Professional Documents
Culture Documents
t
t t
t
P
P P
R (3)
Where:
t
R is the return oI a stock in the period t;
t
P and
1 t
P is the closing price level oI
share or index in the period t and t-1 respectively.
Thus, monthly return oI each security and the index are calculated based on equation (2),
and these returns will be used Ior (i) computing betas (ii) examining the relationship
between return and beta.
Moreover, as mentioned above, all 80 sample shares will be evenly sorted into 8
portIolios based on ranking their individual beta in ascending order, hence, calculating
and ranking beta is necessary. In general, the pre-ranking betas are estimated using 60
monthly returns. According to Arnold (2002), beta which indicates the sensitivity oI a
security to general market movements is:
2
,
) (
m
m i
i
R R Cov
o
| = (2)
Where: ) (
, m i
R R Cov is the covariance oI security i with the market portIolio, and
2
m
o is
the variance oI the market portIolio.
58
AIter calculating realized monthly return Ior each security, as well as Ior portIolio and
index, and also the beta oI each share and the portIolio in required diIIerent periods. The
next stage is to use the two-pass regression model that is Irom Fama and MacBeth (1973),
with modiIications suggested by Pettengill et al. (1995), to examine whether there is a
signiIicant positive risk premium on beta.
For the unconditional approach oI the beta-return relationship, Iollowing regression
model is used in order to test iI there is a positive risk-return relationship:
it i t t it
e R + + = |
1 0
(4)
The test oI the model is based on the mean oI the coeIIicients oI the monthly regressions.
The CAPM implies
ft t
R =
0
and ) (
1 ft mt t
R R = . Where
it
R is the return on asset i at
time t;
ft
R is the risk-Iree rate;
i
| represents the estimated beta oI asset i;
mt
R is the
market indexes return (market portIolio return) at time t; ) (
ft mt
R R is the market risk
premium; and Iinally
it
e is a random error term.
By using equation (4), the market risk premium can be estimated. II the risk premium, in
this case, the value Ior
t 1
is greater than zero (i.e.
ft mt
R R > ), hence, the predicted return
includes a positive risk premium that is proportionate to beta. On the other hand,
predicted return would contain a negative risk premium that is proportionate to beta iI the
risk premium is negative. Consequently, iI the realized return is more than the risk-Iree
59
return with a signiIicant t statistic, a positive unconditional relationship between beta and
return is supported. Such relationship provides important implications concerning about
the empirical test Ior a systematic relationship between beta and return.
Furthermore, the tested relationship between beta and return can be plotted on the
Security Market Line (SML), which shows return as a Iunction oI beta. However, it is
necessary to adjust the SML by some technique skills and statistical methodology, Ior
example, separate the SML during the diIIerent positive or negative market periods.
According to the Iormer experiences oI Pettengill et al. (1995) and Fletcher (1997, 2000),
there should be a positive beta-return relationship during the bullish market time interval,
whereas a negative relationship between beta and return should exist under the condition
that market is bearish. Pettengill, Sundaram and Mathur (1995) divided the monthly risk
premium estimates into two sub-sample periods that taking account oI whether the excess
market return is positive or negative. Jonathan Fletcher (1997, 2000) also divided the
sample period in his research into up market and down market periods based on the
criterion oI whether market return exceeded the risk-Iree return.
As a result, in order to test the conditional relationship, the total sample period in this
study should be divided into two types oI periods with the Iirst type is up periods when
market risk premium is positive, and the second type is down period when market risk
premium is negative. The separate analysis oI the two sub-sample periods can oIIer two
liner regression lines.
60
In respect to Figure 3.1 that has been presented in Chapter 3, the SSE Index had a
Iluctuated rising trend Irom 2000 to the middle oI 2001, achieving the top index oI
2245.42 on 29
th
June 2001, and then declined continuously with only 1161.05 in
December 2005. ThereIore, the market index in this study could be described as a very
short bullish market and a relatively long bearish market.
Similarly, in accordance with Pettengill et al. (1995) and Fletcher (1997, 2000), the
sample period oI this empirical study is split into two periods according to iI the market
return is positive or negative, and each period covers monthly closing prices oI each
security and the market indices as well.
Regarding to the positive and negative market return, Pettengill et al. (1995) adjusted the
Fama and MacBeth (1973) approach oI the unconditional beta-return relationship to
examine the conditional relationship between beta and return by arguing that studies
Iocusing on the beta-return relationship should take account oI the Iact that ex post rather
than ex ante returns are used in the empirical tests. It is obviously that investors would
expect greater return Irom high beta portIolio than low beta portIolio because oI the
compensation Ior additional risks. On the contrary, when the market return is lower than
the risk-Iree rate oI return, i.e. down market, investors would expect that the realized
return on a low beta portIolio would be greater than the return on a high beta portIolio
since high beta portIolio is more sensitive to the market and tend to loss more.
Consequently, the implication is that there should be a positive beta-return relationship
61
when excess market return is positive; and a negative beta-return relationship when the
excess market return is negative.
ThereIore, in the second stage, the cross-sectional regression should be estimated Ior the
conditional relationship between beta and return:
it i t i t ot it
e D D R + + + = | | ) 1 (
2 1
(5)
Where D is a dummy variable that equals to one iI the excess market return is positive
(i.e. D1 iI ) (
ft mt
R R ~0), and equals to zero iI the excess market return is negative (i.e.
D0 iI ) (
ft mt
R R 0);
t 1
and
t 2
is the monthly risk premium estimates in up market
months (positive excess market returns) and down market months (negative excess
market return) respectively. Equation (5) indicates that either
t 1
and
t 2
will be
estimated in a given month depending on the sign oI the excess market return.
The hypotheses predicated by Pettengill et al. (1995) are as Iollows:
H1:
1
is the average value oI the coeIIicient
t 1
which should be positive when it is
estimated in periods with positive excess market returns.
0 :
0 :
1
1 0
>
=
a
H
H
H2:
2
is the average value oI the coeIIicient
t 2
that should be negative when it is
estimated in period with negative excess market returns.
62
0 :
0 :
2
2 0
<
=
a
H
H
The above two hypothesis can be tested by using the standard t-statistic. Pettengill et al.
(1995) pointed out that two conditions are necessary Ior a positive tradeoII between risk
and return. II the conditional CAPM is valid, a systematic conditional relationship
between beta and realized return is supported in both cases, where the null hypothesis
should be rejected.
The last stage is to evaluate the regression and the Security Market Line (SML) based on
the results Irom the unconditional and conditional beta-return relationship analysis. SML
can be drawn on the basis oI the relationship between risk and return, where risk can be
measured by beta. Moreover, is should not been ignored that the calculation Iunction oI
Excel is used as a very important instruments through all processes.
63
4.2 Empirical Test Results and Discussions
Through literature review (Chapter 2), data collection and methodology presentation
(Section 4.1), it can be easier and clearer to test and analyze both the unconditional and
conditional relationship between beta and return.
4.2.1 General Findings
The sample period Ior this study extends Irom January 2001 to December 2005. The
monthly return oI stocks and the market index can be obtained Irom equation (2), with
the purpose to help to calculate the mean monthly return. Both the standard deviation and
beta oI securities are received on the basis oI equation (3) by using Excel. All above
results are shown in Appendix 1 that includes 80 shares chosen at random Irom A-class
shares in the SSE. The range oI mean monthly returns Ior these 80 securities are between
-37.90 (600120) and 11.50 (600002), the standard deviations are between 0.0899
(600052) and 0.5038 (600104), and the betas range are between -0.1523 (600132) and
2.5108 (600071).
In the scatter plot in Figure 4.1, Ior the total sample period Irom Jan 2001 to Dec 2005,
there is no obvious relationship between the risks estimated by standard deviation and
realized returns among 80 stocks. All observations have had negative returns except one
case (600002) that has the highest return without having the greatest value oI beta. The
correlation coeIIicient Ior the entire data sample is 0.4134. It can be seen though not
signiIicant Irom the plot data that lower risk stocks have had relative lower return
whereas high risk shares have got relatively higher return.
