Journal of Economics and Business

G Model
JEB-5773; No. of Pages 16 ARTICLE IN PRESS

Journal of Economics and Business xxx (2017) xxxxxx
Contents lists available at ScienceDirect
Journal of Economics and Business
Research paper
Regionally integrated asset pricing on the african stock

markets: Evidence from the fama french and carhart models
Nicholas Addai Boamah , Edward Watts, Geoffrey Loudon
Department of Applied Finance and Actuarial Studies, Faculty of Business and Economics, Macquarie University, Sydney, Australia
a r t i c l e i n f o a b s t r a c t
Article history: The study explores regionally integrated asset pricing on the African Stock Markets (ASMs)
Received 9 June 2016 via the Fama-French model. It investigates the ability of the model to capture African equity
Received in revised form 11 April 2017 returns at the regional level. It also explores the question of absolute versus relative mea-
Accepted 18 April 2017
sures of size and BM when securities are pooled across markets. Absolute proxies may
Available online xxx
have potential confounding effects or may be appropriate in integrated markets. The study
achieves this by scaling the size and BM of each rm by its cross-sectional means or cross-
Keywords:
sectional standard deviations for the securitys market on the portfolio formation date. The
Size
analysis employs the Fama-French model in a cross-country setting. Evidence is provided
Book-to-Market
African stock markets that both size and BM effects exist on the pooled ASMs. The BM effect, however, appears
Integrated asset pricing stronger than the size effect. Also, the pricing errors are higher when size and BM are mea-
Africa sured relative to their individual country averages rather than to that of the entire pooled
sample. We observe lower mispricing when size and BM are estimated in relation to their
cross-sectional standard deviations for the respective countries. Evidence is provided that
asset pricing on the ASMs is largely not regionally integrated. The evidence shows that the
effect of illiquidity in describing the returns of a pooled sample could be overstated by
uncontrolled disparities between the characteristics of the pooled markets if the markets
exhibit some degree of segmentation.
2017 Elsevier Inc. All rights reserved.
1. Introduction
We investigate whether asset pricing on the African Stock Markets (ASMs) is conducted in a regionally integrated way
or is national in character and its implications for the integration and potential diversication gains across the ASMs. We
explore this through the size and book-to-market (BM) effects in equity returns. The market capitalisation of equities has
been found to correlate negatively with average stock returns (Banz 1981; Reinganum, 1981). High BM equity rms have
also been observed to earn higher returns on average than low-BM equity rms (Fama & French, 1992; Rosenberg, Reid, &
Lanstein, 1985). Fama and French (1993) suggest a three-factor model to capture the size and BM effects in US stock returns.
Fama and French (1996) note that by excluding the Jegadeesh and Titman (1993) momentum effects, the model captures
much of the Capital Asset Pricing Models (CAPM) anomalies. As a result, Carhart (1997) proposes a four-factor model that
augments the Fama-French model with a momentum factor.
Corresponding author.
E-mail addresses: addaianas@yahoo.com (N.A. Boamah), geoff.loudon@mq.edu.ay (G. Loudon).
http://dx.doi.org/10.1016/j.jeconbus.2017.04.002
0148-6195/ 2017 Elsevier Inc. All rights reserved.
Please cite this article in press as: Boamah, N. A., et al. Regionally integrated asset pricing on the african
stock markets: Evidence from the fama french and carhart models. Journal of Economics and Business (2017),
G Model
2 N.A. Boamah et al. / Journal of Economics and Business xxx (2017) xxxxxx
While there have been numerous tests of these models around the world, there has been little investigation of the models
in Africa. The scant evidence on these models in Africa is understandable, given that the ASMs are largely small. In many
of the countries the small number of listed companies makes it difcult to apply the portfolio methodology that underlies
the models. Yet, it could be argued that it is important to the current and future growth of the ASMs that the models be
investigated in Africa. It would seem important to the development of these markets that we can answer such fundamental
questions as: does a rm size or book-to-market effect exists? Are there any consequences of these effects for regional level
nancial market integration?
Prior African studies such as Basiewicz and Auret (2010) and Bundoo (2011) and applied the Fama-French model to explain
average stock returns on some ASMs. These studies may lack power owing to the small sample size. In order to address the
small sample size problem that characterises many of the ASMs, this paper pools capital markets across countries within
Africa. This approach enables us to accumulate a large enough sample to allow a reasonable application of the portfolio
methodology that underpins the models. The idea of pooling across capital markets is comparable to the estimation of
global factors, or even comparable to the estimation of the factors in a single country characterised by a highly diverse
industry base. In effect, the factors estimated in a single countrya country that happens to have a large enough sample
size to facilitate the portfolio methodologyare an average of individual industry factors. Similarly, our pooling approach
provides estimates of the average factors across the selected ASMs.
The pooling technique also enables us to explore the question of whether assets on the ASMs are priced in a regionally
integrated manner or at the individual country level. If nancial markets are efcient and integrated then there should
be only one set of risk factors that describe equity returns (Fama & French, 2012; Hou, Karolyi, & Kho, 2011). We expect
a regional models mispricing to be insignicant if assets are priced in a regionally integrated way, and if the model is
correct. The approach is consistent with Fama and French (2012), Grifn (2002), Heston et al. (1995), Korajczyk (1996) and
Levine and Zervos (1996), who employ the mispricing from a global factor model to make inferences about nancial market
integration. Stock markets are integrated if international factors describe asset returns (Beckers, Connor, & Curds, 1996;
Heston, Rouwenhorst, & Wessels, 1995; Schotman & Zalewska, 2006; Sewell, Stansell, Lee, & Below, 1996). If the ASMs are
regionally integrated, then regional factors should describe equity returns.
The study makes a contribution to the extant literature on integrated asset pricing. Fama and French (2012); Grifn (2002)
and Hou et al. (2011), have examined integrated asset pricing in the developed markets. We contribute to these studies by
examining whether the size and BM effects are captured by an African regional asset pricing model. This is important, as the
choice of a local, regional or global factors model could signicantly inuence the cost of capital for valuing international
rms, for risk management and the performance evaluation of global portfolio managers (Hou et al., 2011). To the best of
our knowledge this is the rst paper to explore this issue at the African regional level.
The study also makes a contribution in providing evidence that both factorssize and BMexplain the cross-section of
average stock returns on the ASMs. This is a good result indicating that risk factors that are priced in developed capital
markets are also priced in the small ASMs. Another contribution of the paper arises from the methodology of pooling across
countries, and a question about the relevance of relative versus absolute size and BM. A reasonable question is whether
rm size, for example, should be compared to the average rm size of the overall pooled sample or compared to the average
rm size of rms in the same country. This paper provides evidence that both size and BM are ranked by investors in relation
to the individual country averages, and not relative to the overall pooled sample. This supports the hypothesis that the
ASMs are less integrated regionally. The paper shows that the Global Financial Crisis (GFC) did not signicantly impact on
regionally integrated asset pricing in the ASMs.
The remainder of the paper is organised as follows: Section 2 reviews the related African literature. Section 3 describes
the cross-country pooling methodology to be employed in the paper. The data is described in Section 4. Section 5 presents
and discusses the evidence of the study. The conclusion is presented in Section 6.
2. Related African literature
Basiewicz and Auret (2010) investigate the applicability of the Fama-French model on the South African market (SAM).
They observe that the model describes returns on the SAM in a signicant manner and that the models mispricings are largely
small. Also, size but not BM predicts returns after risk adjustment with the model. This may be an indication of incomplete
risk adjustment, or that the model does not fully capture the size effect on the SAM. They attribute the persistence of the
size premium post risk adjustment to market microstructure effects.
Bundoo (2011) applies the Fama-French model to the Mauritius market. In order to control for the small sample size
problem, Bundoo constructed the Fama-French factors at the intersection of two size and BM portfolios. Bundoo also used
these size and BM portfolios and the two BM portfolios as test assets. This notwithstanding, the cross-section of assets and
the time span of the data used in the study were small and therefore likely to introduce small sample bias. Bundoo nds
evidence of the size and BM effects and argues that the model captures these effects on the Mauritius market.
Hearn (2011), Hearn and Piesse (2010a), and Hearn et al. (2010) relied on a three-factor model to explain the cross-
section of equity returns in some African markets. Hearn nds evidence that the size effect was least signicant in Morocco,
and signicantly higher in Egypt and Tunisia. Hearn and Piesse document the rm size effect in a sample spanning West
African, London and Paris equity markets. Hearn et al. note the size effect in a sample consisting of East African and London
markets. These studies controlled for the small sample problem by relying on a universe of stocks across their sampled
G Model
N.A. Boamah et al. / Journal of Economics and Business xxx (2017) xxxxxx 3
Table 1
Characteristics of the African stock markets.
Stock market Number of stocks Mean (US$000) Mean book-to-market
Initial sample After ltering Marketcapitalisation Book value
Botswana 23 18 222763 57415 0.50

