Professional Documents
Culture Documents
Market Efficiency
Market Efficiency
Allocational Operational Informational (also Capital Market Efficiency)
Market Efficiency
Allocational
When marginal rates of return for all producers and lenders are equal
Operational
Transactions cost of transferring funds is zero
Public Information
Past Prices
Information Generated by
Exchanges Investors, Analysts, Companies
Public Private
11/9/11
Information Incorporation
Behavioral finance: Some phenomena are best explained if some agents are not rational
But lets stick with the rational for a little longer.
2008 Patrick J. Kelly 9 2010 Patrick J. Kelly 10
11
12
11/9/11
14
Time
2008 Patrick J. Kelly 15 2010 Patrick J. Kelly 16
Semi-Strong?
Weak?
18
11/9/11
19
20
______
____
21
22
______
____
23
______
____
24
11/9/11
Efficient Markets are essentially a Platonic ideal: a perfection toward which we can strive but cant actually be obtained. What we can compare is whether
Some stocks are more efficient than others Some markets in some countries are more efficient than others
25
2008 Patrick J. Kelly 26
Statistical Tests
Variance Ratios
Large: Low return persistence or reversal
Economic test
Mid-term momentum Short-term reversal
Griffin, Kelly, Nardari (2010)
27
28
29
30
11/9/11
Contrarian Profits
Are these profits meaningful?
Note: In the U.S. the contrarian strategy earns 30 basis points per week
60 40 20 0
-20 -40
The long-short strategy in the U.S. earns only 30 basis points per week before accounting for the cost of buying and selling stock.
Is 30 basis points of return per week enough to cover the cost of a high turn over strategy that buys, sells, shorts and covers stocks each week? With $1,000,000 in assets, 30 basis points generates $3000 in profit ignoring trading costs.
Cyprus China Taiwan Egypt Turkey Hong Kong Thailand Chile Bangladesh Philippines Poland Brazil Sri Lanka Indonesia Malaysia Singapore India South Africa Israel Pakistan Argentina Zimbabwe South Korea Portugal Italy Spain Denmark Switzerland Netherlands Greece United Kingdom Austria Germany Finland Belgium New Zealand United States Sweden Norway Japan France Australia Canada Developed Emerging Difference
Limits to arbitrage
If it is too costly to trade on an anomaly a seemingly easy way to profit the anomaly will continue to exist
31
350
94
95
96
97
98
99
00
01
02
03
04 20
19
19
19
19
19
19
20
20
20
20
20
33
Six-Month Momentum
Buys past six-month winners, shorts past six-month losers
100 80
05
Argentina Turkey Malaysia Cyprus Pakistan Singapore Sri Lanka Indonesia China Brazil Egypt Taiwan Thailand Philippines Israel Hong Kong India Chile Poland South Africa Bangladesh Japan Spain Canada Portugal Austria South Korea Greece Finland Netherlands United States Italy Sweden Switzerland France Australia United Kingdom Belgium Denmark Norway New Zealand Germany Developed Emerging Difference
35
11/9/11
60%
Default premium predicts Chen, Roll and Ross (1986)
50%
Keim and Stambaugh (1986) Term premium predicts Campbell (1987), Fama (1984) Keim and Stambaugh (1986) Harvey (1988) Short rate predicts Fama and Schwert (1977) Fama (1981) Geske and Roll (1983) Predictability illusory? Ang and Bekaert (2007) Cochrane (2008) Goyal and Welch (2003, 2008) Valkanov (2003)
40%
30%
Random walk Fama (1965, 1970)
Predictability debatable Goetzmann and Jorion (1993) Hodrick (1992) Kim and Nelson (1993) Richardson and Stock (1989)
20%
10%
Fig. 2. The time-series of predictability research. The literature on stock return predictability follows closely the availability of recession data as a cumulative proportion of the total data in CRSP which originally started in 1962. Shown are the percentages of recession data as a percentage of the available data at a given date, as measured by NBER (solid line) and RSVAR (dashed line) dates. Both the NBER and RSVAR samples show similar proles, although RSVAR recession probabilities represent a much larger proportion of the data. Many seminal, and rst, papers on return predictability were published just after the peaking of the proportion of recession data to total available data in 1985 and are followed by a decline in the proportion of recession data thereafter. The citations are representative for expository purposes and are not intended to be indicative of initial research, nor a comprehensive literature survey (Ang and Bekaert, 2007; Campbell, 1987; Chen et al., 1986; Cochrane, 2008; Fama, 1965, 1970, 1981, 1984; Fama and Schwert, 1977; Geske and Roll, 1983; Goetzmann and Jorion, 1993; Goyal and Welch, 2003; Harvey, 1988; Hodrick, 1992; Keim and Stambaugh, 1986; Kim and Nelson, 1993; Richardson and Stock, 1989; Rozeff, 1984; Shiller, 1981; Valkanov, 2003; Welch and Goyal, 2008).
