The Belgian Census

Belgium is a really cool country, and the Belgian census is a fascinating subject.

I figure I might get some flak for saying that, but at the same time I am posting this on a blog called "The Map Nerd". You should have realised...

Belgium is a curious little wedge of a country, stuck between a rock and a low place, divided in the present day into three administrative regions (Flanders, Wallonia and the Brussels-Capital region), three linguistic communities (Flemish-speaking, French-speaking and German-speaking), and led by a king who is probably the European monarch who intervenes the most in politics since he has to nominate a new prime minister about every two weeks.

Belgium is a federal and multilingual country, where the three official languages are equally respected by the law. On the ground, the historical truth is that French was for a long time the language of the aristocracy and upper class, the prestige language, and Flemish was a dialect spoken in half the country. Until 1878 French was in fact the only official language of Belgium. In recent times however, French-speaking Wallonia has suffered economically and Flanders has prospered... leading to quite a bit of bitterness on both sides.

Starting in 1878, Belgium became an officially bilingual country. Either language could be used. By 1921 they realised it wasn't working, so a linguistic border was drawn across the country. In Flanders only Flemish was to be used in public administration, and likewise for French in Wallonia. Only communities near the border with a significant proportion of speakers of the "other" language are required to provide services to speakers of the minority language in their native tongue. Brussels and the surrouding province of Brabant remained officially bilingual.

Not content to create a complicated system and let it work itself out, the government also decided to plan ahead and make the bilingual zone near the border move based on the result of each decennial census. If a border town in Flanders gained a large proportion of francophones, it became bilingual with the result that all civil servants needed to speak both languages in order to serve the public in both.

The census of 1930 resulted in roughly the same number of French- and Flemish-only border towns becoming bilingual. The 1940 census didn't happen... and when it did happen in 1947 the results showed 14 Flemish-only towns would need to provide services in French, and only one single French-only town became bilingual.

After some hemming and hawing, it was decided that:
1) Some border towns that had changed substantially from Flemish to French or vice versa would be swapped between Flemish and French provinces (creating two strange exclaves in the modern-day map of Belgium's two principal regions).
2) The language border would be fixed once and for all.
3) The country would be split into three administrative regions with the power to manage things related to physical infrastructure, and three linguistic communities (Three? Oh, right. There were some Germans here at one point... Wie geht's, mon ami ? Goed? Great!) with the power to manage things related to people's lives... education and all that.
4) The singularly bilingual province of Brabant would be split into neat French and Flemish parcels, and a capital city region in the middle would remain the only truly bilingual part.
5) Linguistic questions would be banned by law from any future census.

It was the 60's and the 70's. Some people were experimenting with drugs, and others with society. Apparently the Belgian government was doing both...

Oh, you came here for maps! Sorry!

So, without further ado, here is the highly controversial, even contraband, animated map of the evolution of bilingualism in the provinces of Belgium between 1866 and 1947.


(Click to enlarge)

R.I.P. Belgian Linguistic Census

Data courtesy of Wikipedia.

Map courtesy of the CDC, modified to return provinces to pre-1962 town transfer borders.

Who's been naughty and who's been nice?

Among other things, the OECD provides statistics on grants and loans made to developing nations by members of the Development Assistance Committee (DAC). While it doesn't cover all development aid paid out by every county, their data does contain some interesting figures.

One of the indicators included is the total Official Development Assistance (ODA) grants and loans paid by each country as a percentage of their Gross National Income (GNI).

The maps below are based on 2007 ODA as a % of GNI, as calculated by the OECD.


(Click to enlarge)

And the same thing again, but with figures in the map...


(Click to enlarge)


Data used
Downloaded from the OECD website, data for Development / Aggregate Aid Statistics / ODA by Donor.

Program used
Epi Info again... unfortunately. Its main advantage over MICRODEM is that I can load the map and then link data from an Access data directly to the shapefile, without having to go through the pénible steps of adding columns to the DBF file that comes with the shapefile, and filling them with data. The thing I dislike most about Epi Info -- the map projection can't be changed, and the one it uses is really ugly.

Census and Privacy

It's interesting to look at the different data collected by the census bureau of different countries, and what's made available to the public.

