You are on page 1of 10

Available online at www.sciencedirect.


Review of Development Finance 3 (2013) 204–213

The impact of technological improvements on developing financial markets:
The case of the Johannesburg Stock Exchange
Mehmet F. Dicle a,∗ , John Levendis b
a Stanford H. Rosenthal Professor of Risk, Insurance and Entrepreneurship, Joseph A. Butt, S.J., College of Business, Loyola University New Orleans, United States
b Dr. John V. Connor Professor of Economics and Finance, Joseph A. Butt, S.J., College of Business, Loyola University New Orleans, United States

Received 24 February 2012; received in revised form 6 May 2013; accepted 12 September 2013

Can a significant technological improvement make an economically justifiable contribution to a financial market’s development? The Johannes-
burg Stock Exchange (JSE) incorporated the SETS system from the London Stock Exchange in 2002. It is certain that SETS is a technologically
efficient trading system, and it would undoubtedly improve trading in the JSE. We test whether SETS represents a structural break by examining
whether there was an increase in the JSE’s liquidity, market efficiency and international integration after the introduction of SETS. While SETS is
certainly a technological improvement with increased liquidity, it is not a sufficient factor to render it efficient. After the incorporation of SETS,
the JSE has become more independent and it now offers better diversification opportunities for international investors.
© 2013 Africagrowth Institute. Production and hosting by Elsevier B.V. Open access under CC BY-NC-ND license.

JEL classification: G14; G15

Keywords: Johannesburg Stock Exchange; Market efficiency; Market development; Structural change

1. Introduction as the one with LSE] have had the simultaneous impact
of exposing the South African capital markets sector to
The mood was certainly buoyant at a celebratory dinner on 17 the rest of the world, bringing world-class services and
May 2002, where South African President Thabo Mbeki gushed infrastructure to the JSE and entrenching the JSE in the
about the Johannesburg Stock Exchange’s (JSE) new partner- mind of the international investor as the gateway into the
ship with the London Stock Exchange (LSE). The partnership African market (Mbeki, 2002).
entailed the incorporation of the LSE’s SETS trading platform
The development of financial markets is important in facilitat-
by the JSE. In obligatory political fashion, Mbeki proclaimed
ing economic development. For countries where access to capital
that the new trading platform,
is of the utmost importance, financial markets play a crucial
will add the necessary impetus in our work of reconstructing intermediary role between savings and investment. If domestic
and developing not only our country but the entire continent and international banks are unable or unwilling to invest in such
of Africa [by encouraging investment]. Strategic partner- economies, capital can become scarce and prohibitively costly.
ships with a number of globally prominent companies [such When developing countries try to develop their financial mar-
kets, five issues arise. (1) Protecting investors seems to be an a
priori prerequisite for financial development (i.e. La Porta et al.,

1998, 2000). (2) If the investment choice offers no diversification
Corresponding author. Tel.: +1 504 858 9342.
benefits, then there is limited reason for international investors
E-mail addresses: (M.F. Dicle),
(J. Levendis). to consider investing. If the major markets are the main drivers
Peer review under responsibility of Africagrowth Institute. of returns, then the developing market does not offer a unique
investment opportunity. While the return to risk ratio might
be appealing, the market’s contribution toward diversifying a
portfolio would be minimal. (3) Questionable market efficiency
hampers the market’s development (Liu, 2010). There will be
1879-9337 © 2013 Africagrowth Institute. Production and hosting by Elsevier
B.V. Open access under CC BY-NC-ND license.
limited interest in the market by foreign investors if information is unavailable or asymmetric, or returns are predictable or