1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
Date
37
0%
38
The cyclical dynamics of risk premiums and of return predictability need not be synchronous, however. Using the framework of Campbell and Cochrane (1999), Li (2007) shows, counterintuitively, that changes in risk aversion alone are insufcient to induce any return predictability at 2. Background and motivation all. In another example, Mele (2007) demonstrates that countercyclical risk premiums do not necessarily imply 2.1. Dynamics of expected returns higher return volatility in bad times. Nevertheless, we account for the possibility of countercyclical return predictability in two ways. First, we decomEarly empirical evidence of countercyclical risk preThe Journal of Financesources of predictability to control for shifts in 800 pose the miums is in Fama and French (1989) and Ferson and market volatility relative to predictor volatility. Second, we Harvey (1991). The basic intuition for a Average of 16 Three-Year link between Test Periods design December upon professional survey data to better countercyclical risk premiums and return predictability is 1933 and tests based 1980 Between January Period: Length of risk the effects of simple and appealing. If investors demand higher Formation distinguish Three Years current conditions from the effects of expectations regarding future economic conditions. premiums in bad times, and volatility is higher in bad Changes in predictability over time could also result times as well, then overall adjustments to discount rates from infrequent, random structural breaks rather than per unit of change in economic state are larger in bad business cycles. Under different assumptions, Pesaran and times. Crucially, pricedividend ratios become more 0.20Timmermann (2002) and Lettau and Van Nieuwerburgh volatile and prices more sensitive to changing expecta(2008) both identify 1991 as one such structural break. tions as conditions worsen. Predictability might, thereSince there have been further National Bureau of fore, be a countercyclical phenomenon. 0. 15-] expands on our empirical approach and is followed by a description of the data in Section 4. Section 5 reports our empirical ndings. Section 6 concludes. Please cite this article as: Henkel, S.J., et al., Time-varying short-horizon predictability. Journal of Financial Economics Loser Portfolio (2010), doi:10.1016/j.jneco.2010.09.008
Anomalies
Next cross-sectional differences in return
Risk or mispricing? From 1936-1975 Small Firms in the US earned higher returns than explained by CAPM
Banz (1981) and Reinganum (1981)
'.10
C A 0.05-i
0.00
~ ~
--
The January Effect small losing firms have high returns in January
Keim (1983) and Reinganum (1983)
s,'u-vq.ej
req p. e e~?~pvi-e
9.,
r sr
v-,
i.e-s
e s-r
r-t
10
I5
20
25
Figure 1. Cumulative Average Residuals for Winner and Loser Portfolios of 35 Stocks (1-36 months into the test period)
While not reported here, the results using market model and Sharpe-Lintner residualsare similar. They are also insensitive to the choice of December as the month of portfolio formation (see De Bondt [7]). The overreaction hypothesis predicts that, as we focus on stocks that go throughmore (or less) extreme return experiences (duringthe formationperiod), the subsequentprice reversals will be more (or less) pronounced.An easy way to generate more (less) extreme observations is to lengthen (shorten) the portfolio formationperiod;alternatively, for any given formation period (say, two years), we may compare the test period performance of less versus more extreme portfolios, e.g., decile portfolios (which contain an average 82 stocks) versus portfolios of 35 stocks. Table I confirms the prediction of the overreaction hypothesis. As the cumulative average residuals (during the formation period) for various sets of winner and loser portfolios grow larger, so do the subsequent price reversals, measured by [ACARL,t - ACARw,,] and the accompanying t-statistics. For a formation period as short as one year, no reversal is observed at all. Table I and Figure 2 further indicate that the overreaction phenomenon is qualitativelydifferent from the January effect and, more generally, from season-
One of the best known anomalies Returns on small firms are higher than those on large firms after controlling for risk Initial study, smallest 20% of NYSE stocks yield 19.8% higher returns than largest -- huge premium!
Some recent evidence suggests the January effect is a pure anomaly: in the UK says that after the small firm effect was publicized, the pattern has reversed
41
42
11/9/11
Empirical tests:
Variation among analyst earnings forecasts or amount of research available about firms was significantly related to strength of small firm effect
43
44
Tests of the EMH the Book to Market Anomaly 2. Book to Market predicts return:
A seemingly powerful predictor of returns across securities (a role that should be played by ) is book-to-market ratio Fama and French (1992): build 10 groups of stocks by book-tomarket ratio.
Group with highest ratio had return of 1.65% per month Group with lowest ratio had return of 0.72% per month
45
46
Week-end-Effects
Returns are reliably negative over weekends from 1953-77 in US (French, 1980)
47
48