Wednesday, January 19, 2005

Depends What You're Measuring

Mark Kleiman has an interesting post concerning the "competitiveness leagues" some organisations like to publish. The one he's concerned about is published by the Wall Street Journal and the Heritage Foundation, which leads neatly to a first point you can probably guess. Without going into any of the arguments from opinion about this, I think it's worth unpacking some of the economics.

It's been a common theme of politics in the last few years that every so often, one of these studies is published and the opposition immediately seizes on it, whilst the government denies its relevance. Usually (as is the case with the WSJ one), the people preparing the study draw up a set of tick-boxes that they consider to be the conditions of economic success, and then either evaluate it themselves or ask a sample of businesspeople to fill in the form. They then rank the scores. Now, this brings up some problems. First, can you really say that countries (as opposed to companies) compete? Certainly, the business environments have an effect on companies' competitiveness, but it's worth asking how powerful this is. If you wanted an example of everything that would make for a terrible business environment (short of warfare, plague and like dramas), postwar Italy would fit the bill. Inflation, corruption, bureaucracy, high tax rates, infrastructure problems, monster public debt, terrorism, poisonous labour relations - it was all there, but it didn't stop northern Italy from enjoying a spectacular industrial boom. In fact, now that some of these have been conquered, the Italian economy looks much less impressive.

Secondly, are your criteria right? There's a serious question of methodology here. The whole idea of a competitiveness survey or study is that there are some characteristics that lead to greater competitiveness (however problematic that notion is). But, of course, as long as you are measuring things you think will be good in the future rather than actual performance, you are dealing in subjectives. This also goes for the sample; if you ask conservative economists who they think is more competitive, don't be surprised if they say "conservatives". (This holds vice versa, naturally.) There's a deeper point, though. Charles Goodhart's famous Law states that "to control is to distort". His point was that trying to measure your success or failure by intermediate figures is foolish, as your own actions will distort them. Far better to use final-goal figures: growth, inflation, unemployment - that sort of thing. After all, nobody would suggest giving the Premiership to the football team a panel of coaches thought would be best next year. Another problem is internal consistency - if you mark down the UK for poor infrastructure, it isn't very consistent to then also mark down for taxes and transport costs: this would mean that solving the infrastructure question would simply knock points off another category.

There might be some use for these surveys if it wasn't for the politically useful practice of treating them as "leagues". This (as in the example) ignores that a league is calculated from results, not opinions. If the notion of competitiveness among countries is sound, then, surely a better guide would be the results? A little project of mine comes in here. Simply, I'm going to look up some of these surveys and cross-check with the growth numbers. If they are valid, high rankings should correlate strongly with high economic growth (at least, high economic growth relative to competitors), either simultaneously or as a leading indicator. Another question that this brings up is which countries we can consider as close competitors. I thought I'd use the eurozone figures, simply in order to get rid of as many structural differences as possible, but I've just realised that there's a good chance they will move together on perception alone and hence we need some extra-European data points. OECD sounds good. Watch this space - I feel a graph coming on.

No comments:

kostenloser Counter