On the one hand, in Canada even the maps of individual census collection blocks is not available free of the charge to the general public. $1000 for a map, StatCan? Really? After I pay almost 50% of my income in taxes? No, thanks.

On the other hand, the US Census Bureau makes all sorts of maps and statistics available to the public online for free. It's almost frightening the level of detail you can get down to.

Take for example, this map of Grand Isle County, Vermont, which shows where there were people who identified as "Asian" in the 2000 census.


(Click to enlarge)

Each orange dot represents one person.

Has the hair on your arms just stood up on end?

The good news is that the data available really requires some serious number crunching and a good computer to be able to create this kind of map. I tried to add the same data for the adjacent counties in New York and Vermont on either side of the lake, and my computer crashed.

Also, the orange dots are not GPS-triangulated locations of actual people. They are just placed at a random location within each census collection block.

Software used
I used a new program this time, called Epi Info. It can be freely downloaded from the CDC website, though I think it's a toned-down version of a bigger piece of GIS software created by ESRI.

♫ If I can (dun, dun) pay rent there, I'll make it (dun, dun) anywhere... ♫

Today's map shows occupancy rates of housing units in Manhattan. The three maps show rate of occupancy by renters, by owners and vacant properties.


(Click to enlarge)

I feel like "housing unit" has a somewhat inconsistent definition with the U.S. Census Bureau. I'll have to research it when I have time. As in my earlier map of Detroit, where a high vacancy rate was observed in the census block containing the airport, the map of Manhattan has two oddities. The Mt. Sinai Hospital, which fills an entire census block by itself, is listed as being owner occupied, while the Metropolitan Hospital Center located a few blocks east, and which similarly fills an entire block, is listed as vacant.

You get the just of it, though.

Demographics, Maps and Problems

One of the things I love about maps is how they can tell you a lot, just by looking at them. Of course, maps based on statistics are just like the statistics themselves -- depending on how you use and frame the data, you can sometimes make them say what you want them to say.

Take for example Detroit. We drove to Detroit last year to see if it really was as bad as what people say. People love to talk, and say things, and sometimes exaggerate because it makes for a better story. We wanted to have the opportunity to see with our own eyes if it really was a failed city.

No map could describe how awful the city was. We drove not only around downtown, home of empty skyscrapers, some of which have been turned into parking garages because no other use could be made of them, but we also drove through the residential areas north of downtown. Abandoned houses abound, most overgrown with weeds and boarded up. On each block, every single block, there was at least one burned out house. Burned by its desperate owner in order to collect an insurance payment and move on?

But I wonder how many houses are actually vacant in Detroit?


(Click to enlarge)

The scale maxes out at 60%, and note that there is actually a part of downtown Detroit that reaches that. It's probably something to do with the airport being in that block. Sometimes the statistics themselves are inconsistent in what they define as a "housing unit", but nonetheless most of downtown is around the 30% vacancy mark. Imagine one out of every three houses on your street being empty.

Los Angeles by comparison is doing pretty well. The parts with high vacancy rates along the coast to the northwest of downtown LA are Malibu. Holiday homes?

Another demographic collected by the U.S. Census Bureau, which I find a bit surprising, is race. "Race" in the simplest, schematised way you could imagine it. Someone who is the 12th generation descendant of slaves brought to the United States before it was the United States, and someone who is on a student exchange from Botswana might both be counted as exactly the same thing by the census, if they were both "black".

But as someone pointed out to me, after discussing the fact that they'd written "Race: White" on a fix-it ticket my husband received while driving in Maryland with one headlight out, only if you track things like "race" can you point out that people with dark skin get more tickets.

Washington, DC is a lovely city. I was blown away the first time I visited, because I had very low expectations. Never have I seen another American city with such grace and vibrance. Manhattan is amazing, but gruff. Boston is decidedly English, in many senses. Los Angeles wears microshorts, and San Francisco wears hemp. DC wears tailored suits and an authentic, but understated Rolex.

But there's something fishy about the parts of DC I've spent most of my time in during my few visits. I thought DC was "black". Where are the African Americans in Dupont Circle and Georgetown?


(Click to enlarge)

Notes: The scale goes from 0 to 100%, and represents the proportion of "black" inhabitants in each census block in DC. The white-coloured areas of the map actually mean "not black", and other "races" are not factored in at all.