2005) and volume. unique opportunities for interna. Dicle. The JSE and its performance another. high the studies conducted after 2002 tend to detect more efficiency number of market participants. volume) finding that the studies differ over methodology. the expectation for market development will fade away. perhaps. these sets of equities should be perfectly information. . and individuals can trade without revealing their correlated. limited impact on the informational aspect of market efficiency. therefore for the development of the South African economy Ferret and Page (1998) analyzed four South African futures overall. Samouilhan hypothesizes that the causation happens at a frequency that is higher than daily data can reveal. More importantly. Bekaert et al. though. Furthermore. Institutions can trade without revealing their tegration. time periods. 1992). its adoption should alleviate conclude that the JSE is not semi-strong form efficient. First. there will be more local companies listed. implies a lack of interest in the market. Finally. Are returns in the JSE correlated with those in 2002. investors. Diversification can reduce risk while maintaining why even more capital does not find its way to Johannesburg? returns. liquidity. M. informationally efficient. return model. (5) A higher number of listed com. Thus. efficiency of the JSE is at best mixed. If investment choices have the and liquidity? same return generating process. and ous African stock exchanges. (Jefferis and Smith. while the five issues listed above for the development of the JSE and Jefferis and Okeahalam (1999) conclude that it is. Then. and generally is SETS is a milestone for the JSE. implying that the market is trading activity (volume). Thompson and Ward’s (2007) show that liquidity is a priced risk and part of the expected (1995) paper reviews the literature on the efficiency of the JSE. less devel. study on equity markets in other developing countries. restricts his sample to daily data on broad aggregate and sectoral oped. the JSE’s adoption of weak and semi-strong form efficiency” (p. walks. Auret and Rudolph (2006) find that the announcement of panies. (3) While SETS has a in the JSE. Ferret and Page find a long-term linear relationship identity. and high volumes will increase than the earlier studies. but they also find that changes of listed companies. markets of more developed countries? In other words. Its liquidity is growing. “the evidence on the interest. Humavindu and Floros (2006) find no such volatility. Jefferis and Smith (2004) market efficiency. particularly regarding In terms of financial market structure. exchanges are highly integrated. while reviewing the earlier empirical research on the JSE’s efficiency. In order to accomplish this. investors will increase liquidity and lower the transaction costs. while Hearn and Piesse (2008) tion. or they have a lead/lag relation- The paper is structured as follows. has the incorporation of a more efficient must be uncorrelated so that changes in one will not necessar- trading platform paid off in terms of increased market efficiency ily offset the changes in another. We describe the institutional evolution of the JSE. Institutional investment in emerging markets may be conditional Is the JSE relatively efficient? Earlier studies of the JSE’s mar- on the availability of liquidity (Chuhan. (1) The anonymity of trades and traders allows higher contracts and their corresponding spot market indices for coin- investor protection. SETS is a system used by not strong-form efficient. Several studies have investigated the integration of the vari- form several comparison-of-means tests of returns. faster trades and cheaper trades will increase CPI has no effect on stock prices. 2004). Piesse and Hearn (2002. 59). In the present study. then diversification cannot be achieved with these related the JSE is of interest as a case study on market efficiency and instruments. in the futures contracts lead those of the spot market by up to With diverse listed companies. and conclusions. we per. (4) A lack of liquidity will hamper the development efforts. increasing. Levendis / Review of Development Finance 3 (2013) 204–213 205 manipulatable. attributing to the incorporation of SETS in and efficiency. it seems that the JSE’s market efficiency has been statistical properties of the JSE – in terms of returns. among the 40 most capitalized individual stocks. J. three days. For example. Ideally. ket efficiency are largely inconclusive.e. This finding points to an exploitable lack of efficiency tional investors will be offered by the JSE. the varying assets And just as importantly. by local companies to be listed due to the associated inefficiency as is its relative size in terms of the world’s market capitalization risks. We conclude with a discussion of the implications of our find some integration only between selected pairs of countries. For example. financial markets in the region. there will be limited interest African stock market by a large margin. whereas the smaller indices are not. (2) With a system that can accommodate a high number between these two sets of equities. potential for diversification. Competition between form efficient. we explain why ship. While the studies before 2002 offered contradictory results. Appiah-Kusi and Pescetto (1998) find volatility spillovers pointing to a likely structural break in 2002. (4) Having more investors and faster trades find that large-cap indices are random walks and are weak- will provide liquidity (cheaper trades). With the diminishing samples. a system that can accommodate high transaction speeds. We follow this by between the JSE and the Namibian stock exchange. (5) a lack of trading activity (i. liquidity. Glass and Smit (1995) the London Stock Exchange. and among It will also allow for the emergence of market makers who in the 40 most liquid small-cap stocks. He The JSE’s development sets an example for other. and structural breaks. we examine the impact of SETS on the Overall. is the How well integrated is the JSE into the rest of the world’s JSE a good source of diversification? Are there inefficiencies equity markets? International markets provide options for diver- in the South African exchange that could perhaps account for sification. The results to date have been mixed. development. 32 seem to follow random turn will provide liquidity. The JSE is the largest indices. Samouilhan (2006) finds that the JSE and LSE move almost simultaneously: if there is a causal effect from one market to 2.F. We also test for structural breaks of the predictability Tyandela and Biekpe (2001) find that the Southern African stock of returns (a test of market efficiency) and international integra. In their words.

