Pricing Southern African Shares in the presence of Illiquidity: A Capital

Asset Pricing Model augmented by Size and Liquidity Premiums

Bruce Hearn
Department of Management
Kings College London
150 Stamford St, London SE1 9HN

Email: bruce.hearn@kcl.ac.uk

Jenifer Piesse*
Department of Management
Kings College London
150 Stamford St, London SE1 9HN

Email: jenifer.piesse@kcl.ac.uk
and University of Stellenbosch, South Africa
Summary:- This paper augments the Fama and French (1993) three-factor model Capital
Asset Pricing Model to take account of company size and liquidity levels. These additional
risks faced by investors have not been addressed in any formal way. The sample includes two
of the largest most developed African markets: the Johannesburg Stock Exchange and
Nairobi Stock Exchange and two of the smallest: Swaziland and Mozambique. The evidence
suggests that while size and liquidity are important valuation factors in the larger markets
these measures are less well defined and not significant in pricing models for the very small
markets.

Keywords: Sub-Saharan Africa, Southern Africa, Capital Asset Pricing Model,

Liquidity, Emerging Financial Markets

Corresponding author

1
Electronic copy available at: http://ssrn.com/abstract=1262956

An Augmented Capital Asset Pricing Model:

Liquidity and Stock Size in Emerging Financial Markets
Summary:- This paper augments the Fama and French (1993) three-factor model Capital
Asset Pricing Model to take account of company size and liquidity levels. These additional
risks faced by investors have not been addressed in any formal way. The sample includes two
of the largest most developed African markets: the Johannesburg Stock Exchange and
Nairobi Stock Exchange and two of the smallest: Swaziland and Mozambique. The evidence
suggests that while size and liquidity are important valuation factors in the larger markets
these measures are less well defined and not significant in pricing models for the very small
markets.
1.

Introduction

Numerous studies have examined the effectiveness of the original Capital Asset Pricing
Model of Sharpe (1964) and Lintner (1965) and most have found that for developing country
markets this is subject to considerable ambiguity. More recently, additional factors have been
proposed to provide a more reliable explanation of the cross section of average returns. These
include firm size, the book to market equity ratio, the price earnings ratio, the cash flow to
price ratio and the performance of the firm in terms of sales growth (see Shum and Tang
(2005) for a full review). A major innovation to asset pricing was proposed by Fama and
French (1993) by introducing a number of changes to previous research, including the set of
asset returns, the variables in the model and the estimation approach, choosing a time series
cross section method rather than simply a cross section dimension.
Tests of the CAPM on markets other than those in OECD countries are somewhat
limited. Shum and Tang (2006) test common risk factors in assessing returns in Asian stock
markets, using a sample of asset listed on the Hong Kong, Singapore and Taiwan Stock
Exchanges.

Their results confirm those of Fama and French for the US when using

contemporaneous market factors, but the augmented model that includes size and book-tomarket ratios reports no significant improvement over the traditional CAPM. Only with past
values of these variables is there any enhanced accuracy of asset pricing in these markets.

2
Electronic copy available at: http://ssrn.com/abstract=1262956

Drew and Veerarachavan (2003) test the Fama and French three-factor model on Hong Kong,
South Korea, Malaysia and the Philippines and find size and value effects can be identified in
these markets using a cross section approach. However, nothing of this kind has been done
for African markets.
This paper incorporates some aspects of the Fama and French method, that is, the time
series approach, and the inclusion of a firm size factor. But it is also the first to incorporate a
measure of illiquidity, following Liu (2006), in the specific context of emerging markets.
Liquidity is a major factor in explaining asset returns and a number of measures have been
suggested. These include the quantity of trades (Datar et al, 1998), the speed of trades (Liu,
2006) and the costs of trading (Amihud and Mendelson, 1986) or by the impact that a trade
has on price (Amihud, 2002; Pastor and Stambaugh, 2003). However, many of these aspects
are difficult to capture in emerging markets and this paper focuses on the fourth of these, the
price effect. The market wide illiquidity factor is constructed following the work of Amihud
(2002) and is based on intraday trading volumes and order flow that impacts stock prices.
The countries in the study are all in southern and eastern Africa and, with the exception of
South Africa, represent some of the most illiquid financial markets in the world. However,
they are also countries that have attracted some interest from international investors and some
multinationals, particularly those in the mining sector, for example, Anglo American, Anglo
Gold, and Anglo Ashanti and the financial sector, such as Old Mutual, Standard Bank,
Standard Chartered, Barclays, Socit General, and BNP Paribas. These companies dominate
the domestic markets and create a very uneven degree of liquidity.
The sample data are for South Africa, Kenya and Mozambique, with two factors
determining the time span of the analysis. The first is availability as data for Swaziland and
Mozambique are only available from 2002-2007, while there are data for 1991 to 2007 for
South Africa and Kenya.

Secondly, it is important to test the model across a range of

emerging markets. Therefore in this study of African markets we have chosen the most
advanced, the one with the potential for ensuring regional integration, a very small market
with a close link to a successful one and finally a market that is just emerging that will be the
focus of a programme of privatisation.
The paper proceeds as follows.

Section 2 describes the characteristics of these

markets, the source of the data and the complex construction of the illiquidity series. Section
3 defines the model and incorporates the illiquidity factor into the CAPM. This presents a
three-factor model with risk elements that is similar in construction to Fama and French (1993)
but augmented with a illiquidity measure proposed by Liu (2006). The results are in Section
4, discussed in total, by country and in two periods in the case of South Africa and Kenya.
The final section concludes.

2.

Data

b) Institutional Characteristics of Emerging Financial Markets

Four emerging markets are examined in this paper, and there are clear differences in their
institutional design, market capitalisation and level of development. The major characteristics
of these markets are summarised (see Piesse and Hearn (2005) for an extended discussion of
African stock markets):
i

South Africa.

The Johannesburg Stock Exchange (JSE) is the largest, the most

developed and best regulated market in Africa. The JSE adopted the order-driven electronic
trading platform used by the London Stock Exchange in 2002. Trading takes place daily and
the market has a pre-opening electronic call auction 8-25am and 9-00am and continuous
trading 9-00am to 4-00pm. Despite being classified as an emerging market there is
considerable institutional investor participation and ownership is high diversified, unlike any
other market in Africa. (Bloomberg LP, 2006). Settlement is through a central depository on

a rolling contractual basis of trade date plus five working days (T + 5) and is largely G30
compliant (STRATE website, 2007).
The South African market has experienced two distinct periods of transition during the
sample period. The first was 1990 to 1995 when the market was closed to foreign investors,
largely due to sanctions by the rest of the world. Also at this time domestic investors had to
comply with the National Partys prescribed assets regulation, which emphasised
investment in domestic equities rather than money or bond market instruments (Grandes and
Pinaud, 2004). The second followed the ending of apartheid in 1995 and the subsequent real
and financial market liberalisation that followed, including the opening up of markets to
foreign institutional investment, the move to electronic trading and the introduction of formal
legislation to ensure international levels of corporate governance. i Further revision of the
Kingly report in early 2000 has led to increased investor confidence and market development
although competitiveness has been hindered by volatility of the domestic currency and high
risk premiums that have a negative impact on overseas investors (Grandes and Pinaud, 2004).
This has also resulted in a loss of liquidity in the domestic market and the tendency for
primary listings to take place on overseas exchanges such as London and New York in
preference to the JSE.
ii

Kenya. The Nairobi Stock Exchange (NSE) is the largest market in the East African

Community (EAC) and is the only one open to foreign investors. ii The policy to enhance
competitiveness in the smaller financial markets relies on regional integration and the East
African centre is in Nairobi. The central depository for the EAC is based in the NSE
building. Trading takes place daily by a central electronic book entry system, and is limited
to the floor of the exchange between 10-00am and 12-00.

