emh empirical

7/30/2019 EMH Empirical

1/57

Efficent

Market

Hypotesis

&

Empirical Evidence

on Security Returns


2/57

Do security prices reflect information ?

Why look at market efficiency

Implications for business and corporate finance

Implications for investment

Efficient Market Hypothesis (EMH)

Random Walk and the EMH

Random Walk - stock prices are random

Actually submartingale

Expected price is positive over time

Positive trend and random about the trend


3/57

Random Walk with Positive Trend

Security

Prices

Time


4/57

Why are price changes random?

Prices react to information

Flow of information is random

Therefore, price changes are random

Random Price Changes

Stock prices fully and accurately reflect publicly

available information Once information becomes available, market

participants analyze it

Competition assures prices reflect information

EMH and Competition


5/57

So, What is market efficiency ?

Prices reflect all available information.

Thus, fin. asset prices are fairprices.

They are neither too high, nor too low.

What is meant by all available information?

Historical trading data

Publicly available information

All (private and public) information


6/57

Three Forms of Efficiency

Strong form of capital market efficiency. Current prices reflect all information that can

possibly be known to anyone.

Semi-strong form of cap mkt efficiency. Current prices reflect all publicly available

information.

Weak form of capital market efficiency. Current prices reflect only the information

contained in past prices.


7/57

Information set of market efficiency

Strong Form

(All information affecting the assets value)

Semi-Strong Form

(All publicly available information)

Weak Form

(Information contained in historical trading

data)


8/57

Implication of Efficiency for Investors

Future market prices cannot be predicted based onavailable information

Random Walk

Investments in these markets have a zero NPV. The expected rate of return equals the required

rate of return.

The expected rate of return compensates the

investor for the risk borne.

Abnormally high returns are earned by pure

chance.


9/57

Frictions in Capital Markets

Frictions in the capital markets prevent thesemarkets from being perfectly efficient.

Frictions include:

Transaction Costs: time, effort, and money

required to make a transaction.

Asymmetric taxes. Asymmetric information.


10/57

Information and Price Movements In an efficient capital market, prices reflect all available

information.

When new information arrives, prices react instantaneously.

Since new information is that which cannot be predicted, it

would arrive at random points in time.

Price movements are random (i.e. cannot be predicted).

Technical Analysis - using prices and volume information to

predict future prices

Weak form efficiency & technical analysis

Fundamental Analysis - using economic and accounting

information to predict stock prices

Semi strong form efficiency & fundamental analysis

Types of Stock Analysis

f k


11/57

Reaction of Stock Price to New

Information in Efficient and Inefficient

MarketsStock

Price

Days before (+) andafter (-) announcement

30

20

10 0 +10 +20 +30

Overreaction andreversion

Delayed response(Underreaction)

Efficient-marketresponse to new information


12/57

Active Management : Security analysis, Timing Passive Management: Buy and Hold, Index Funds

Implications of Efficiency for Active

or Passive Management

Market Efficiency

and Portfolio Management

Even if the market is efficient a role exists for

portfolio management:

Appropriate risk level

Tax considerations

Other considerations


13/57

Event studies

Assessing performance of professional managers

Testing some trading rule

Empirical Tests of Market Efficiency

How Tests Are Structured?

1. Examine prices and returns over time


14/57

Returns Over Time

0 +t-t

Announcement Date


15/57

2. Returns are adjusted to determine if they are abnormal

Market Model approacha. Rt = at + btRmt + et

(Expected Return)b. Excess Return =

(Actual - Expected)et = Actual - (at + btRmt)

How Tests Are Structured (contd)

c. Cumulate the excess returns over time:

0 +t-t


16/57

Magnitude Issue

Selection Bias Issue

Lucky Event Issue

Possible Model Misspecification

Issues in Examining the Results

What Does the Evidence Show? Technical Analysis

Short horizon Long horizon

Fundamental Analysis

Anomalies Exist


17/57

Small Firm Effect (January Effect)

Neglected Firm

Market to Book Ratios

Reversals

Post-Earnings Announcement Drift

Market Crash of 1987

Anomalies

Mutual Fund and Professional Manager Performance

Some evidence of persistent positive and negative

performance Potential measurement error for benchmark returns

Style changes

May be risk premiums

Superstar phenomenon


18/57

Why? Perfect Capital Markets?

No barriers to entry.

Perfect competition.

each participant is sufficiently small andcannot affect prices by her/his actions.