when the or percentage spread fall five or more standard deviations from government eliminated the exchange rate controls. p. on the other hand includes price reaction only the usual burden of exchange rate controls. then it was of huge value to South Pi. the JSE was burdened by several structural inefficiencies which diminished its ability to allocate capital. Results from the restricted sample are available from the authors upon request. which reported price-relevant informa- When we turn to detecting Granger causality from other mar- tion in real time.t RCCi.t−1 similar trading platforms. we opted for the larger sample. they find a break significantly earlier than the legal We calculate percentage bid and ask spreads of stock i at time date of liberalization. when an elected national-unity gov. In the span of less than 10 years. Expanding our sample did not change (BBC. kets. (2000) in filtering our data to remove the JSE from 1991 to 1996. December 2007. Understanding these helps isolate the effects 1990s in South Africa. the deck is certainly stacked against the importance of basis. structural breaks may be then the market can be manipulated by those institutions. (2) April 27. be a source of market inefficiency. In a similar manner. Michello (2001) examines evidence equities whose price puts them at the tails of the distribution. We restrict our attention to ordinary stocks. this might not blocks of shares discreetly.t is the dividend and split-adjusted price of stock i Makina and Negash (2005) investigate the possibility of a at time t.t refer to the daily closing ask and bid ing toward fully electronic trading. 690). for structural changes on three events: (1) April 1. 2002). 4. of stock i on day t. the sample began at year 2000. J. Given that these Returns can be calculated using either close-to-close prices seemingly important changes had no appreciable effect on the or open-to-close prices. so at the suggestion of a referee. lead to information. Levendis / Review of Development Finance 3 (2013) 204–213 3.t = ln (1) Africa. the JSE began using the London Stock Exchange’s SETS electronic trading system. Data One such example was its dual exchange rate system – a type of exchange rate control – which was eliminated in 1995. If individual investors follow so that the market before these changes has little relationship to institutional investors. where Pi. to the information during the trading hours.t and Smith. . at the middle of this range. 1991.. any observations whose returns. Institutional stealth trading has received consider- misleading conclusions (Dicle and Dicle.1 For. its effect should be accounted for in order trading platform also has the potential to allow dual primary to avoid omitting variables. not find any structural breaks on these dates. In light of this. The authors argue that the market had t as: already anticipated the forthcoming reforms by December 1992. Until the 1990s.t − Bidi. service was introduced. its performance afterwards.t PRSi. and (3) March 13. therefore. Using data on real aggregate calculated analogously. Blau et al. ernment was established. An institutional herding effect can be appropriate. whom they presume are better informed. the JSE has become “one of Aski. many researchers simply even small investors. 1994. as there may be structural changes in the series. Open-to-close returns are denoted ROCi. If SETS marks able attention in the literature (ex. The data for the study are from Thomson’s Reuters. According to Rus- sel Loubser. stock prices.3 Returns. the JSE began transition- where Aski. Open-to-close returns. this technologically advanced 2 In an earlier version of the paper. The new a true structural break.t .F. respectively.0%. These swings add to opening. Insti- confused econometrically with non-stationarity.2 The incorporation of SETS into the JSE sits other reforms. increased efficiency. it allows for anonymous trading. and are computed as: SETS.206 M. 3 Close-to-close returns include price reaction to information during the trad- 1 Dual exchange rates are prone to wild swings in the relative popularity of ing hours and during the period between previous day’s closing and daily the currencies. and are structural break in the JSE in 1995. Moreover. 2009). if it were found that SETS had an impact   on the JSE’s development. so that all trading platform is “more functional and reliable and stable” of the tumultuous 1990s were excluded.5% to excluded. when Thus. are denoted by RCCi. stocks whose price is 100 times above the daily average are the tax rate on marketable securities was dropped from 1. Dicle. we restrict our attention to January 1997 of SETS by not including the possibly confounding effects of to December 2007. the head of the JSE.t . excluded. which the results in any meaningful way. 2010). 2004). Moreover. stocks whose price is <1% of the daily average are also 1.t = (2) the most technologically advanced emerging markets” (Jefferis Pi. and allows larger investors to trade larger opt for the longest time series available. Michello did the mean are excluded from the sample. 2004. 1995. a phenomenon known as Gresham’s law. This is because anonymity gives comfort to In determining their sample period. However. In 1996. calculated on a close-to-close JSE. an electronic news prices. Looking at changes in the bid-ask spreads of 135 stocks on We follow Chordia et al. we must restrict ourselves to looking at market averages On 13 May 2002.t and Bidi. and it could be argued that other countries should adopt Pi. through this liquidity. using the eign ownership and limited liability corporate membership were Quotecenter application for the period between January 1997 to allowed in 1995 (Jefferis and Smith. There have been several structural reforms in the JSE over the Given the important changes that occurred throughout the past two decades. Not accounting tutions also become reluctant to invest to avoid revealing their for the possibility of structural breaks could. Structural breaks and the evolution of JSE should increase the market’s liquidity and. In 1997. listings in several exchanges.