The market is dominated by

blockholders and smaller retail investors with free float percentages of shares available to the
public being typically low. Order flow to the market is precipitated through a small network

of licensed stock brokers and their regional affiliates where investors are required to establish
both a trading account with stockbroker as well as a separate individual account at the central
depository. The traders on the exchange floor that operate the trading workstations and input
orders to the electronically held system are representatives of the individual licensed brokers.
The dissemination of market sensitive announcements and real-time prices takes place
through an investor relations officer inside the exchange and this is then passed to the
financial press. Public releases of shares in the primary market and IPOs are managed
through local investment banks, with Capital Markets Authority responsible for regulation
and supervision. There is no formal corporate governance regime for this market, although
larger companies try to follow best practice in the Cadbury Report, adhering to at least some
of the core principles, such as disclosure of directors holdings. In a market dominated by the
informal sector, and where so few companies in the formal sector can afford the stringent
listings fees and ongoing regulatory costs, strictly following good governance is prohibitive
and a considerable deterrent to listing.
iii

Swaziland. The Swaziland stock exchange in Mbabane was incorporated in July 1990

and since inception has been hindered by both a lack of infrastructure and formal regulation.
Trading was initially conducted by a single broker until a second brokerage house was
established as an over-the-counter trader in1998. The Securities Markets Regulation Bill has
still to be ratified by parliament and not surprisingly, this has been a major barrier to the
growth and development of the market. However, in 2000 the Central Bank of Swaziland
extended the existing Financial Services Regulation to cover the stock market, and has also
provided some trading facilities within the Bank premises. To minimise costs, Central Bank
staff provide both trading and building maintenance duties. Trading takes place daily 1000am and 12-00 midday as a call auction where orders are pooled and then matched and
executed at the price that most accurately reflects all information available to market at time

of trade. Interesting, although liquidity is very low, Swaziland is part of an economic and
currency union with South Africa and therefore domestic currency premiums are directly
comparable to those of the rand.

Consequently several cross listings between the two

exchanges have taken place, particularly since the establishment of an asset management
trust, The African Alliance and South Africas Interneuron Asset Manager. Asset allocations
in favour of Swazi listed instruments in excess of those in South Africa and other regional
centres is very low, often less than 5% of portfolio inventories (African Alliance, 2006).
Ownership is highly concentrated and free float capitalisations available to the public are low
(Swaziland Stock Exchange, 2007). Settlement is by physical transfer of assets between the
holdings of local banks with sufficient capitalisation to act as market custodians.
iv

Mozambique. The Maputo Stock Exchange, the Bolsa de Valores de Maputo (BVM),

is a fledgling market with only one listed stock, the former national brewing company
Cervejas de Mocambique.

This listing followed a joint venture with South African

Breweries, which was part of the privatisation programme recommended by the international
financial institutions and development agencies. Despite its small size this market has been
well designed from inception, with assistance from the New York and Lisbon exchanges.
However, regulation is weak, particularly with respect to the protection of small and minority
investors. Trading is predominantly government treasury bills and bonds with maturities of
up to 3 years, as the exchange acts as a second outlet for debt issues outside the Central Bank
auctions. Despite a network of licensed brokers including recently privatised commercial
banks, trading activity in the single equity remains tightly focussed on providing the interbank
market for debt with an additional instrument to broaden holding portfolios. Trading is by a
delocalised electronic order book system where trades are matched on a price-bargain basis.
Licensed brokers have access to an online trade entry system run from the Stock Exchange
website. Levels of illiquidity are severe with trades taking place only once every several

months and the time from submission of an order to market execution can be 25 working
days. Official trading hours are 10-00am to 12-00 midday on Tuesday, Thursday and Friday
(BVM, 2007). Free float percentages are some of the lowest recorded and have fallen since
listing in 2001 from 6-5% to the current level of 4% (BVM, 2007). Ownership is highly
concentrated and settlement by local market custodians similar to Swaziland. Trading and
settlement activity for the single listed equity is dominated by the South African broker
Standard Bank.

This is a clear indication of the lack of investor confidence in local

institutions, which is not surprising given the 1994 financial crisis when two newly formed
commercial banks failed due to bad debt overhang from the previous socialist era.
All four markets have poor liquidity compared with developed world markets, but
particularly Swaziland and Mozambique, which are two of the smallest markets in Africa.
They both lack transparency, liquidity and fall far short of best practice in terms of corporate
governance. In Swaziland there is a complex state ownership structure of cross holdings
between various listed entities. In Mozambique the very small formal sector has only recently
adopted auditing techniques into national accounting practices (Standard Bank, 2007). They
are both a very long way from the highly regulated standards in South Africa and the growing
improvement in Kenya. These markets all present varying degrees of risk and illiquidity that
make the sample an excellent focus for a study of a risk-adjusted capital asset pricing model.

c) Data: Sources and Series Construction

Values of the daily total returns are from Datastream for each stock held within the
constituent list of the overall market indices for South Africa and Kenya.

These were

supplemented with daily stock price levels and trading volumes to generate liquidity factors.
These measures are used to sort stocks into portfolios, following Amihud (2002).

