Ok, time to take a break from the South Asia forecasts and talk about something a bit more global: The 2010 FIFA World Cup. In anticipation of the world’s greatest sporting event, three major investment banks have put their stakes on the table. Will the quant models of JP Morgan Chase, Goldman Sachs, and UBS trump your gut feeling? Let’s take a look.
JP Morgan Chase
The JPMC report is based on “a simple Quant methodology” in which analysts looked at four categories of variables that they normally use in studying companies: Valuation, Market Sentiment, Fundamentals, and Price Trends.
Each category was broken into specific proxy metrics, and countries were ranked for those indicators. FIFA World Ranking and Bookmakers’ Odds, for example, were the two metrics for Valuation. Price Trends were estimated by looking at historic trends of those two indicators.
To come up with a composite score, the analysts weighted (admittedly arbitrarily) each category: Valuation 40%, Sentiment 15%, Fundamentals 15%, and Trend 30%. After this, the strongest teams to emerge were Brazil, Spain, Netherlands, and England.
The next step was particularly innovative. To account for penalty shootouts in tied games, the report measured “Ability to Score” (=goals scored/games played) and “Goalkeeper Ability” (=goals conceded/games played). Brazil ranked very poorly in this metric, not even in the top ten.
The analysts plugged in the model score of each team to predict the outcome of each game; in case of draws (teams of similar strength), the penalty shootout model predicted the winner. This throws out Brazil, the overall strongest team, in a quarterfinal penalty shootout against Netherlands. The other surprise is that Slovenia, on the back of recent trends, advances to the semifinals. In the final, England beats Spain on penalties based on their much better shootout record.
In contrast to JPMC, the GS report does some macro analysis. It notes a weak relationship (-0.17) between GNP/capita and FIFA ranking. But there’s a stronger and positive correlation between improvement in FIFA ranking since 2006 and the overall economic growth environment of a country, especially if the two outliers, Brazil and Argentina, are excluded. In short, if emerging markets want to improve their chances in the World Cup, they better invest in growth and human development and improve rule of law. Equity market performance is not a good indicator of World Cup performance.
GS predicts that Brazil is going to be favorites for a long time to come. Not only is its past performance (64-14-14) better than any other country (Germany is second, at 55-19-18), but its growth environment is healthy and its demographics, esp. number of males, are stronger than European countries. It is also a large football market.
Now, the main question: Who will win? The narrative in GS’s report reads mostly like a journalistic analysis, purposely devoid of numbers. It looks through the groups, and predicts England, Argentina, Brazil, and Spain as the semi-finalists. The report then introduces a simple probability model. Like JPMC, the main inputs into the probability model are official FIFA rankings and odds from different bookmakers. (I suspect a high multicollinearity between these two variables.) Arbitrary weighting is another problem, and GS admits: “Our model-probabilities are intuitive to a large extent” (p. 65). While JPMC explains its method in detail, GS does not. The model predicts Brazil as the winner, and Spain with the next highest odds. Overall, GS opts for a more conservative outcome than JPMC.
GS’s report has a couple of other gems. One, it polls its clients (N=an astonishing 2,955) to come up with a Dream Team. It also lists odds for the golden boot award: Villa, Rooney, and Messi top the list.
Unlike GS’s global review, UBS focuses on Africa, with an eye to promoting the continent to investors and philanthropists. The report was published by its Wealth Management Research group.
Once the selling is out of the way, UBS considers three criteria in making its prediction. The first is past performance. Unlike the other two banks, UBS uses Elo ratings, a system developed by Hungarian-American Physicist Arpad Elo. This apparently determines the strengh of teams better than FIFA ranking does.
The second consideration is whether or not a team is the host nation. UBS gives home ground advantage much more prominence than does JPMC or GS. UBS notes that 63% of the time the hosts have reached semifinals. Now it’s doubtful whether South Africa can add to this trend, but most people didn’t consider South Korea in 2002 either.
The third element is — yep you guessed it — a proprietary model, described as “an objective quantitative measure that assesses the strength of each team three months before the start of the World Cup.” Fine. UBS also claims flat out that socio-economic factors and GDP growth have “no explanatory power when it comes to forecasting the performance of a specific team.” That also makes sense. But disappointingly, UBS does not reveal the workings of its model.
Historically five teams (Argentina, Brazil, France, Germany, and Italy) have accounted for 53% of all semifinal places. But UBS notes that of late, weaker teams on paper are posing more surprises, and so there will be at least one rank outsider in the semifinals. But the winner will definitely come from usual powers. And competition will be intense. Based on Elo scores, this World Cup will see “the strongest Spanish team ever to go to a World Cup, the strongest English team since 1970, the strongest Dutch team since 1978, [and] the strongest Brazilian and German teams since 1998.”
In the end, Brazil emerges on top, with 22% probabilty to win the World Cup. UBS puts Germany (this was before Ballack’s injury) at 18%. England, JPMC’s favorite, has only 4% chance of winning according to UBS. So does fancied Spain, even though if you look at bookmakers’ odds, Spain is numero uno (4/1), edging out Brazil (9/2). All the more reason to wish UBS exposed details about its model.
Who is right?
Is one prediction better than the other? Let’s just take the bankers’ playbook and do a meta-analysis using “an objective quantitative measure.” Let’s assume, with good reason, that the overall predictive power of the quant models of each bank is reflected in its relative stock price trend. After all, if their models perform consistently well, their trading desks will make more money, and investor confidence will drive up the stock price. In short, although multiple factors exist, we simply assume that relative stock price movements subsume all such factors into one neat proxy metric.
The chart below shows the relative stock price movements of JPMC, GS, and UBS for the last three years (2007=0).
As you can see, all three ticker indices are down since 2007. But JPMC (-20% approx.) has outperformed GS (-40% approx.) and UBS (-80% approx.) pretty consistently for the last 2.5 years. Therefore, JPMC’s prediction is more reliable.
So what’s next?
Does all this mean you should get out your excel sheet and hash out some kind of model? Chances are, you already know all you need to know from just following football. So my advice: just stay put on your couch. After all, even UBS admits, “one needs to be humble about the predictive power of one’s models,” before concluding that forecasting football is more art than science. That’s good news for me, because none of these analyses made me change my mind. I’m rooting for Spain.