Egypt 331 197 316090 146838 0.92
Ghana 39 35 158181 160743 0.92
Ivory Coast 41 36 159130 32445 0.90
Kenya 69 56 127777 51195 1.13
Morocco 103 82 616059 201600 0.56
Mauritius 57 52 50075 38577 1.26
Nigeria 232 174 540934 140808 0.83
South Africa 1035 848 730275 306263 1.14
Tunisia 66 62 260920 104134 0.93
Total 1996 1560 3182204 1240016
Descriptive statistics of the African markets are provided in this table. The mean market capitalisation, book value and book-to-market are the averages
over the sample period of their respective means for each market at the end of June of each year Yt . The book values at the end of June of Yt are the scal
end year book values for year Yt1 . The book-to-market at the end of June of Yt is the scal year end book values for Yt1 divided by the December end
market capitalisation forYt1 . The number of stocks is the total number of rms that each market contributes to the sample over the entire sample period.
The sample period is 31/01/1996 to 30/04/2012.
markets; however, by pooling securities across different markets without controlling for variations in market capitalisation,
they potentially confounded the size effect on the individual markets. Their cross-country tests may therefore be measuring
variation in rm size across the markets studied, rather than for the pooled sample. This would even be more severe in the
case of Hearn and Piesse, and Hearn et al., who combined securities across developed and small emerging African markets.
Soumar et al. (2013) compare the ability of the CAPM and the Fama-French models to describe returns of individual
stocks on the Bourse Rgionale des Valeurs Mobilires (BRVM).They argue that the FamaFrench model better describes
returns on the BRVM than does the CAPM. The generally low adjusted R2 from their tests (the highest is 20.40%) shows that
the explanatory power of both models is low on the BRVM. This may be because the aggregated factors are less relevant for
individual stocks. Soumar et al. study also suffers from small sample size problems.
The evidence from studies such as Hearn (2011) and Hearn and Piesse (2010a) suggests a cautious application of these
models on the ASMs. Small sample size is a major constraint for most of the African studies. Basiewicz and Aurets sample
of 893 stocks appears to be the largest data ever to have been used for an asset pricing test on the ASMs, but this, arguably,
is small. Small sample size may likely not provide reliable estimates of the rm attributes effects in tests based on a model
such as the Fama-French or Carhat. Our cross-country approach, whilst controlling for variations in market capitalisation,
enables us to overcome the small sample problem.
3. Cross-country pooling methodology
The study avoids the small sample problem of prior African studies by pooling securities across 10 ASMs (see Table 1).
The pooling approach enables us to construct more diversied test assets and consequently improve the power of our tests.
Individual country portfolios are less diversied, and their returns contain a high idiosyncratic component (Fama & French,
1998; Fama & French, 2012; Harvey, 1991). The approach enables us to explore the relative versus absolute size or BM
question. The methodology used in this study is similar to Barry et al. (2002), Fama and French (1998, 2012) and Grifn
(2002) who carried out cross-country studies. Our approach is different from Fama and French in that while we measure
the size and BM of each rm in relation to the pooled sample mean, or relative to the rms local market mean or standard
deviation, and employed the scaled variables to rank stocks across countries, they ranked stocks within region and pooled
the ranks across regions to form the global BM portfolios. Grifn also constructs the international or global factors as the
weighted average of the country factors. Barry et al. measures size and BM relative to each local market mean; however,
whilst we employ the Fama and French(Eq. (1)) and Carhart (Eq. (2)) models, they rely on cross-sectional regression of
portfolio returns on size, BM and market beta.

rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + it (1)

rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + ei WMLt + it (2)
r it = return to the proftfolio i at time t, rft = the risk free rate at time t,
MKTRt = return to the market protfolio at time t,
SMBt = portfolio that is short in small capitalisation stocks and long in big capitalisation stocks,
G Model
WMLt = portfolio that is long in the past 12 months winners and short in the past 12months losers, it = error term.
Continuously compounded stock returns (in US$) were estimated from monthly stock prices adjusted for stock splits and
cash dividend payment. All the returns are in excess of the risk-free rate. We use the monthly South African Treasury bill
rate (in US$) as a proxy for the risk-free rate. This was motivated by the dominance of South African stocks in the sample.
Also, it was the only short-term instrument from the sampled markets with data available over a longer period.
We measure rm size as the market value of shares outstanding at the end of each June. On each portfolio formation date,
we estimate the relative market value for each rm by scaling the rms size by the cross-sectional standard deviation, or
cross-sectional mean of the market capitalisation of listed rms on the given exchange. Similarly, we estimate the BM ratio
for each rm in a given country relative to the cross-sectional mean or standard deviation of the BM of all equities on that
market. The BM ratios were estimated at the end of each June as follows:
BMi,t = BVi,t1 /MVi,t1 (3)
BM i,t = book value to market value ration of stock i in June year Yt,
BM i,t1 = fiscal year end book value of stock in yearY t1 ,
BM i,t1 = market value of stock i in December of year Y y1 .
Scaling rm size and BM by the cross-sectional standard deviation or mean is important, as it ensures that rms are
measured as big or small, for instance, based on their ranks on their respective markets and not on their ranking in the
pooled sample. This is essential in controlling for any potential confounding effects of the size and BM effects in the pooled
sample. Two rms A and B of the same market capitalisation, for instance, which are listed on markets C and D respectively,
with A and B correspondingly ranked big and small on their respective markets, should appropriately not be ranked as being
of the same size in the pooled sample. Rather, their ranks in the pooled sample should reect their respective ranks on their
markets; otherwise, the tests may incorrectly measure variations in capitalisations across markets C and D. The scaling,
however, is only necessary if markets in the pooled sample are segmented. In integrated markets, size and BM should be
measured relative to the pooled sample mean.
Thus, relative market capitalisation enables us to control for any bias that may be induced by variation in rm size
across the markets. The relative BM ratios also help in reducing the impact of any potential variations in accounting systems
across the African markets. The standardisation of size and BM also ensured that the number of rms in each portfolio was
distributed across all sampled markets based on the proportion of rms that they each contribute to the study sample.
Relative measures provide an appropriate mechanism for comparing size and BM across the African markets, since prior
studies suggest some degree of segmentation amongst them (e.g. Alagidede, 2010; Kodongo & Ojah, 2011; Piesse & Hearn,
2005; Sugimoto, Matsuki, & Yoshida, 2014).
The universe of securities listed on the sampled ASMs was used in forming the HML, SMB, and WML. Firms were sorted into
the size-based portfolios using the median market capitalisation as breakpoint. The size portfolios were balanced annually
at the end of each June of year Yt . The rankings at the end of June of Yt were used to assign rms to one of the size-based
portfolios from July of Yt to June of Yt+1 . Similarly, the June of year Yt BM ratios were used to assign stocks into one of three
BM-sorted portfolios from July of Yt to June of Yt+1 based on breakpoints of 30% and 70%. The BM portfolios were balanced
at the end of June of each year. This ensured that the book values were available to the market for at least six months, and
thus controlled for any potential look-ahead bias (see Banz & Breen, 1986). Following Fama and French (2012), at the end
of each month, the SMB was constructed as the difference between the average of the three small-rm portfolios and the
average of the three large-rm portfolios across the BM tercile. Also, at each month end the difference between the average
returns of the two highest BM portfolios and the average returns of the two lowest BM portfolios within the two size groups,
constituted the HML.
Following prior studies, we form the WML based on the t-2 to t-12 months cumulative returns. The WML is formed at the
intersection of the two size and three momentum portfolios. The momentum portfolios are formed using breakpoints of
30% and 70%. Firms that fall within the bottom 30% and the top 30% of the past 12 months cumulative returns are the losers
and winners respectively. The difference between the returns of the two extreme portfolios is the WML.
The return on the market portfolio was estimated as the value-weighted return of the market portfolio returns of all the
markets being investigated. The market indices of the various markets were used as surrogates for their respective market
portfolios. The market return was estimated as:
10
MKTRt = wj,t Rj,t (4)
j=1
W t = total market capitalization of market j as a proportion of the total market capitalization of all the sampled
markets at time t,
G Model
Rj,t = return to the market portfolio of market j at time ft.
Also, nine value-weighted test portfolios are formed by tri-sorting on both size and BM.1 The size of our sample makes
nine portfolios most appropriate; however, this is not expected to bias the outcome of the study. The portfolios are rebalanced
annually. As before, the size portfolios are constructed at the end of June of year Yt , and the rankings used to assign rms
into one of the three size-based portfolios from July of Yt to June of Yt+1 . Also, the BM portfolios were balanced annually at
the end of June of Yt , and the ranking employed to assign rms into one of the three BM portfolios from July of Yt to June
ofYt+1 . Both the size and BM-sorted portfolios are constructed based on breakpoints of 33.3% and 66.6%.
The estimated SMB, HML, MKTR, WML, and the nine test portfolio returns are regional factors. The investigation is carried
out at the African regional level, rather than individual country level. This approach enables us to overcome the small samples
problem in prior African studies, and to also test whether assets are priced in a regionally integrated way on the African
markets.
4. Data
The data are from DataStream, augmented with Bloombergs data. The data spans the period 31 January 199630 April
2012. The data were available for most of our sampled markets only over this period. The sample duration enables us to test
the impact of signicant global events, such as the GFC, on regional level segmentation of the ASMs. The cut-off date of 2012
is informed by the evidence in prior studies (see e.g. Boamah, Loudon, & Watts, 2016; Sugimoto et al., 2014) that the effect
of the GFC persisted into 2012. All rms that are listed on the selected African markets are sampled. It has been argued that
some of the investigations into the rms characteristics effect suffer from survivorship bias that may distort the ndings of
such studies (e.g. Banz & Breen, 1986; Kothari, Shanken, & Sloan, 1995). Delisted rms are thus included in the sample up
to the delisting date to control for survivorship bias problems. We assume the last trading day price as current price when
estimating delisting returns. The initial sample consists of 1996 stocks across 10 ASMs (see Table 1).
The sampled markets (out of about 21) are the ones for which data were available on most rms over a relatively long
period of time (15 years). Firms that had less than 12 price observations (168) and rms without accounting and number of
shares outstanding data (75) were excluded from the sample. Excluding rms with less than 12 price observations helps in
improving the power of the tests. We also exclude rms that trade for less than 20% of possible trading months (157) to help
reduce the impact of the documented thin trading problem on the ASMs (see e.g. Appiah-Kusi & Menyah, 2003; Mlambo &
Biekpe, 2005). Following Fama and French (1993), negative book equity rms (36) are not included in the study.
The nal sample consists of 1560 stocks; this, arguably, is the largest sample ever to have been applied in any African asset
pricing study. Table 1 shows that South Africa contributed the largest number of stocks (54.36%), with Botswana contributing
the smallest number of stocks (1.15%) to the sample. The proportion of South African equities dominates the sample in each
year of the sample period. This has the potential to bias the results towards the South African market. We address this issue in
a robustness test by conducting the analysis separately for the South African and non-South African markets in sub-section
2.6.3.
5. Presentation and analysis of results2
5.1. Descriptive statistics of the SMB, HML, MKTR and WML portfolio returns
This sub-section presents the mean and standard deviations of the risk factors. Table 2 indicates that the average monthly
returns to the SMB, HML, MKTR and the WML are, respectively, 0.52%, 1.2%, 0.62% and 0.93%. The results are consistent with
Barry et al. (2002), and Basiewicz and Auret (2010), who document signicant SMB and HML returns in emerging markets.
The market return of 0.62% seems signicant in economic terms; however, it is not signicant statistically. The SMB, HML,
and WML returns are both statistically and economically signicant.
5.2. Returns analysis
The average portfolio size is in the range of 31.5 and 121.3 stocks. None of the portfolios had less than 12 stockseven
in the early years of the sampleand are therefore likely to be less noisy compared to prior African studies. The size of
1
The size of our sample potentially enables us to form test portfolios based on a 3 4 or 4 4 sorts on size and BM. This, however, was not feasible since
some portfolios in the early years of the sample had less than 8 stocks. Such portfolios were quite noisy. A 3 3 sort provided portfolios with at least 10
stocks, even in the early years of the sample. The study thus relies on a 3 3 sort in forming the test assets.
2
Illiquidity is a major concern on all of the ASMs (see e.g., Appiah-Kusi and Menyah, 2003; Hearn and Piesse, 2013a; Hearn, 2014). We employ the Dimson
and Marsh (1983) approach to control for illiquidity. The results, which are not recorded, show that the technique had minimal impact on the outcome
of the study. The results with and without Dimson and Marsh adjustment generally corroborate each other, although the Dimson and Marsh results are
associated with slightly higher pricing errors and GRS statistic. The ndings show the effectiveness of this papers thin trading adjustment technique. The
unrecorded results are available from the authors on request.
G Model
the portfolios provides some sureness that the impact of noise-induced bias on this study is limited. However, due to the
generally small size of the portfolios, the evidence from the study may have to be interpreted with caution. Table 3 shows
that the monthly returns for the nine portfolios range from 0.88% to 2.51% (Panel A) with standard deviation within the
range of 6.2% 7.5% (Panel B). The observed means and standard deviations for the excess returns are generally large. This
is consistent with the observations by Kenny and Moss (1998) and Claessens et al. (1995) that stock markets in emerging
markets offer dramatic returns and are volatile. The results indicate that the African markets offer an opportunity for higher
returns but at the cost of higher variability.
Panel A of Table 3 shows that there exists a decrease in returns from the small-rm to the big-rm portfolio, while
holding BM constant. Similarly, when size is held constant, returns increase from the low-BM to the high-BM portfolios. For
instance, within the low-BM and high-BM portfolios, small rms outperform big rms by 0.7% and 1.0%, respectively, on a
monthly basis. Similarly, within the small and big-rm portfolios, the excess monthly returns of the high-BM portfolio over
the low-BM portfolio were, respectively, 0.95% and 0.59%. This is consistent with studies such as Banz (1981), Fama and
French (1992, 1998) and Reinganum (1981).
The results indicate that while controlling for size, the BM effect still persists. Likewise, the size effect is not accounted for
by the BM effect. The returns are highest for the small, high-BM group (2.5% per month) and low for the big, low-BM group
(0.9% per month), suggesting some interaction between the size and BM effects. The small-value rm returns are signicant
both statistically and economically. Notice, however, that the independent size and BM effects are, respectively, 0.78% and
0.97% per month and are both signicant statistically. The size and the BM effect therefore appear to be independent and do
not subsume each other. This result is consistent with Basiewicz and Auret (2010), Fama and French (1992, 1993) and Van
Rensburg and Robertson (2003).
Consistent with Miles and Timmerman (1996) for the UK market, and Fama and French, the BM appears stronger than
the size effect on the African markets. The value premium is statistically signicant at the 5% level of signicance across
the small and medium size groups. However, the size premium is only statistically signicant at the 5% level of signicance
within the high-BM portfolios. It thus appears that the bulk of the size effect occurs in the high-BM tercile. The BM effect
is strong within the big-rm portfolios (0.6% per month), though weaker than that of the small-rm portfolio (1.0% per
Table 2
Summary statistics of the monthly returns of the risk factors.
SMB HML MKTR MKTR-rf WML
Mean 0.0052 0.0120 0.0062 0.0051 0.0093