For

Swaziland and Mozambique, where total returns indices for stocks are not available through

local or international media sources, they have to be constructed from data collected directly
from the markets. Corporate actions, such as stock splits and rights issues, stock prices,
number of shares outstanding, dividend payments history, daily trading volumes and market
capitalisation data were also from the Stock Exchanges. These data were used to form
individual stocks total returns indices following the method used by Standard & Poors to
construct total returns indices. Companies that are delisted are not deleted from the sample
prior to their delisting to prevent survivorship bias.
All data series were converted to US$ in order to present the US investor perspective,
which removes the effects of high and volatile local currency premiums in the calculation of
excess returns. The exchange rate data is from Datastream, Bloomberg and the South African
Reserve Bank. The one-month US-Treasury Bill yield rate represents the risk free rate
although this is adjusted to take account of monthly excess returns as opposed to the quoted
equivalent annualised rates. The conversion of the total returns series and prices into US$
exchange rates and use of US-Treasury yield rate assumes long term parity between
individual domestic currencies and the US$. US-Treasury Bill yield data is from Federal
Reserve Bank.
A critical factor in the portfolio sorting is that all information is known in the year
preceding the annual stock sorting and portfolio rebalancing that is at end of December in
each year. The size factor is simply the value of a stocks market capitalisation in December
of each year, calculated from the product of the number of shares outstanding with the dollar
price per share in the case of Swaziland and Mozambique. In addition, since the Amihud
(2002) liquidity factor depends on the positive modulus of stock price returns in order to
assess dollar traded impact on price, it is necessary to use the absolute value of the returns.
Stock price returns are calculated on a daily basis and then divided by daily dollar trading
volumes and a mean of this factor for each month is calculated creating monthly values of the

Illiquidity factor for each stock. Because the markets in this sample include some of the most
illiquid stocks in the world and have highly variable illiquidity factor profiles, the mean of all
the monthly illiquidity values is taken to represent the annual aggregated average of the
illiquidity factor that can be used in the end of year portfolio sorting process. It is necessary
to use caution in the interpretation of the monthly time series of the Amihud (2002) illiquidity
factors for these highly illiquid markets, particularly Swaziland and Mozambique, but also
some of the individual stocks in Kenya and South Africa. These markets and stocks are
frequently so illiquid that trading only takes place once a quarter and because of this
exceptionally low frequency the calculation involved in generating this factor treats
chronically illiquid periods as periods of zero values. This illiquidity measure can reflect the
very low levels of this factor attained for highly liquid stocks, causing a false interpretation.
For each month t, each company j is ranked by the market value of equity at the end of
December. Firms are then classified into 3 portfolios based on market value, from the
smallest to the largest. For each size portfolio, stocks are further sorted into 3 separate
illiquidity ranked portfolios in accordance to their annualised generated illiquidity factor
values in ascending order.

Nine size-illiquidity portfolios are then constructed and are

rebalanced annually. The equally weighted monthly returns on portfolios are computed each
month from December to the following December. Repeating this procedure for every year
results in 199 equally weighted monthly returns from January 1991 to August 2007 with the
final portfolio rebalance taking place in August 2007 as opposed to December as in previous
years. In addition to these portfolios rebalanced and sorted to reflect size and illiquidity state
factors, four additional equally weighted portfolios are generated for stocks local to each
domestic component market within the overall sample, that is, country portfolios for South
Africa, Kenya, Swaziland and a single stock series for Mozambique. The market excess
returns variable is generated as the aggregate average returns each month across the market.

10

Shum and Tang (2005) form a market returns variable from both an equally weighted as well
as market capitalisation weighted average but in this paper the equally weighted average of
returns is adopted as the market portfolio. This is because the JSE dominates all of the
African equity markets and therefore a market capitalisation weighted portfolio would impose
a high level of bias that reflects the characteristics of South African stocks. Equally, other
methods commonly used in the literature to determine the market variable, such as a regional
investment index proxy, for example the Standard & Poors or MSCI range of indices, are
complicated by the lack of such benchmarks for Sub Saharan Africa.
The monthly size factor SMB (small minus big) is the difference between the average
returns on the three small stock portfolios and the average returns on the three big stock
portfolios. The monthly liquidity factor (ILLIQ) is the difference between the average returns
on the three high-illiquidity portfolios and the average returns on the three low-illiquidity
portfolios.

3.

Model

Intuitively, investors in small emerging markets with low levels of development may be
attracted to large, well-known companies rather than smaller ones as these are considered
safer investment opportunities with more reliable dividend payouts. These larger blue-chip
companies may either be the better domestic parastatals and former state owned enterprises
that have been privatised or large privately owned companies or multinationals. All appear to
represent profitable investments, partly because of investor confidence that they will comply
with corporate governance standards that smaller companies would find more costly to
implement. In addition, it is well documented that investors implicitly price a liquidity
premium into valuations and expected returns, although the literature documenting methods
of measurement of liquidity premiums remains scarce.

11

Although a number of variables have been constructed to capture or effectively proxy

liquidity in the recent literature each has its own shortcomings depending on what
fundamental trading statistics are used to assess liquidity. Some originate from an analysis of
market micro-structural, some price determination and others order flow. The illiquidity
measure originally proposed by Amihud (2002) has been used successfully by Martinez et al
(2005) to analyse liquidity premiums in pricing models applied to the Spanish stock market.
This method is used here. The measure captures the price impact as the response associated
with one US$ of trading volume. In particular, illiquidity for a given stock on a given day is
the ratio of the absolute value of the percentage price change per US$ of trading volume.
This resembles similar measures developed from a market trading volume order flow
perspective and states that the illiquidity of stock j in month t is
ILLIQ jt

1
D jt

D jt

R jdt

V
d =1

(1)

jdt

where Rjdt and Vjdt are the return and US$ trading volume on day d in month t and Djt is the
number of days with observations in month t of stock j. If a particular stock has a high
ILLIQjt this indicates that the price moves a lot in response to trading volume and therefore
the stock is considered illiquid. The market-wide cross sectional liquidity risk factor is
simply an aggregation of this measure across all stocks expressed
ILLIQt

1
Nt

Nt

ILLIQ
j =1

jt

(2)

where Nt is number of stocks available in month t in the sample.

Martinez et al (2005) state that when this factor increases it should interpreted as an
adverse shock to aggregate liquidity. Stocks that tend to pay lower average returns when this
measure increases (negative betas relative to this factor) do not provide desirable hedging
behaviour to investors and therefore extra compensation is required to those holding these
stocks. This implies that the premium associated with this liquidity factor in a cross section
12

should be negative.

Shum and Tang (2005) cite previous work documenting that smaller

market value portfolios have been found to produce higher average returns.
Following this reasoning the three factor model of Fama and French (1993) to capture
CAPM average-return anomalies can be adjusted to apply to emerging markets. In this paper
the specific characteristic of these markets is the high level of illiquidity and this should be
incorporated in an emerging market asset pricing model. Thus, in addition to the market
excess returns the model is augmented by the size factor excess returns, SMB, and the excess
returns attributed to the Illiquidity factor, ILLIQ. This restates the three factor CAPM as the
expected return on a risky portfolio p, in excess of the risk free rate E(Rp) Rf is a function of
(i) the excess return on the market portfolio, Rm Rf, (ii) the difference between the return on
a portfolio of small-size stocks and the return on a portfolio of large-size stocks, SMB (small
minus big); and (iii) the difference between the return on a portfolio of high illiquidity stocks
and the return on a portfolio of low illiquidity stocks, ILLIQ. Therefore the expected excess
returns on a portfolio p of emerging market stocks can be written as
E(R p ) - R f = p [E(R m ) - R f ] + Sp E(SMB) + Sp E(ILLIQ)

(3)

The equilibrium relation of Fama and French (1993) three factor model is stated in
terms of expected returns. In order to test the model with historical data, it is necessary to
transform (3) to the following estimating equation
R pt - R ft = p + p (R mt - R ft ) + SpSMB t + H p ILLIQ t + pt

(4)

where the variables are described above and p, t is an iid disturbance term. The factor
sensitivities or loadings, p , Sp , Hp are the slope coefficients in the time series regression. In
addition to the nine time series regressions for each size-illiquidity portfolio within both
sample periods: 1991 to 2001 and 2002 to 2007, pooled regressions are run for individual
aggregate country portfolios representing the four markets.

