Friday, July 21, 2006

Even a broken clock is right twice a day...

There is an old saying among economists. When making a prediction, give either a number or a date, but never both.

The 60 or so economists who respond to Bloomberg's monthly survey can't resist the temptation however. Participants are asked to predict various figures in the investment markets, such as the level on the 10-year Treasury, at each of the next six quarter-ends.

I have been tracking their success since June 2000, looking at the median predicted level on the 10-year Treasury approximately 4 quarters forward versus the actual level. In short, their record has been abysmal. First of all, the median survey level has predicted higher 10-year rates one year forward in every survey since 2000. Rates actually rose 57% of these periods. The median prediction deviated from the actual by an average of 78bp. That's 1 1/2 standard deviations from the actual rate! Here is the chart of the predicted versus the actual.

Perhaps the most damning thing about this survey is that the correlation between the median prediction and the actual 10-year rate 4 quarters hence is only 0.03. However the correlation between the predicted 10-year rate and the prevailing rate at the time of the survey is 0.88!

At first the survey's ugly record may seem surprising. But if you think about the concept of market efficiency, you realize a survey can't possibly be anything but noise. If the economists surveyed are representative of actual market participants, then their outlook is already priced into the market. In other words, if you think rates are going to rise, you have already positioned your portfolio for rates to rise. In other words, you either don't own any bonds, you have shorted bonds, or you are underweight bonds. If literally everyone had the same prediction, everyone would have already sold their bonds. There wouldn't be anyone left to sell to drive prices lower.

In the real world, we'd never have it where everyone had the same prediction, so the previous sentence can't be literally true. What is true, and I think what we witnessed with the 30-year in 2004 and 2005, is that when there is a ton of cash on the sidelines because sentiment is overwhelmingly bearish, the market is likely to move the other way. The market becomes thin and any buying moves prices higher (rates lower) rapidly. Then those holding cash get nervous, because they are facing lower and lower investment rates. Some capitulate and start buying, which pushes rates even lower, causing others to capitulate, and so on...

For the curious, the median prediction for the end of 2Q 2007 is 5.25%, according to the survey conducted between 6/30 and 7/10. Leave a comment if you have any questions about my methodology.


Accrued Interest said...

Go start your own stats blog.

Anonymous said...

Here are some links that I believe will be interested