There are virtually no inhabitants who self identify as African American in the government and fancy neighbourhoods of DC.

But I don't like cutting off a map at an administrative border, and excluding immediately adjacent areas. I wonder what it's like in neighbouring counties in the states of Virginia and Maryland...


(Click to enlarge)

Oh, dear.

Data used
U.S. Census Bureau Tiger/LINE shapefiles downloaded from the ESRI website. I used Census Blocks and Census Block Demographics (SF1) for the 2000 census.

Maps and Me, A Sordid Love Affair

As long as I can remember I've been fascinated with maps. Travel as well, but especially maps. My friends and I used to play role playing games (never Dungeons and Dragons specifically, but similar ones), and at one point I got a huge sheet of paper to be able to map out a big fantasy world we could use for our games.

These days I'm more interested in making maps of the real world. I currently use a free program called MICRODEM, which I'd originally downloaded to convert satellite images to maps for the game Transport Tycoon... but that's a different story altogether. The maps I've been making use shapefiles showing various census and administrative divisions in Canada and the US, and census data.

The current fling started because I wanted to map out population density on both sides of the Canada/US border, covering Montréal, Québec (city) and the adjacent parts of New York and New England. The idea was to get a real image of how the population is laid out on both sides of the border, but not actually plotting the border as a line across the map.

I looked all over the place and never once managed to find anything like it. Most maps I found were either exclusively of the US or exclusively of Canada, or didn't really have the level of detail I was looking for.

So I made this.


(Click to enlarge)

The colour scale maxes out at 100 inhabitants per square kilometre on purpose. I wanted to make sure that sparsely populated rural areas, and the slight differences between them, would show up. The side effect of this is that even smallish cities like Burlington, Vermont (the larger of the two dark spots south of Montréal) seem bigger than they really are, because the black can represent 101 people per square kilometre, or 4000 people per square kilometre.

One issue was getting a hold of comparable data on both sides. The issue of metric vs. imperial units (people per square kilometre in Canada, and people per square mile in the US) was easy enough to get past by converting the US data to metric. But it was a little more difficult to get data broken down into similarly-sized parcels of land on both sides. In the US, I ended up using Census County Subdivisions, which are somewhat regular in size. In Canada I used Census Subdivisions, which vary quite a bit in size, as they sometimes correspond exactly with the limits of a municipality, and other times correspond to a non-governmental découpage of rural territory.

Data used
From Statistics Canada / Statistique Canada,
Census Subdivision Boundary Files and Population and dwelling counts, both for the 2006 census.
From the U.S. Census bureau, Cartographic Boundary Files of Census 2000 County Subdivisions and population density by census county subdivision from the oh-so-user-friendly American FactFinder... To repeat what I did, simply go to Data Sets, Decennial Census, 2000 Census Summary File 1, Geographic Comparison Tables, select "State", select a state, select "Place and (in selected states) County Subdivision", include data from the group "GCT-PH1-R.", view and download it.

Linking the data and maps
Canadian census boundary files can be linked to the data through the Geographic Code / CSDUID, which is a unique ID that should match perfectly 1-to-1 in both files for the same census year.
US census data files contain a field called "place" which will contain something like "Bridgeport town, Fairfield County". Truncate off the county, and you will have something you can compare to the two fields named "NAME" and "LSAD_TRAN" and that will contain "Bridgeport" and "town" respectively. Watch out for some of the overlapping data areas in the data file -- cities split across two counties and recorded once as a city, then twice as city areas in each of the two counties, etc. Also in some larger states like New York, there are cities with exactly the same name in two counties, so the county info can't just be thrown away as with smaller states.

(You bored yet?)

The end result is pretty different from what I expected. In my mind, the border had a big effect on population density. Northern New York and Vermont feel so desolate in comparison to the area of Québec just on the other side of the border, that I was really expecting to see the border through the difference in population density. But while you can indeed make out where the border is because the divisions on either side of it don't line up, the population density on both sides is quite similar. You can almost make out a corridor of lightly-populated areas that run uninterrupted from Montréal to Plattsburgh, NY and Burlington, VT.

If you stand back and stare blankly at the map from a distance, a 3-D dinosaur also pops out at you.