13

Prior to estimation, time series diagnostic tests check for autocorrelation and
heteroskedasticity were run given the sensitivity of the disturbance terms to normality
assumptions in the distribution properties of the data. Tests for heteroskedasticity using the
White test (White, 1980) and the Durbin-Watson test (Durbin and Watson, 1950 and 1951)
for autocorrelation found significant heteroskedasticity and autocorrelation. These test results
are not reported here but suggest the t-test in the OLS regressions are unreliable and Newey
and West (1987) methods were used and the tests repeated. It should noted that this adjusts
the only the standard errors and not the regression estimates.

4.

Results and Discussion

a) Preliminary Data Analysis

The descriptive statistics for all nine size-illiquidity factor sorted portfolios and the zero-cost
SMB and ILLIQ portfolios are in Table 1. In general, the average returns means increase
considerably from small to large size stock portfolios. This is also reflected in the measure of
volatility, where standard deviations increase dramatically from larger size firm to smaller
size firm portfolios. Average returns in small size stock portfolios tend to be more risky than
in larger stock portfolios, measured by higher standard deviations, but also have higher
potential returns, that is, higher means. Although there is little discernable difference between
the cross section of low to high liquidity portfolio means, there is an increase in volatility and
standard deviations from low illiquidity to high illiquidity stock portfolios. Even on a less
liquid market this result is expected in terms of the impacts of sudden erratic order flow on
stock prices that reflect significant adjustments in value where there is occasional trading
activity. It is harder to interpret the coefficient of variation as the average return means are
close to zero, which makes the value very large. However, the coefficient of variation tends

14

to be larger for larger size stock portfolios than small size stock portfolios, confirming the
results for South East Asian markets (Shum and Tang, 2005).
Table 1
Table 2 presents results on the annual average number of stocks in the monthly
portfolios for the nine size-illiquidity constructed portfolios. The relative thinness of the
markets in this study is highlighted by the small number of companies in these portfolios
despite the universe of companies selected from each individual market All Share Index. This
is a significant issue when interpreting results in the context of emerging markets compared
with the numerous studies of the CAPM type pricing models using developed country data.
Panel 1 of Table 2 shows the average number of non-South African stocks in each of the 9
size-illiquidity sorted portfolios. Given both the small size of Kenyas formal sector and the
small size of the companies relative to those on the JSE, the majority of Kenyan stocks fall
within the small size portfolios where they are evenly distributed with respect to illiquidity.
However the Kenyan stocks that are included in larger size portfolios are concentrated in the
higher illiquidity portfolios, with few if any in lower illiquidity portfolios.
Table 2
The results for Swaziland and Mozambique should be interpreted with considerable
caution given the earlier discussion of the difficulties in constructing a suitable illiquidity
measure for such small markets.

Not surprisingly, the single equity in Mozambique is

concentrated in the high illiquidity portfolio of the medium size group, while the results for
Swaziland are ambiguous with a relatively even spread of stocks across all illiquidity levels
despite the minimal trading activity on this exchange compared.

b) Illiquidity Factors

15

Figure 1 shows the cross sectional market aggregate average illiquidity factors. Again caution
is necessary in interpreting these data as market-wide indicators of liquidity because of the
sample bias that results from the simple equally weighted average of individual stocks
illiquidity. These illiquidity distributions are highly skewed given some stocks are the most
illiquid in the world. However, they do highlight the variance in the liquidity profiles of the
markets and reflect the differences in institutional, regulatory and macroeconomic
environments that exist in this group of emerging markets.
Figure 1
All markets are characterised by very large spikes in illiquidity and this profile is
characterised in the extreme cases of Swaziland and Mozambique where the illiquidity profile
a function of a zero or near-zero trading activity. A small period of illiquidity for South Africa
during the early 1990s is attributed to the closed nature of these markets immediately prior to
the end of National Party control and the establishment of democratic rule. The later period of
significant illiquidity around the beginning of 2000 reflects the general downturn in
developed country financial markets that led fund managers to transfer holdings out of
emerging markets to less risky investment, having recently experienced the 1997 Asian
currency crisis, the 1998 Russian debt crisis, and the 2000/2001 depreciation of the Rand.
Quite difference factors influenced the market in Kenya. The period of high illiquidity prior
to 1994 ended with the exchange rate liberalisation and the ease of access to foreign investors
that was encouraged by the introduction of tighter market regulation and better protection of
minority shareholders. The gradual increase in illiquidity again from late the 1990s to 2003
can be attributed to the general loss of value and stagnation of the stock exchange in Kenya
during this period.

c) Early sample period: 1991 to 2001

16

Table 3 reports the results from the grouped pooled regression. As expected, the Jensen
alpha, p, is significantly different from zero is about 50% of the cases. All the low illiquidity
portfolios, the large size medium illiquidity and the small high illiquidity portfolios indicate
hidden unrealised value in this investment segment in relation to the rest of the market. The
coefficients on portfolio excess returns against both the market excess return (p) and the size
factor-mimicking portfolio (Sp) are large and significant in almost all cases. The coefficients
on the illiquidity factor (Hp) are smaller in the majority of cases but are nonetheless
significantly different from zero. The illiquidity factor mimicking portfolio also tended to be
larger with respect to smaller portfolios than larger ones providing further evidence of the
illiquidity of small company stocks.

The increased explanatory power of these models

illustrates the importance of the augmented CAPM, in this case, specifically designed for
these highly illiquid markets, compared with the original linear model of Sharpe (1964) and
Lintner (1965). In the Table, the first adjusted R2 (ADJ R2 (1)) is the result from regressing
the expected return on risky portfolio p, in excess of the risk free rate E(Rp) Rf as a function
of the excess return on the market portfolio, Rm Rf,. The second adjusted R2 (ADJ R2 (3)) is
the result from regressing the size and illiquidity augmented three-factor model. In all size
and illiquidity groups there is substantial improvement, in many cases by more than 100%.
Table 3

d) Latter sample period: 2002 to 2007

Although this period was smaller and had fewer data points than the initial period the results
are similar to the full period discussed above. The size and illiquidity factor-mimicking
portfolios have a more polarised effect on the significance of the coefficients and their relative
magnitude.

