Hypotheses of the CAPM Tests of the CAPM

Lecture 3.4: Empirical Evidence on the CAPM

Investment Analysis
Fall, 2012
Anisha Ghosh
Tepper School of Business
Carnegie Mellon University
November 15, 2012
Hypotheses of the CAPM Tests of the CAPM
Some Hypotheses of the CAPM
1
Higher risk (Beta) should be associated with a higher level of return.
2
Return is linearly related to Beta for every unit increase in Beta,
there is the same increase in return.
3
The difference in average returns between two assets should be
entirely driven by difference in their Betas there should be no added
return for bearing nonmarket risk
4
If the CAPM holds, investing should constitute a fair game with respect
to it deviations of a security or portfolio from equilibrium should be
purely random and there should ne no way to use these deviations to
earn an excess prot.
Hypotheses of the CAPM Tests of the CAPM
Some Hypotheses of the CAPM
1
Higher risk (Beta) should be associated with a higher level of return.
2
Return is linearly related to Beta for every unit increase in Beta,
there is the same increase in return.
3
The difference in average returns between two assets should be
entirely driven by difference in their Betas there should be no added
return for bearing nonmarket risk
4
If the CAPM holds, investing should constitute a fair game with respect
to it deviations of a security or portfolio from equilibrium should be
purely random and there should ne no way to use these deviations to
earn an excess prot.
Hypotheses of the CAPM Tests of the CAPM
Some Hypotheses of the CAPM
1
Higher risk (Beta) should be associated with a higher level of return.
2
Return is linearly related to Beta for every unit increase in Beta,
there is the same increase in return.
3
The difference in average returns between two assets should be
entirely driven by difference in their Betas there should be no added
return for bearing nonmarket risk
4
If the CAPM holds, investing should constitute a fair game with respect
to it deviations of a security or portfolio from equilibrium should be
purely random and there should ne no way to use these deviations to
earn an excess prot.
Hypotheses of the CAPM Tests of the CAPM
Some Hypotheses of the CAPM
1
Higher risk (Beta) should be associated with a higher level of return.
2
Return is linearly related to Beta for every unit increase in Beta,
there is the same increase in return.
3
The difference in average returns between two assets should be
entirely driven by difference in their Betas there should be no added
return for bearing nonmarket risk
4
If the CAPM holds, investing should constitute a fair game with respect
to it deviations of a security or portfolio from equilibrium should be
purely random and there should ne no way to use these deviations to
earn an excess prot.
Hypotheses of the CAPM Tests of the CAPM
Some Hypotheses of the CAPM
1
Higher risk (Beta) should be associated with a higher level of return.
2
Return is linearly related to Beta for every unit increase in Beta,
there is the same increase in return.
3
The difference in average returns between two assets should be
entirely driven by difference in their Betas there should be no added
return for bearing nonmarket risk
4
If the CAPM holds, investing should constitute a fair game with respect
to it deviations of a security or portfolio from equilibrium should be
purely random and there should ne no way to use these deviations to
earn an excess prot.
Hypotheses of the CAPM Tests of the CAPM
Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes (1972))
Cross Sectional Tests (Fama and Macbeth (1973))
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972))
Black, Jensen, and Scholes (1972) were the rst to conduct an in-depth
time series test of the CAPM
They took as their basic time series model:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
If the CAPM holds
i
= 0
The estimated
i
measures the systematic departure of the return on
asset i from the CAPM benchmark
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972))
Black, Jensen, and Scholes (1972) were the rst to conduct an in-depth
time series test of the CAPM
They took as their basic time series model:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
If the CAPM holds
i
= 0
The estimated
i
measures the systematic departure of the return on
asset i from the CAPM benchmark