We included First.0650 0. the expectation is toward a more efficient value of one if the company is in the top two quintiles and zero market. we conduct simple comparison-of-means tests to see the quoted and effective spreads to show the spread magnitudes. 5% and 1% level.0000 0.t  improved demand for South African listed securities.0004 0.13 *** *** Liquidity Illiquidity 0.0360 53.and post.99 *** *** Proportional quoted spread 0. The binary variable large is assigned a adoption of SETS. The binary variable small is assigned a value of one if Hypothesis 1.0548 −3.4. measure of liquidity is preferred since JSE listed stocks have increased in price post-SETS.t Qi. μpre μpost t pre > post pre = / post pre < post Daily returns 0. . This is especially necessary for smaller = λ0 + λ1 ε2t−1 + λ2 σt−1 2 (5) exchanges. Fama (1970) argues that autocorrelated returns indicated The variable SETSt is a binary variable that is equal to 1 if an inefficient market. J. dependent upon lagged returns. “*”. quoted spreads are decreasing. the proportional quoted spread and proportional Value-weighted returns are: effective spread have decreased.vs. 5 Since the price levels are increasing in the JSE. In the first row. therefore toward less predictable returns.791 6. Since us to examine the effect of SETS on certain stock characteris.0522 −4. However. SETS was in place during period t. liquidity and volume have changed pre. given that prices and volume exchange c: have both increased. are to capture JSE is inefficient. (2010) also employ an autocor- the size effect. Griffin et al. We report these results in Table 1.SETS. While higher returns of all JSE stocks.t−1 become more or less liquid after SETS. implying that the JSE has   Pi. σt2 ) σt2 evaluate individual stocks as well as portfolios for their mar- ket efficiency tests. Levendis / Review of Development Finance 3 (2013) 204–213 207 Table 1 Comparison of pre. where returns of each stock in market c are weighted by the value Our analysis continues with formal econometric testing of of its shares outstanding (price times quantity) as a percentage returns. or value-weighted) returns higher after the incorporation of SETS.0008 0. using the Quotecenter application for the period January 1997 until December 2007. walks). we see that daily returns are. Evaluation of individual stocks also allows 4 Amihud’s measure of illliquidity (row 2). even though the proportional which list stocks with similar characteristics. Open-to-close returns characterized by unpredictable returns (most likely as random are calculated similarly. We calculate equal-weighted.8118 48. Empirical results WRt = (β0 + β1 WRt−1 + β2 WRt−2 + β3 PRSt + β4 PRSt−1 ) Thoroughly analyzing the effect of SETS on the JSE requires looking at two different levels of data: (1) the overall market · (1 − SETS) + (γ0 + γ1 WRt−1 + γ2 WRt−2 + γ3 PRSt level.t = Pi.t (4) Trading activity (volume) has nearly doubled for JSE after SETS Pi. Efficient markets are of the listing markets’ total capitalization. (2010) + γ4 PRSt−1 ) · (SETS) + εt εt ∼N(0. relation test of returns to evaluate market efficiency. and 0 if otherwise.F. with the appropriate modifications. has increased after SETS. respectively. on average.0001 2.e. otherwise.t   become more liquid for international institutional investment. SETS had an impact on overall market returns. With the italization into quintiles. liquidity and market independence.51 ** *** Quoted spread 32.73 *** *** Proportional effective spread 0. i. EWRCCc. we examine whether the JSE has Pi.0004 −2.t = ln (3) In rows two through six. the quoted spreads are increasing. “**” and “***” refer to statistical significance at the 10%.t Qi. Amihud’s illiquidity ratio is affected by this lack of trading activity. since 1997. large and small. nearly three times One may calculate (equal-weighted.0642 0. M.41 *** *** Volume Volume 251. Griffin et al.72 *** *** (rather than causality between thousands of individual stocks).940 475. they can returns as the simple sum of the returns from each stock i in also imply higher risk.6188 3. This would be of interest for exchanges other than the JSE activity.t ln / Pi. Data are through Reuters.77 *** *** Effective spread 30. close-to-close imply higher compensation for market participants. If today’s returns in the JSE are predictable.t−1 i∈c i∈c compared to the period before SETS. and (2) the level of individual stocks. Dicle.5 VWRCCc. We sort the companies based on their market cap. the more likely cause for the increase is an   Pi.71 *** *** Relative volume 0. We test whether SETS had any impact on the returns process of the JSE using the GARCH(1.1) model below: 5. the company is in the bottom two quintiles and zero otherwise. The proportional whether returns.0015 3. two binary variables. Two (inverse) measures i∈c of liquidity. the data includes many non-trading days and illiquid stocks with limited trading tics. then this is evidence that the Finally.7889 3. as what might appear to be a market-wide phe- nomenon might actually be the result of a few high-valued and/or high volume stocks.