Across all models, risky stock portfolios excess returns series have large

coefficients on the market portfolio excess returns as well as the two augmented risk factor

17

portfolios. These results are in Table 4. The size of the coefficients on the illiquidity factor is
considerably greater for small size stock portfolios while those on the size mimicking
portfolios are generally monotonically increasing in values from the small size to large size
portfolios.

This is further reflected in the t-statistics, which indicate the value of the

coefficients is significantly different from zero.

The t-statistics also confirm the null

hypothesis that the illiquidity factor mimicking portfolio results are an important component
in asset pricing in the high and low illiquidity portfolios but not for only for small sized
medium illiquidity portfolios.
Table 4

e) Individual country average returns

i

South Africa. The difference in the estimated coefficients between the two periods

reported in Table 5 shows how the combined effects of market liberalisation and institutional
development have changed the efficiency of this market. Jensens indicates the excess
returns of a risky portfolio, in this case the nine size-illiquidity portfolios, over the required
rate of return designated by the portfolios beta to the market excess returns, or risk premium.
This measure is significantly different from zero in the first period and while diminishing by
the second in two of the four models indicating that the models explanatory power has
increased. In the first period the original CAPM does not adequately price the assets in these
portfolios, the size effect is a useful adjustment but including both size and illiquidity results
in a better-performing model. The importance of the illiquidity factor has diminished by the
second sample period indicated by lower betas and reduced t-statistic values. Equally the
explanatory power of model has improved by second period reflected in substantially higher
adjusted R-squared terms. The size of the risk premium has increased by the second period as
this market is more highly regulated and has liberalised allowing the active participation of

18

foreign investors. The size effect has increased due to greater market concentration through
merger and acquisition activity but inclusion of the illiquidity factor has improved the
explanatory power of the model. The illiquidity and size covariance risk premiums for South
Africa are represented in Figure 2 for comparison. Owing to the lack of significance of these
premiums over the course of the second period the results are shown only for the first period:
1991 to 2001.
Table 5
ii

Kenya. The differences between the periods for Kenya in Table 6 reflect those of

South Africa, with an improved regulatory regime and increased openness, although these
changes are based in an overall African policy framework for increased market integration
rather than the move to a democratic government. The increase in levels of market efficiency
has reduced Jensens and a less tightly managed monetary policy has lowered the average
risk premium by the second period. In both periods the size factor is important but more in
the second, but the illiquidity factor improved the fit of the model overall. Although there is a
slight drop overall of Adj R2 in the second period there is a marked increase in the illiquidity
coefficient.

The coefficient on the size factors increased as did the level of statistical

significance. However in the second period the combined three factor model overall exhibits
little dependence on the illiquidity factor in contrast to the size and market factors. This is a
change from the first period where all three explanatory factors were important and statistical
significance. Figure 2 outlines the size and illiquidity covariance risk in basis points for
Kenya for the initial period of 1991 to 2001 owing to the lack of significance of these factors
during second period. The magnitude of the premiums in Kenya is much greater than those of
South Africa and this is especially visible in the scale of the size and liquidity premiums. The
magnitude of these premiums in Kenya in contrast to South Africa can be explained by Kenya
being a smaller market where these company characteristics would be taken into account

19

within valuation more than for South Africa. Equally despite Kenyas small size it does not
suffer from the severity of measurement issues exhibited in the very small markets of
Swaziland and Mozambique and as such premiums relating to size and liquidity are both
measurable and significant in valuation.
Table 6
iii

Swaziland. Not surprisingly, the CAPM is any form does not perform well in pricing

assets in this market, as is shown in Table 7, Panel 1. The adjusted R2 terms are consistently
below 5% in all models and the size and illiquidity factors are not statistically different from
zero. The risk premium is high and significant but neither stock size nor the high illiquidity
that exists are important in the model. This market is a very long way below the required
level of activity to ensure price discovery and investors have reasons to be wary. However
the lack of a significant liquidity factor in the presence of such severe illiquidity is a serious
issue as rationale investors would likely demand compensation for liquidity risk in such a
small market. The lack of suitable valuation tools to cope with illiquidity would further
compound problems in attracting overseas investment needed for development.
Table 7
iv

Mozambique.

A single listed stock is unlikely to perform well and this one clearly

does not, see Table 7, Panel 2. While the privatisation programme is focused on equity
markets it may be difficult for Mozambique to provide a successful environment for
conventional investment through its stock exchange particularly as investors will face
significant hurdles in valuation. Investors subject to rational expectations are unlikely to
invest in such a market.
Figure 2

5.

Conclusion

20

This paper proposes a size and liquidity-augmented capital asset pricing model specifically
focussing on emerging markets. Four African markets are used, the well regulated and active
JSE, the NSE as the regional centre of market integration in the East Africa Community and
two fledgling markets, Swaziland and Mozambique. Illiquidity series were constructed on a
time series cross section basis and incorporated into the Fama French (1993) risk adjusted
CAPM.
Results show this model is superior to both the Sharpe/Linter and the Fama and
French models as illiquidity is both a priced and consistent characteristic in the two larger
emerging markets. The two small markets demonstrate markedly different characteristics and
owing to the liquidity factor only being poorly defined due to the severity of illiquidity both
the liquidity and size premiums lack significance in valuation models.

The differences

between the periods for South Africa and Kenya reflect political and economic events that
influence markets. However, the use of such models for Swaziland and Mozambique when
analysed separately are less positive. The illiquidity in these markets is too extreme that any
form of CAPM fails to predict excess returns with any degree of confidence. This evidence
suggests that while the firm size factor is as important in pricing assets as in developed
markets the major risk component in emerging markets is illiquidity.

21

References
African Alliance (2006). "Request for Swaziland Fund Holdings." E-mail correspondence to
Nozipho Dlamini. 2 October 2007.
Amihud, Y. (2002). Illiquidity and stock returns: cross section and time series effects. Journal
of Financial Markets, 5, 31-56.
Amihud, Y., & Mendelson, H. (1986). Asset pricing and the bid ask spread. Journal of
Financial Markets, 5, 31-56.
Bloomberg L.P. (2006). Institutional investor stock holdings for South Africa 1/1/90 to
1/12/06. Retrieved December 2, 2006 from Bloomberg terminal.
BVM (2007). Boletim de Cotacoes. Bolsa de Valores de Maputo, Mozambique
Datar, V., Naik, N., & Radcliffe, R. (1998). Liquidity and asset returns: an alternative test.
Journal of Financial Markets, 1, 203-220.
Drew, M. E., & Veerarachavan, M. (2003). Beta, firm size, book to market equity and stock
returns in the Asian region. Journal of the Asia Pacific Economy, 8, 354-379.
Durbin, J., & Watson, G. S. (1950). Testing for serial correlation in least squares regression: I.
Biometrika, 37, 409-428
Durbin, J., & Watson, G. S. (1951). Testing for serial correlation in least squares regression:
II. Biometrika, 38, 159-177
Durbin, J., & Watson, G. S. (1951). Corrections to part I: Testing for serial correlation in least
squares regression: I. Biometrika, 38, 177-178
Fama E and French K (1993). Common risk factors in the returns on stocks and bonds,
Journal of Financial Economics, 33, 3-56
Grandes, M., & Pinaud, N. (2005). Which policies can reduce the cost of capital in Southern
Africa. OECD Development Centre Policy Brief, 25, 2-27.
Liu, W. (2006). A Liquidity-augmented capital asset pricing model. Journal of Financial
Economics, 82, 631-671.
Lintner, J. (1965). The valuation of risk assets and the selection of risky investments in stock
portfolios and capital budgets. Review of Economics and Statistics, 17, 13-37.
Martinez, M. A., Nieto, B., Rubio, G., & Tapia, M. (2005). Asset pricing and systematic
liquidity risk: An empirical investigation of the Spanish stock market. International
Review of Economics and Finance, 14, 81-103.
Newey, W. K., & West, K. D. (1987). A simple, positive semi-definite, heteroskedasticity and
autocorrelation-consistent covariance matrix, Econometrica, 55, 703-708.