64
Figure 4.1
TotaI SampIe Periods
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0 0.1 0.2 0.3 0.4 0.5 0.6
Risk (STD)
R
e
t
u
r
n
Stock
On the other hand, there is no precisely linear relationship between betas and realized
returns among 80 securities either, as demonstrated in Figure 4.2 that is also a scatter
diagram. Nevertheless, the positive trend with the climbing betas does exist in the total
sample period, although it is not explicit. The correlation coeIIicient between beta and
return Ior the total sample period is 0.2693.
Figure 4.2
TotaI SampIe Periods
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
-0.5 0 0.5 1 1.5 2 2.5 3
Beta
R
e
t
u
r
n
Stock
65
The total sample period is Irom Jan 2001 to Dec 2005, which is 60 months. These 60
months are Iurther divided into two sub-periods with the Iirst sub-period Irom Jan 2001
to Jun 2003, and the second sub-period Irom Jul 2003 to Dec 2005.
Figure 4.3 Figure 4.4
Figure 4.3 and Figure 4.4 represents the relationship between betas and realized returns
Ior all 80 shares in the Iirst and second sub-period respectively. There is an observation in
Figure 4.3 that has had the extreme high realized return but moderate value oI beta. In
contrast, a stock has got the highest return associated with the biggest beta as showed in
Figure 4.4. The correlation coeIIicients Ior the two sub-periods are 0.3094 and 0.3163. In
Iact, though there is still no linear beta-return relationship exactly, the positive
relationship between beta and return is somewhat obvious, especially despite outliers oI
some extreme observations.
Based on these betas estimated Irom the total sample period and two sub sample periods
respectively, 80 stocks are sorted into 8 portIolios and each portIolio involves 10 stocks.
Sub SampIe Periods 2001.1-2003.6
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
-1 -0.5 0 0.5 1 1.5 2 2.5 3 3.5
Beta
R
e
t
u
r
n
Stocks
8
Sub SampIe Periods 2003.7-2005.12
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
-2 -1 0 1 2 3 4 5
Beta
R
e
t
u
r
n
Stocks
66
PortIolio 1 contains shares with the lowest betas; hence, portIolio 8 includes stocks with
the highest betas. Furthermore, in the total sample period and two sub-periods, the
portIolio betas are estimated. Table 4.1 exhibits the mean monthly return, standard
deviation and beta oI the 8 portIolios in the whole sample period.
1able 4.1: Portfolios-1otal Sample Period
PortfoIio Return Std. Deviation Beta ()
1 -0.1654987 0.0886911 0.1074482
2 -0.2578897 0.0899062 0.4302563
3 -0.2235533 0.1042360 0.6006156
4 -0.1772967 0.1198148 0.7494762
5 -0.2000875 0.1581201 0.8938922
6 -0.1860875 0.1726019 1.0405933
7 -0.1741028 0.2139060 1.3143666
8 -0.1282828 0.2912941 1.8587302
Certainly, it could be seen Irom Table 4.1 that the risk (Std. deviation) basically climbs
up when the beta oI portIolios increases. Although there are some variations among
realized returns associate with rising betas, in general, the mean monthly returns Irom
portIolio 2 to portIolio 8 go up as well. However, one special case should be noticed here
is that portIolio 1 with the lowest beta but the second highest mean monthly return.
Figure 4.5
TotaI SampIe Periods
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0 0.1 0.2 0.3 0.4
STD
R
e
t
u
r
n
Portf olios
67
Figure 4.
TotaI SampIe Periods
-0.3
-0.2
-0.1
0
Beta
R
e
t
u
r
n
Portfolios
Portfolios -0.165 -0.258 -0.224 -0.177 -0.2 -0.186 -0.174 -0.128
0.11 0.43 0.6 0.75 0.89 1.04 1.31 1.86
In addition, both Figure 4.5 and Figure 4.6 illustrate the average realized returns Ior 8
portIolios in combine with their standard deviation and beta respectively (with portIolio 1
consisting oI the lowest betas, as well as standard deviations) in the total sample period.
From Figure 4.5, it is clear that portIolio with higher standard deviation has had higher
rate oI realized return except portIolio 1. Meanwhile, Figure 4.6 illustrates that there is no
absolutely positive relationship between beta and realized return as the return oI
portIolios Iluctuating, but is has a climbing trend generally (despite portIolio 1).
Figure 4.7
Sub Period 1 (2001.1-2003.6)
-0.3
-0.2
-0.1
0
Beta
R
e
t
u
r
n
Portfolio
Portfolio -0.197 -0.182 -0.179 -0.199 -0.124 -0.098 -0.041 -0.133
0.05 0.35 0.55 0.7 0.84 1.1 1.33 2.02
68
In the two sub-periods, Figure 4.7 revealed that there is almost a clear positive trend
between beta and return Ior these 8 portIolios in the Iirst sub-period (2001.1-2003.6)
despite portIolio 8, as its return declines substantially. At the same time, the second sub-
period (Figure 4.8) has got similar situation compared with the total sample period, in
which portIolio 1 shows the third highest return with the lowest beta value. There is no
exact relationship between betas and realized return in the second sub-period (2003.7-
2005.12).
Figure 4.8
Sub Period 2 (2003.6-2005.12)
-0.4
-0.3
-0.2
-0.1
0
Beta
R
e
t
u
r
n
Portfolio
Portfolio -0.207 -0.293 -0.275 -0.236 -0.261 -0.203 -0.243 -0.134
-0.4 0.34 0.53 0.69 0.87 1.06 1.27 2.12
According to Brigham (1985), the standard deviation is oIten used by investors to
measure the risk oI a security or a portIolio. In Iinance, the basic concept is that the
standard deviation is a measure oI volatility; a volatile stock would have a higher
standard deviation when compared to a stable stock. Similarly, beta is commonly used to
measure the risk oI a security or a portIolio as well as standard deviation. Nevertheless,
beta is regarded as the tendency oI a security`s return in response to swings in the market.
It assesses the volatility, or systematic risk oI a stock or a portIolio in comparison to the
market as a whole.
69
In general, in respect to stocks in the Shanghai Stock Exchange, shares with higher risk
examined by standard deviation has got higher realized return than lower risk, and there
is also a positive but not explicit trend between systematic risk (beta) and return.
However, it is somewhat diIIicult to conclude that the clear positive beta-return
relationship Ior stocks exist in the SSE.
Moreover, as to portIolios, higher standard deviation, that is, higher risk has got relative
higher realized return generally, while similar situation applies to the relationship
between beta and return. But there are some extreme cases Ior all three sample periods,
Ior instance, portIolio 1 in the total sample period and sub-period 2, portIolio 8 in the Iirst
sub-period. Nevertheless, despite these extreme cases, there is a positive trend in most
time oI total sample period and the Iirst sub-period, in which higher betas are associated
with higher realized return.
4.2.2 Regression Analysis
Through the basic data collection and general Iindings, I run two regression models on
the basis oI Equation (4) and (5) to analyze the unconditional and conditional relationship
between beta and realized return respectively. The average monthly realized returns are
regressed on the betas, while the number oI observations in the cross-sectional
regressions is equal to the number oI stocks and portIolios respectively (80 stocks or 8
portIolios). The coeIIicients in the regressions are averaged, and hypothesis tests are on
the basis oI these averages. All regression results are illustrated in Appendix 2.
70
4.2.2.1 The Unconditional Relationship
The unconditional relationship between beta and return in Equation (4) is initially
examined over the whole sample period and then the two sub-periods. In respect to a
traditional test, it ignores the conditional nature oI the diIIerences between positive and
negative market risk premium. Pettengill et al. (1995) in his paper divided the whole
sample period (1936-1990) into three approximately equal sub-periods and tested
whether or not each oI three sub-periods have consistent results. Similar studies have
been Iound Irom both Jonathan Fletcher (1997) and Elsas et al. (2003) in which equal
sub-periods were used to test the beta-return relationship empirically. In this study, the
sub-periods are between Jan 2001 to Jun 2003, and Jul 2003 to Dec 2005 which include
30 months respectively. The coeIIicients estimated in the monthly cross-sectional
regressions (Equation 4) are averaged. Following the standard procedure, a t-test is then
used to determine whether the mean oI the coeIIicients is signiIicantly diIIerent Irom zero.