Standard deviation (%) 3.04 3.64 6.89 6.89 5.11
t-statistic 2.26 4.39 1.19 0.98 2.64
SMB is the portfolio that is long in low capitalisation stocks and short in high capitalisation stocks; HML is the portfolio that is long in high-BM and short
in low-BM stocks; MKTR is the market return estimated as the weighted average of the market returns of all of the markets included in the study; WML
is the portfolio that is long in the past 12-months winners and short in the past-12 months losers. rf is the risk free rate estimated as the 3-months South
African Treasury bill rate expressed in monthly rates and in US dollars. All returns are expressed in US dollars. SMB and HML were formed by sorting on
two size and three BM portfolios.
Table 3
Descriptive statistics of the monthly returns of the test assets.
Size Book-to-market
Panel A: Mean monthly returns

Low Medium High High-low
Small 0.0156 (2.78) 0.0208 (3.84) 0.0251 (4.83) 0.0095 (2.22)
Medium 0.0088 (1.90) 0.0165 (3.40) 0.0224 (4.38) 0.0137(3.48)
Big 0.0089 (1.64) 0.0145 (2.82) 0.0148 (2.88) 0.0059 (1.74)
Small-big 0.0067 (1.63) 0.0063 (1.95) 0.0103 (2.52)
Independent size effect = 0.0078(2.92)
Independent book-to-market effect = 0.0097(3.71)
Panel B: Standard deviation of returns (%)
Low Medium High Average
Small 7.46 7.20 6.90 7.19