0711*** 0. Data are through Reuters.0003 −0.0613*** 0.0011 −0.9057*** 0.0094*** −0.0403 0.0009 −0.8705*** 0.0161*** 0.0003 −0.0861*** 0.1099*** 0.0974*** 0.0746** 0.1836*** 0.0000 0.1742*** 0.0066*** 0.0000 0.1934*** 0.0005** 0.0396 0.0011*** 0.0082** −0.0031** −0.0005 −0.0000*** 0.2195*** 0.0004 −0.0288 0.0050*** 0. Levendis / Review of Development Finance 3 (2013) 204–213 2007.0046 Post-SETS γ0 0.1014*** 0. respectively.8927*** 0.6529*** −0.0000*** 0.0070*** 0.1544*** λ2 0.0861*** 0.0009*** β1 0.0012*** 0.0058** β4 0.1352*** 0.2281*** 0.0000*** 0.0852*** 0.F.0017*** 0.0700*** 0.2143*** 0.1327*** 0.0165*** −0.8856*** 0.0118*** −0.0000*** λ1 0.0002 −0.0071*** −0.0027** 0.6559*** .0000 0.3331*** 0.3150*** 0.0009 −0.1013*** 0. PRS refers to the percentage spread as the measure of liquidity.0008*** −0.0836*** 0.0500 0.0024** 0.0021 0.1496** β3 −0.0016 γ4 0.1087*** 0. “**” and “***” refer to statistical significance at the 10%.0019 −0.0212 0.1807*** β2 0.0002 0.1916*** 0.0003 0. using the Quotecenter application for the period January 1997 until December M.0009*** 0.0000*** 0.0013*** 0.0029* −0.0006*** γ1 0.0079*** −0.0017*** 0.0001 0.2227*** 0.0014** 0. W refers to either equally weighted (EW) or value weighted (VW).0000 −0.0022* −0.0001 −0. The results provided in this table are obtained using Eq.0895*** 0.0000*** 0. δ2t ) and δ2t = λ0 + λ1 ε2t−1 + λ2 δ2t−1 .0008 −0. “*”.0641 0.0915** 0.0006*** 0.0066*** −0. 5% and 1% level.0022 0.0027** −0.9812*** 0. R refers to either close-to-close (RCC) or open-to-close (ROC) returns. 5 years before and after SETS 3 years before and after SETS 1 year before and after SETS RCC ROC RCC ROC RCC ROC VW EW VW EW VW EW VW EW VW EW VW EW Pre-SETS β0 −0.0067*** −0.0625** 0.2613*** γ2 0.0000*** 0. likewise.0009*** 0.0032 0.0821*** 0.0023 0.0010*** 0.1105*** 0.0617 0.0082*** −0.9183*** 0.0011 −0.0014*** 0.0010*** 0.1077*** 0.0921 0.0744 0.0115 0.0000*** 0.0156 0. J.0000** 0.1930*** 0. Dicle. (5): WRt = (β0 + β1 WRt−1 + β2 WRt−2 + β3 PRSt + β4 PRSt−1 )•(1 − SETS) + (γ 0 + γ 1 WRt−1 + γ 2 WRt−2 + γ 3 PRSt + γ 4 PRSt−1 )•(SETS) + εt where εt ∼N(0.0014 −0.0000** 0.0625 0.0810 0.0007 −0.0059*** −0.0102 0.0525 0.0003* 0.0000** 0.2214*** 0.0009*** 0.1117*** 0.0004 −0.9169*** 0.0000*** 0.1003** 0.0059*** −0.1102*** 0.0008 −0.0021 δ2t λ0 0.0013 −0.8681*** 0.7731*** 0.0404 0.1372*** 0.0009*** 0. 208 Table 2 Evaluation of market level returns for the Johannesburg Stock Exchange.1089*** 0.0031* 0.0816 γ3 0.0016*** 0.