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972))
Black, Jensen, and Scholes (1972) were the rst to conduct an in-depth
time series test of the CAPM
They took as their basic time series model:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
If the CAPM holds
i
= 0
The estimated
i
measures the systematic departure of the return on
asset i from the CAPM benchmark
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972))
Black, Jensen, and Scholes (1972) were the rst to conduct an in-depth
time series test of the CAPM
They took as their basic time series model:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
If the CAPM holds
i
= 0
The estimated
i
measures the systematic departure of the return on
asset i from the CAPM benchmark
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972))
Black, Jensen, and Scholes (1972) were the rst to conduct an in-depth
time series test of the CAPM
They took as their basic time series model:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
If the CAPM holds
i
= 0
The estimated
i
measures the systematic departure of the return on
asset i from the CAPM benchmark
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972)) contd.
Black, Jensen, and Scholes (1972)) performed the time series
regression on portfolios (instead of individual securities).
In order to maximize the spread in Betas across portfolios, they sorted
stocks into 10 portfolios based on their Betas.
Procedure:
Use 5 years of monthly data to estimate the Beta of each stock
and sort them into deciles
Compute monthly returns on the 10 portfolios over the next (i.e.,
sixth) year
Use data from second to sixth year to rank stocks and form
deciles that are considered portfolios for the seventh year, ...
This process is repeated until deciles and the return for each
decile is computed for 35 years
Each of the 10 portfolios is then used to perform the time series
regression
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972)) contd.
Black, Jensen, and Scholes (1972)) performed the time series
regression on portfolios (instead of individual securities).
In order to maximize the spread in Betas across portfolios, they sorted
stocks into 10 portfolios based on their Betas.
Procedure:
Use 5 years of monthly data to estimate the Beta of each stock
and sort them into deciles
Compute monthly returns on the 10 portfolios over the next (i.e.,
sixth) year
Use data from second to sixth year to rank stocks and form
deciles that are considered portfolios for the seventh year, ...
This process is repeated until deciles and the return for each
decile is computed for 35 years
Each of the 10 portfolios is then used to perform the time series
regression
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972)) contd.
Black, Jensen, and Scholes (1972)) performed the time series
regression on portfolios (instead of individual securities).
In order to maximize the spread in Betas across portfolios, they sorted
stocks into 10 portfolios based on their Betas.
Procedure:
Use 5 years of monthly data to estimate the Beta of each stock
and sort them into deciles
Compute monthly returns on the 10 portfolios over the next (i.e.,
sixth) year
Use data from second to sixth year to rank stocks and form
deciles that are considered portfolios for the seventh year, ...
This process is repeated until deciles and the return for each
decile is computed for 35 years
Each of the 10 portfolios is then used to perform the time series
regression
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972)) contd.
The model explains excess returns well (high values of correlation
coefcients) support for the linear equation as a good explanation of
security returns
However, the s vary quite a bit from zero.
Hypotheses of the CAPM Tests of the CAPM
Time Series Tests (Black, Jensen, and Scholes
(1972)) contd.
The model explains excess returns well (high values of correlation
coefcients) support for the linear equation as a good explanation of
security returns
However, the s vary quite a bit from zero.
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
First-pass regression: Form 20 portfolios and estimate Betas using time
series regression:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
Second-pass regression: For each month subsequent to the estimation
period, perform the following cross-sectional regression:
R
it
=
0t
+
1t