Medium 6.15 6.44 6.82 6.47
Big 7.22 6.85 6.85 6.97
Average 6.94 6.83 6.86
This table presents the mean monthly returns of nine portfolios sorted on 3-size and 3-book-to-market ratio. The size and BM portfolios were constructed
based on 33.3% and 66.6% breakpoints. Firms within the bottom 33.3%, above the 66.6%, and in-between the breakpoints of the market capitalisation were
respectively considered small, big and medium size rms. Similarly, rms within the top, the bottom and middle BM tercile were deemed high, low and
medium BM rms respectively. All returns are expressed in US dollars. All portfolios were formed at the end of each June and balanced annually. Size and
BM of a security were measured relative to the cross-sectional standard deviation of the market capitalisation and BM of the securitys exchange at the end
of each June. The size effect independent of the BM effect is the average of the excess small rm monthly returns over the big rms across all BM groups.
Similarly, the independent BM effect was estimated as the average of the value effect across all size groups. T-statistics are in parenthesis.
G Model
Table 4
Fama-French regression on size and book-to-market sorted portfolios.
Size Book-to-market Book-to-market
Low Medium High Low Medium High
a b
Small 0.0096(2.58) 0.0102(2.2) 0.0092(3.06) 0.8244(15.9) 0.8633(17.84) 0.8765(20.9)
Medium 0.0044 (1.95) 0.0083(3.3) 0.0093(3.01) 0.7740(24.7) 0.8105(21.25) 0.8550(20.0)
Big 0.0077 (5.06) 0.0070(3.0) 0.0032(1.24) 0.9605(45.5) 0.9323(29.02) 0.8593(23.7)
GRS: a1. . .. . ..= a9 = 0 = 3.93/0.0147**
c d
Small 0.8049(6.72) 0.8574(7.7) 0.9676(10.01) 0.2908(2.97) 0.0560(0.61) 0.4404(5.56)
Medium 0.5495(7.6) 0.6946(7.9) 0.2854 (2.9) 0.2906(4.91) 0.0544 (0.75) 0.5212(6.45)
Big 0.1829(3.76) 0.2790(3.6) 0.0961 (1.15) 0.3187 (8.0) 0.0234(0.390 0.5519(8.07)
R2
Small 0.6402 0.6633 0.7258
Medium 0.8067 0.7386 0.7079
Big 0.9366 0.8360 0.7919
This table presents the results of estimating the model below for the nine test portfolios (ri ) formed at the intersection of 3-size and 3-BM sorted portfolios.
MKTRt rf , SMB and HML are correspondingly the excess return to the market portfolio, the difference between the small and big rm portfolio returns, and a
portfolio of high-BM minus a portfolio of low-BM stocks. MKTR is estimated as the weighted average of the market returns of all of the studied markets. SMB
and HML were formed by sorting on 2-size and 3-BM portfolios. rf is the South African 3-months Treasury bill rates expressed in monthly rates. All portfolios
were constructed at June end and balanced yearly. All returns are expressed in US dollars. Size was estimated as the market capitalisation of an equity divided
by the cross-sectional standard deviation of market capitalisation of equities on the securitys exchange on the portfolio formation date. Similarly, book
value is equitys scal year end book value scaled by the cross-sectional standard deviation of scal year ends book values of asecuritys market for a given
year. Book-to-market was estimated at December end each year. T-statistics are in parenthesis. rit rt = ai + bi MKTRt rft + ci SMBt + di HMLt + it . **
indicates probability. GRS is the F-test of Gibbons et al. (1989).
month). This is consistent with Fama and French (1993, 2006), and Loughran (1997), who all observe a weaker BM effect
within the big-rm stock portfolio relative to that of the small-rm portfolio. The evidence indicates that the big, high-BM
rms outperform the big, low-BM rms by 7.08% per annum. The value premium on the ASM is therefore unlikely to be a
small-rm phenomenon.
Panel B indicates that the monthly average standard deviations increase slightly from the big stock (6.97%) to the small
stock (7.19%) portfolios and from the high-BM (6.86%) to the low-BM (6.94%) portfolios. This suggests that the small and
high-BM stocks may not be riskier compared to the big and low-BM rms. The observed small and high-BM stocks premium
may therefore not be a compensation for higher variability. This is consistent with Haugen and Baker (1996).The monthly
portfolio volatilities observed in this study are lower than what has been observed by prior emerging market studies, such
as Bekaert and Harvey (1997). This does not, however, imply that the African markets are less volatile than the emerging
markets investigated by prior studies. The low portfolio standard deviations documented in this study may be the outcome
of pooling securities across several markets, and hence a benet from international diversication. This is consistent with
Kodongo and Ojah (2011) and Sugimoto et al. (2014), who observe that the ASMs are partly segmented among themselves.
Segmented or less correlated markets provide higher risk diversication gains. The correlations between the returns to the
market portfolios of the sampled ASMs are in the range of 0.07 to 0.52. These are relatively low. This is consistent with
Collins and Biekpe (2003), who observe correlations of between 0.17 and 0.339 amongst the ASMs.
5.3. Results from estimating the Fama-French three-factor model3
The Fama-French regression (Eq. (1)) is estimated for the test assets and the results presented in Table 4. The nine
regressions are estimated as a system of equations using seemingly unrelated regressions (SUR). The SUR methodology
controls for heteroscedacity, and also adjusts for the cross-sectional correlations in the residual returns across assets. The
approach also enables us to carry out a joint test of signicance of the regression intercepts.
In general, the intercepts do not increase monotonically from the big to the small rms and from the low-BM to the high-
BM stocks. Thus, the size and BM effects in Table 3 are generally not obvious from the regression intercepts. However, the
intercepts for the small rm portfolios (averaging 0.97% per month) are bigger relative to those of the large rms portfolios
(averaging 0.60% per month). The intercepts for the high- and low-BM portfolios each average 0.72% per month. The results
show higher pricing errors for the small-rm relative to the big-rm portfolios. The spread between the pricing errors for
the small and big-rm portfolios appears economically meaningful (0.37% per month). The intercepts suggest that the Fama-
French factors partly explain the returns to these portfolios. Fig. 1 further illustrates these. The gure shows that the size and
3
Lewellen et al. (2010) argue that asset pricing tests based on size-BM portfolios lead to misleading inferences since the portfolios show strong factor
structure. They propose extending the test assets beyond the size-BM portfolios to include, for instance, industry portfolios. We implement the Lewellen
et al. proposal by augmenting the 9 size-BM test portfolios with 10 industry test portfolios. This, however, had minimal impact on the tests in this study.
For brevity, the results are not reported, but are available from the authors on request.
G Model
Fig. 1. Absolute mispricing from the Fama-French model and mean returns.
presentsthe absolute pricing errors from the model below. It also indicates the mean returns of the size and BM sorted portfolios. rit rt =
The gure
ai + bi MKTRt rft + ci SMBt + di HMLt + it .
BM effects are only partially captured by the model. Within the medium-size rm group, the intercepts rise monotonically
from the low to the high-BM group, and also the mispricing decreases in a monotonic manner from the small to the big-rm
group within the medium-BM tercile, which are consistent with the pattern in returns.
Merton (1973) observes that the intercepts should be statistically indistinguishable from zero if a model meaningfully
explains the dependant returns. An inspection of the intercepts indicates that of the nine intercepts, only two are statistically
indistinguishable from zero. The F-test of Gibbons, Ross, & Shanken, 1989 suggests that the intercepts are jointly signicantly
different from zero. This indicates that the Fama-French factors do not fully describe the returns to the size and BM-sorted
portfolios on these markets. The average annual mispricing of 9.24% is signicant statistically and economically, suggesting
the inadequacy of the model. This may be the outcome of constructing diversied test assets that consequently increased
the power of our test. It may also be an indication of regionally segmented African markets, assuming the model is correct.
The signicant mispricing is the expected result if the African markets are not completely integrated. As observed by
Korajczyk (1996), market segmentation produces signicant pricing errors in a global factor model due to its constraining
effect on cross-market arbitrage. Errunza and Losq (1985) and Errunza et al. (1992) show that nancial markets are most
likely partially integrated, but not completely integrated or segmented. The signicant mispricing observed in this study is
thus expected, given that the ASMs are unlikely to be completely integrated. The mispricing thus contains useful information
about the regional level segmentation of the ASMs.
The market betas are all positive and signicantly different from zero, but are generally less than one. Consistent with
Fama and French (1993, 1996), the high-BM and low-BM portfolios correspondingly load positively and negatively on the
HML, whilst the small and big rms portfolios, respectively, have positive and negative loadings on the SMB. This suggests the
existence of size and value effects in the markets under study. An exception is the big medium-BM rms that unexpectedly
load positively on the SMB.
Within the context of Fama and French (1996) risk-based interpretations, the implication of the ndings is that distress
risk decreases with rm size, and rises with BM equity. However, within the irrational pricing hypothesis (see e.g. Ali,
Hwang, & Trombley, 2003; Black, 1995) the ndings suggest market mispricing of securities on the ASMs. Investors may
hold erroneous expectations of the market or may under-utilise available information (Daniel, Hirshleifer, & Subrahmanyam,
2001). This may be particularly so in the African context where the markets are still developing and where the investment
analyst industry is largely underdeveloped. Investors ability to process information may therefore be limited.
The R2 rises from the small-rm to the big-rm portfolios. This suggests that the Fama-French factors explain the big rm
returns better than the small rm returns. With the exception of the small-rm portfolios, where the R2 increases from the
low to the high-BM portfolio, the R2 decreases moving from the low to the high-BM portfolios, while holding size constant.
Excepting the small rm tercile, the R2 for the low-BM portfolios are higher than those of the high-BM portfolios. These
suggest that the model better describes the returns to the low-BM than it does the high-BM portfolios. The ndings suggest
that the size and BM effects may not be due to the peculiar characteristics of the developed markets, but may be pervasive
even in small and volatile emerging markets. The overall nding does not support the hypothesis of regionally integrated
asset pricing.
5.4. Results from estimating the Carhart four-factor model
The results from estimating the Carhart model (Eq. (2)) are presented in Table 5. The table indicates that, with the
exception of the big and high-BM rm group, and the medium-size and low-BM rm portfolios, the mispricings of the model
are statistically distinguishable from zero. The GRS F-statistic rejects the joint test that the regression coefcients are zero.
The evidence indicates that the Carhart model is inadequate in describing the returns to the size and BM-sorted portfolios
G Model
Table 5
Carhart regression on size and book-to-market sorted portfolios.
Size Book-to-market Book-to-market
a b
Small 0.0085(2.26) 0.0092(2.62) 0.0078(2.6) 0.8410(16.1) 0.8778(18.0) 0.8977(216)
Medium 0.0038(1.65) 0.0077(2.75) 0.0103(3.32) 0.7793(24.5) 0.8129(20.9) 0.8400(195)
Big 0.0077(4.99) 0.0061(2.63) 0.0032(1.21) 0.9605(44.8) 0.9445(29.2) 0.8597(233)
GRS: a1 =. . .. . .. . .= a9 = 0 = 3.63/0.0195**
c d
Small 0.7469(6.1) 0.8060(7.0) 0.8923(9.1) 0.2864(2.9) 0.0597(0.7) 0.4460(5.8)
Medium 0.5247(7.0) 0.6706(7.4) 0.3366(3.3) 0.2872(4.9) 0.0548(-0.8) 0.5171(6.5)
Big 0.1835(-3.6) 0.2353(3.1) 0.0983(1.1) 0.3187(7.9) 0.0262(0.4) 0.5517(8.0)
e R2
Small 0.1392(1.8) 0.1225(1.7) 0.1789(2.9) 0.647 0.6689 0.7387
Medium 0.0480(1.0) 0.0485(0.9) 0.1233(1.9) 0.8075 0.7386 0.714
Big 0.0014(0.05) 0.1031(2.2) 0.0044(0.08) 0.9365 0.8403 0.7919
MKTRt rf , SMB, HML and WML are correspondingly the excess return to the market portfolio, the difference between the small and big rm portfolio
returns, a portfolio of high-BM minus a portfolio of low-BM stocks and a portfolio of the past 12-months winners minus the portfolio of the past 12-months
losers. MKTR is estimated as the weighted average of the market returns of all of the studied markets. SMB and HML were formed by dual-sorting on market
capitalisation and tri-sorting on BM. rf is the South African 3-months Treasury bill rates expressed in monthly rates. All portfolios were constructed at June
end and balanced yearly. All returns are expressed in US dollars. Size was estimated as the market capitalisation of an equity divided by the cross-sectional
standard deviation of the market capitalisation of equities on the securitys exchange on the portfolio formation date. Similarly, book value is equitys scal
year end book value scaled by the cross-sectional standard deviation of scal year ends book values of asecuritys market for a given year. Book-to-market
was estimated at December end each year. T-statistics are in parenthesis. rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + ei WMLt + it .
** indicates probability. GRS is the F-test of Gibbons et al. (1989).
Fig. 2. Absolute mispricing from the Carhart model and mean returns.
The gure presents the absolute pricing errors from the model below. It also indicates the mean returns of the size and BM sorted portfolios. rit rft =

ai + bi MKTRt rft + ci SMBt + di HMLt + ei WMLt + it
on the ASMs. Compared with the evidence in Table 4, the GRS statistic (3.63) and the average absolute mispricing (8.57%
per annum) indicate that the regional four-factor model better describes equity returns on the ASMs than does the regional
three-factor model. Both the GRS statistic and the average mispricing are lower for the four-factor model relative to the
three-factor one.
Consistent with the evidence in Table 4, the mispricings are higher for the small-rm portfolio relative to the big-rm
group, though the increase in the mispricings from the small to the big-rm group is non-monotonic. The average absolute
mispricing is larger for the small-rm portfolio (0.57%) than the big-rm one (0.31%); the spread is 0.26%. Also, the spread
between the average absolute mispricing for the high-BM (0.71%) and the low-BM (0.67%) portfolios is 0.04%. The pricing
errors, however, portray the size effect observed in Table 3 that indicates the inadequacy of the African regional model in
describing the returns of the pooled assets. Also, excluding the medium-size rm group, the models mispricing increases
from the high-BM group to the low-BM portfolio, which is an indication of a reversal of the BM effect. Fig. 2 represents a
monotonic surge in the pricing errors from the low-BM to the high-BM group within the medium-rm portfolio. Also, the
mispricing increases in a monotonic manner from the big to the small-rm portfolio within the medium-BM group. These
results are consistent with the returns pattern.
The market betas are all positive and statistically signicant. The loadings on the SMB and the HML generally have the
expected signs. With the exception of the medium-size and high-BM portfolios, the loadings on the momentum factor have
positive signs. The R2 in Table 5 are higher for the big-rm portfolios relative to the small-rm portfolios, and are larger for
the low-BM portfolios in relation to the high-BM portfolios, excluding the small-rm tercile. There is no signicant variation
G Model
Table 6
Fama-French regression on size and book-to-market sorted portfolios using size and BM measured relative to the mean of the pooled sample and that of
the individual markets.
Size Book-to-market
Panel A: Absolute size and BM Panel B: Size and BM relative to the mean of a securitys market