5 years. Bekaert et al. The difference between β0 and γ 0 is not statistically signif. in all our tests of market efficiency.t + γ7 PRSi. We also consider the value-weighted market averages. in order to control for the overall market’s The relationship between the returns and liquidity is statis. trading costs. In sum. it may be possible to take advantage of overnight returns et al.) The above model could also be tested using a simple take advantage of return predictability. in general form. higher since there are more smaller stocks than larger stocks. This GARCH model is also equivalent to a sim- that would be likely to affect returns. For the sake of simplicity. This would imply that. the empirical results point to the impact of SETS on overnight That is. Griffin SETS. (2007) show the impact individual stock returns after SETS. = λ0 + λ1 ε2t−1 + λ2 σt−1 2 (6) icant. As robustness checks Table 3 repeats + γ3 VWRett + γ4 VWRett−1 + γ5 VWRett−2 the test using other liquidity measures. Levendis / Review of Development Finance 3 (2013) 204–213 209 This type of regression specification allows the data gener. and compares the process pre. Columns information releases. positive impact on intraday returns. after the financial literature. for both equal-weighted and a open-to-close and close-to-close basis).t−2 + β3 VWRett In Table 2 we explore all of the possible permutations of these choices: value-weighted and equal-weighted returns. (2009) find commonality in liquidity for a not worry about value-weighting or equal-weighting returns for sample of countries that include emerging markets. (2010) employ variance ratio tests as part of their market for smaller stocks. we provide the estimating equation Hypothesis 2. preference for continual liquidity for smaller stocks. this relationship. and predictive power of liquidity on returns in emerging mar.and post-SETS. proportional effective spread and proportional quoted + γ6 PRSi. including effective spread. Intraday (3) and (4) re-estimate the first two columns. and one year) before and after SETS. Coefficients are allowed to vary pre. SETS had a its own returns. variables.t−2 with the varying windows. Column (2) esti- Overnight returns may be affected by overnight news and mates the same model. M. returns. Reti. and two dummy variables indicating whether the stock is rela- Fama (1970) argues that market efficiency is violated with tively small or large. While there cannot be a definite explanation for this Column (1) of Table 4 estimates the panel model conditional persistence. after SETS. the weight of smaller stocks is tions. ating process to vary pre. we condition the past volatility and implement it alternative liquidity measures. calculated + β4 VWRett−1 + β5 VWRett−2 + β6 PRSi. especially considering t-test to compare returns before SETS and returns after SETS.e. each for equal-weighted market averages.t . Brockman et al. (Including only one with value weighted returns.t−1 + β8 PRSi. + β7 PRSi. the return of a stock listed on the JSE is a function of returns to be different for large and small stocks. after SETS. persists contemporaneously as well as for the Results are shown in Table 4. possibly international news. SETS had an impact on individual stock returns. After SETS. using the dynamic panel data model below: R refers to either close-to-close (RCC) or open-to-close (ROC) returns.t on open-to-close and close-to-close prices. J. except of equal-weighted open-to-close market averages. That is. Intraday returns are less predictable after SETS. As this is a panel data model. However. Also. first lag. The results are largely consistent across all specifica- With equal-weighted returns. By employing The results remain qualitatively similar when we employ GARCH(1. The dependent variable is individual stock returns.t−2 + γ9 Smalli spread.t−2 + β9 Smalli + β10 Largei ) We consider three different windows (i. important for small and large stocks. only cratic return βi (and γ i post-SETS).vs..and post-SETS. A Hausman test with 1% the contemporaneous return-liquidity relationship exists. sta. before SETS. for equal-weighted than fixed effects. . This implies that liquidity is effect of stock size on the returns-generating process. 3 years. we argue that the evidence may reflect investors’ on open-to-close market returns without controls for stock size. SETS dummy additively would restrict the betas and gammas to Lower coefficients on lagged returns would make it harder to be equal. That is. likewise. the weight of larger stocks is higher. but only for larger stocks conditional heteroscedasticity on returns is well established in (value-weighted market returns). On the statistical significance indicates that the random effects. the overall (value-weighted) market’s returns. it is essentially the same. The results are consistent across the different measures + γ10 Largei ) · (SETS) + εt εt ∼N(0. deviations from random walks. for each of the equations W refers to We test whether SETS had an impact on individual returns either equal-weighted (EW) or value-weighted (VW). a sign of market inefficiency.t−1 + γ2 Reti. smaller stocks might behave differently than large stocks. σt2 ) σt2 of liquidity. but add the size returns may be affected by movements in correlated markets.t = (βi + β0 + β1 Reti. we employ a GARCH(1.1). The results are consistent ·(1 − SETS) + (γi + γ0 + γ1 Reti. especially for a develop. market averages. ple t-test to compare individual stock returns before SETS to ing market such as JSE.1) model. so we need kets. However. variant of the model should be estimated. but uses close-to-close returns. Reti. rather other hand.F. liquidity is stock assumes its own idiosyncratic risk. we condition on value-weighted returns (measured on tically significant before SETS. Dicle. with its own idiosyn- important for smaller stocks only. espe- tistical significance between returns and liquidity exists only cially in emerging markets.t−1 + β2 Reti. implying similar return levels before and after SETS. efficiency evaluation across emerging markets.t−1 + γ8 PRSi. since the effect of this is also true for overnight returns. We therefore evaluate factors post-SETS. However. In effect.

