327. Global crisis and one more plot to think about
Uncategorized — By Dmitry Podolsky on March 29, 2009 at 4:05 pm
Save This Post as PDF
Dmitry Podolsky has got his PhD from Landau Institute for Theoretical Physics. He currently works as postdoc at Case Western Reserve University. He is also one of the editors of NEQNET.
The following quantities are plotted here – total US GDP, investments, SP500 index and so called treasury yield curve,
By definition, the yield curve is the relation between the interest rate and the time to maturity of the debt, and the yield curve above is the one for US treasury securities and US dollar interest rates.
Would you like to discuss a bit the physics of the yield curve? The reason why it should be interesting is surprising anti-correlation with SP500 data as you can see from the plot, that is, the yield curve can be used to predict the time of crisis’ take-off. What is the nature of this anti-correlation?
Via Schegloff.
9 Comments
>> What is the nature of this anti-correlation?
The short term yield is mostly determined by the Fed funds rate.
In times of crisis the Fed tries to lower the short term rates to
stimulate the economy and when times are good they raise the short term rate to fight inflation…
Thus the slope of the yield curve is anti-correlated with the other indicators.
Hi Wolfgang,
however I notice that the yield curve starts to react a bit earlier than actual fall-off of, say, SP500 happens. So, in principle, following the dynamics of the yield curve I can predict the crisis’ take-off, can’t I? Does not this small time delay contradict your simple and very nice explanation?
Cheers,
Dmitry.
Are these quantities also corrected by inflation?
Yes, they are (“SP500 deflated”), but this is not important in this particular case – we are looking for _phase_ correlation.
Dear Dmitry, I am sorry but I don’t see your “lag” on the graph you gave us. Quite on the contrary, e.g. for the dotcom bubble burst around 2000, it seems that SP500 began to fall earlier.
But even if it were the other way around and you could show that such a lag is statistically significant, which I don’t believe, I wouldn’t be shocked. First, I don’t think that SP500 measures the “crisis”: GDP growth is the indicator of depressions, and for the GDP growth, your “lag” will be even more inconclusive.
On the other hand, the spreads and stocks are about traders’ expectations. The money-market traders might indeed be reacting faster than the stock traders because money is being changed more quickly than stocks.
But I doubt that these observations are still unused by metatraders. Have you tried to analyze your hypotheses about lags quantitatively? With Mathematica, you could get the best fit etc. easily. Imagine that you want some linear fit, with some lag, and optimize for these variables.
Dear Lubos,
As it seems to me, it started falling later, but then it should be a matter of quantitative analysis, not qualitative, as you said. Maybe, I’ll indeed use statistics…
Oh, I am sure I am not discovering something new in economics, just trying to learn a bit
If the effect is there, professional most certainly use it.
Dear Dmitry, I was not really trying to say that you have discovered the wheel. Instead, I was suggesting that the reliable predicting effect can’t exist, otherwise it would have already been used by some traders.
Imagine that you can quasi-reliably predict that SP500 will go down in 2 months, from the spread. Clearly, some big fish would short the SP500 stocks already now. The net effect of this big fish is that the effect and supposedly reliable lagged correlation disappears.
There are so many effects, countereffects, and people’s desires to benefit from them that I think it is more sensible to say that all these lagged correlations are coincidences and unpredictable.
Trackback responses to this post