Sunday, October 26, 2008

Which countries does the NYT cover most and least?   posted by agnostic @ 10/26/2008 09:16:00 PM
Share/Bookmark

Greg Cochran left the following comment in a Matt Yglesias blog entry:

What you need is a map of the world in which the sizes of the countries are adjusted to the number of column-inches they get in the New York Times and the Washington Post. I think it would be illuminating.


Well, I've done something close enough. I only looked at the NYT, and I made a bubble chart instead of one of those distorted cartograms. Also, I used number of articles rather than column inches -- but these must correlate highly. It's not as if Tonga gets a few 10,000-word articles, while Iraq gets many 50-word articles. At any rate, let's see what the results look like.

nytimescountries.jpg
Here are the results for the 192 members of the United Nations. Move the mouse over an unlabeled blob to see who it is, or search for a specific country. The results cover 2000 to the present, and are standardized by dividing by the number of articles for the entire period. To ensure that the graphing algorithm would pick up order-of-magnitude differences, I multiplied the fractions -- which ranged in order from 10^(-5) to 0.1 -- by 10^5, so that they range in order from 1 to 10,000. Some countries I had to estimate rather than get the exact number, since their names are shared with other things, like Turkey (see Note).

The first thing you notice is a few big blobs and lots of tiny blobs, in accord with a Power Law. Rather than futz around with getting my pictures to post here, I'll simply list the frequency distribution, where the first column is the fraction of all NYT articles devoted to some country, binned by order of magnitude:

Order-----Frequency
0.00001--0.046875
0.0001---0.354167
0.001----0.5
0.01-----0.09375
0.1------0.005208

The one country in the 0.1 bin is the US. Everyone else is lucky to get something on the order of a percent in coverage. Still, the modal country gets mentioned on the order of once every thousand articles -- not too shabby if you're Qatar. Here is the full dataset, in case you want to download and play around with it yourself.

How do we infer the level of insanity in our foreign policy implied by these data? Looking at the countries from greatest to least emphasis, the low-ranking ones make sense -- they belong to the parts of the world you've never heard of, and will not have reason to hear about within your lifetime, such as Tuvalu and Bhutan.

But there are some funny ones at the top. For example, it takes the top 9 to discover all 5 permanent members of the UN Security Council. The remainder of the top 9 are Germany and Japan -- which at least are G8 countries -- but also Iraq and Israel. Speaking of the G8, it takes the top 12 to discover them, which adds another lesser country to this elite list -- Mexico (China is not G8 but is still important). Afghanistan, Iran, and Pakistan also rank pretty high.

This is a perfectly rational outcome -- our foreign policy may obsess over these places, but by placing criteria on them like "permanent member of UN Security Council" or "member of G8," we can see which ones don't deserve the attention. They represent the parts of the world, like Iraq, where we're wasting a bunch of money to squat over an over-glorified sandbox, hoping that our colonial piss will transform it into a lush oasis. Or they're the places, like Mexico, where we're importing a large illiterate peasant underclass from. This seems like a useful way to change our foreign policy: see who we're obsessed with, but who don't really matter, and cut them loose (relatively speaking).

By the way, the Many Eyes website has a global map feature, but it only allows an additive scale for bubble size, with the three smallest orders-of-magnitude collapsed into one bubble-size. So it didn't look very good. Maybe at some point I'll screen-capture the bubble chart, and cut and paste each bubble onto a picture of a world map, but that probably won't happen.

Note: I used the common English names for countries -- e.g., Syria rather than Syrian Arab Republic -- and made the following modifications to make sure I picked up the country rather than something else by that name:

Chad: added "Africa" to search
Georgia: added "Tbilisi" -- probably an undercount, but not my much
Guinea: subtracted "Equatorial Guinea," "Guinea-Bissau," and things like "guinea pig"
Jordan: added "Israel" -- again, an undercount, but not by much
Palau: subtracted "Barcelona" and "Catalonia" (it means "palace" in Catalan)
Turkey: subtracted "Thanksgiving" -- probably an overcount, but not by much
United States: searched "America," and subtracted "Latin America," "South America," and "Central America"

Labels: ,