Financial assets are infinitely divisible. No transaction costs.

All information is fully available to everyparticipant, at no cost.

No tax asymmetries. No restrictions on trading.


19/57

Some Imperfections

Asymmetric taxes

These change the zero-sum nature of capital

market transactions.

Asymmetric information

Information is not equally (and costlessly)available to all market participants.

Transaction costs

Generally less important an imperfection.Frictions in the capital markets prevent markets

from being perfectly efficient


20/57

STRATEGY IN AN EFFICIENT MARKET

Diversify; select suitable asset allocation

Dont try to time security price movement

Keep tax consideration in mind

Passive investing: Index Fund, Dollar cost

averaging


21/57

Empirical Evidenceon Security Returns


22/57

Overview of Investigation Tests of the single factor CAPM or APT Model

Tests of the Multifactor APT Model Results are difficult to interpret

Studies on volatility of returns over time

Tests of the Single Factor Model

Tests of the expected return beta relationship

First Pass Regression

Estimate beta, average risk premiums, unsystematic risk

Second Pass: Using estimates from the first pass to determine

if model is supported by the data

Most tests do not generally support the single factor model


23/57

Single Factor Test Results

Return %

Beta

Predicted

Actual


24/57

Rolls Criticism

Only testable hypothesis is on the efficiency of the

market portfolio

Benchmark error

Measurement Error in Beta Statistical property

If beta is measured with error in the first stage

Second stage results will be biased in the directionthe tests have supported

Test results could result from measurement error


25/57

Tests of the Multifactor Model

Chen, Roll and Ross 1986 Study. Factors:

Growth rate in industrial production

Changes in expected inflation

Unexpected inflation

Changes in risk premiums on bonds

Unexpected changes in term premium on bonds

Study Structure & Results Method: Two -stage regression with portfolios constructed by

size based on market value of equity

Findings:

Significant factors: industrial production, risk premium on

bonds and unanticipated inflation

Market index returns were not statistically significant in the

multifactor model


26/57

Anomalies LiteratureIs the CAPM or APT Model Valid?

Numerous studies show the approach is not valid Why do the studies show this result

Other factors influence returns on securities

Statistical problems prohibit a good test of the model

Fama and French Study

Size and book-to-market ratios explain returns on securities

Beta is not a significant variable when other variables areincluded

Study results show no support for the CAPM or APT

Researchers Responses


27/57

Researchers Responses

to Fama and French Utilize better econometric techniques

Improve estimates of beta

Reconsider the theoretical sources and implications of the

Fama and French-type results

Return to the single-index model, accounting for nontraded

assets and cyclical behavior of betas

Jaganathan and Wang Study Included factors for cyclical behavior of betas and human

capital

When these factors were included the results showed returns

were a function of beta

Size is not an important factor when cyclical behavior and

human capital are included

S h i V l ili


28/57

Stochastic Volatility

Stock prices change primarily in reaction to

information New information arrival is time varying

Volatility is therefore not constant through time

Stock Volatility Studiesand Techniques

Pagan and Schwert Study

Study of 150 years of volatility on NYSE stocks Volatility is not constant through time

Improved modeling techniques should improve results of

tests of the risk-return relationship

GARCH Models to incorporate time varying volatility


29/57

IS THE MARKET EFFICIENT ?

According to Eugene Fama

According to Grossman-Stiglitz

According to Fischer Black

According to You ?


30/57


31/57

Efficient markets: intuition

Expectation

Time

Price

Realization

Price change is

unexpected


32/57

Weak Form Efficiency

Random-walk model: Pt-Pt-1 = Pt-1 * (Expected return) + Random error

Expected value (Random error) = 0

Random error of period t unrelated to random component of anypast period

Implication: Expected value (Pt) = Pt-1 * (1 + Expected return)

Technical analysis: useless

Empirical evidence: serial correlation Correlation coefficient between current return and some past

return

Serial correlation = Cor (Rt, Rt-s)


33/57

S&P500 Daily returns

-0.08

-0.06

-0.04

-0.02

0

0.02

0.04

0.06

0.08

-0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08

Return day t

Returndayt+1


34/57

Semi-strong Form Efficiency

Prices reflect all publicly available information

Empirical evidence: Event studies

Test whether the release of information influencesreturns and when this influence takes place.

