The 2018 World Cup is now reaching its climax, and some of the most fancied nations have already returned back home. But World Cup upsets aside, what makes a country good at football? I thought it would be worth updating a regression model I built some years ago in which I found that the strengths of national football teams are largely determined by social and economic factors. The research was originally published in Social Sciences Quarterly back in 2007 and you can request a free copy of the paper here.
The factors influencing national team strengths were – and indeed still are -:
Size of the Talent Pool
In the 2007 article I used FIFA ratings to measure team strength. This time, I used the unofficial World Football Elo ratings, partly because the FIFA ratings have been much criticised, and partly because FIFA announced they are going to abandon their existing system after the world cup and use Elo-based ratings instead. I also updated the GDP and other data to reflect the new time-frame.
Despite these changes the results were very much the same as I found 11 years ago. Then, the factors explained 70% of the variance in ratings; this time they explain 73%. In a way, that is not surprising; national football rankings haven’t tended to change much over the years. The correlation between the Elo ratings in 2018 and the Elo ratings 10 years ago is 0.95, and FIFA rankings are just as stable. Of course there have been movers; the Philippines (+383), Iceland (+346), Afghanistan (+329), Belgium (+320) and Luxmbourg (+304) have all increased their Elo ratings by over 300 points, and on the other hand the British Virgin Islands (-302), Angola (-294), Singapore (-290 ) and Sir Lanka (-288 ) have seen considerable drops.
So what can we say about the factors driving national team performance?
What I mean by money is national wealth, which economists measure by gross domestic product (GDP). In wealthy countries, individuals are free to support and to engage in leisure activities and competitive sports to a greater degree than in poorer countries. And the easier access to medical and advanced training resources in wealthy countries probably enables players to perform more consistently at higher levels than would otherwise be possible. So we would expect high-GDP countries to have stronger teams. Rather than the raw GDP value, however I use the logarithm. The point here is that when a country has a small GDP, they are a lot more sensitive to small changes in GDP than they are when their GDP is already large. The higher the GDP they have, the less sensitive they are to small or even moderate changes in GDP. Using the logarithm means that countries are placed on a ratio scale of GDP, so that any two countries the same distance apart have the same ratio of GDP. It is also appropriate to use GDP per capita, to account for the fact that the resources accruing from GDP are distributed across populations of different sizes.
Size of Talent Pool
Countries with a large pool of talent to select from will produce stronger teams than those with a smaller pool to select from; more populous countries
should therefore outperform smaller ones. The data in my 2007 paper was taken from a FIFA survey of its member associations carried out in 2006, and I showed that the number of registered male football players was much more predictive of team success than the total population. (Because the national team consists of a very small number of players, the talent pool operates according to a law of diminishing returns; increasing the talent pool from say 1 million to 2 million gives the same increase in team strength as increasing it from 2 million to 4 million. A mathematical proof is given in the paper.)
Unfortunately, the FIFA survey has not been repeated since. However, it contained estimates of the number of youth (under 18) male players, and it seemed reasonable to use these numbers as an approximation for the size of talent pool producing international football players in the current era. In fact the 2006 youth player numbers correlate a bit more strongly with the 2018 team ratings than the 2006 registered player numbers do.
Countries with a history of fielding international football teams are likely to gain the kinds of know-how competitive intelligence that will improve their performance. A country’s football tradition was measured by the number of years since its national football association was affiliated to FIFA, with appropriate adjustments for countries such as those of the former Soviet Union, whose national affiliation to FIFA dates from the break-up of the Soviet Union, but whose effective tradition begins earlier.
Over one third of players representing their country play their club football abroad; and most of these play in a wealthier and/or higher-ranked country. These expatriate players have access to advantages like top-level coaching, and advanced sports science that is not available in their home countries, and can bring individual skills into their national teams. The results showed that – as we might have predicted – playing abroad is much more beneficial to poor countries than to wealthy ones.
Once again, the data for this section was kindly provided by Benjamin Strack-Zimmermann and Florian Meyer over at National Football Teams. You can visit their website by clicking the picture below, where you will find a wealth of information about national football teams past and present.
A game of football makes considerable physical demands on the players, and the way it is played is influenced by the weather. For example, in hot conditions, games may be played at a slower pace. However, raw measures of climate don’t have much predictive power, so I clustered the countries into three groups and used group membership as a predictor. Here are the climate groups.
|Tropical-subtropical||High||High||High||Brazil, Sri Lanka, Cuba|
|Medium Pressure||Low||High||Medium||Portugal, Uruguay, Gambia|
|Temperate||Medium-Low||Low||Low||England, Sweden, USA|
When analysed along these lines, being in a Medium-Pressure zone (like Costa Rica, Portugal and Colombia) does confer some advantage after the other factors are taken into account. The main suspect is likely to be vapour pressure (which measures the amount of water in the air and is related to humidity). When the vapour pressure is too high, as in Tropical countries, the body cannot sweat enough to get rid of the excess heat, and when it is too low as in the Temperate countries, sustained physical effort can induce respiratory problems.
In addition, the climate zones interact with GDP; the positive effects of GDP are lower for Tropical and Subtropical countries, but I don’t have a good explanation for this.
What Can Countries Do to Improve National Team Performance?
Obviously, having a good coach, a good camp, and building team spirit have big effects on performance. But many of the socio-economic factors that influence team success are pretty firmly fixed. You can’t change your weather or your footballing tradition. But there are two elements in the model that countries can change.
First, the size of the talent pool can be expanded by encouraging more youngsters to play football, and identifying the most promising talents (I know, Matthew Syed is going to kill me again.) And that is exactly what Iceland have done to very good effect, as England found out when Iceland knocked them out of the Euros in 2016. Second, national federations could assist more players to gain skills abroad, though that of course is not without its risks to the domestic game. And there are other more innovative solutions. Iceland overcame its weather disadvantage by building indoor pitches. The Philippines’ improvement was based on stretching their definition of nationality quite considerably, not a tactic everyone would want to copy, but they also employed good foreign coaches.
Over and Under-Performing Nations
The model described here explains a substantial amount of the differences in country performance, and considerably outperforms the Economist’s recent model described here. A graphic of the results is shown at the top of the post, where the model’s predictions and the actual Elo ratings are plotted against each other. Countries above the diagonal regression line outperform the model’s prediction, and countries below the line under-perform. Perhaps surprisingly to some, England’s long-term performance is somewhat higher than predicted by the model. It is often forgotten that England have consistently done well in the qualifying stages of major tournaments. It is in the later stages of tournaments where England have tended to disappoint.
If you are interested, you can find more information in my paper along with a bunch of references to other work in this area.