a a
Small 0.0129(3.3) 0.0094 (2.7) 0.0123(4.0)0.0117(3.0) 0.0133(3.2) 0.0139(4.5)
Medium 0.0048(1.9) 0.0030(1.2) 0.0113(4.4)0.0040(1.6) 0.0117(4.2) 0.0075(2.5)
Big 0.0098(6.0) 0.0064(3.3) 0.0046(1.6)0.0055(3.5) 0.0087(3.7) 0.0039(1.4)
GRS: a1=. . .. . .. . ..=a9 = 0 = 5.59/0.0040** a1=. . .. . ..= a9 = 0 = 4.73/0.0075**
b b
Small 0.7447(11.6) 0.9279(16.3) 0.9556(19.4)
0.8544(14.6) 0.8751(14.1) 0.8476(18.2)
Medium 0.7995(20.4) 0.9282(23.8) 0.8345(20.3)
0.8017(20.9) 0.7581(17.8) 0.7799(17.3)
Big 0.8052(30.5) 1.0216(32.5) 0.8477(17.8)
1.0012(41.93) 0.9321(26.1) 0.8936(20.6)
c c
Small 0.6016(4.6) 0.7933(6.7) 1.0711(10.5)
0.6667(5.8) 0.8007(6.5) 0.6449(7.0)
Medium 0.4910(6.1) 0.6135(7.6) 0.3842(4.5)0.5782(7.6) 0.5560(6.6) 0.3611(4.1)
Big 0.2133 (3.9) 0.0659 (1.0) 0.2003 (2.0)
0.1063 (2.3) 0.2000(2.8) -0.2688(-3.1)
d d
Small 0.1334 (1.3) 0.0761(0.81 0.3257 (-3.5)
0.5486(6.8) 0.0618 (-0.6) 0.2434(3.3)
Medium 0.1465 (2.3) 0.1865(2.9) 0.2574 (4.2)
0.5269(7.8) 0.1337 (1.9) 0.5291(7.4)
Big 0.3460(8.0) 0.0972(1.9) 0.2197(5.8)
0.6498(8.3) 0.0827(1.5) 0.5169(7.5)
R2 R2
Small 0.4515 0.6116 0.7036 0.5671 0.5356 0.6643
Medium 0.7257 0.7823 0.7678 0.7249 0.6500 0.6814
Big 0.8993 0.9045 0.7988 0.9305 0.8180 0.7940
MKTRt rf , SMB and HML are correspondingly the excess returns to the market portfolio, the difference between the small and big rm portfolio returns,
and a portfolio of high-BM minus a portfolio of low-BM stocks. MKTR is estimated as the value weighted average of the market returns of all of the studied
markets. SMB and HML were formed by dual-sorting on market capitalisation and tri-sorting on BM. rf is the South African 3-months Treasury bill rates
expressed in monthly rates. All portfolios were constructed at June end and balanced yearly. All returns are expressed in US dollars. Panel A presents the
results when size and book values are estimated in absolute terms. In panel B, size was estimated as the market capitalisation of an equity divided by the
mean market capitalisation of equities on its exchange on the portfolio formation date. Similarly, book value is equitys scal year end book value scaled
by the cross-sectional mean of scal year ends book values of a securitys market for a given year. Book-to-market was estimated at December end each
year. T-statistics are in parenthesis. GRSis the F-test of Gibbons et al. (1989).
** indicates probability .rit rt = ai + bi MKTRt rft + ci SMBt + di HMLt + it
in the regression ts from the three and four-factor models. The evidence in Table 4 is generally consistent with that in
Table 5. The ndings from Tables 4 and 5 reject the hypothesis of regionally integrated asset pricing of the African capital
markets.
5.5. Regional level integration of the African stock markets
This sub-section explores the integration of the ASMs amongst themselves, through the asset pricing process. The pre-
ceding sub-sections provide evidence in support of the applicability of the Fama-French and Carhart models for the pooled
African sample, though the models do not fully capture the size and BM effects. The evidence, however, provides some
basis for the adoption of the models in examining the integration of the sampled markets. This is examined by relying on
versions of the models that are based on absolute and relative measures of the ranking variables. A fundamental assumption
underpinning the absolute measures of size and BM is that the pooled markets are highly regionally integrated. We therefore
expect that the model based on absolute size and BM measures should be associated with lower mispricing relative to those
relying on relative size and BM proxies if the markets show a higher level of integration regionally.
Eq. (1) is also estimated based on test assets, SMB and HML constructed from size and BM that were measured in absolute
terms and relative to their cross-sectional means for the securitys local market. The results are presented in Table 6. The
average absolute intercept coefcients are 9.24%, 9.96%, and 10.68% (per annum), respectively, when size and BM of a security
are measured relative to its cross-sectional standard deviations for the securitys market (RSD, Table 4), in absolute terms
(ABS), and in relation to the cross-sectional means for the individual markets (RM).The variation between the yearly average
standard errors is 1.44% between RSD and RM, 0.72% between ABS and RM, and 0.72% between RSD and ABS. These mispricing
differences are signicant economically. In all the models estimated, the GRS F-statistic rejects the joint hypothesis that the
intercepts are zero. These indicate that none of the models fully describes equity returns on the ASMs. The GRS statistics are
3.93, 4.73 and 5.59, respectively, for the RSD, RM and ABS. A lower GRS is preferable. The results show that the RSD proxies
for size and BM are superior to their RM counterparts, and the RM proxies superior to the ABS proxies.
We estimate the Carhart model (Eq. (2)) by employing ABS and RM-based SMB, HML and test assets. For brevity, the
full results are not presented. The mispricing and the R2 are presented in Table 7. The GRS test rejects the hypothesis that
the intercepts are jointly indistinguishable from zero. The GRS statistic indicates that tests based on the RSD better price
G Model
Table 7
Carhart regression on size and book-to-market sorted portfolios using size and BM measured relative to the mean of the pooled sample and that of the
individual markets.
Size Book-to-market
Panel A: Absolute size and BM Panel B: Size and BM relative to the mean of a securitys market

a a
Small 0.0127(3.14) 0.0091(2.53) 0.0107(3.53) 0.0115(2.92) 0.0119(2.88) 0.0122(3.98)
Medium 0.0050(2.05) 0.0026(1.06) 0.0121(4.69) 0.0037(1.42) 0.0111(3.86) 0.0074(2.44)
Big 0.0091(5.55) 0.0058(2.93) 0.0042(1.4) 0.0057(3.54) 0.0076(3.23) 0.0034(1.16)
GRS: a1 = . . .. . .. . .= a9 = 0 = 05.07/0.0058** a1 = . . .. . .. . .= a9 = 0 = 4.08/0.0129**
R2 R 2
Small 0.4518 0.6134 0.7184 0.5673 0.544 0.6828

Medium 0.7279 0.7834 0.7717 0.7219 0.653 0.6813
Big 0.9026 0.9066 0.7995 0.9305 0.8247 0.7952
This table presents the results of estimating the model below for the nine test portfolios (ri ) formed by tri-sorting on both size and BM. MKTRt , SMB, HML
and WML are correspondingly the excess return to the market portfolio, the small minus big rm portfolio returns, a portfolio of high-BM minus a portfolio
of low-BM stocks and a portfolio of past 12 months winners minus the past 12 months losers. MKTR is estimated as the weighted average of the market
returns of all of the studied markets. SMB and HML were formed by dual-sorting on market capitalisation and tri-sorting on BM. rf is the South African
3-months Treasury bill rates expressed in monthly rates. All portfolios were constructed at June end and balanced yearly. All returns are expressed in US
dollars. Panel A presents the results when size and book values are estimated in absolute terms. In panel B, size was estimated as the market capitalisation
of an equity divided by the mean market capitalisation of equities on its exchange on the portfolio formation date. Similarly, book value is equitys scal
year end book value scaled by the cross-sectional mean of scal year ends book values of a securitys market for a given year. Book-to-market was estimated
at December endeach year. T-statistics
are in parenthesis. ** indicates probability. GRS is the F-test of Gibbons et al. (1989)
rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + ei WMLt + it .
assets on the ASMs than the RM based tests, and that the RM performs better than the ABS. The evidence in this paper
suggests higher mispricing for the absolute measures of size and BM models, indicating a relatively higher level of regional
segmentation of the African markets.
5.6. Impact of the Global Financial Crisis (GFC)
We explore the inuence of the GFC on the rm attributes effect on the African markets. We employ the mispricings from
Eq. (4) to make inference about the changing degree of ASMs integration amongst them following the GFC. The mispricing
should be statistically insignicant if assets are priced in a regionally integrated way and if the model is correct (Fama and
French, 2012). We thus expect that if the level of regional integration of the ASMs has improved, then the models pricing
errors post-GFC should be insignicant or smaller relative to the pre-GFC pricing errors. We use October 2008 as the cut-off
date for the GFC.