The results oI all 80 sample shares, both Ior the total sample period, and two sub-periods
with equal length, are shown in Table 4.2.
1able 4.2 Unconditional 1est Results--Stocks
Period
0
1
t-statistic P-vaIue
Total Period
(2001-2005)
-0.22482 0.040848 2.469934 0.015697
Sub-Period 1
(2001-2003.6)
-0.1908 0.051467 2.87352 0.005228
Sub-Period 2
(2003.7-2005)
-0.2742 0.05283 2.944585 0.00426
Note: A monthly cross-sectional regression is estimated by the mean month oI the index returns oI 80
shares and the estimated betas oI each share. The beta oI each security is estimated over the whole sample
period relative to the SSE index. The table includes the mean oI the monthly coeIIicients oI the intercept
(
0
) and slope (
1
). The t statistics are similar to the Fama and MacBeth (1973) t statistics and P-value is
Ior a t-test oI the null hypothesis in which the mean value oI the coeIIicient is 0. *SigniIicant at 1
71
The results in Table 4.2 demonstrates that there is a insigniIicant (t-test 2.4699)
relationship between beta and realized return in the Shanghai stock market Irom Jan 2001
to Dec 2005 at 1 level oI signiIicance, where the P-value is 0.015697 that is more than
1. The adjusted R square is 0.0725 (Appendix 2), which means only about 7.25 oI the
observed variability in total return can be explained by the independent variable beta. The
remaining 92.75 oI security`s total return variation is the non-systematic component.
On the contrary, diIIerent results showed over the two sub-periods as there is a relatively
signiIicant relationship between beta and realized return, where t-test is 2.87 and 2.94, P-
value is 0.005 and 0.004, and R square is 0.096 and 0.10 respectively. In addition, it is
obvious that the relationship between beta and realized return over the whole sample
period and two sub sample periods is positive (
1
).
The above Iinding is inconsistent with Fama and MacBeth (1973) in which a signiIicant
positive relationship between beta and monthly return were Iound in the US market. It is
possible to conclude that there is a positive but insigniIicant relationship between beta
and realized return over the whole sample period. However, a statistically positive
signiIicant beta-return relationship does observed in the two sub-periods. Consequently,
the null hypothesis ( 0
1
= ) oI strong signiIicant relation between beta and realized
return cannot be rejected Ior the total sample period, but can be rejected Ior the two sub-
periods.
Apart Irom the objects observed, in this case, the total 80 securities, the portIolio sorted
by beta oI each individual security is assessed as well. AIter dividing 80 shares into 8
72
portIolios with 10 shares in each portIolio in accordance with the ascending trend oI beta,
the Iirst portIolio hence has the lowest beta, and in turn, the last portIolio has the highest
beta. In the same way as did Ior total 80 stocks, the research period include the total
sample period and two sub sample periods. Table 4.3 illustrates the results oI the
regressions Ior portIolios.
1able 4.3 Unconditional 1est Results--Portfolios
Period P
0
P
1
t-statistic P-vaIue
Total Period
(2001-2005)
-0.22598 0.042177 1.775994 0.126074
Sub-Period 1
(2001-2003.6)
-0.19426 0.057545 2.026042 0.089151
Sub-Period 2
(2003.7-2005)
-0.26098 0.036466 1.563429 0.168983
Note: A monthly cross-sectional regression is estimated by the mean monthly returns oI portIolios and the
estimated betas oI each portIolio over 3 periods. The beta oI each portIolio is estimated over the whole
sample period and the two sub-periods relative to the SSE index. The table includes the mean oI the
monthly coeIIicients oI the intercept (P
0
) and slope (P
1
). The t statistics are similar to the Fama and
MacBeth (1973) t statistics and P-value is Ior a t-test oI the null hypothesis in which the mean value oI the
coeIIicient is 0. *SigniIicant at 1
Likewise, results in Table 4.3 shows that there is a weak insigniIicant relationship
between portIolio beta and realized return in the SSE in the period during Jan 2001 to
Dec 2005 at 1 level oI signiIicance as the t-test result is 1.7760 and corresponding P-
value is 1.26. The R square Irom Appendix 2 reaches 0.3446, which means 34.46 oI
observed variability in total rate oI return can be explained by portIolio beta. Unlike the
test Ior stocks that demonstrates a signiIicant beta-return relationship, diIIerent results
come Irom the two sub-periods. There is only a weak insigniIicant relationship between
portIolio beta and its realized return in both the Iirst and second sub-period. The t-tests
are 2.026 and 1.563, P-values are 0.089 and 0.169, and R squares are 40.62 and 28.95
73
correspondingly. Nevertheless, the slopes (P
1
) Ior the total sample period and two sub-
periods are all positive that indicates there is a positive relationship between portIolio
beta and realized return. As a result, only a positive but insigniIicant relationship between
portIolio beta and realized return Ior all the three sample periods has been identiIied.
Hence, the null hypothesis oI signiIicant relation between portIolio beta and realized
return cannot be rejected Ior the total sample period, either the Iirst sub-period or the
second sub-period.
In a word, there is a positive signiIicant relationship between stock beta and realized
return in the SSE over the two sub sample periods, whereas only a weak insigniIicant
beta-return relation exists in the total sample period. This is inconsistent with Fama and
MacBeth (1973), Richard and Dowen (1998), and Clare et al. (1998), where the empirical
examinations showed a positive signiIicant beta-return relationship in the US and UK
stock markets respectively. Regarding to portIolios, there is only a positive but weak
insigniIicant relationship between portIolio beta and realized return Ior all three sample
periods.
Although all sample shares and portIolios show a positive relationship between beta and
return (
1
~0, i.e. the risk premium is a little more than zero), the test oI the unconditional
relationship between beta and return cannot be supported because oI the identiIied
insigniIicant beta-return relationship. In addition, this result is consistent with most oI the
empirical tests implemented in other emerging markets. For example, the test oI Taiwan
stock market by Lee (1988), the test oI Korean market by Bark (1991), and the rigorous
74
test oI the CAPM Ior Hong Kong, Korea, Malaysia and the Philippines by Drew and
Veeraraghavan (2003). Evidences have been shown that beta alone is not quite adequate
Ior describing the cross-section oI stock returns, which is similar to my empirical Iindings.
4.2.2.2 The Conditional Relationship
The test Ior unconditional relationship oI the CAPM conducted in the Iormer sub-section
does not take into consideration oI the conditional nature that the relation between beta
and return is actually two sidesunder one situation where market risk premium is
positive and the other situation where there is negative market risk premium. ThereIore,
the beta-return relationship will be estimated by Equation (5) rather than Equation (4) in
the Iollowing part.
Market risk premium is deIined by the market return
mt
R and the risk-Iree rate oI
return
ft
R , that is, market risk premium equals to ) (
ft mt
R R . The market risk premium is
positive when ) (
ft mt
R R is more than zero, whereas negative market risk premium occurs
when ) (
ft mt
R R is less than zero. Through calculations by using Excel, there are 13
months in which the market risk premium is positive, while negative market risk
premium involves 47 degressive months in the total sample period. Similar to the
unconditional test, 8 portIolios are sorted by ascending beta oI 80 securities as
observations. However, portIolios in this conditional test contain diIIerent shares Irom
the portIolios in the unconditional test.
The result oI conditional relationship between portIolio beta and realized return are
75
shown in Table 4.4.
1able 4.4 Conditional 1est Results--Portfolios
Positive Market Risk Premium Negative Market Risk Premium
Period
1
t-statistic P-vaIue
2
t-statistic P-vaIue
Total Period
(2001-2005)
0.033512 2.576590 0.041961 -0.009644 -0.371917 0.722736
Sub-Period 1
(2001-2003.6)
0.146530 5.599902 0.001381 -0.016570 -0.663065 0.531921
Sub-Period 2
(2003.7-2005
0.068274 3.084228 0.021546 -0.017875 -0.662980 0.531972
Note: A conditional cross-sectional regression is estimated by the mean monthly returns oI portIolios and
the estimated betas oI each portIolio over 3 periods. The beta oI each portIolio is estimated over the whole
sample period and the two sub-periods in relative to the SSE index. The table includes the mean oI the
monthly risk premiums in up market months (positive excess market returns)
1
(the slope oI beta) and in
down market months (negative excess market returns)
2
(also the slope oI beta). The t statistics, which
are similar to the Fama and MacBeth (1973) t statistics, and P-value test whether the mean values oI
1
and
2
are signiIicantly positive and negative respectively. *SigniIicant at 1
Table 4.4 shows that the relation between beta and realized return is insigniIicant in
either up market or down market Ior the whole sample period at the signiIicance level oI
1, though up market has got relatively more signiIicance oI beta-return relationship
compared to down market. The t statistic is 2.577 in up market and the corresponding P-
value is 4.196, which is clearly higher than the signiIicant value. In the positive market
oI total sample period, although both t-stats and P-value indicates poor signiIicance oI
beta-return relation, the R square Irom Appendix 3 still reaches 0.5253, which implies
about 52.53 oI observed variation in mean rate oI return can be explained by the
portIolio beta. On the other hand, t-value and P-value Ior the down market is -0.3719 and
72.27 respectively. Particularly, the P-value is much higher than the 1 signiIicance
level. R square Ior the down market is only 0.0225 (Appendix 3) in the whole sample
period, that is, only 2.25 oI cross-sectional variation in average return can be explained
by the portIolio beta. Thus, it is obvious that the conditional relationship between realized
76
return and portIolio beta is not signiIicant in up market as well as in down market Ior the
total sample period. Furthermore, there is a positive (
1
= 0.0335) but insigniIicant
relationship between realized return and portIolio beta in positive market risk premium,
whereas a negative (
2
= -0.0096) insigniIicant relationship exists between realized return
and portIolio beta under the condition that market risk premium is negative. For the total
sample period, portIolios with higher beta tend to have higher rate oI return when market
premium is positive, while there is a negative trend between beta and return when the
market risk premium is negative, though this negative trend is relatively Ilat.
Hence, despite the positive and negative beta-return relation observed in up market and
down market respectively, the relationship between realized return and portIolio beta is
not signiIicant in both oI the two market conditions. The results oI the conditional test in
the Iull sample period can only be weakly supported Ior the conclusion that betas are
related to realized returns in the way predicted by the CAPM theory. Figure 4.9 below
illustrates the beta-return relation in up and down market correspondingly.
Figure 4.9
TotaI SampIe Period-Up Market & Down Market
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
Beta
R
e
t
u
r
n
Portfolio-Up Market -0.008 -0.103 -0.05 -0.059 0.0324 -0.006 -0.011 0.1309
Portfolio-Down Market -0.207 -0.204 -0.227 -0.268 -0.273 -0.251 -0.283 -0.197
0.07 0.48 0.74 0.84 0.91 1.06 1.27 1.86
77
Figure 4.9 conIirms the results. It shows the average realized return Ior the 8 beta-sorted
portIolios (with portIolio 1 that contains 10 lowest beta shares, thus portIolio 8 includes
10 largest beta shares) separately Ior total 60 months with positive and negative market
risk premium. Higher beta portIolios generally have got higher rate oI return than lower
beta portIolios under the positive market excess return condition. While under the
condition that market risk premium is negative, low beta portIolios tend to have higher
rate oI return compared to high beta portIolios. Nevertheless, such negative beta-return
relation under the negative market premium condition is relatively Ilat in comparison
with the relationship between portIolio beta and realized return in the up market.
In order to Iurther analyze the conditional relation, regressions are conducted among the
8 beta-sorted portIolios Ior the two sub sample periods. The results are also shown in
Table 4.4.
In the Iirst sub-period, there is a signiIicant relationship between portIolio beta and
realized return in up market whereas only an insigniIicant beta-return relation exists in
down market at the 1 signiIicance level. The Iormer has a t-test oI 5.6000 and 0.0014 oI
the corresponding P-value. R square reaches 0.8394 (Appendix 3), which indicates about
83.94 oI the observed variability in total return can be explained by the independent
variable beta, only 16.06 oI security`s total return variation is the non-systematic
component. In contrast, the latter has only got -0.6631 Ior the t-test with corresponding P-
value 53.19, the R square is only about 0.0683 (Appendix 3). Moreover, a positive
relationship between portIolio beta and realized return is observed in up market, while
78
there is a negative beta-return relation under the condition oI negative market risk
premium (
1
= 0.1465 and
2
= -0.0166 respectively). It is certain the test result oI up market
Ior the Iirst sub-period provide the strongest result so Iar.
On the contrary, there is an insigniIicant relationship between realized return and
portIolio beta either when the market risk premium is positive or the market risk premium
is negative (t-test 3.0842 and -0.6630, P-value 0.0215 and 0.5320, R square 0.6132
and 0.0683 respectively) Ior the second sub-period. However, similar to Iindings Ior the
total sample period and the Iirst sample period, up market has got a positive relationship
between realized return and portIolio beta, whereas a negative beta-return relation exists
in the down market.
To sum up, it is obvious that there is positive relationship between return and beta in up
market over the whole sample period and two sub-periods, while there is negative but
relatively Ilat beta-return relation in down market over the three diIIerent sample periods.
Nevertheless, the positive relation in up market is not signiIicant despite Ior the Iirst sub
sample period, and the negative relation in down market is extremely weak and
insigniIicant across all sample periods.
Consequently, in general, Ior the hypothesis oI H1 examined by Pettengill et al. (1995),
the null hypothesis oI no relationship between realized return and beta during the periods
oI positive excess market returns can be rejected in Iavor oI an expected positive
relationship at 1 level. Hence, the alternative hypothesis is true that there is a positive
79
relation in up market, though the positive relationship is relatively insigniIicant. On the
other hand, in the hypothesis oI H2, although there does exists a negative beta-return
relation, the negative relationship in down market is rejected as the null hypothesis
between realized return and portIolio beta cannot be reject across all sample periods
when the excess market return is negative, since the beta-return relation is extremely
weak and insigniIicant. As a result, the conditional relationship between beta and return
can not be supported.
4.2.3 Security Market Line (SML) Analysis
From Equation (1), it is clear that the CAPM states there should be a linear relationship
between a security`s or a portIolio`s return and beta. In theory, the relationship between
risk as measured by beta and the expected return is graphed by the security market line
(SML). According to the CAPM, all shares and portIolios lie on the security market line
with their positions determined by their beta exactly. However, as Arnold (2002)
indicated, in Iact, iI ten portIolios with diIIerent levels oI beta are created, the
relationship between beta and return is not exactly as described by the SML Ior these ten
portIolios. Although the plot points may not be precisely on the SML, it would be
reasonable to conclude that higher beta portIolios tends to earn higher rate oI return than
lower beta portIolios. Moreover, one extra point to note here is that iI a regression line
was Iitted to the observed data, its shape would be Ilatter than the SML passing through
the market portIolio plot points. Such results were approved by Brigham (1985) in the
early years.
80
In this part, the theoretical and empirical SML is estimated and compared under the
conditional CAPM Ior the total sample period Jan 2001 to Dec 2005. The conditional
linear regression tested by empirical evidence in the whole sample period was already
calculated in the sub-section 4.2.2 regression analysis. Here, the results oI empirical SML
can be represented by the Iollowing equations:
Up Market:
it i i
e R + + = | 03351 . 0 04581 . 0 (6)
Down Market:
it i i
e R + = | 00964 . 0 23015 . 0 (7)
The above equations show that there is a positive relationship between portIolio beta and
realized return when the market risk premium is positive, whereas a negative beta-return
relation exists under the condition that market risk premium is negative.
Figure 4.1 Figure 4.11
Up Market
- 0.2
- 0.1
0
0.1
0.2
0.3
0.4
0.5
0.6
-2 -1 0 1 2 3 4
Bet a
Portfolio
Down Market
-0.3
- 0.25
-0.2
-0.15
-0.1
- 0.05
0
0 0.5 1 1.5 2
Bet a
Portfolio
Figure 4.10 and Figure 4.11 below illustrates the relationship between realized return and
portIolio beta in up market and down market respectively Ior he total sample period. It
81
could be seen that all portIolios almost lie on a straight line empirically in both the up
market and the down market. This Iinding is consistent with the results revealed by
Richard Roll (1978), Brigham (1985), and Arnold (2002) etc. Furthermore, it is obvious
that portIolio with higher beta has got higher rate oI return compared to portIolio with
lower beta in the up market. Conversely, when under the condition that market risk
premium is negative, that is, down market, the portIolio with higher beta tends to have
lower rate oI return than lower beta portIolio.
However, in order to draw the theoretical SML, the original CAPM model (Equation 1)
has to be used to calculate the expected return. The saving deposit interest rate declared
by China Central Bank stands Ior the proxy oI the risk-Iree rate oI return
f
R . The
Shanghai Stock Exchange Index is regarded as the market rate oI return
m
R . Then the
linear regression estimated by the theoretical data can be calculated by using the Excel.
Consequently, there are two linear regressions to explain the relationship between
expected return and portIolio beta under the condition oI positive market risk premium
and negative market risk premium respectively. (Appendix 4)
Up Market:
it i i
e R + + = | 127719 . 0 000704 . 0 (8)
Down Market:
it i i
e R + = | 145598 . 0 000634 . 0 (9)
Similarly, Equation (8) and (9) demonstrates that there is a positive relationship between
expected return and portIolio beta when the market risk premium is positive. In contrast,
82
negative relationship between expected return and portIolio beta is observed when the
market risk premium is negative.
Figure 4.12
Up Market
-0.1
0
0.1
0.2
0.3
0 0.5 1 1.5 2 2.5
Beta
R
e
t
u
r
n
Theoretical SML Empirical SML
Figure 4.13
Down Market
-0.4
-0.3
-0.2
-0.1
0
0.1
0 0.5 1 1.5 2 2.5
Beta
R
e
t
u
r
n
Theoretical SML Empirical SML
Figure 4.12 and Figure 4.13 shows the contrast between the theoretical SML and
empirical SML in up market and down market respectively. It is clear that the slope oI
the empirical SML in either up market or down market is less than the slope oI the
theoretical SML, which implies a consistency with Brigham (1985) and Arnold (2002)`s
results as the slope oI the relationship between realized returns and systematic risk
deIined by beta is usually less than that oI the relationship between expected return and
83
beta predicted by the CAPM.
Moreover, it is clear that both the slopes oI empirical SML in up market and down
market is Ilatter than theoretical SML distinctly; especially under the condition that
market risk premium is negative, which indicates that investors in Chinese stock market
have got little conIidence. In respect to up market, both low and high beta portIolios tend
to have smaller rate oI return than theory suggests, whereas diIIerence is shown Ior the
down market where high beta portIolios has got higher rate oI return than the CAPM
predicts. The intercept value Ior SML in either up market or down market is negative,
thus realized rate oI return with beta equals to zero is lower than the risk-Iree rate oI
return; investors not only have not got reasonable compensations Ior taking risks in the
stock market but also to obtain a rate oI return that is lower than the risk-Iree deposit
interest rate declared by the central bank. Such abnormal phenomenon Iurther indicates
the lack oI conIidence Irom investors regarding to Chinese stock market, since the
compensations are too low Ior investors to eIIiciently make proIits or avoid losses in up
market and down market correspondingly. Investors would like to put money into the
bank rather than invest in the stock market.
Furthermore, the estimated SML during the up market (slope 0.0335) by realized return
and portIolio beta is steeper than that oI the estimated SML when market risk premium is
negative (Slope -0.0097). The volatility oI the portIolio beta in up market inIluences the
change oI the realized return more intensively than that in the down market. ThereIore,
the estimated risk premium Ior the up market and down market is asymmetric with the up
84
market risk premium greater than the down market.
Thus, according to above analysis, the empirical SML is not equal with the theoretical
SML. The conditional CAPM cannot be supported.
Conclusion
The empirical test results oI the CAPM by using Fama and MacBeth`s (1973) approach
with modiIications suggested by Pettengill et al. (1995) have been examined and
discussed in this chapter. Consistent with most Iindings in the emerging markets, the
beta-return relationship is Ilat in Chinese stock market, beta alone is not suIIicient to
describe the cross-section rate oI returns Ior stocks. Hence, neither unconditional nor
conditional CAPM can be supported.
Overall, although obviously there are limitations oI this empirical study such as the
deviation between the estimated risk-Iree rate and the real risk-Iree rate which may partly
lead to the unIavorable result, however, the inadequacy oI the CAPM may mainly be
attributed to market ineIIiciency as the Chinese stock market is certainly immature. For
instance, it has been long concerned that A-class shares in Chinese stock market have
little relation with Iundamentals Iactors, capital gains are more likely to be achieved by
non-Iundamental reasons such as Iads and market manipulation. Moreover, since the
typical characteristics oI Chinese stock market that the state owned capital and companies
occupy a large proportion, so that the central government can intervene in the stock
market directly. These excessive interventions Irom the government also results in the
85
many abnormalities oI the stock market.
86
CHAPTER 5 SUMMARY AND CONCLUSION
To summarise, the main objective oI this paper is to test the validity oI the CAPM in
Chinese stock market by examining the unconditional and conditional relationship
between realized return and beta in Shanghai Stock Exchange during the period Irom Jan
2001 to Dec 2005, and compares the empirical SML with theoretical SML using the
conditional CAPM approach.
BeIore conducting the empirical test, I Iirst give a review in chapter 2 regarding to the
CAPM theoretical model and previous empirical Iindings Irom both developed markets
and emerging markets with the aim to provide some general ideas about the applicability
oI the CAPM in practice. Despite early evidences such as Fama and MacBeth (1973) that
seem to in support oI the CAPM model, evidences thereaIter give largely unIavourable
results as variables other than beta (e.g. size, E/P ratio etc.) are Iound be to signiIicant in
capturing the variance oI average returns.
In chapter 3, in order to provide background knowledge that would be helpIul to
understand the research base, market institutions oI the Chinese stock markets especially
the Shanghai stock exchange are described and discussed in detail. In general, other than
the common Ilaws such as high market volatility and less market oriented as a whole in
every emerging market, some other typical characteristics are identiIied regarding to the
Chinese stock markets. For instance, irrational investors instead oI institutional investors
are the dominant party. In particular, there are very limited tradable shares as most shares
that belong to the state owned enterprises are protected and restricted Irom trading by the
87
government, and share prices are more likely to be driven by some qualitative Iactors
such as policies rather than market Iundamentals. Hence, it seems that the CAPM
assumptions are largely violated in the Chinese stock market, and there is strong tendency
to believe that the CAPM will be invalid in the Chinese stock market.
Chapter 4 then empirically tested the CAPM by applying the approach Irom Fama and
MacBeth (1973), with modiIications suggested by Pettengill et al. In consideration oI
data availability and time constraint, the CAPM is limited in the Shanghai Stock
Exchange (SEE) Ior the period Jan 2001 Dec 2005 since the SSE is the largest and
representative stock exchange in the mainland China when compared to Shenzhen Stock
Exchange (SZSE).
Inconsistent with Fama and MacBeth (1973), there is a weak insigniIicant relationship
between stock beta and return over the whole sample period, whereas positive signiIicant
beta-return relation is Iound to exist only Ior the two sub-periods. Regarding to portIolios,
a positive but weak insigniIicant relationship between portIolio beta and realized return
Ior all three sample periods have been identiIied. Thus, the unconditional test Iailed to
support the validity oI the unconditional CAPM model in the SSE.
Moreover, Iollowing the conditional approach proposed by Pettengill et al. (1995), results
indicate that there is a positive relationship between portIolio beta and realized return in
up market over the whole sample period as well as two sub-periods, but only signiIicant
Ior the Iirst sub-period. However, when the market risk premium is negative, there only
88
exists an extremely weak negative and insigniIicant beta-return relation. Consequently,
inconsistent with the idea that there should be a signiIicant positive relationship between
beta and return in up market and a signiIicant negative relation in down market suggested
by Pettengill et al. (1995), the applicability oI the conditional CAPM is doubt in the SSE.
Further, the theoretical SML estimated by ex-ante expected returns compare to the
empirical SML estimated by ex-post realized returns are analyzed. Although results show
a consistency with Brigham (1985) and Arnold (2002) as the slope oI the relationship
between realized returns and systematic risk deIined by beta is less than that oI the
relationship between expected return and beta predicted by the CAPM, the intercepts Ior
the SML in both up and down market are less than zero, so that the realized return with
beta equal to zero is lower than the risk-Iree rate oI return. In addition, the estimated risk
premium Ior the up market and down market is asymmetric with the up market risk
premium statistically larger than that oI the down market, that is, the estimated SML is
steeper positively in the up market and Ilatter negatively in the down market.
As a result, I draw a conclusion that the CAPM is invalid in the Chinese stock market
since: (1) Overall, no signiIicant relationship between beta and return can be observed. In
the Fama-MacBeth regression, beta-return relation is positive but insigniIicant; Ior the
conditional CAPM approach, the relationship between beta and return is still
inconspicuous, though positive beta-return relationship are observed in up market
whereas negative relationship in down market. (2) Beta alone is not quite adequate Ior
describing the cross-section oI stock returns, as the explanatory ability oI R squares are
89
low in most oI the tests.
Nevertheless, there are major limitations that may inIluence the result oI this study.
Firstly, the sample may be too small, testing period may not be suIIicient compared to
studies that have been reviewed in Chapter 2. Secondly, the CAPM assumes a semi-
strong eIIicient market whereas the Chinese stock market is an obvious immature market,
Irom using such an immature market to implement the CAPM test empirically, great
deviations may be resulted. In addition, despite these limitations, Iinancial models may
not be applicable to every market since diIIerent markets have got diIIerent
characteristics, investor behaviour and political environment.
In conclusion, neither unconditional nor conditional CAPM can be supported or applied
eIIiciently in Chinese stock market. On the one hand, it is important Ior decision-makers
to bear in mind when making investment decisions that rather than solely rely on the
results oI the Iinancial models, diIIerent markets obviously have diIIerent characteristics
such as the Chinese stock market. On the other hand, the enormous economic
development in China over the past twenty years certainly produced the abnormal
Iinancial market subsequently. The Chinese government and the CSRC should emphasis
on improving the quality oI the listed Iirms, reIorm the market structure, and enhance the
ability to supervision, so that the Chinese stock market, especially the SSE could run to
the world level.
90
BIBLIOGRAPHY
Arnold, G., (2002). Corporate Financial Management. 2
nd
edition. Prentice Hall.
Banz, R. W., (1981). The Relationship between Return and Market Value oI Common
Stocks. Journal oI Financial Economics, pp 3-18.
Bark, H. K., (1991). Risk, Return, and Equilibrium in the Emerging Markets: Evidence
Irom the Korean Stock Market. Journal of Economics and Business, Vol. 43, pp 353-362.
Basu, S., (1977). Investment PerIormance oI Common Stocks in Relation to Their Price-
Earnings Ratios: A Test oI the EIIicient Market Hypothesis. The Journal of Finance, Vol.
32.
Basu, S., (1983). The Relationship between Earnings` Yield, Market Value and Return
Ior NYSE Common Stocks: Further Evidence. Journal of Financial Economics, pp 129-
156.
Bentzen, E., and Hansson, B., (2005). Systematic variations in January. Copenhagen
Business School, and Lund University
Black, F., (1972). Capital Market Equilibrium with Restricted Borrowing, Journal of
Business, Vol.45, pp.444-445
Black, F., Jensen, M., and Scholes, M., (1972). The Capital Asset Pricing Model: Some
Empirical Results. Studies in the Theorv of Capital Markets, Praeger Publishers., 1972
Blume, M. E., and Friend, I., (1973). A New Look at the Capital Asset Pricing Model.
Journal of Finance, Vol. 28, No. 1, pp1933.
91
Brealey, R.A., Myers, S.C., and Allen, F., (2005). Corporate Finance. 8
th
edition.
McGraw Hill.
Chan, K., Menkveld, A. J., and Yang, Z., (2002). Evidence on the Foreign Share
Discount Puzzle in China: Liquidity or InIormation Asymmetry? Hong Kong University
oI Science and Technology, Vrije University Amsterdam, and Tsinghua University
Chan, Y. C., (1997). Multivariate Testing oI the Capital Asset Pricing Model in the Hong
Kong Stock Market. Applied Financial Economics, Vol. 7, pp 311-316.
Cheung, Y. L., and Wong, K. T., (1992). An Assessment oI Risk and Return: Some
Empirical Findings Irom the Hong Kong Stock Exchange. Applied Financial Economics,
Vol. 2, pp 105-114.
Cheung, Y. L., Wong, K. A., and Ho, Y. K., (1993). The Pricing oI Risky Assets in Two
Emerging Asian Markets-Korea and Taiwan. Applied Financial Economics, Vol. 3, pp
315-324.
Clare, A. D., Priestley, R., and Thomas, S. H., (1998). Reports oI Beta`s Death are
Premature: Evidence Irom the UK. Journal of Banking and Finance, pp1207-1229.
DeFusco, R.A., McLeavey, D.W., Pinto, J.E., and Runkle, D.E., (2004). Quantitative
Methods Ior Investment Analysis. 2
nd
edition. United Book Press
Dowen, R. J., (1988). Beta, Non-Systematic Risk and PortIolio Selection. Applied
Economics, Vol. 20, pp 221-228.
Drew, M. D., and Veeraraghavan, M., (2003). Beta, Firm Size, Book-to-Market Equity
and Stock Returns: Further Evidence Irom Emerging Markets. Journal of the Asia Pacific
Economv, Vol. 8, pp 354379.
92
Elsas, R., El-Shaer, M., and Theissen, E., (2003). Beta and Returns Revisited Evidence
Irom the German Stock Market. Journal of International Financial Markets, Institutions
and Monev, Vol.13, pp 1-18.
Fama, E. F., and French, K. R., (1992). The Cross-Section oI Expected Stock Returns.
Journal of Finance, Vol. 47, pp 427-465.
Fama, E. F., and French, K. R., (1996). MultiIactor Explanations oI Asset Pricing
Anomalies. Journal of Finance, Vol. 51, pp 5584.
Fama, E. F., and French, K. R., (1996). The CAPM is Wanted, Dead or Alive. The
Journal of Finance, Vol. 51, pp 1947-1958.
Fama, E. F., and French, K. R., (2004). The Capital Asser Pricing Model: Theory and
Evidence. Journal of Economic Perspectives, Vol. 18, pp 25-46.
Fama, E. F., and MacBeth, J. D., (1973). Risk, Return and Equilibrium: Empirical Tests.
Journal of Political Economv, Vol. 81, pp 607-636.
Fletcher, J., (1997). An Examination oI the Cross-Sectional Relationship oI Beta and
Return: UK Evidence. Journal of Economics and Business, Vol. 49, pp 211-221.
Fletcher, J., (2000). On the Conditional Relationship between Beta and Return in
International Stock Returns. International Review of Financial Analvsis, Vol. 9, pp 235-
245.
Gultekin, M. N., and Gultekin, N. B., (1983). Stock Market Seasonality: International
Evidence. Journal of Financial Economics, pp 469-481.
Harvey, C. R., (2000). Asset Pricing in Emerging Markets. National Bureau of Economic
Research, Cambridge, MA. Working Paper, Duke University, Durham.
93
Heston, S. L., Rouwenhorst, K. G., and Wessels, R. E., (1999). The Role oI Beta and Size
in the Cross-Section oI European Stock Returns. European Financial Management, Vol.
5, pp 9-27.
History Trend oI the SSE Index, 2001-2006. |Online|. Available Irom: http://c
n.Iinance.yahoo.com/q/bc?s000001.SS&tmy&lon&zm&ql&c ||Accessed 28
August 2006|
Huang, Y. S., (1997). An Empirical Test oI the Risk-Return Relationship on the Taiwan
Stock Exchange. Applied Financial Economics, Vol. 7, pp 229-239.