0013 0.0000 0.0005 0.0000*** 0.0000 0. W refers to either equally weighted (EW) or value weighted (VW).0009*** 0.0005 −0.3170*** 0.3175*** 0.0005 −0. “*”.1013*** 0.2226*** 0. 5% and 1% level.0007 −0.1434*** 0.0000 0.0005 −0.0011 0.0000 −0.1926*** β2 0.1106*** β3 −0.0112 0. (5): WRt = (β0 + β1 WRt−1 + β2 WRt−2 + β3 PRSt + β4 PRSt−1 )•(1 − SETS) + (γ 0 + γ 1 WRt−1 + γ 2 WRt−2 + γ 3 PRSt + γ 4 PRSt−1 )•(SETS) + εt where εt ∼N(0. J.0115 0.0000*** 0.2198*** γ2 0.0000*** 0.0031** −0.3383*** 0. “**” and “***” refer to statistical significance at the 10%.3386*** 0.0003 0.0023*** −0.0006 0.0013 0.0009*** 0. Dicle. δ2t ) and δ2t = λ0 + λ1 ε2t−1 + λ2 δ2t−1 .0070*** −0.0002 0.0009 −0.1075*** 0.0837*** 0.0066*** −0.0411 0.0010 −0.0004 −0. Data are through Reuters.0026** −0.0010*** 0.0000*** 0.1105*** 0.0704*** 0.0001 −0.2239*** 0.9196*** 0.1993*** 0.0032* δ2t λ0 0.0867*** 0.8668*** 0.0834*** 0.0003** −0.0010*** 0.0402 0. proportional effective spread and proportional quoted spread.1370*** 0.3152*** 0.2312*** 0.0833*** 0.1508*** 0.8681*** 0.9200*** 0.1120*** 0.0000* 0. R refers to either close-to-close (RCC) or open-to-close (ROC) returns.0012*** 0.9127*** 0.1018*** 0.3330*** 0.0010*** 0.2066*** 0.0000*** 0.2037*** 0.0000*** 0.0047*** γ4 0.0011 0.0002 0.0008 −0.8284*** 0.0000*** λ1 0. The results provided in this table are obtained using Eq.0700*** 0.0401 0.0009*** 0.0892*** γ3 0.0009*** 0.1325*** 0.1124*** 0.0009*** 0.0012*** 0. PRS refers to the measure of liquidity: effective spread.0000*** 0.0008 0.1417*** 0.0000 0.0003** −0.0867*** 0.0018* 0. 210 Table 3 Evaluation of market level returns for the Johannesburg Stock Exchange.0002 −0.0005 −0.2297*** 0.1933*** 0.9183*** 0.0009*** γ1 0.9129*** 0.0073*** Post-SETS γ0 0. likewise.0025** 0.0010 −0.0005 −0.0012*** 0.0000*** 0.0576** 0.1126*** λ2 0.0000 0.0576** 0.0000*** 0. Levendis / Review of Development Finance 3 (2013) 204–213 2007.0006*** 0.F.0000* 0.9169*** 0.1999*** 0.0002 β1 0.0009 −0.0625** 0.0006 −0.0160*** β4 0.0009*** 0.0006 0.0700*** 0.0006*** 0.0711*** 0.1375*** 0.1112*** 0.0009*** 0.8298*** 0. Effective spread Proportional effective spread Proportional quoted spread RCC ROC RCC ROC RCC ROC VW EW VW EW VW EW VW EW VW EW VW EW Pre-SETS β0 −0.1525*** 0.0002 −0.0006*** 0.8696*** .1088*** 0.0751*** 0.0110 0.0009*** 0.1372*** 0.2232*** 0. using the Quotecenter application for the period January 1997 until December M.0036*** −0.8668*** 0.0001 −0.1085*** 0.0015** 0.0752*** 0.0000*** 0.0000*** 0. respectively.1001*** 0.0004 −0.1102*** 0.0059*** −0.