i
+
2t

2
i
+
3t
S
ei
+
it
CAPM implies
1
E (
3t
) = 0
2
E (
2t
) = 0
3
E (
1t
) > 0
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
First-pass regression: Form 20 portfolios and estimate Betas using time
series regression:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
Second-pass regression: For each month subsequent to the estimation
period, perform the following cross-sectional regression:
R
it
=
0t
+
1t

i
+
2t

2
i
+
3t
S
ei
+
it
CAPM implies
1
E (
3t
) = 0
2
E (
2t
) = 0
3
E (
1t
) > 0
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
First-pass regression: Form 20 portfolios and estimate Betas using time
series regression:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
Second-pass regression: For each month subsequent to the estimation
period, perform the following cross-sectional regression:
R
it
=
0t
+
1t

i
+
2t

2
i
+
3t
S
ei
+
it
CAPM implies
1
E (
3t
) = 0
2
E (
2t
) = 0
3
E (
1t
) > 0
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
First-pass regression: Form 20 portfolios and estimate Betas using time
series regression:
R
it
R
Ft
=
i
+
i
(R
Mt
R
Ft
) +
it
Second-pass regression: For each month subsequent to the estimation
period, perform the following cross-sectional regression:
R
it
=
0t
+
1t

i
+
2t

2
i
+
3t
S
ei
+
it
CAPM implies
1
E (
3t
) = 0
2
E (
2t
) = 0
3
E (
1t
) > 0
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
contd.
Main ndings:
The average value of
3t
is small and not statistically different from
zero residual risk has no effect on the expected return of a
security
The average value of
2t
is small and not statistically different from
zero
The average value of
1t
is statistically signicantly positive, albeit
smaller than the sample market risk premium support for linear
relation between expected return and Beta
The average value of
0t
is greater than the sample average of the
riskless rate R
F
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
contd.
Main ndings:
The average value of
3t
is small and not statistically different from
zero residual risk has no effect on the expected return of a
security
The average value of
2t
is small and not statistically different from
zero
The average value of
1t
is statistically signicantly positive, albeit
smaller than the sample market risk premium support for linear
relation between expected return and Beta
The average value of
0t
is greater than the sample average of the
riskless rate R
F
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
contd.
Main ndings:
The average value of
3t
is small and not statistically different from
zero residual risk has no effect on the expected return of a
security
The average value of
2t
is small and not statistically different from
zero
The average value of
1t
is statistically signicantly positive, albeit
smaller than the sample market risk premium support for linear
relation between expected return and Beta
The average value of
0t
is greater than the sample average of the
riskless rate R
F
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
contd.
Main ndings:
The average value of
3t
is small and not statistically different from
zero residual risk has no effect on the expected return of a
security
The average value of
2t
is small and not statistically different from
zero
The average value of
1t
is statistically signicantly positive, albeit
smaller than the sample market risk premium support for linear
relation between expected return and Beta
The average value of
0t
is greater than the sample average of the
riskless rate R
F
Hypotheses of the CAPM Tests of the CAPM
Cross Sectional Tests (Fama and Macbeth (1973))
contd.
Main ndings:
The average value of
3t
is small and not statistically different from
zero residual risk has no effect on the expected return of a
security
The average value of
2t
is small and not statistically different from
zero
The average value of
1t
is statistically signicantly positive, albeit
smaller than the sample market risk premium support for linear
relation between expected return and Beta
The average value of
0t
is greater than the sample average of the
riskless rate R
F
Hypotheses of the CAPM Tests of the CAPM
Fama and French (1992)
Approach:
Sort stocks by Beta (
i
), size (ln(ME
i
)), and the ratio of the book
value of equity to the market value of equity (B/M
i
)
Estimate Beta for each stock and run the following cross-sectional
regression:
E (R
it
) =
0
+

i
+
size
ln (ME
i
) +
B/M
(B/M)
i
+
i
Hypotheses of the CAPM Tests of the CAPM
Fama and French (1992)
Approach:
Sort stocks by Beta (
i
), size (ln(ME
i
)), and the ratio of the book
value of equity to the market value of equity (B/M
i
)
Estimate Beta for each stock and run the following cross-sectional
regression:
E (R
it
) =
0
+

i
+
size
ln (ME
i
) +
B/M
(B/M)
i
+
i
Hypotheses of the CAPM Tests of the CAPM
Fama and French (1992)
Approach:
Sort stocks by Beta (
i
), size (ln(ME
i
)), and the ratio of the book
value of equity to the market value of equity (B/M
i
)
Estimate Beta for each stock and run the following cross-sectional
regression:
E (R
it
) =
0
+

i
+
size
ln (ME
i
) +
B/M
(B/M)
i
+
i
Hypotheses of the CAPM Tests of the CAPM
Size Effect
Beta explains almost nothing!
Size matters: Small rms have, on average, earned higher returns than
that implied by the CAPM
Hypotheses of the CAPM Tests of the CAPM
Size Effect
Beta explains almost nothing!
Size matters: Small rms have, on average, earned higher returns than
that implied by the CAPM
Hypotheses of the CAPM Tests of the CAPM
Book-to-Market Effect
Value stocks (i.e. stocks that trade at less than their fundamental) tend
to outperform Growth or glamor stocks (i.e. the trendy ones on the
market perceived to have high growth potential)
Hypotheses of the CAPM Tests of the CAPM
A Horse Race: