Statistical Cherry-Picking

While surfing the other day I came across this gem of an article on football as a metaphor for globalization by Branko Milanovic, Lead Economist at the World Bank's research department. It's an interesting argument about how the free movement in labour in football - as in all other areas - can lead to higher overall welfare (i.e. better top-level competitions watched globally) while at the same time leading to localized losses of welfare (i.e. African competitions starved of local talent) due to what he cleverly calls "leg drain". Basically - more money + more mobility = more concentration of talent. Which is great if you happen to support the richer teams and bad if you don't.

Milanovic's thesis is interesting, but unfortunately he relies on three statistical examples which are - in my humble opinion - totally bogus.

The first two observations relates to the competitiveness of the European Cup/Champions League. He notes that since the start of the big-money high-mobility era (which, can be dated from pretty much anywhere in between the birth of the Premiership in 1992 and the Bosman ruling in 1995), the concentration of CL quarter-finalists has increased (that is, there are fewer of them over any given stretch of time), and that even among the quarter-finalists, there is also greater concentration of semi-finalists, finalists and winners. Fewer teams, in other words, are getting to the top at each level.

I won't bore you with the critique of his use of gini coefficients: suffice to say that he has managed to create a statistical argument in which 1998-2002 (the article was written in 2003) appears to be the period of the greatest concentration of talent in history despite the fact that no team won the CL two years in a row in this period. Other periods of clear single-team dominance - such as Real Madrid's five straight in the 50s, the back-to-back three-peats of Ajax and Bayern in the 1970s, and Liverpool's incredible run in the late 1970s and early 1980s - all apparently were more competitive eras than the present.

Even if this were true, there is presumably a confounding factor in his analysis; namely, the collapse of the east bloc after 1990. Yes, the 1990s were a time of increasing money and mobility in Europe. But in European competitions, the effect was magnified by the fact that certain footballing nations were in the midst of economic turmoil and unable to compete financially. Only now do we see a reversal of this trend, with big Russian (and to a lesser extent Ukranian) clubs starting to compete for big money players. Now that eastern teams can genuinely compete for the signatures of marquee players like Neri Castillo and Vagner Love, one assumes the effect will gradually be reversed.

His third statistical argument is even shakier. Using the more money + more mobility = less competition thesis, he hypothesizes that within countries with large income inequalities, over time good teams will tend to converge in the richer areas of the country. His illustration of this is that, over time (up until 2002) the number of teams in Serie A from the country's poorer south was decreasing.

This is nonsense. First, the disappearance of southern teams from Serie A in 2002 was an extremely temporary phenomenon: the number is up again - significantly - and Palermo even managed to qualify for Europe two seasons in a row. Second, the geographic concentration of football clubs is highly path-dependent and contingent upon historical factors. The concentration of English football in Lancashire and the midlands cannot be predicted on the basis of economics, but follows relatively smoothly from the fact that all thirteen of the original members of the League were from these areas. The near-total absence of good football in Paris is certainly not a result of the capital's lack of money, either.

Part of the problem here - and indeed, part of the problem of any econometric analysis of football - is that while the value of football's human inputs - (player quality as measured in transfer dollars, statistical measurements or whatever) can be be shoved into econometric equations with relative ease, the arrangement of those inputs cannot. Teams are more than the sum of their parts. Mourinho's Porto, Clough's Forest, Del Nido's Chievo - even Lobanovsky's Dynamo - are all examples of teams playing collectively way above their level because of the way they play as an integrated unit. As Perez found out at Madrid, you can't simply buy success.

There is also the issue of tactics. Teams with limited abilities (e.g. Liverpool under Benitez) can gain success simply by signing a lot of tall players and pinging the ball around in the air a lot. The resulting football can sometime resemble "shit on a stick" (best Jorge Valdano quote ever) but it is undeniably effective - and extremely difficult to model statistically.

(for the record, I'm with Brian Clough: if football were meant to be played in the air, God would have put grass in the sky.)

Anyways, have a look at the Milanovic piece. It's entertaining and the basic points are interesting even if they are contestable: and it's a good economics lesson, anyway.