Monday Blognado: Does Size Matter?
Welcome to the first day of Blognado – an experiment where I try to shake myself out of my blogging drought (admittedly I’ve never been all that prolific, but still I’d like to try and get the juices flowing a bit) by writing a blog post each day, just to see if I can. If you have any suggestions as to what I should cover, do let me know.
Today I’m going to riff off a comment I made as an aside in a post last year where I said in passing:
is there a feasible upper a limit to a functional state, given existing information and institutional technology, and if so how big is it? Maybe some countries would be better off breaking themselves up or at least delegating their decisions down a bit so as to avoid clogging up the system. Maybe I should run a regression analysis of population size against the corruption index (probably a decent proxy for quality of governance) some time.
Well, today seems as good a time as any, so I did exactly that, using population data from Wikipedia (my data will be a few months out of date now, but that won’t change things much) and Corruption Perceptions Index data for 2010 (the 2011 results hadn’t been announced when I grabbed the data) from Transparency International.
Now before I get into the results, a couple of extra disclaimers on top of my regular set. I obtained this data from public sources as a private citizen. I analysed the data using software licensed to me personally on my own computer in my own time. Just to be clear – government resources were not used to obtain or analyse this data in any way.
Now, having got that out of the way, here’s what I did (this gets a bit technical, so if you’re not familiar with the ins and outs of regression analysis you might want to skip the green text):
I regressed the corruption Index (a score out of 10 where higher means less corrupt) for each country where I had a population value and a corruption score by the population and a set of dummy variables: 1 for each region defined by Transparency International, except I have collected the 5 Anglosphere countries (USA, Australia, UK, Canada and New Zealand) into their own region. The point of using the regional dummies is to try and correct for the cultural differences in different parts of the world. All things being equal a typical country in sub-Saharan Africa will be more corrupt than a similarly sized country in Western Europe. The dummy variables mean that each region has a different intercept, but the same slope (I did try allowing for different slopes as well, but it turns out the differences in slopes between regions aren’t statistically significant).
I looked at three likely specifications: linear (Corruption index vs. population + the region dummies), quadratic (Corruption index vs. population and population squared + the region dummies), and linear-log (Corruption index vs. natural log of population + the region dummies). Of the three, the linear-log model has the best fit, and a good fit too, with an R-squared of 0.89.
OK, math over, here’s a graph of the relationship between population and corruption index for each region:
You can see a negative relationship between population and score on the corruption index: a 1% increase in population reduces the corruption index score by 0.002 out of 10. The regional differences aren’t all that surprising either: The Anglosphere is the star performer, followed by Western Europe, after a decent gap. After a similar gap you get the Americas, Asia-Pacific and Middle East & North Africa, then finally Sub-Saharan Africa and Eastern Europe in last place.
So we have some preliminary evidence (and it is preliminary, if I were doing this for real, I’d be running a battery of additional tests and checking for other confounds) that the governments of larger countries tend to be more corrupt than the governments of smaller ones. But how important is this effect? Well let’s compare the actual scores of the Anglosphere countries with the fitted scores from the model:
There seems to be a lot going on here aside from the size of the countries. While the best performer (New Zealand) may be the smallest and the worst (USA) may be the largest, there are unanswered questions. For one thing the US and UK seem to do worse than its size would suggest, while New Zealand, Canada and to a lesser extent Australia do better. Less than a quarter of the difference between the US and Canada can be explained by Canada’s smaller size. It gets hard to say more than that with only 5 points of data, but cultural or institutional differences within the Anglosphere are clearly still playing a big role.
So, to go back to the question that originally sparked this digression: How much good does federalism do for the quality of government decision-making? The answer would appear to be some, but not all that much. At the end of the day, centralising government functions may still improve decision-making so long as the new authority has a good institutional framework.
Now ensuring that it has a good institutional framework, that’s the tricky part.