|
|
|
|
 |
Daily Speculations The Web Site of Victor Niederhoffer & Laurel Kenner Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter; a forum for us to use our meager abilities to make the world of specinvestments a better place. |
Write to us at:
(address is not clickable)
Paper: "Financial Variables and the Predictability of Stock and Bond Returns: An Out-of-Sample Analysis"
Abstract
Most studies of the predictability of stock and bond returns
rely on in-sample tests. In this paper, we test the ability of
ten financial variables that have appeared in the extant
literature to predict S&P 500 and CRSP equal-weighted stock
returns out-of-sample over horizons of 1-10 years. We also
test the ability of two financial variables, the term and
default spreads, to predict long-term corporate bond real
returns out-of- sample. For S&P 500 returns, we identify three
variables with out-of-sample predictive ability: the equity
share in total new debt and equity issues, term spread, and
market value-to-net worth ratio ( Fed q ). For CRSP
equal-weighed returns, we find that the dividend yield,
price-earnings ratio, Fed q, and equity share all exhibit
significant out-of-sample predictive power. In addition, the
default spread exhibits significant out-of-sample predictive
power for long-term corporate bond returns. As out-of-sample
tests of predictive ability raise the bar relative to
in-sample tests, our results strengthen the case for stock and
bond return predictability.
http://faculty.washington.edu/~ezivot/econ512/WoharReturns3.pdf
Review by
Alex Castaldo, PhD:
This paper is
interesting methodologically for two reasons. First, the
regressions are estimated during one period (1927-1963) and
then tested during another period (1964-1999). This is
helpful. However, it is not the way an investor would
actually operate, they would probably re-estimate the
regressions each year, a more laborious procedure than the
professors'.
The second interesting point is the use of recently developed statistical tests for forecast assessment and references to the literature thereon that the scholarly might be well advised to read.
The main conclusion for S&P prediction is:
In predicting one year returns only the Equity Share in New Issues of Equity and Debt of Baker and Wurgler (recently featured on the DailySpeculations Web site) was found to work. Three other variables, the Dividend Yield, the Price Earnings Ratio, and the Q ratio computed by the Fed only worked at horizon of 8 years or more. (As will be familiar to readers of PracSpec, such long-horizon-only predictability is of limited speculative value and is suspect due to overlap; it also raises the question of why it would take 8 years for the market to adjust to such information).
When applied to the CRSP EW index, which includes smaller stocks than the S&P500, additional variables are found to have predictive value. However this finding may have limited practical import because the CRSP EW is not really an investable index.
Stopping the study at the end of 1999, when 4 full years have already elapsed, is annoying to the practically minded, but is fairly standard for academic work.
Overall an interesting paper. (March 2004)
Comment
from Vic:
I don't believe any
forecasts of 25 yearly returns with 8 and 10 years overlap
could possibly be significant. And whatever tests have been
developed for same are totally flawed. The stopping point is
arbitrary, the methods have been selected retrospectively,
they're all deflated by some crazy CPI adjustment
(retrospective or prospective) that introduces biases, and all
the results would have led to randomness on top of randomness,
a hodgepodge, like a consensus of mutual back slapping of each
other's Web site and professorial affiliation.
Let us hope that one of the big fund-of-fundists backs this method as an augmentation to the long wait that may be forthcoming to profit from the derivative expert's work during this environment of asset diminution and seminar speaking and fee income for all.
Comment
from Dr. L:
Nice response. Here's a somewhat shorter reply:
Q: How many out-of-sample backtests does it take before they
become in-sample?
A: Two.
Comment
from Alix Martin:
I agree with Dr Lo's
comment. The very process of academic publication actually
makes the data not so out of sample, as we wouldn't have heard
about it if it didn't concur with the in-sample hypothesis.
The issue is stability of certain predictive relationships. I don't think using only two time-periods is likely to tackle this issue the most conclusively.
To try to get a picture of such stability, I graphed a trailing 10-year correlation between PER and next year real SP return. The correlation happily swings from -60% to +60%, with a sobering effect on any thought of using it as a market timing indicator. I would like to try that with the supposedly more predictive variables if I can find the data.
Besides, Rapar and Wohar's point doesn't seem to be that they can predict the market, but rather that they used the right statistics. I certainly can't reject this hypothesis.