Daily Speculations
Volatility
(Jan. 6, 2004) Much recent talk about volatility shows the danger of using long moving averages, non contemporaneous data, and subjective retrospective but non predictive descriptions of 2 or 3 turning point., and fuzzy non predictive framing of questions. aside from the non contemporaneous trap,. These are endemic to almost all of technical analysis. and we should attempt to eschew same, even when the Spec is away practicing his world's worst violin and dancing.
To be talking about the incredible increase in volatility,
when it is currently at a 7-year low both historical and
implied (see the quant stuff on empirical volatility the Specs
painstakingly counted in their penultimate posts before
hegira), shows that something is amiss. But it can quickly be
rectified by the study of survival statistics in the presence
of immunity, a trait which long episodes in the market without
a devastating calamity, appear to be well modeled by. -- Vic
(Jan. 6,
2004) That model also explains/describes the contraction in
credit spreads, which has occurred despite rate hike fears,
lots of issuance, and distant
disasters like Parmalat. 00 George Zachar
(Dec. 14, 2003 14:32:45)
Volatility in the
market has to be lowest in 7 years. Stands at 16.4. Yearly
levels of the VIX at year-end:
2002 29 2001 24
2000 27 1999 23
1998 23 1997 24
1996 21 1995 13
1994 13 1993 12
1992 13 1991 19
1990 26
As mentioned more than 9 months has gone by without a 25-point one-day decline and this compares to a previous 10-year max of 6 months. Why did this happen? Does it foretell much lower returns in the next year I would think so.
To understand why returns and volatility should be negatively correlated, it would be good to be familiar with the Capital Asset Pricing Model, which states that excess returns in an asset are a linear function of its non-diversifiable volatility. These modern contributions can occasionally be useful even to those who studied before the 1970's era when all this became the standard. A good one-page discussion of the relations between stock market return and volatility is contained in "Does Stock Market Volatility Forecast Returns," Monetary Trends Feb 2003 , Federal Reserve Bank of St. Louis.
Volatility tends to
persist. A regression of volatility on excess returns one
quarter in future gives only a 1% r2. The confounding
variable seems to be liquidity in investors hands and this
hypothesis is tested in Guo, "Limited Stock Market
Participation and Asset Prices,"" working paper 2000-031b,
Federal Reserve Bank of St. Louis, Jan 2002.
"Intuitively if investors have excess liquidity, they might be
willing to hold stocks when expected return is low, even
though expected volatility is high." Guo believes that the
secret missing variable is the consumption wealth ratio, a
measure of liquidity. It correlates about 30% with sub
one-quarter returns 1952 -2002 with quarterly data. By
including both variables in regression they find about a 40%
correlation between predicted return and actual the next
quarter. Pretty good, especially if no retrospective bias
involved. And I hereby ask someone to find some reasonable
data on consumption wealth now and past. Wealth would seem
highly coterminous or, as they say, multicollinear or part
whole with stocks so consumption would seem to be key. Doesn't
sound overly likely; but more important, you don't have to be
an academic to figure out why high subsequent return and high
risk are like the horse and carriage or love and
marriage. The reason is that when you're scared out of your
wits to invest, like we all were in the beginning of 2003, the
only way the entrepreneurs can induce you to invest in new
things ( and therefore hold old things), is to promote you
with 100% annual returns, and guess what? People generally get
what they expect, i.e. it's a good bet that the a priori
and the a posteriori are equal.
Mr. Tom Downing, formerly chief bottle washer to the
incomparable Mr. Samuel Eisenstadt at Value Line, adds:
"In case you missed it, the
Guo article you cite says (page 5) '...real-time investors
might obtain the same information about future stock returns
from alternative sources. In particular, Guo and Savickas
(2003) show that the value-weighted idiosyncratic volatility,
which is directly observable and not subject to
revisions, exhibits very similar forecasting patterns for
stock market returns to those of cay [Consumption Wealth
Ratio].'
"Idiosyncratic Volatility apparently correlates -0.52 with
Consumption-Wealth Ratio. Direct calculation of the
Consumption-Wealth seems quite difficult and is subject to
possibly revised data, so this appears be a more tractable
and realistic approach."
"Idiosyncratic Volatility,
Stock Market Volatility, and Expected Stock Returns" by Hui
Guo and Robert Savickas
Abstract: Theories suggest that both stock market risk and
idiosyncratic risk are important determinants of the equity
premium, but empirical evidence is inconclusive. In this
paper, we show that the (value-weighted) idiosyncratic stock
volatility in conjunction with aggregate stock market
volatility exhibits strong predictive abilities for excess
stock market returns. Consistent with the CAPM, we find a
positive relation between stock market risk and returns.
However, contrary to the conventional wisdom, the
idiosyncratic volatility is negatively related to future stock
returns in the data. This puzzling result reflects the fact
that the idiosyncratic volatility is negatively correlated
with the consumption-wealth ratio, which, as argued by some
recent authors, is a proxy for the liquidity premium due to
non-traded human capital.
See paper at http://tinyurl.com/z9bz
Tom
Thank you, Tom, for your erudite reading of the relevant papers. It would seem that a good and much faster method of computing the idiosyncratic risk would be to look at the distribution of the returns of individual stocks within a year, and compute the old GINI coefficient or some such. But the semi interquartile range would do. Such data is available under IRR on Bloomberg and one will compute it directly. -- Vic