22

Pastor, L., & Stambaugh, R. (2003). Liquidity risk and expected stock returns.
Journal of Political Economy, 111, 642-685.
Piesse, J., & Hearn, B. (2005). Regional Integration of Equity Markets in Sub-Saharan Africa.
South African Journal of Economics, 73(1), 36-52.
Sharpe, W. F. (1964). Capital asset prices: A theory of market equilibrium under conditions of
risk. Journal of Finance, 19, 425-442.
Shum, W. C., & Tang, G. Y. N. (2005). Common risk factors in returns in Asian emerging
stock markets. International Business Review, 14, 695-717.
Standard Bank (2006). Standard Bank Research Economics: Mozambique. By Yvonne
Mhango, Standard Bank Johannesburg. January 2006
White, H. (1980). A heteroskedasticity-consistent covariance matrix estimator and a direct
test for heteroskedasticity. Econometrica, 48, 817-838.

23

Table 1: Summary statistics for equally weighted monthly excess returns on 9 portfolios formed on size and illiquidity for two sample periods,
1991 to 2001 and 2002 to 2007, and entire sample period 1991 to 2007
Illiquidity
Zero-cost Portfolios
High
Medium
Low
High
Medium
Low
High
Medium
Low
SMB
ILLIQ
Mean
Standard Deviation (SD)
Coefficient of Variation (CV)
Panel A: Summary Statistics for sample period: 1991 2001
Big
0.0012
-0.0029
-0.003
0.022
0.01
0.012
18.66
-3.01
-4.05
-0.0106
0.0062
Mean
Medium
0.0124
0.0022
-0.004
0.065
0.04
0.012
5.31
18.55
-3.08
0.058
0.047
SD
Small
0.0099
0.0054
0.012
0.0819
0.06
0.079
8.28
11.62
6.75
-5.45
7.63
CV
Panel B: Summary Statistics for sample period: 2002 2007
Big
0.0046
0.0006
-0.0012
0.014
0.013
0.015
3.17
22.11
-12.31
-0.0118
0.0102
Mean
Medium
0.0146
0.0091
0.0074
0.028
0.019
0.016
1.91
2.08
2.15
0.0218
0.019
SD
Small
0.0234
0.0102
0.0057
0.041
0.032
0.017
1.76
3.13
2.99
-1.85
1.88
CV
Panel C: Summary Statistics for overall period: 1991 2007
Big
0.0023
-0.0017
-0.0024
0.0196
0.011
0.013
8.44
-6.23
-5.54
-0.0110
0.0076
Mean
Medium
0.0131
0.0045
-0.0001
0.0557
0.035
0.014
4.24
7.67
-124.36
0.0484
0.0401
SD
Small
0.0145
0.0070
0.0097
0.0708
0.054
0.065
4.89
7.71
6.72
-4.407
5.291
CV
Notes:
For each year, t, every company is ranked by its market capitalisation of equity and the end of December in year t. Stocks are then classified into 3 portfolios based on
market value, from the smallest to the largest. For each size portfolio, stocks are further sorted into 3 Illiquidity portfolios based on individual stocks Illiquidity ranking in
ascending order. Nine size-illiquidity are so formed and rebalanced annually. The equally weighted monthly returns on portfolios are computed each month from January to
the following December. Repeating this procedure for every year results in an overall sample set of 199 equally weighted monthly returns from January 1991 to December
2006 with values in 2007 being rebalanced in August 2007. Two sample periods are chosen, namely 1991 to 2001 and 2002 to 2007 and an overall period of 1991 to 2007.
Data from Mozambique and Swaziland are available only within the more recent sample period while the initial period is exclusively Kenya and South Africa. Additionally
for each sample time period two zero cost portfolios, SMB(ILLIQ) representing long small size (high illiquidity) portfolios and short large size (low illiquidity) portfolios.
Size

24

Table 2: Average number of stocks in each of the 9 size-illiquidity portfolios by year in period: 1991-2007
Year*

Portfolio Portfolio Portfolio Portfolio

Portfolio Portfolio
Portfolio Portfolio Portfolio
1
2
3
4
5
6
7
8
9
B/H
B/M
B/L
M/H
M/M
M/L
S/H
S/M
S/L
1991
18
19
19.50
19.25
19
18.92
19
19.08
20.58
1992
19
19.17
22
21
19
18
20
20
23.67
1993
20
21
21.67
21
20.50
21
22
22
21
1994
19.25
21
22.42
20.17
21
24
21
20.25
26.17
1995
21.83
23
24
21
19.25
17.92
25
25
25
1996
22.42
24.17
25.83
23.83
24.08
23.75
26
23
23
1997
23.92
25.50
28.67
23.75
26.50
23.67
27.17
28
24.17
1998
29.25
30.83
31.17
30.83
30
24.67
30.50
29.17
26.67
1999
31.75
31.75
36.50
33.08
32.25
33.25
36.58
36.33
29.50
2000
34.42
35
35.50
35.42
36.58
34.75
35.92
35.17
32
2001
36
35.75
36.17
36.08
36
34.67
37
35.08
35
2002
36
37
37
36.92
36
36.83
37.50
38
37
2003
36
38
36
38
38
35
38
38
35.58
2004
37
36
37
37
37
37
35.83
37
40
37
38
35
37
37
40
2005
36
36
38
2006
36
36
38
37
36
37
36
36
45
2007
36
36
38
37
36
37.67
35.33
36
45
Panel 1: Average number of non-South African stocks in average annual portfolios
B/H
B/M
B/L
M/H
M/M
M/L
S/H
S/M
S/L
1991
0
0
0
9
0
1
12
11
11
1992
0
0
0
12
2
0
13
6
12
1993
1
0
0
14
1
0
14
8
11
1994
2
0
0
13
2
1
14
10
8
1995
1
0
0
9
3
0
15
13
9
1996
0
0
0
10
4
2
10
9
17
1997
0
0
0
12
4
1
11
14
11
1998
1
0
0
14
4
0
13
10
13
1999
0
0
0
10
5
2
14
9
15
2000
3
0
0
13
4
2
6
14
14
2001
2
0
0
12
8
4
3
11
17
2002
2(1SW)
0
0
7(1MZ)
3(1SW)
6(2SW)
2
14
23
2003
3
0
0
7(1SW/1MZ)
8
4(2SW)
13
8
14
2004
0(1SW)
0
0
6(1MZ)
6(2SW)
6(2SW)
3
18
18
7(1SW/1MZ) 7(2SW)
3(1SW)
4
15
20
2005
1(1SW)
0
0
2006
1(1SW)
0
0
9(1SW)
4
7(1SW/1MZ)
8
8
23(2SW)
2007
1
0
0
6(1SW/1MZ)
5
7(1SW)
12
8
23(3SW)
Notes
*Annual rebalancing takes place annually every December except for last year when it is in August
**Figures within panel 1 indicate numbers of Kenyan stocks, while those values in parentheses are Swaziland stocks (SW) or Mozambique stocks (MZ).