Jagannathan, R., and Wang, Z., (1996). The Conditional CAPM and the Cross-Section oI
Expected Returns. Journal of Finance, Vol. 51, pp 3-53.
Javed, A.Y., (2000). Alternative Capital Asset Pricing Models: A Review oI Theory and
Evidence. Pakistan Institute oI Development Economics.
Kothari, S. P., Shanken, J., and Sloan, R. G., (1995). Another Look at the Cross-Section
oI Expected Stock Return. The Journal of Finance, Vol. 50, No. 1, pp 185-224.
Lam, K. S. K., (2001). The Conditional Relation between Beta and Returns in the Hong
Kong Stock Market. Applied Financial Economics, Vol. 11, pp 669-680.
Linter, J., (1965). The Valuation oI Risk Assets and the Selection oI Risky Investments in
Stock PortIolio and Capital Budgets. Review of Economics and Statistics, Vol.47, pp.13-
37
Lo, A. W., and MacKinlay, A. C., (1990). Data-Snooping Biases in Tests oI Financial
Asset Pricing Models. The Review of Financial Studies, Vol.3, Number 3, 1990
94
MacKinlay, A. C., (1994). MultiIactor Models Do Not Explain Deviations From the
CAPM. National Bureau of Economic Research, Working Paper No. 4756.
Michailidis, G., Tsopoglou, S., Papanastasiou, D., and Mariola, E., (2006). Testing the
Capital Asset Pricing Model (CAPM): The Case oI the Emerging Greek Securities
Markets. International Research Journal of Finance and Economics.
Miller, E. M.,(2001). Why the Low Return to Beta and Other Forms oI Risk. Journal of
Portfolio Management, Vol. 27, pg 40.
Mossin, J., (1966). Equilibrium in a Capital Asset Market. Econometrica, Vol. 34, pp
768-783.
Nimal, P. D., (2006). An Empirical Analvsis of Capital Asset Pricing Model. Thesis
(PhD). Shiga University.
Pettengill, G.. N., Sundaram, S. S., and Mathur, I., (1995). The Conditional Relation
between Beta and Returns. Journal of Financial and Quantitative Analvsis, Vol. 30, pp
101-116.
Reinganum, M. R., (1981). A New Empirical Perspective on the CAPM. The Journal of
Financial and Quantitative Analvsis, Vol. 16, No. 4, pp. 439-462.
Roll, R., (1978). Ambiguity when PerIormance is Measured by the Securities Market
Line. The Journal of Finance, Vol. 33, No. 4, pp. 1051-1069.
RozeII, M. S., and Kinney, W. R., (1976). Capital Market Seasonality: The Case oI Stock
Returns. Journal of Financial Economics, Vol. 3, pp 379-402.
Sharpe, W. F., (1964). Capital Asset Prices: A Theory oI Market Equilibrium under
Conditions oI Risk. The Journal of Finance, Vol. 19, pp 425-442.
95
Stockstar. |Online|. Available Irom: http://www.cnstockstar.com/ |Accessed 17 August
2006|
Strong, N., and Xu, X. G., (1997). Explaining the Cross-Section oI UK Expected Stock
Returns. British Accounting Review, Vol. 29, pp 1-23.
Tan, L., (2005). Empirical Analvsis of Chinese Stock Market Behavior. Evidence from
Dvnamic Correlations, Herding Behavior, and Speed of Adfustment. Thesis (PhD),
Drexel University.
Tang, G. Y. N., and Shum, W. C., (2003). The Conditional Relationship between Beta
and Returns: Recent Evidence Irom International Stock Markets. International Business
Review, pp 109-126.
The Shanghai Stock Exchange Historical Database |Online|. Available Irom: http://ww
w.sse.com.cn/sseportal/enus/ps/md/shbhd.jsp |Accessed 28 August 2006|
Tinic, S. M., (1984). Risk and Return: January vs. the Rest oI the Year. Journal of
Financial Economics, pp 561-574.
Wong, K. A., and Tan, M. L., (1991). An Assessment oI Risk and Return in the
Singapore Stock Market. Applied Financial Economics, Vol. 1, pp. 11-20.
Xu, R., (2001). The CAPM Test in China Stock Market. with Subsidarv Tests on
Fundamentals and Returns. Thesis, University oI Nottingham
Xu, Y., (2003). Diversification in the Chinese Stock Market. School of Management, The
University oI Texas at Dallas and Shanghai Stock Exchange
96
Xu, X., and Wang, Y., (1997). Ownership Structure, Corporate Governance, and Firms
Performance. The Case of Chinese Stock Companies. Amherst College and The World
Bank
Yahoo Finance, |Online| Available Irom: http://cn.Iinance.yahoo.com/ |Accessed 17
August 2006|
97
APPENDIX 1: Data of 80 Listed Securities and Market Index
Average Return STD Beta ()
Index -0.0857306 0.14218
600001 -0.1225053 0.165338 0.819189
600002 0.1150044 0.43349 2.109887
600051 -0.2001906 0.224524 1.161518
600052 -0.2697894 0.089868 0.234299
600054 -0.064976 0.198222 0.100265
600055 -0.207481 0.217827 0.989723
600056 -0.2200162 0.253816 0.126581
600058 -0.1631152 0.280144 1.531694
600059 -0.1176502 0.201274 0.070316
600060 -0.1453038 0.161262 0.822352
600061 -0.2620495 0.322928 1.543396
600062 -0.045114 0.446575 1.799526
600063 -0.1817788 0.175348 0.890285
600064 -0.243508 0.099372 0.526345
600065 -0.2755321 0.26471 1.523908
600066 -0.1253568 0.166474 0.710594
600067 -0.1807627 0.245824 0.964469
600068 -0.1707816 0.154392 0.897912
600069 -0.1647041 0.189194 0.993553
600070 -0.2054272 0.242684 1.223579
600071 -0.1588251 0.442077 2.510768
600072 -0.1653189 0.211165 0.909145
600073 -0.135665 0.230733 0.045411
600074 -0.1974682 0.285206 1.565186
600075 -0.0820416 0.212133 0.706889
600076 -0.3132966 0.211751 1.19976
600077 -0.2662736 0.138067 0.523948
600078 -0.1722122 0.18936 0.3523
600079 -0.3001498 0.139707 0.399461
600080 -0.2575314 0.123079 0.682131
600081 -0.2485913 0.358632 1.458743
600082 -0.195291 0.283167 1.12971
600085 -0.0238068 0.143714 0.011888
600086 -0.201922 0.491222 2.305648
600087 -0.2479332 0.156649 0.094657
600088 -0.1615714 0.19595 0.785991
600089 -0.1411185 0.26866 0.785808
600091 -0.2258528 0.122062 0.766998
600092 -0.3096973 0.157667 0.698904
600093 -0.2301242 0.312245 1.00268
98
600094 -0.2548895 0.118543 0.391817
600095 -0.2608316 0.112069 0.607419
600096 -0.0253819 0.224996 1.189997
600097 -0.1674138 0.202442 0.512695
600098 -0.1524816 0.214585 1.004989
600099 -0.1625716 0.325127 1.823876
600100 -0.249703 0.192166 0.522613
600101 -0.1838691 0.20322 0.929028
600102 -0.1418042 0.310153 1.623069
600103 -0.2365285 0.149757 0.510227
600104 -0.0649626 0.503828 1.774251
600105 -0.2122563 0.134558 0.681679
600106 -0.1799095 0.125192 0.541485
600107 -0.2971204 0.117701 0.373626
600108 -0.2560447 0.185348 0.922211
600109 -0.1850002 0.218071 0.901783
600110 -0.1446337 0.366293 1.517573
600111 -0.1715894 0.197074 1.069939
600112 -0.1169609 0.216054 1.270366
600115 -0.0802257 0.237285 1.11454
600116 -0.1907271 0.204585 0.904875
600120 -0.3790036 0.174347 0.328225
600121 -0.3099105 0.160178 0.59192
600125 -0.1331032 0.17254 0.646039
600126 -0.1068201 0.227314 1.350948
600129 -0.1041932 0.478356 1.247274
600132 -0.0943236 0.237201 -0.15228
600148 -0.2949247 0.138085 0.220144
