strange results: all seat totals are even because one seat out of every two is reserved for ‘workers and peasants’ In the nominal tier, every district has 2 seats. According to the links in your post, in the PR tier every district has 4 seats, except in 10 governorates with a number of seats not divisible by 4, where there is one district of 6 seats. At least there seems to be a rule behind it, even if the magintude is low for PR and high for majoritarian.
Fair districting is always a tradeoff between population equalization and preservation of existing administrative units. To judge to what extent the districter was free to gerrymander, you have (1) to compare population (or electorate) figures at the district level and (2) to judge to what extent districts are created out of existing administrative units.
This raises several questions: what are your population sources, an dare they also available for the district level? The subunits (police stations) used to define the districts, are they the usual subdivision of a governorate, or does this definition by police stations cut across another tier of administratieve units? Are the districts of both tiers are always nested, meaning: is every nominal district within a single PR district?
In any case, choosing a small-M-system means is choosing extra workload and debatable decisions (districting) before the elections… and isn’t that something to avert in a situation like Egypt now?
And what’s the system and the districting for the upper house?
– So setting districts around Police stations is not how it’s done in the rest of the world?
– You know a lot of the Cairo residents don’t actually vote in Cairo…in my family we’re three who live in Cairo, but only 1 of us votes in Cairo, the rest are still registered on their Alexandria address. Since most of Cairo/Giza are people migrating from other cities and villages, they would probably go back to their original homes to vote right?
– I am worried that they are not discussing the upper house at all! There’s very little info about it, no timeline for that. A lot of people are calling for cancelling the upper house all together – are there countries that work with one house only?
Pingback: Egypt’s potential district boundaries, Ctd. Workers and farmers « Ahwa Talk