Ethiopia and the ethnicity rat race - part 1: geography

On the first anniversary of the start of this horrible war,  let's take a moment to examine the concept that bedevils the country.  Obviously, there are much more urgent things going on right now, but this is something that's been bothering me for more than half my life, and my cup runneth over.  For a simple minded engineer, if something is broken, you debug it. And like many many other Ethiopians, I think there's a bug in the operating system of the country, a bug which was introduced by a system update about 30 years ago.  And this bug made it vulnerable to infection by catastrophic viruses,  which are now threatening a system crash.  

I'm referring of course to the idea that the ethnic group, rather than the person, is the fundamental unit of society. In this design, the country is a collection of ethnic groups. The name of your administrative region, the name of your political party,  your ID card, everything is based on ethnicity. That's a bug. Whether you like it or not is irrelevant, it just doesn't work. To see why, let's look at just one aspect: geography. In a follow-up post I will look at it from historical perspective. (Update: part 2 focusing on history is posted)

A few months ago, I tried to answer a basic question: is it feasible to separate ethnic groups into different geographic regions in Ethiopia? I gave it the fancy title of ethnogeography and posted a snippet at the time. What follows is a more detailed explanation.

 

If you are not technically inclined, feel free to skip the next two paragraphs and go to Results.

Metric: To make the point objective, let us introduce a precise definition of  "ethnic diversity".  For example, if region A has three ethnic groups each representing 1/3 of the population, and region B has 4  groups with 55%, 15%, 15%, 15%, which one is more "diverse"?  You could say B is more diverse since it has 4 groups and A has only 3. Or you could say B is less diverse because there's a clear majority group.  So what's a good metric of diversity? You could ask similar question about anything that is based on a probability distribution. For example for income inequality, they use the Gini-coefficient. Here, I decided to use the following metric: if you take two random people from a population, what's the probability that they belong to the same group? This captures the following intuitive idea:  how likely is it that your neighbor, or your classmate, or a person you meet on the street is from another group? This metric has a simple formula, which is 1 - Σ_ip_i², where p_i represents the fraction of the population that belongs to group i.

Data: For the data, we can use the Ethiopian census of 2007 which you can find on statsethiopia.gov.et. There are also copies on other websites run by the UN, World Bank and the US government.  This data is not ideal (more on that below) but at least it's consistent, I've cross-checked the different copies and archives and it is the real data. To be on the safe side, I've also saved a copy of the data here. Unfortunately the data is in PDF files! So I had to write a  custom parser (code at the link) to extract it into usable form.

Results:  Now we can look at ethnic diversity at each available administrative region level: country, province ("k'l'l") or district ("zone"). A diversity score of 0 means there's no diversity at all, everyone in the region is from one group.  A diversity score of 1.0 means every person belongs to a different ethnic group.  A score of 0.25 means there's a 25% chance that any two people chosen at random are from different ethnic groups. Here's what it looks like at the zone level. The darker regions are more diverse.

This gives us a sense of diversity spatially.  But of course, this doesn't tell the whole story. Some tiny regions have huge populations and vice versa. So to get a better sense we can view diversity and population size together as follows:

Keep in mind the colors represent the degree of diversity, not the ethnic groups themselves, of which there are more than 80.  So if the country could be neatly divided into ethnic regions, it would all be light yellow. If it was completely mixed everywhere, it would all be dark brown.  For details, including the code to generate these results, see this notebook on github.

Note we can assume this data understates the degree of diversity.  One hint is the number of people classified as "Ethiopian National of different parents".  For example, according to this data, there are only 20,724 such people in Addis Abeba, less than 1% of the population, which is absurdly low.  One explanation for this is that the census was done in 2007, when ethnicism was the governing philosophy, and the 2005-2007 period political repression and fear were at a peak. We can assume that most people who are mixed just chose one ethnicity, out of convenience or necessity.  I had a couple of personal experiences during that regime which corroborate this -- one before this period and one after this period.  Both times,  I was asked for my ethnicity when interacting with the government (one instance was local, the other was federal). When I said I am a mix of three different ethnicities, both times the officials refused to accept the answer, and demanded that I choose one. In the end, both times I just told them to choose for me.  To this day, I'm not sure what they filled in.  Of course my anecdotes don't  prove anything about the census data, but given the political environment of the last few decades, the real number of individuals with mixed parentage is most likely higher than the above shows.

What we see in the above pictures is a very very diverse country. Not just in the superficial sense of having 80 different ethnic groups. But in our precise sense that they are geographically deeply intertwined. Some highlights: In Addis Abeba, if you pick two random people, there's a 71% chance that they are of different ethnic groups! In Dire Dawa it's 69%. But it's not just the major cities that have a deep ethnic mix. In Mezhenger, it's 81%; Metekel -- 76%.   Looking more closely, the Awash river, the Rift Valley lakes, the Omo river, and the Nile valley are apparent in the pattern of colors, even though they are not drawn on the map.  If you think about the history of civilization, it kind of makes sense. People need water, so  rivers and lakes attract populations, and over hundreds of years, those areas will become more mixed, while desert and mountainous area populations remain relatively isolated. (I could be totally wrong about this explanation, happy to learn more!)

Most importantly, it's clear that in most areas,  separating ethnic groups geographically is practically impossible.  Behind our technical metric is a grim truth, almost too horrific to contemplate. When we say the diversity score is 0.71, what we are saying is that if your region had to become ethnically homogeneous, there's a 71% chance that either you or your neighbour are not gonna make it.  And homogeneity is the inexorable direction of ethnic political parties in ethnic regions. Now look at that map again, and think of the tens of millions of people in the brown parts.  For many people, the dividing lines would not be just between neighbours,  they would be inside the house, in the bedroom, inside my own body!

It's like the story of Procrustes from Greek mythology. Procrustes owned a hotel, and he was very proud of a key feature -- he had designed the perfect bed. Then, guests of different height  started showing up. Some too tall, some too short for his bed. But Procrustes was so convinced of the perfection of his bed that he decided the problem was the guests. So he insisted on making them fit by chopping or stretching their body to fit his design.

Ethnic federalism is the Procrustean bed of Ethiopia. 

And that's the bug.  It's even right there in the Amharic name for the administrative regions: ክልል (k'l'l) . I had never really thought about  until I heard it broken down by an erudite Ethiopian (not sure who it was, citation needed!) but the word is not neutral like province, state, etc.  The verb መከለል is to put a barrier, a shield, a fence etc.  It implies protection from the other, assumes hostility and need for separation, apart-ness. When combined with ethnicity,  the closest analogue we can find in another language is ... apartheid.  We must fix this bug. 

The changes that started in 2018, including (re)formation of some political parties on a non-ethnic basis, the removal in some regions of ethnicity from IDs etc.  Those were hopeful signs! But the bug is still there, and it has never been more threatening than today.

Oh what a rat race!

Some a gorgon-a, some a hooligan-a, some a guinea-gog-a

In this here rat race

....

Political violence fill your cities

When you think there's peace and safety,

A sudden destruction!

Collective security for surety? Yeah

...

Don't forget your history

Know your destiny

In the abundance of water

The fool is thirsty

...

Oh it's a disgrace

to see the human race

in a rat race.

- Bob Marley