600157 -0.2556032 0.135829 0.38765
600166 -0.1876237 0.317804 0.636578
600172 -0.1859018 0.144737 0.323199
600180 -0.258649 0.187901 0.575758
600181 -0.3471129 0.222209 1.030599
600186 -0.2047745 0.2895 0.827763
600189 -0.1311023 0.187429 1.105731
600191 -0.2134098 0.114121 0.646834
600192 -0.2019877 0.133016 0.715907
600196 -0.2296556 0.227888 0.844208
600198 -0.2363312 0.194572 0.552099
600218 -0.2329242 0.192605 0.911712
99
APPENDIX 2: The Unconditional Regression Test Results
(a) Stocks (Total Sample Period)
Regression Statistics
Multiple R 0.269331
R Square 0.072539
Adjusted R Square 0.060648
Standard Error 0.077904
Observations 80
ANOVA
df SS MS F Significance F
Regression 1 0.037024 0.037024 6.100573 0.015697
Residual 78 0.473381 0.006069
Total 79 0.510405
Coefficients Standard Error t Stat P-value
ntercept -0.22482 0.016882 -13.3174 8.8E-22
X Variable 1 0.040848 0.016538 2.469934 0.015697
(b) Stocks (First Sub-Period 2001.1-2003.6)
Regression Statistics
Multiple R 0.309397
R Square 0.095727
Adjusted R Square 0.084134
Standard Error 0.101055
Observations 80
ANOVA
df SS MS F Significance F
Regression 1 0.084323 0.084323 8.257116 0.005228
Residual 78 0.796547 0.010212
Total 79 0.88087
Coefficients Standard Error t Stat P-value
ntercept -0.1908 0.019064 -10.0083 1.22E-15
X Variable 1 0.051467 0.017911 2.87352 0.005228
100
(c) Stocks (Second Sub-Period 2003.7-2005.12)
Regression Statistics
Multiple R 0.316292
R Square 0.100041
Adjusted R Square 0.088503
Standard Error 0.119723
Observations 80
ANOVA
df SS MS F Significance F
Regression 1 0.124281 0.124281 8.670582 0.00426
Residual 78 1.118025 0.014334
Total 79 1.242306
Coefficients Standard Error t Stat P-value
ntercept -0.2742 0.01973 -13.8979 8.36E-23
X Variable 1 0.05283 0.017941 2.944585 0.00426
(i) Portfolios (Total Sample Period)
Regression Statistics
Multiple R 0.586992
R Square 0.34456
Adjusted R Square 0.23532
Standard Error 0.034139
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.003676 0.003676 3.154153 0.126074
Residual 6 0.006993 0.001165
Total 7 0.010669
Coefficients Standard Error t Stat P-value
ntercept -0.22598 0.024019 -9.40843 8.19E-05
X Variable 1 0.042177 0.023748 1.775994 0.126074
101
(ii) Portfolios (First Sub-Period 2001.1-2003.6)
Regression Statistics
Multiple R 0.637358
R Square 0.406225
Adjusted R Square 0.307263
Standard Error 0.04637
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.008826 0.008826 4.104845 0.089151
Residual 6 0.012901 0.00215
Total 7 0.021728
Coefficients Standard Error t Stat P-value
ntercept -0.19426 0.029604 -6.56193 0.0006
X Variable 1 0.057545 0.028403 2.026042 0.089151
(iii) Portfolios (Second Sub-Period 2003.7-2005.12)
Regression Statistics
Multiple R 0.538017
R Square 0.289462
Adjusted R Square 0.171039
Standard Error 0.045517
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.005064 0.005064 2.444309 0.168983
Residual 6 0.012431 0.002072
Total 7 0.017495
Coefficients Standard Error t Stat P-value
ntercept -0.26098 0.02478 -10.532 4.31E-05
X Variable 1 0.036466 0.023324 1.563429 0.168983
102
APPENDIX 3: The Conditional Regression Test Results
Up Market
(a) Portfolios (Total Sample Period)
Regression Statistics
Multiple R 0.724757
R Square 0.525272
Adjusted R Square 0.446151
Standard Error 0.052057
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.017991 0.017991 6.638817 0.041961
Residual 6 0.01626 0.00271
Total 7 0.034251
Coefficients Standard Error t Stat P-value
ntercept -0.04581 0.023264 -1.96905 0.096477
X Variable 1 0.033512 0.013006 2.57659 0.041961
(b) Portfolios (First Sub-Period 2001.1-2003.6)
Regression Statistics
Multiple R 0.916185
R Square 0.839396
Adjusted R Square 0.812628
Standard Error 0.074627
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.174645 0.174645 31.35891 0.001381
Residual 6 0.033415 0.005569
Total 7 0.208061
Coefficients Standard Error t Stat P-value
ntercept -0.07194 0.035635 -2.01891 0.090037
X Variable 1 0.14653 0.026167 5.599902 0.001381
103
(c) Portfolios (Second Sub-Period 2003.7-2005.12)
Regression Statistics
Multiple R 0.78308
R Square 0.613214
Adjusted R Square 0.54875
Standard Error 0.090222
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.077432 0.077432 9.512462 0.021546
Residual 6 0.04884 0.00814
Total 7 0.126272
Coefficients Standard Error t Stat P-value
ntercept -0.13866 0.035251 -3.93362 0.007681
X Variable 1 0.068274 0.022136 3.084228 0.021546
Down Market
(d) Portfolios (Total Sample Period)
Regression Statistics
Multiple R 0.150114
R Square 0.022534
Adjusted R Square -0.14038
Standard Error 0.036603
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.000185 0.000185 0.138322 0.722736
Residual 6 0.008039 0.00134
Total 7 0.008224
Coefficients Standard Error t Stat P-value
ntercept -0.23015 0.026802 -8.58701 0.000137
X Variable 1 -0.00964 0.02593 -0.37192 0.722736
104
(e) Portfolios (First Sub-Period 2001.1-2003.6)
Regression Statistics
Multiple R 0.261291
R Square 0.068273
Adjusted R Square -0.08701
Standard Error 0.039858
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.000698 0.000698 0.439655 0.531921
Residual 6 0.009532 0.001589
Total 7 0.010231
Coefficients Standard Error t Stat P-value
ntercept -0.19285 0.027212 -7.08698 0.000396
X Variable 1 -0.01657 0.024991 -0.66306 0.531921
(f) Portfolios (Second Sub-Period 2003.7-2005.12)
Regression Statistics
Multiple R 0.26126
R Square 0.068257
Adjusted R Square -0.08703
Standard Error 0.049518
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.001078 0.001078 0.439543 0.531972
Residual 6 0.014712 0.002452
Total 7 0.01579
Coefficients Standard Error t Stat P-value
ntercept -0.25319 0.028607 -8.85065 0.000116
X Variable 1 -0.01788 0.026962 -0.66298 0.531972
105
APPENDIX 4: The Theoretical Test Results for the Conditional
Relationship between Expected Return and Beta
Up Market
Regression Statistics
Multiple R 1
R Square 1
Adjusted R Square 1
Standard Error 1.15E-17
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.261322 0.261322 1.99E+33 8.6E-99
Residual 6 7.88E-34 1.31E-34
Total 7 0.261322
Coefficients Standard Error t Stat P-value
ntercept 0.000704 5.12E-18 1.37E+14 1E-83
X Variable 1 0.127719 2.86E-18 4.46E+16 8.6E-99
Down Market
Regression Statistics
Multiple R 1
R Square 1
Adjusted R Square 1
Standard Error 4.43E-17
Observations 8
ANOVA
df SS MS F Significance F
Regression 1 0.042241 0.042241 2.15E+31 6.79E-93
Residual 6 1.18E-32 1.96E-33
Total 7 0.042241
Coefficients Standard Error t Stat P-value
ntercept 0.000634 3.25E-17 1.95E+13 1.22E-78
X Variable 1 -0.1456 3.14E-17 -4.6E+15 6.79E-93