Abnormal return AR : ARt = Rt - Rmt

Cumulative abnormal return:

CARt = ARt0 + ARt0+1 + ARt0+2 +... + ARt0+1


35/57

Efficient Market Theory

-16

-11

-6

-1

4

9

14

19

24

29

34

39

Days Relative to annoncement date

Cum

ulativeAbnorm

alReturn

(%)

Announcement Date


36/57

Example: How stock splits

affect value

0

5

10

15

20

25

30

35

40

Month relative to split

Cumulativeabnormal

return %

-29 0 30

Source: Fama, Fisher, Jensen & Roll

E t St di Di id d


37/57

Event Studies: Dividend

OmissionsCumulative Abnormal Returns for Companies Announcing

Dividend Omissions

0.146 0.108

-0.72

0.032-0.244

-0.483

-3.619

-5.015-5.411

-5.183-4.898-4.563-4.747-4.685-4.49

-6

-5

-4

-3

-2

-1

0

1

-8 -6 -4 -2 0 2 4 6 8

Days relative to announcement of dividend omission

Cumulative

abnormalreturn

(%)

Efficient market

response to bad news

S.H. Szewczyk, G.P. Tsetsekos, and Z. Santout Do Dividend Omissions Signal Future Earnings or Past Earnings?Journal of Investing

(Spring 1997)


38/57

Strong-form Efficiency

How do professional portfolio managers perform?

Jensen 1969: Mutual funds do not generate abnormalreturns

Rfund - Rf= + (RM - Rf)

Insider trading

Insiders do seem to generate abnormal returns

(should cover their information acquisition activities)


39/57

What moves the market

Who knows?

Lot of noise:

1985-1990: 120 days with | DJ| > 5%

28 cases (1/4) identified with specific event(Siegel Stocks for the Long Run Irwin 1994 p 184)

Orange juice futures (Roll 1984)

90% of the day-to-day variability cannot explained by

fundamentals

Pity financial journalists!


40/57

PhD 01-1 |40

Trading Is Hazardous to Your Wealth(Barber and OdeanJournal of Finance April 2000)

Sample: trading activity of 78,000 households1991-1997

Main conclusions:1. Average household underperforms benchmark by

1.1% annually2. Trading reduces net annualized mean returns

Infrequent traders: 18.5% Frequent traders: 11.4%

3. Households trade frequently (75% annual turnover)

4. Trading costs are high: for average round-trip trade4%

(Commissions 3%, bid-ask spread 1%)


41/57

PhD 01-1 |41

US Equity Mutual Funds 1982-1991(Malkiel, Journal of Finance June 1995)

Average Annual Return Capital appreciation funds 16.32%

Growth funds 15.81%

Small company growth funds 13.46%

Growth and income funds 15.97% Equity income funds 15.66%

S&P 500 Index 17.52%

Average deviation from benchmark -3.20%

(risk adjusted)


42/57

PhD 01-1 |42

: Excess Return

Excess return = Average return - Risk adjusted

expected return

Risk

Return Expected return

Average

return

Risk

Jensen 1968 Distribution of t values for


43/57

PhD 01-1 |43

Jensen 1968 - Distribution of t values for 115 mutual funds 1955-1964

0

5

10

15

20

25

30

35

-5 -4 -3 -2 -1 0 1 2 3 4

Not significantly

different from 0


44/57

US Mutual Funds

Consistency of Investment Result

Successive Period Performance

Initial Period Performance Top Half Bottom Half

Goetzmann and Ibbotson (1976-1985)

Top Half 62.0% 38.0%

Bottom Half 36.6% 63.4%

Malkiel, (1970s)

Top Half 65.1% 34.9%


Malkiel, (1980s)

Top Half 51.7% 48.3%


Source: Bodie, Kane, Marcus Investments 4th ed. McGraw Hill 1999 (p.118)

D iti f M t l F d


45/57

Decomposition of Mutual Fund

Returns(Wermers Journal of Finance August 2000)

