Can a butterfly in Brazil really cause a tornado in Texas?
This post is about Nate Silver's book The Signal and the Noise (which I am currently reading with great pleasure) and chaos theory and how it relates to the dynamic between economics and equity markets. This blog will be a follow-up to my latest analysis on the S&P 500 highs that spurred some comments about the uselessness of economics in relation to financial markets.
Similarities between economics and the weather
In 1969, Edward Norton Lorenz, coined the term butterfly effect which later transformed into Does the flap of a butterfly's wings in Brazil set off a tornado in Texas as the title of a presentation in 1972. The idea is that certain systems are very sensitive to initial conditions so that a small change in one place in a deterministic nonlinear system can result in large final outcomes at a later stage. In Nate Silver's book, the example used is when Lorenz and his team are running different forecasts for the weather in Kansas and get totally different scenarios despite using the same model and data. Later, the team discovered that one technician hAD truncated the barometric pressure data in one corner of HIS grid so the data reading was 29.517 instead of 29.5168. This small difference in the starting point before running the weather model created big differences in the forecast. This is the butterfly effect.
In many regards the economy is similar to the weather. It is a very nonlinear and dynamic system that constantly changes through time and where cause and effect seems to change through time as well. You can easily get spurious relationships and one leading indicator may not be so in all business cycles. In short it is a mess and the fact that so few predicted the last recession is testimony to the paucity of our knowledge of the economic system. To muddle things up even more, what is a leading indicator at one point of the business cycle may be a lagging at another point. Consumer confidence and housing starts are obvious recent cases.
Equity markets: a leading or lagging indicator on the economy?
In his book, Nate Silver's discusses whether the Conference Board's leading indicator of the US economy is in fact leading or not. The book says that it was a poor gauge in the Great Recession of 2007-2009. You can argue that indeed it did predict a slowdown in the economy ahead of the actual start of the recession, but the model severely underestimated the depth of the downturn. This has also later led to some profound changes of the leading index's composition. Previously, the S&P 500 was a bigger component in the leading index but that has now been scaled down to 3.9 percent. So the S&P 500 is actually perceived to be a leading indicator on the US economy. Interesting given the recent rally. But is it always a leading indicator?
The chart above shows all the recession in the US economy since 1960 versus the six-month rolling drawdown in the S&P 500. What is clear from this 52-year period is that the changes in S&P 500 give economic forecasters a lot of false positives. Since 1960 there have been six false positives defined as a 15 percent drawdown within the last six months; more specifically in 1962, 1966, 1987, 1998, 2002 and 2011. Six false positives are many false signals given we have only had eight recessions since 1960. To make things even worse, in some cases the S&P 500 peaked a few months before the recession started making is difficult to interpret a small decline from the highs as a signal of recession.
Given that equities are more a leading indicator on the economy than the other way around, it is close to being useless to use most economic indicators as arguments for where equities should trade. Changes in risk premium and specifically relative risk premium between asset classes may say much more about the future direction (in probabilistic terms) or level of equities.
Maybe it is because of the avoidance of mixing up economics and investing that Warren Buffett has been so successful. Other successful traders in equities such as Ed Thorp did not use economics as input but instead focused on statistical arbitrage.