rit rft = ai + 2 ai + 1 bi MKTRt rft + ci SMBt + di HMLt + 2 bi MKTRt rft + ci SMBt + di HMLt + it (4)
2 = 1 1
1 is an indicator variable that takes the value of 0 for the post October 2008 period and1otherwise, r it
= return to the portfolio i at time t,
r ft = the risk free rate at time t, it = error term.
We do not specify a crisis window, but instead split the sample into the pre and post-GFC eras. The contagion literature
indicates that tests examining the impact of crisis are sensitive to the choice of crisis window and that arbitrarily chosen
crisis periods result in misleading inferences (see e.g., Billio & Pelizzon, 2003; Caporale, Cipollini, & Spagnolo, 2005; Yarovaya
& Lau, 2016). Our approach enables us to avoid arbitrariness in the selection of the crisis length. Additionally, the evidence
in Boamah et al. (2016) shows that the only global event that is relevant to all of the ASMs is the GFC. They show that the
impact of the GFC on the ASMs persisted to at least the end of 2012. Sugimoto et al. (2014) provide corroborative evidence
on the persistence of the crisis. The authors nd that the GFC merged into the Eurozone sovereign debt crisis and that the
two crises are inseparable on the ASMs. Sugimoto et al. ndings suggest longer crisis duration. Testing for only the effect of
the GFC and the splitting technique adopted is thus appropriate.
Table 8 shows that 4 of the 9 mispricings are statistically insignicant in the pre-GFC era. Excluding the medium-size
and the big, low-BM rm groups, the mispricings post-crisis are lower in relation to the pre-GFC pricing errors. The average
absolute pricing errors are, respectively, 6.69% and 5.79% (per annum) pre and post-GFC. Also, aside from the small and
low-BM group, the shift between the pre and post-GFC mispricing is statistically insignicant. The pricing errors suggest
that the regional Fama-French model does not price African equities better in the post-GFC relative to the pre-GFC era. This
G Model
Table 8
Fama-French regression on size and book-to-market sorted portfolios for pre- and post-GFC eras.
Size Book-to-market
a a*
Small 0.0105(2.3) 0.0093(2.3) 0.0096(2.4) 0.0179(2.0) 0.0049(-0.6) 0.0020(0.3)
Medium 0.0006(0.2) 0.0043(1.4) 0.0073(2.0) 0.0031(0.6) 0.0025(0.4) 0.0007(0.1)
Big 0.0020(1.0) 0.0050(1.9) 0.0016(0.5) 0.0025(0.7) 0.0034(-0.7) 0.0010(0.2)
MKTRt rf , SMB and HML are correspondingly the excess return to the market portfolio, the difference between the small and big rm portfolio returns,
and a portfolio of high-BM minus a portfolio of low-BM stocks. MKTR is estimated as the value weighted average of the market returns of all the studied
markets. SMB and HML were formed by dual-sorting on market capitalisation and tri-sorting on BM. rf is the South African 3-months Treasury bill rates
expressed in monthly rates. All portfolios were constructed at June end and balanced yearly. All returns are expressed in US dollars. Size was estimated as
the market capitalisation of an equity divided by the cross-sectional standard deviation of market capitalisation of equities on the securitys exchange on
the portfolio formation date. Similarly, book value is equitys scal year end book value scaled by the cross-sectional standard deviation of scal year ends
book values of a securitys market for a given year. Book-to-market was estimated at December end each year. T-statistics are in parenthesis. GRS is the

F-test of Gibbons et al. (1989).
rit rft = ai + 2 ai + 1 bi MKTRt rft + ci SMBt + di HMLt + 2 bi MKTRt rft + ci SMBt + di HMLt + it, 2 = 1 1 , 2 = 11 , 1 is an indicator
variable that takes te value of 0 for the post October 2008 period and 1 otherwise.
suggests that the GFC has not led to regionally integrated asset pricing on the ASMs. Thus, the relevance of regional factors
in pricing assets on the ASMs has not changed post-GFC. The evidence may be due to limited power of the test or a lower
level of economic engagement amongst the African countries. The R2 of the test are quite high, indicating that the test has
power. This paper argues that the results are driven by the low level of economic activity amongst the African countries.
Bekaert and Harvey (1995), Bekaert et al. (2011) and Eiling et al. (2012) observe that economic integration impacts on the
degree of nancial markets segmentation.
Taken together, the mispricings in Tables 4 and 8 suggest some degree of parameter instability over the sample period.4
The average absolute mispricing of 9.24% (full sample, Table 4), 6.69% (Pre-GFC, Table 8) and 5.79% (Post-GFC, Table 8) per
annum are dissimilar. The results are largely driven by the big and medium-size rm groups. The spread between the full
sample and the pre-GFC period average absolute mispricings for the big, medium and small-rm groups are, respectively,
0.31%, 0.32%, and 0.01% per month. Similarly, the difference between the full sample and the post-GFC era average absolute
pricing errors for the big, medium and small-rm portfolios are 0.38%, 0.15%, and 0.32% per month, respectively. The evidence
in Table 8 should thus be cautiously interpreted. The parameter instability may be driven by various global and country-
specic economic episodes over the duration of the study. It may also be an indication of time variation in the ASMs regional
degree of integration, which is consistent with Bekaert and Harvey (1995). This is supported by the pre and post-crisis average
absolute mispricing difference.
For brevity, the full results of a test of the impact of the GFC using ABS and RM formed risk factors and test assets are not
presented. The results are consistent with the evidence in Table 8.
5.7. Robustness to the South African and ex-South African markets
We test whether the performance of the Fama-French model is different for the SAM relative to the ex-SAM, using Eq.
(5). We investigate if the pooled results were pervasive or driven by the dominant SAM. Arguably, the SAM is more efcient
than the ex-SAM. It is therefore likely that the results for the SAM may differ from that of the ex-SAM.

rit rft = ai + ai (Dxsa ) + bi MKTRt rft Dsa + bi MKTRt rft Dxsa
+ci SMBt (Dsa ) + ci SMBt (Dxsa ) + di HMLt (Dsa ) + di HMLt (Dxsa ) + it (5)
Dxsa = dummy variable that takes the value of 1 for the exSAM and 0 otherwise,
Dsa = dummy variable that takes the value of 1 for the SAM and 0 otherwise,
For brevity, we present only the pricing errors from the model based on size and BM of an equity that are estimated
relative to their cross-sectional standard deviations for the given market in Table 9. The results show that the pricing errors
for the SAM are higher than those of the ex-SAM. In seven out of nine cases, the pricing errors of the SAM exceed those of the
ex-SAM. This may indicate that the model explains the test portfolios return for the ex-SAM better than the SAM. Table 9,
however, shows that the difference between the pricing errors for the SAM and the ex-SAM is statistically signicant in only
4
We thank an anonymous reviewer for drawing our attention to the parameter stability issues.
G Model
Table 9
Fama-French regression on size and book-to-market sorted portfolios for the South African and ex-South African Markets.
Size Book-to-market
a a*
Small 0.0083(1.7) 0.0097(2.0) 0.0115(3.1) 0.0044(0.6) 0.0031(0.5) 0.0017(0.3)
Medium 0.0023(0.8) 0.0010(0.3) 0.0073(2.2) 0.0021(0.5) 0.0103(2.4) 0.0030(0.7)
Big 0.0061(2.6) 0.0060(2.0) 0.0030(0.7) 0.0015(0.5) 0.0056(1.4) 0.0038(0.7)
This table presents the results of estimating the model below for nine test portfolios (ri ) formed at the intersection of 3-size and 3-BM sorted portfolios.
MKTRt rf , SMB and HML are correspondingly the excess return to the market portfolio, the difference between the small and big rm portfolio returns,
and a portfolio of high-BM minus a portfolio of low-BM stocks. MKTR for South Africa is the South African FTSE/All Share Index whilst the MKTR for the
ex-South African markets is the value weighted average of the market returns of the sampled ex-South African markets. SMB and HML were formed by
dual-sorting on market capitalisation and tri-sorting on BM. rf is the South African 3-months Treasury bill rates expressed in monthly rates. All portfolios
were constructed at June end and balanced yearly. All returns are expressed in US dollars. Size was estimated as the market capitalisation of an equity
divided by the cross-sectional standard deviation of the market capitalisation equities on the securitys exchange on the portfolio formation date. Similarly,
book value is equitys scal year end book value scaled by the cross-sectional standard deviation of scal year ends book values of a securitys market for
was estimated
a given year. Book-to-market at December
endeach year. T-statistics are in parenthesis. GRS is the F-test of Gibbons et al. (1989).
rit rft = ai + ai (Dxsa ) + bi MKTRt rft Dsa + bi MKTRt rft Dxsa + ci SMBt (Dsa ) + ci SMBt (Dxsa ) + di HMLt (Dsa ) + di HMLt (Dxsa ) + it .Dxsa = dummy vari-
able that takes the value of 1 for the ex SAM and 0 otherwise,
Dsa = dummy variable that takes the value of 1 for the SAM and 0 otherwise.
one out of nine cases. This suggests that the performance of the model for the SAM may not be different from that of the
ex-SAM. The results for the SAM and ex-SAM are generally consistent with those of the entire African sample. The average
absolute intercepts are, respectively, 0.77%, 0.61% and 0.59% for the ASM, SAM and ex-SAM models.
The average intercepts suggest that the models performance appears weaker for the full ASM relative to the SAM and
ex-SAM samples. This may signal segmented or partially segmented ASMs. In regionally integrated ASMs we expect that
African factors should fully explain African test assets. It may also be an indication that the ASM test is more powerful than
the SAM and ex-SAM tests. The standard errors for the SAM and ex-SAM models are higher than those of the ASM model.
This may conrm the small sample bias problem in the small individual ASM studies.
5.8. Effects of market liquidity5
The ASMs including the SAM are largely illiquid (see e.g., Appiah-Kusi & Menyah, 2003; Hearn & Piesse, 2010a; Hearn
& Piesse, 2013a, 2013b). Hearn and Piesse (2013a, 2013b) show that institutional factors and rm ownership structure
impact on illiquidity on the ASMs. For instance, State participation enhances liquidity, whereas, extended family networks
involvement raise illiquidity on the ASMs. The existing evidence shows that liquidity is a signicant driver of nancial market
integration. Garca-Herrero et al. (2009) argue that illiquidity constrained regional level integration of Asian nancial markets
post the Asian nancial crisis. Similarly, Hearn and Piesse (2010b) argue that illiquidity restricts the ability of investors to
exploit diversication opportunities, and also, undermines the price recovery process and creates informational inefciency.

rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + fi LMHt + it (6a)

rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + ei WMLt + fi LMHt + it (6b)
LMH t = portfolio that is long in less liquid stocks and short in more liquid stocks.
The question then arises as to the relevance of the observed illiquidity to the level of segmentation amongst the ASMs, and,
of course, the evidence in this paper. We examine these issues using Eq. 6. Equations (6a) and (6b), respectively, augment the
Fama-French and Carhart models with a liquidity factor (LMH). Liquidity is measured by the t-2 to t-12 cumulative turnover
ratio. We construct the LMH as the difference between the returns of a portfolio of low liquidity and a portfolio of high
liquidity stocks; this technique is similar to that of Hearn and Piesse (2010a), Lam and Tam (2011) and Liu (2006). The LMH
is formed from a 2-by-3 sort on size and liquidity. The LMH was created using breakpoints of 30% and 70%, with rms falling
within the bottom 30% and the top 30% of the liquidity proxy, respectively, classied as less liquid and more liquid.
The results from estimating Eq. (6a) are presented in Table A1 (Appendix). We do not present the results of Eq. (6b)
since it is generally consistent with that of Equation (6a). Panels A and B (Table A1) present the results for the absolute and
relative size and BM-based models. Panel A indicates an average R2 , average absolute mispricing, and GRS of 70.92%, 14.76%
(per annum) and 4.89, respectively. The evidence indicates that the liquidity-augmented Fama-French model does not fully
describe the test portfolio returns. Generally consistent with Hearn and Piesse (2010a), all of the LMH loadings (except one)
are statistically signicant. This suggests the relevance of liquidity in pricing assets on the ASMs. Comparing the results in the
5
We thank an anonymous reviewer for drawing our attention to the liquidity related problems.
G Model
panel with the corresponding results in Panel A (Table 6), the GRS suggests that the liquidity augmented model performed
better than the Fama-French model; however, the average absolute mispricing for the Fama-French model is lower than
that of the liquidity-augmented version of the model.
Panel B (Table A1) shows that the GRS, annual average pricing error, and average R2 are, respectively, 3.88, 9.48% and
70.41% for the relative-based liquidity-augmented Fama-French model. Unlike the absolute-based model, all of the liquidity
coefcients excluding one are statistically indistinguishable from zero. Compared with the corresponding Fama-French
model (Table 4), the GRS and the pricing errors suggest that the performance of the two relative-based models is about
equal. That is, augmenting the Fama-French model with a liquidity factor does not improve the performance of the model in
a signicant manner. This contradicts the results for the absolute-based models. An important question is, what drives the
signicant disparity between the performance of the absolute and relative-based liquidity-augmented models? A critical look
at the construction and the underlying assumptions of the absolute and relative size and BM-based models provides some
insights. The answer lies in the variations between the sampled markets characteristics and how the disparity is controlled.
Hearn (2013), for instance, nds signicant disparity in the disclosure quality across the ASMs; this has implications for the
liquidity of the ASMs. The variations in institutional quality expose the ASMs to varying degrees of illiquidity. How does this
disparity impacts on the results in Table A1?
The absolute measures approach implicitly assumes ASMs integration and does not control for the disparities in the
sampled markets characteristics, whereas the relative models approach assumes ASMs segmentation and thus controls for
the variations across the sampled markets. If the ASMs are not integrated, then the absolute measures are basically measuring
the variations in the characteristics of the sampled markets. We argue that the signicant LMH coefcients (for the absolute-
based model) is not evidence of a signicant role for liquidity in describing the pooled sample returns, but rather evidence
of a signicant variation in liquidity across the sampled markets. This validates our earlier concern, that pooling securities
across markets without controlling for the disparities in market characteristics would measure the variations across the
markets if the pooled markets are not fully integrated.
The relevance of the SMB and HML are not undone with the inclusion of the LMH for both the absolute and relative
versions of the models. We argue that this is the outcome of our screening technique which adequately controlled the
illiquidity problem on the ASMs. Additionally, our approach for estimating the relative size and BM adequately captures the
disparities in the characteristics of the sampled markets, and also distributes stocks in the various portfolios proportionately
across the sampled markets based on their respective contribution of stocks to the sample.
6. Conclusion
The existence of the size and BM effects has been investigated within an African continent-wide context via the Fama-
French and Carhart models. The GRS test indicates some degree of mispricing for both models. We observe no signicant
difference between the pricing errors for the pre-GFC and post-GFC periods. This implies that the GFC has not led to integrated
asset pricing on the ASMs. Also, the results show that the average absolute mispricing is highest for the pooled African sample,
followed by the South African sample, and then ex-South African markets. This may be an indication that the ASMs are less
regionally integrated. This is corroborated by the evidence that the models based on relative measures of size and BM
describe African equity returns better than their counterparts that are based on absolute measures of the same variables. If
the ASMs are regionally integrated, then the models based on absolute measures of size and BM should better capture the
returns to the test assets than their relative counterparts. The evidence in this paper suggests that asset pricing on the ASMs
is not regionally integrated. Regional factors may therefore be less relevant in pricing securities on the individual markets.
Also, there are potentially high diversication gains by forming a portfolio of African securities.
The R2 are generally high, though less than what Fama-French (1993, 1996) observe in the US market. Given the fact
that the cross-section of stocks employed in this study is less than what was used in Fama and French, this result is not
surprising. In general, the ndings indicate that the Fama-French and Carhart models partly describe the returns to the size
and BM-sorted portfolios on the African capital markets. The Carhart model, however, is associated with lower levels of
mispricing and a lower GRS statistic than the Fama-French model.
We observe that small rms outperform large rms, and that high-BM equity rms generate superior returns to low-BM
equity stocks. We note that the size and BM effects are independent of each other on the sampled markets. The results are
generally consistent with the evidence from the developed markets and other emerging markets. The markets investigated
are generally small and highly volatile, underdeveloped, and perceived to be partly segmented from the developed and
other emerging markets. The evidence therefore suggests that the size and BM effects may be pervasive in international
markets. Evidence is provided that the institutional differences between markets could impact on the results of cross-
country studies, and that the relevance of liquidity in describing pooled sample returns, for instance, could be seriously
questioned if institutional characteristics are uncontrolled and if the markets are somehow segmented.
G Model
Appendix A.
Table A1
Liquidity augmented Fama-French regression on size and book-to-market sorted portfolios.
Size Book-to-Market
Panel A: Absolute size and BM Panel B: Size and BM relative to the standard
deviation of a securitys market
Low Medium High Low Medium High Low Medium High Low Medium High
a R2 a R2
Small 0.0143*** 0.0500** 0.0149*** 0.4684 0.6109 0.6288 0.0020 0.0065* 0.0218** 0.5031 0.6285 0.6754
Medium 0.0064** 0.0015 0.0104*** 0.723 0.6698 0.7035 0.0055** 0.0105** 0.0114*** 0.7904 0.6145 0.6422
Big 0.0079*** 0.0025 0.0031 0.8839 0.8651 0.8293 0.0093*** 0.0036 0.0003 0.8825 0.8446 0.7558
GRS: a1=. . .. . .. . ..= a9 = 0 = 4.89*** a1=. . .. . ..= a9 = 0 = 3.88**
b b
Small 0.6675*** 0.9471*** 0.7525*** 0.8667*** 0.8821*** 1.1061***
Medium 0.7330*** 1.2404*** 0.7756*** 0.8751*** 1.0425*** 0.9529***
Big 0.9976*** 1.0451*** 0.9682*** 0.9393*** 1.0497*** 0.9568***
c c
Small 0.6275*** 0.0767 0.5777*** 0.5284*** 0.7101*** 0.9408***
Medium 0.2929*** 0.8595*** 0.2327*** 0.5075*** 0.5965*** 0.5127***
Big 0.1120*** 0.0198 0.0175 0.1524*** 0.0611 0.2232**
d d
Small 0.4814*** 0.6060 0.4043*** 0.0964 0.0789 0.0222
Medium 0.2121*** 0.1686 0.4185*** 0.2860*** 0.0018 0.2877***
Big 0.2949*** 0.0327 0.6977*** 0.3437*** 0.0379 0.7281***
f f
Small 0.3477*** 0.0221 0.3034*** 0.1382 0.0624 0.2865*
Medium 0.1536*** 0.3207*** 0.1520*** 0.0436 0.0012 0.0663
Big 0.0591** 0.0457* 0.0916*** 0.0393 0.0123 0.0024
MKTRt rf , SMB, HML and LMH are correspondingly the excess returns to the market portfolio, the difference between the small and big rm portfolio
returns, a portfolio of high-BM minus a portfolio of low-BM stocks and the difference between the less liquid and more liquid portfolio returns. MKTR is
estimated as the value weighted average of the market returns of all the studied markets. SMB and HML were formed by dual-sorting on market capitalisation
and tri-sorting on BM. LMH is formed from 2- size and 3-liquidity portfolios. rf is the South African 3-months Treasury bill rates expressed in monthly rates.
All portfolios were constructed at June end and balanced yearly. All returns are expressed in US dollars. Panel A presents the results when size and book
values are estimated in absolute terms. In panel B, size was estimated as the market capitalisation of an equity divided by the mean market capitalisation
of equities on its exchange on the portfolio formation date. Similarly, book value is equitys scal year end book value scaled by the cross-sectional mean of
values ofa securitys market for a given year. Book-to-market was estimated at December end each year. T-statistics are in parenthesis.
scal year ends book
rit rft = ai + bi MKTRt rft + ci SMBt + di HMLt + fi LMHt + it . *, **, and *** respectively indicates 10%, 5% and 1% signicance level. GRS is the F-test of
Gibbons et al. (1989).
References
Alagidede, P. (2010). Equity market integration in Africa. African Review of Economics and Finance, 1(2), 88119.
Ali, A., Hwang, L., & Trombley, M. A. (2003). Arbitrage risk and the book-to market anomaly. Journal of Financial Economics, 69(2), 355373.
Appiah-Kusi, J., & Menyah, K. (2003). Return predictability in african stock markets. Review of Financial Economics, 12(3), 247270.
Banz, R. W., & Breen, W. J. (1986). Sample-dependent results using accounting and market data: Some evidence. Journal of Finance, 41, 779793.
Banz, R. W. (1981). The relationship between return and market value of common stocks. Journal of Financial Economics, 9, 318.
Barry, C. B., Goldreyer, E., Lockwood, L., & Rodrigueza, M. (2002). Robustness of size and value effects in emerging equity markets, 19852000. Emerging
Markets Review, 3, 130.
Basiewicz, P. G., & Auret, C. J. (2010). Feasibility of the Fama and French three factor model in explaining returns on the JSE. Investment Analysts Journal, 71,
2010.
Beckers, S., Connor, G., & Curds, R. (1996). National versus global inuences on equity returns. Financial Analysts Journal, 52(2), 3139.
Bekaert, G., & Harvey, C. R. (1995). Time-Varying world market integration. Journal of Finance, 50(2), 403444.
Bekaert, G., & Harvey, C. R. (1997). Emerging equity market volatility. Journal of Financial Economics, 43, 2977.
Bekaert, G., Harvey, C. R., Lundblad, C. T., & Siegel, S. (2011). What segments equity markets? Review of Financial Studies, 24(12), 38413890.
Billio, M., & Pelizzon, L. (2003). Contagion and interdependence in stock markets: Have they been misdiagnosed? Journal of Economics and Business, 55(5),
405426.
Black, F. (1995). Estimating expected return. Financial Analysts Journal, 51(1), 168171.
Boamah, N. A., Loudon, G., & Watts, E. J. (2016). Structural breaks in the relative importance of country and industry factors in African stock returns. The
Quarterly Review of Economics and Finance, 63, 7988.
Bundoo, S. K. (2011). Asset price developments in an emerging stock market: The case of Mauritius, AERC research paper 219. African Economic Research
Consortium [Nairobi January 2011].
Caporale, G. M., Cipollini, A., & Spagnolo, N. (2005). Testing for contagion: A conditional correlation analysis. Journal of Empirical Finance, 12(3), 476489.
Carhart, M. M. (1997). On persistence in mutual fund performance. The Journal of Finance, 52(1), 5782.
Claessens, S., Dasgupta, S., & Glen, J. (1995). Return behaviour in emerging stock markets. The World Bank Economic Review, 9(1), 131151.
Collins, D., & Biekpe, N. (2003). Contagion and interdependence in african stock markets. The South African Journal of Economics, 71(1), 181194.
Daniel, K., Hirshleifer, D., & Subrahmanyam, A. (2001). Overcondence, arbitrage and equilibrium asset pricing. Journal of Finance, 56, 921965.
Dimson, E., & Marsh, P. R. (1983). The stability of UK risk measures and the problem of thin trading. Journal of Finance, 38(3), 753783.
G Model
Eiling, E., Gerard, B., & De Roon, F. A. (2012). Euro-Zone equity returns: Country versus industry effects. Review of Finance, 16(3), 755798.
Errunza, V., & Losq, E. (1985). International asset pricing under mild segmentation: Theory and test. Journal of Finance, 40(1), 105124.
Errunza, V., Losq, E., & Padmanabhan, P. (1992). Tests of integration, mild segmentation and segmentation hypotheses. Journal of Banking and Finance, 16,
949972.
Fama, E. F., & French, K. R. (1992). The cross-section of expected stock returns. Journal of Finance, 67, 427465.
Fama, E. F., & French, K. R. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics, 33(1), 356.
Fama, E. F., & French, K. R. (1996). Multifactor explanations of asset pricing anomalies. Journal of Finance, 51(1), 5584.
Fama, E. F., & French, K. R. (1998). Value versus growth: The international evidence. Journal of Finance, 53, 19752000.
Fama, E. F., & French, K. R. (2006). The value premium and the CAPM. Journal of Finance, 61(5), 21632185.
Fama, E. F., & French, K. R. (2012). Size, value, and momentum in international stock returns. Journal of Financial Economics, 105, 457472.
Garca-Herrero, A., Wooldridge, P., & Yang, D. Y. (2009). Why dont asians invest in asia? The determinants of cross-border portfolio holdings. Asian
Economic Papers, 8(3), 228246.
Gibbons, M. R., Ross, S. A., & Shanken, J. (1989). A test of the efciency of a given portfolio. Econometrica, 57, 11211151.
Grifn, J. M. (2002). Are the Fama and French factors global or country specic? Review of Financial Studies, 15(3), 783803.
Harvey, C. R. (1991). The world price of covariance risk. Journal of Finance, 46(1), 111157 [(Mar. 1991].
Haugen, R. A., & Baker, N. (1996). Commonality in the determinants of expected stock returns. Journal of Financial Economics, 41, 401439.
Hearn, B., & Piesse, J. (2010a). Modelling size and illiquidity in west african equity markets. Applied Financial Economics, 20(13), 10111030.
Hearn, B., & Piesse, J. (2010b). Barriers to the development of small stock markets: A case study of Swaziland and Mozambique. Journal of International
Development, 22(7), 10181037.
Hearn, B., & Piesse, J. (2013a). A reassessment of stock market integration in SADC: The determinants of liquidity and price discovery in Namibia. Applied
Financial Economics, 23(2), 123138.
Hearn, B., & Piesse, J. (2013b). Firm level governance and institutional determinants of liquidity: Evidence from Sub Saharan Africa. International Review of
Financial Analysis, 28, 93111.
Hearn, B., Piesse, J., & Strange, R. (2010). Market liquidity and stock size premia in emerging nancial markets: The implications for foreign investment.
International Business Review, 19(5), 489501.
Hearn, B. (2011). Modelling size and liquidity in north african industrial sectors. Emerging Markets Review, 12, 2146.
Hearn, B. (2013). The institutional determinants of IPO Firm prospectus length in a developing context. Research in International Business and Finance,
27(1), 5265.
Hearn, B. (2014). The political institutional and rm governance determinants of liquidity: Evidence from North Africa and the Arab Spring. Journal of
International Financial Markets. Institutions & Money, 31, 127158.
Heston, S. L., Rouwenhorst, K. G., & Wessels, R. E. (1995). The structure of international stock returns and the integration of capital markets. Journal of
Empirical Finance, 2(3), 173197.
Hou, K., Karolyi, G. A., & Kho, B. C. (2011). What factors drive global stock returns? Review of Financial Studies, 24(8), 25272574.
Jegadeesh, N., & Titman, S. (1993). Returns to buying winners and selling losers: Implications for stock market efciency. Journal of Finance, 48(1), 6591.
Kenny, C. J., & Moss, T. J. (1998). Stock markets in africa: Emerging lions or white elephants? World Development, 26(5), 829843.
Kodongo, O., & Ojah, K. (2011). Foreign exchange risk pricing and equity market segmentation in africa. Journal of Banking & Finance, 35, 22952310.
Korajczyk, R. A. (1996). A measure of stock market integration for developed and emerging markets. The World Bank Economic Review, 10(2), 267289.
Kothari, S. P., Shanken, J., & Sloan, R. G. (1995). Another look at the cross-section of expected stock returns. Journal of Finance, 50(1), 185224.
Lam, K. S., & Tam, L. H. (2011). Liquidity and asset pricing: Evidence from the Hong Kong stock market. Journal of Banking & Finance, 35(9), 22172230.
Levine, R., & Zervos, S. (1996). Stock markets, banks, and economic growth. The American Economic Review, 88(3), 537558.
Lewellen, J., Nagel, S., & Shanken, J. (2010). A skeptical appraisal of asset pricing tests. Journal of Financial Economics, 96(2), 175194.
Liu, W. (2006). A liquidity-augmented capital asset pricing model. Journal of Financial Economics, 82, 631671.
Loughran, T. (1997). Book-to-market across rm size exchange, and seasonality: Is there an effect? Journal of Financial and Quantitative Analysis, 32(3),
249268.
Merton, R. C. (1973). An intertemporal capital asset pricing model. Econometrica, 41, 867887.
Miles, D., & Timmerman, A. (1996). Variations in expected stock returns: Evidence on the pricing of equities of a cross-section of UK companies.
Econometrica, 63, 369382.
Mlambo, C., & Biekpe, N. (2005). Thin trading on african stock markets: Implication for market efciency testing. Investment Analyst Journal, 61, 2940.
Piesse, J., & Hearn, B. (2005). Regional integration of equity markets in Sub-Saharan Africa. South African Journal of Economics, 73(March (1)) [2005].
Reinganum, M. R. (1981). Misspecication of capital asset pricing: Empirical anomalies based on earnings yields and market values. Journal of Financial
Economics, 9, 1946.
Rosenberg, B., Reid, K., & Lanstein, R. (1985). Persuasive evidence of market inefciency. Journal of Portfolio Management, 11, 917.
Schotman, P. C., & Zalewska, A. (2006). Non-synchronous trading and testing for market integration in central european emerging markets. Journal of
Empirical Finance, 13, 462494.
Sewell, S. P., Stansell, S. R., Lee, I., & Below, S. D. (1996). Using chaos measures to examine international capital market integration. Applied Financial
Economics, 6(2), 91101.
Soumar, I., Amnounv, E. K., Diop, O., Mit, D., & Nsougan, Y. D. (2013). Applying the CAPM and the FamaFrench models to the BRVM stock market.
Applied Financial Economics, 23(4), 275285.
Sugimoto, K., Matsuki, T., & Yoshida, Y. (2014). The global nancial crisis: An analysis of the spillover effects on African stock markets. Emerging Markets
Review, 21, 201233.
Van Rensburg, P., & Robertson, M. (2003). Size, price-to- earnings and beta on the JSE securities exchange. Investment Analysts Journal, 58, 716.
Yarovaya, L., & Lau, M. C. K. (2016). Stock market comovements around the global nancial crisis: Evidence from the UK: BRICS and MIST markets.
Research in International Business and Finance, 37, 605619.