25

Table 3: Time series regressions using equally weighted monthly contemporaneous market excess returns for 9 portfolios formed on size and illiquidity
for period: 1991 2001, for Kenya and South Africa.
Size

Low

Medium

High

0.007023
-0.003701
-0.001534

-0.004273
0.002453
-0.002067

-0.004380
0.004204
0.001335

1.689640
0.657157
0.805528

0.386033
1.661897
0.452678

0.410829
1.513702
1.083518

-0.627773
0.202980
0.346711

-0.605716
0.530933
0.198686

-0.708242
0.117529
0.455414

-1.288713
-0.079306
-0.112015

0.301649
-0.106339
-0.052716

0.852501
0.637041
0.050398

0.366309
0.216902
0.009196

0.639150
0.151887
0.004112

0.730588
0.627463
0.181308

0.885913
0.383024
0.483791

0.704686
0.232777
0.288546

0.910498
0.786907
0.516693

R p t - R ft = p + p (R m t - R ft ) + S p S M B t + H p IL L IQ t + p t

p
Small
Medium
Big
p
Small
Medium
Big
Sp
Small
Medium
Big
Hp
Small
Medium
Big
Adj R2 (1)
Small
Medium
Big
Adj R2 (3)
Small
Medium
Big

Low
T(p)
2.93
-4.30
-1.97
T(p)
7.47
8.09
10.97
T(Sp)
-6.13
5.51
10.42
T(Hp)
-21.66
-3.70
-5.79

Notes: Newey-West HAC Standard Errors & Covariance, standard errors are used in the t-tests

26

Medium

High

-1.40
0.77
-3.17

-1.99
1.54
1.34

1.34
5.57
7.36

1.98
5.91
12.76

-4.64
3.93
7.13

-7.56
1.01
11.42

3.98
-1.35
-3.25

15.66
9.44
1.76

Table 4: Time series regressions using equally weighted monthly contemporaneous market excess returns for 9 portfolios formed on size and illiquidity
for period: 2002 2007, for Kenya, South Africa, Swaziland and Mozambique.
Size

Low

Medium

High

0.002454
0.002068
-0.002508

-0.004410
0.001795
-0.001017

-0.000636
0.001711
0.000937

0.688669
1.043540
1.050078

1.354412
1.143766
0.882326

0.747128
1.177493
0.857564

-0.279335
0.015051
0.457469

-0.849908
0.057794
0.435625

-0.519163
0.254456
0.457817

-0.556696
-0.302097
-0.192019

-0.639466
-0.140860
-0.047894

1.150543
0.605011
0.193743

0.143646
0.547373
0.318057

0.427586
0.587840
0.368167

0.461687
0.570533
0.483527

0.376552
0.635056
0.783690

0.669633
0.599133
0.815401

0.804696
0.667651
0.811408

R p t - R ft = p + p (R m t - R ft ) + S p S M B t + H p IL L IQ t + p t

p
Small
Medium
Big
p
Small
Medium
Big
Sp
Small
Medium
Big
Hp
Small
Medium
Big
Adj R2 (1)
Small
Medium
Big
Adj R2 (3)
Small
Medium
Big

Low
T(p)
1.24
1.47
-2.40
T(p)
4.94
10.53
14.19
T(Sp)
-3.07
0.23
9.47
T(Hp)
-5.00
-3.82
-3.25

Notes: Newey-West HAC Standard Errors & Covariance, standard errors are used in the t-tests

27

Medium

High

-1.64
1.02
-1.22

-0.23
0.72
1.021

7.14
9.23
15.01

3.97
7.09
13.20

-6.87
0.71
11.36

-4.23
2.34
10.79

-4.22
-1.42
-1.02

7.66
4.56
3.73

Table 5: Time pooled cross-section regression for equally weighted monthly excess returns on 9 portfolios formed on size and illiquidity for
South Africa
Explanatory Variables

T()

T()

Panel 1: Period of 1991 to 2001

Excess Market alone
Excess Market and SMB
Excess Market and ILLIQ
All Three Factors

-0.002778
-0.001921
-0.002631
-0.001558

-3.71
-3.21
-3.58
-2.94

0.097755
0.519913
0.141184
0.650114

3.68
10.11
4.62
13.02

Panel 2: Period of 2002 to 2007

Excess Market alone
Excess Market and SMB
Excess Market and ILLIQ
All Three Factors

-0.002843
-0.000994
-0.002315
-0.001016

-2.43
-1.11
-1.97
-1.11

0.564473
0.757382
0.657261
0.752276

8.00
13.08
7.91
11.68

R p t - R ft = p + p (R m t - R ft ) + S p S M B t + H p IL L IQ t + p t

Notes: Newey-West HAC Standard Errors & Covariance, standard errors are used in the t-tests

28

T(S)

0.225166

8.98

0.255956

11.32

0.291176

7.53

0.294138

6.99

-0.048832
-0.081490

-0.126097
0.009606

T(H)

Adj R2

-2.70
-6.20

0.09
0.43
0.13
0.56

-2.00
0.18

0.48
0.72
0.50
0.72

Table 6: Time pooled cross-section regression for equally weighted monthly excess returns on 9 portfolios formed on size and illiquidity for
Kenya
Explanatory Variables

T()

T()

R p t - R ft = p + p (R m t - R ft ) + S p S M B t + H p IL L IQ t + p t

Panel 1: Period of 1991 to 2001

Excess Market alone
Excess Market and SMB
Excess Market and ILLIQ
All Three Factors

0.007599
0.005185
0.007167
0.004135

3.13
2.48
2.99
2.14

3.681069
2.492207
3.553220
2.115075

42.69
13.89
35.70
11.62

Panel 2: Period of 2002 to 2007

Excess Market alone
Excess Market and SMB
Excess Market and ILLIQ
All Three Factors

