I am not a soccer fan, I prefer the American version of football. That said, I am admittedly actively following the 2010 World Cup. While watching the opening match between South Africa and Mexico, I thought it would be fun to ask the question, “Do free countries produce better football teams?”
So, I quickly combined the FIFA World Rankings and Press Freedom Index for all of the countries participating in the 2010 World Cup, and came up with this:

Note, the Press Freedom Index goes from 0.00 (most free) to 115.00 (least free), and the FIFA point totals increase as the teams overall quality improves, so we might expect a negative relationship. Also, I removed North Korea from the data set because it was such an extreme outlier on the press freedom dimension. So, we find basically a null result. There is a slight negative relationship, but it is essentially random.
The peak and valley of the smoothed fit curve is a bit interesting. For the worst teams, as freedom goes down the quality of the teams go up, but around a freedom score of 20, that relations inverses and as the quality of the teams increases so does the level of freedom—until we reach the best team, Brazil and Spain.
Automatically Generated Related posts:
- Democracy, Freedom of the Press, and Government Data Requests to Google
- 2009 IC Post-doc Competition Announced
- New Dataset: The Location and Purpose of Wars Around the World
- What Countries are ‘Pulling their Weight’ for Haiti?
- A Brief Who’s Who in the World of Computational Analysis for National Security




Hi, nice post. Can you please post the code for the same?
[Reply]
Drew Conway Reply:
June 13th, 2010 at 9:15 pm
I will add it to the ZIA Code Repository
[Reply]
The average Press Freedom Index value seems to be quite low overall. Could the fact, that these countries have made it to the World Cup, cause a selection bias?
[Reply]
Drew Conway Reply:
June 13th, 2010 at 9:16 pm
It is an enormous selection bias, but my predictions is that there might be slightly higher R-squared values for the full sample but nothing close to a statistically significant finding. Any relationship would likely be dubios.
[Reply]
I agree with Shreyas, it’s a very nice graph and I would very much like the code to replicate it!
[Reply]