The US stock market has dropped over 6% in 2016. One question that keeps investors up at night is whether the economy is currently falling into a recession. A recent publication by Roger Farmer, economics professor at UCLA, suggests that the equity market crash caused the Great Recession and finds a stable connection between the real value of the stock market and the unemployment rate. (Click here to access this paper.) In light of these interesting findings, we would like to share some thoughts on this study.
First, this paper uses data from 1953 to 2011, but only finds significant result during sub-period 1979-2011. The author argues it’s due to the “Volcker disinflation”. When we replicate the test using an even more comprehensive dataset, the overall result became very weak. Therefore, the finding may be subject to sample selection and underlying economic condition, which could undermine the generalizability of this study.
The foundation of the paper is the Granger causality test, which statistically confirms the causal relationship between two events. In the sub-period between 1953 and 1979, Farmer finds that the stock market downturn causes the BAA spread to widen, which causes the unemployment rate to rise. This finding recalls a classic debate, which we can simplify this way: “If A causes B, and B causes C, does A cause C?” Maybe. However, Granger causality isn’t necessarily transitive, since econometric analysis relies heavily upon statistical significance, which may be weakened from A to C. Thus, without direct evidence of the causal relationship between the stock market and recession, Farmer’s thesis may only apply to a very limited period of history.
So, did the stock market actually cause the Great Recession? Sometimes the answer is right in front of our eyes. Let’s walk down memory lane and take a look at the financial crisis one more time.
As we can see from the graph above, the peak and trough of credit spreads proceed ahead of unemployment. For example, credit spreads reached their highest point in December 2008, before both the bottom of the S&P 500 and the peak of unemployment. While this order of events tends to confirm the finding that growing credit spreads cause unemployment to rise, it does not strongly support the argument that the stock market is the fundamental cause – the evidence would be more convincing if the stock market had bottomed first, before credit spread ticked upward.
From the perspective of empirical research, Farmer’s work lacks key control variables. Although the author controls for real GDP, real investment spending, the risk-free rate, and CPI, he excludes other important factors such as housing. If we added more control variables to the Granger test, would the result still persist? This paper doesn’t tell us. But a study on the causes of the Great Recession that doesn’t look at housing is truly a hard sell. It is likely that both the housing and credit market crises caused the stock market to decline during the Great Recession. The stock market crash could have been a by-product rather than the root cause.
Farmer’s model design also raises some questions. The S&P 500 is deflated by the author’s proprietary variable money wage. But as we learned in Econ 101, wages are significantly related to unemployment. Although the purpose of the data transformation performed in the paper is to meet the stationarity requirement, could it be that money wage deflation rather than the stock market plunge caused the unemployment rise in the Granger causality test? The author didn’t rule out this possibility. We replicated this test using more comprehensive data and took out the money wage deflator, we did not find significant causality between the stock market and rising unemployment.
In summary, Farmer’s paper would be a more convincing story if we had direct evidence of causality between stock market performance and the recession. However, this question has a long history of going unanswered. As Paul Samuelson once quipped, the stock market has forecast nine of the last five recessions.