0.011959
0.003016
0.009226
0.003008

2.20
0.75
1.70
0.74

3.067358
2.134097
2.586886
2.132204

9.38
8.27
6.77
7.43

Notes: Newey-West HAC Standard Errors & Covariance, standard errors are used in the t-tests

29

T(S)

-0.634100

-7.25

-0.723286

-8.77

-1.408657

-8.18

-1.407559

-7.51

0.143756
0.236039

0.652950
0.003562

T(H)

Adj R2

2.44
4.92

0.93
0.95
0.93
0.95

2.26
0.02

0.56
0.78
0.59
0.77

Table 7: Time pooled cross-section regression for equally weighted monthly excess returns on 9 portfolios formed on size and illiquidity for
Swaziland and Mozambique over sample period: 2002 to 2007
Explanatory Variables

T()

T()

R p t - R ft = p + p (R m t - R ft ) + S p S M B t + H p IL L IQ t + p t

Panel 1: Swaziland
Excess Market alone
Excess Market and SMB
Excess Market and ILLIQ
All Three Factors

0.010020
0.012319
0.012416
0.013373

1.23
1.44
1.48
1.55

0.910853
1.150852
1.331999
1.401970

1.83
2.08
2.24
2.29

Panel 2: Mozambique
Excess Market alone
Excess Market and SMB
Excess Market and ILLIQ
All Three Factors

0.019706
0.025393
0.025259
0.027766

1.06
1.32
1.33
1.43

0.416428
1.009917
1.392434
1.575789

0.37
0.81
1.04
1.14

Notes: Newey-West HAC Standard Errors & Covariance, standard errors are used in the t-tests

30

T(S)

0.362252

0.98

0.216609

0.54

0.895808

1.08

0.567614

0.63

-0.572327
-0.472393

-1.326369
-1.064496

T(S)

Adj R2

-1.27
-0.97

0.034073
0.033479
0.043010
0.032501

-1.31
-0.97

-0.013030
-0.010629
-0.002265
-0.011635

Amihud Liquidity Factor (x10^6)

3,000

2,500

2,000

1,500

Amihud Liquidity Factor (x10^6)

3,500

200

150

100
01
/0
1/
90
01
/0
1/
91
01
/0
1/
92
01
/0
1/
93
01
/0
1/
94
01
/0
1/
95
01
/0
1/
96
01
/0
1/
97
01
/0
1/
98
01
/0
1/
99
01
/0
1/
00
01
/0
1/
01
01
/0
1/
02
01
/0
1/
03
01
/0
1/
04
01
/0
1/
05
01
/0
1/
06
01
/0
1/
07

Amihud Liquidity Factor (x10^6)

4,000

Amihud Liquidity Factor (x10^6)

01
/0
1/
90
01
/0
1/
91
01
/0
1/
92
01
/0
1/
93
01
/0
1/
94
01
/0
1/
95
01
/0
1/
96
01
/0
1/
97
01
/0
1/
98
01
/0
1/
99
01
/0
1/
00
01
/0
1/
01
01
/0
1/
02
01
/0
1/
03
01
/0
1/
04
01
/0
1/
05
01
/0
1/
06
01
/0
1/
07
5,500

01
/1
2
01 / 01
/0
3
01 / 02
/0
6/
01 02
/0
9
01 / 02
/1
2
01 / 02
/0
3/
01 03
/0
6
01 / 03
/0
9
01 / 03
/1
2
01 / 03
/0
3/
01 04
/0
6
01 / 04
/0
9/
01 04
/1
2
01 / 04
/0
3
01 / 05
/0
6
01 / 05
/0
9
01 / 05
/1
2
01 / 05
/0
3
01 / 06
/0
6
01 / 06
/0
9
01 / 06
/1
2/
01 06
/0
3
01 / 07
/0
6
01 / 07
/0
9/
07

01
/0
2/
01 97
/0
8/
01 97
/0
2/
01 98
/0
8/
01 98
/0
2/
01 99
/0
8/
01 99
/0
2/
01 00
/0
8/
01 00
/0
2/
01 01
/0
8/
01 01
/0
2/
01 02
/0
8/
01 02
/0
2/
01 03
/0
8/
01 03
/0
2/
01 04
/0
8/
01 04
/0
2/
01 05
/0
8/
01 05
/0
2/
01 06
/0
8/
01 06
/0
2/
01 07
/0
8/
07

Figure 1: Aggregated Market Illiquidity Factors, by Country

South Africa Market Aggregate Amihud Liquidity Factor: 1990 to 2007

700
Kenya Market Aggregate Amihud Liquidity Factor: 1990 to 2007

5,000

4,500

600

500

400

300

200

1,000

500

100

300
Swaziland Market Aggregate Amihud Liquidity Factor: 1997 to 2007

30000
Mozambique Market Aggregate Amihud Liquidity Factor: 2002 to 2007

250

25000

20000

15000

10000

50

5000

Illiquidity factors constructed according to Amihud (2002) techniques outlined in equations (1) and (2). Larger absolute values are interpreted as higher levels of aggregate illiquidity (lower
levels of liquidity

31

Figure 2: Covariance Risk Premiums for South Africa, Kenya and Swaziland
Kenya Risk Premiums (Total, Market, Size and Liquidity)
750
700
650
600
550
500

Premiums (Basis Points)

450
400
350
300
250
200
150
100
50
0
-50
-100
-150

KENYA: SIZE

KENYA: LIQUIDITY

KENYA: MARKET

01
/0
1/
07

01
/0
1/
06

01
/0
1/
05

01
/0
1/
04

01
/0
1/
03

01
/0
1/
02

01
/0
1/
01

01
/0
1/
00

01
/0
1/
99

01
/0
1/
98

01
/0
1/
97

01
/0
1/
96

01
/0
1/
95

01
/0
1/
94

01
/0
1/
93

01
/0
1/
92

01
/0
1/
91

-200

KENYA: TOTAL

South Africa Risk Premiums (Total, Market, Size and Liquidity)

140
120
100

Premiums (Basis Points)

80
60
40
20
0
-20
-40
-60
-80

SA: SIZE

SA: LIQUIDITY

SA: MARKET

32

SA: TOTAL

01
/0
1/
07

01
/0
1/
06

01
/0
1/
05

01
/0
1/
04

01
/0
1/
03

01
/0
1/
02

01
/0
1/
01

01
/0
1/
00

01
/0
1/
99

01
/0
1/
98

01
/0
1/
97

01
/0
1/
96

01
/0
1/
95

01
/0
1/
94

01
/0
1/
93

01
/0
1/
92

01
/0
1/
91

-100

The Kingly Report that regulates corporate governance practices in South Africa is very similar to the UK
Cadbury Report and the US Sarbanes Oxley Act.
ii
Countries in the East African Community are Kenya, Tanzania and Uganda.

33