Journal of Economics and Business

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Journal of Economics and Business

Caricato da

Copyright:

Formati disponibili

G Model

JEB-5773; No. of Pages 16 ARTICLE IN PRESS

Contents lists available at ScienceDirect

Journal of Economics and Business

Regionally integrated asset pricing on the african stock

2. Related African literature

Stock market Number of stocks Mean (US$000) Mean book-to-market

Initial sample After ltering Marketcapitalisation Book value

Botswana 23 18 222763 57415 0.50

3. Cross-country pooling methodology

MKTRt = return to the market protfolio at time t,

BMi,t = BVi,t1 /MVi,t1 (3)

BM i,t1 = fiscal year end book value of stock in yearY t1 ,

BM i,t1 = market value of stock i in December of year Y y1 .

Rj,t = return to the market portfolio of market j at time ft.

5. Presentation and analysis of results2

5.2. Returns analysis

SMB HML MKTR MKTR-rf WML

Mean 0.0052 0.0120 0.0062 0.0051 0.0093

Panel A: Mean monthly returns

Panel B: Standard deviation of returns (%)

Low Medium High Average

Small 7.46 7.20 6.90 7.19

Size Book-to-market Book-to-market

Low Medium High Low Medium High

5.3. Results from estimating the Fama-French three-factor model3

5.4. Results from estimating the Carhart four-factor model

Size Book-to-market Book-to-market

Low Medium High Low Medium High

Low Medium High Low Medium High

5.5. Regional level integration of the African stock markets

Low Medium High Low Medium High

Small 0.4518 0.6134 0.7184 0.5673 0.544 0.6828

5.6. Impact of the Global Financial Crisis (GFC)

= return to the portfolio i at time t,

r ft = the risk free rate at time t, it = error term.

Low Medium High Low Medium High

5.7. Robustness to the South African and ex-South African markets

Low Medium High Low Medium High

5.8. Effects of market liquidity5

Potrebbero piacerti anche