Sample: 1,758 funds 1976-1994

Benchmark 14.8%

+1%

Gross return 15.8%

Expense ratio 0.8%

Transaction costs 0.8%

Non stock holdings 0.4%

Net Return 13.8%

Stock picking +0.75%

No timing ability

Deviation from benchmark +0.55%

Funds outperformbenchmark

Not enough to cover

costs

E i i l h ll


46/57

Empirical challenges

Explaining the cross section of returns

Explaining changes in expected returns

Beta


47/57

Beta

NoDu

Durb

Oil

Chem

Manu

TelcUtil

ShopMone

Other

MktPort

RF

0.00

2.00

4.00

6.00

8.00

10.00

12.00

14.00

16.00

18.00

0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60

Beta

Averagereturn

NoDu

Durb

Oil

ChemManu

Telc

Util

Shop

Mone

Other

MktPort

RF

Fama French


48/57

Average return vs market beta for 25 FF stock portfolios 1926-2004

Mkt

RF

S1S2

S3

S4

S5

GovB

CorpB

BM1

BM2

BM3

BM4

BM5

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

-0.20 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1.80

Beta

Meanretu

rn

Size S: S1 smallest - S5 biggest

B/M: BM1 lowest - BM5 highest

Average monthly returns

Small Big

LowB/M 0.91 1.01 1.08 1.01 0.92 0.99

1.29 1.33 1.26 1.10 0.92 1.18

1.50 1.46 1.30 1.30 1.03 1.32

1.69 1.51 1.40 1.35 1.11 1.41

HighB/M 1.83 1.64 1.53 1.46 1.34 1.56

1.45 1.39 1.32 1.24 1.07

Fama French


49/57

PhD 01-1 |49

Size and B/M

12

34

5

Low B/M

S2

S3

S4High B/M

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

m

e

a

n

m

o

nt

h

l

y

r

e

t

u

r

n

%

Size

Value

Low B/M

Series2

Series3

Series4

High B/M

Small

Big

Fama French


50/57

Small Big

Av.Ret. 1.45 1.39 1.32 1.24 1.07

St.Dev. 9.47 7.59 7.03 6.70 6.45

Beta 1.35 1.17 1.15 1.13 1.05

SIZE

Low High

Av.Ret. 0.99 1.18 1.32 1.41 1.56

St.Dev. 7.38 6.91 6.75 7.10 8.85Beta 1.17 1.12 1.10 1.13 1.32

B/M

Based on monthly data 192607 200411

File: 25_Portfolios_5x5_monthly.xls

Fama French

Fama French


51/57

Fama FrenchFama French Factors - Annual

-60

-40

-20

0

20

40

60

80

1927

1930

1933

1936

1939

1942

1945

1948

1951

1954

1957

1960

1963

1966

1969

1972

1975

1978

1981

1984

1987

1990

1993

1996

1999

2002

RM SMB HML

Predictability: Interest Rates and Expected Inflation


52/57

Predictability: Interest Rates and Expected Inflation

Sample period (Sample Size)

1831-2002 (2,053) -2.073

(-3.50)

1831-1925 (1,136) -3.958

(-4.58)

1926-1952 (324) 0.114

(0.03)

1953-1971 (228) -5.559

(-2.57)

1972-2002 (357) -1.140(-1.08)

Schwert, W., Anomalies and Market Efficiency,WP October 2002

http://ssrn.com/abstract_id=338080

tftmt RR

Predictability: D/P


53/57

Predictability: D/P

Price/dividend ratio

0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

90.00

100.00

1926

1928

1930

1932

1934

1936

1938

1940

1942

1944

1946

1948

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

Predictability


54/57

Predictability

Nobs 77

R(t+1)=a+b*R(t)+e(t+1)

Mean StDev Slope Standerro t R

Stock 0.1190 0.2050 0.03 0.1154 0.27 0.001

Tbill 0.0421 0.0350 0.92 0.0465 19.79 0.838

Excess 0.0769 0.2083 0.04 0.1155 0.31 0.001

Excess(t+x) = a + b (D/P)(t) + e

Horizon

1 year 4.17 1.60 2.61 0.0822 year 8.13 2.26 3.60 0.147

3 years 11.27 2.62 4.30 0.200

4 years 13.69 2.95 4.64 0.228

5 years 15.02 3.21 4.67 0.233


55/57

ER(+5)=a+b*(D/P)(t)+e

-1

-0.5

0

0.5

1

1.5

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

D/P

ExcessReturn+

5

Econometrician wanted


56/57

Econometrician wantedExcess Return + 5 : Residuals

-1.5

-1

-0.5

0

0.5

1

1.5

1926

1928

1930

1932

1934

1936

1938

1940

1942

1944

1946

1948

1950

1952

1954

1956

1958

1960

1962

1964

1966

1968

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998


57/57

Reference:

Investment, 2008. Bodie-Kent-Markus.

Lecture Handout - Prof. Roy Sembel (2008)

International Investment, Prof. Andr Farber,Solvay Business School Universit Libre de Bruxelles

emh empirical

Documents