0008** Post-SETS γ0 0. Specifically. SETS exhibited a significant impact on JSE market Granger-caused the Johannesburg market prior to or fol- stocks.t−2 ) · (SETS) Hypothesis 3.4421*** 0.0469*** −0. We perform this test for 30 different The idea is that a more efficient trading system would allow for exchanges. that SETS had a differential impact on stocks.0005*** 0.0334*** γ −0. tained in Table 5.0008** β1 −0.0338*** 0.F.0003 −0.0002** β9 0.0205*** γ 0.0005*** 0. σt2 ) σt2 = λ0 + λ1 ε2t−1 + λ2 σt−1 2 (7) Columns (3) and (4) of Table 4 allow for the investigation where Foreign refers to market returns in the foreign mar- of Hypothesis 3.0017*** −0. In order to investigate this hypothesis.0001** −0.0280*** 0.0386*** 0. not Summary of the results from thirty foreign markets is con- statistically significant. Accordingly. δ2t ) and δ2t = λ0 + λ1 ε2t−1 + λ2 δ2t−1 . the intraday predictabil- Hypothesis 4. Dependent variable (1) (2) (3) (4) ROC RCC ROC RCC Pre-SETS β0 −0. so that their employed instead of a t-test since GARCH is superior for mod- returns would be higher.0385*** 0.0258*** 0.1) is greater investments in smaller.t−1 + λ2 Reti.0281*** 0. however.0006*** 0. the lagged returns of individual WRt = (β0 + β1 WRt−1 + β2 WRt−2 + β3 PRSt + β4 PRSt−1 stocks increase and change sign.0020*** γ1 −0.0013*** γ 10 −0. This would imply that JSE + β5 Foreigni. The coefficients on the “small” variable.t−2 + λ9 Smalli + λ10 Largei )•(SETS) + εt where εt ∼N(0.2662*** 0. but lowing the introduction of SETS.0101* γ 0.0799*** 0. Data are through Reuters.2662*** γ4 0.0002 Market returns have both contemporaneous and lagged reduced that of overnight returns.0001 β8 −0. Dicle.0001* −0.0338*** 0.0240*** −0. and this effect is smaller for overnight returns.0004*** β7 −0.0000 0.0004*** −0.t−1 + λ8 PRSi. SETS had a positive impact on smaller stocks’ returns.t + λ7 PRSi.0584*** 0. The results provided in this table are obtained using Eq.0010*** 0.t−2 + β3 VWRett + β4 VWRett−1 + β5 VWRett−2 + β6 PRSi.t−2 + β9 Smalli + β10 Largei )• (1 − SETS) + (λi + λ0 + λ1 Reti.t−2 + λ3 VWRett + λ4 VWRett−1 + λ5 VWRett−2 + λ6 PRSi.t−1 + β2 Reti.0038 γ −0. (6): Reti.0334*** −0.0001 0. level or at the individual stock level.0000 0.0001 0.0616*** 0.0462*** 0.3275*** 0.0462*** β −0.0001 −0. it reports the . respectively.t = (βi + β0 + β1 Reti. increased after SETS (the coefficient almost doubles).0006** 0. we estimate the fol- the lagged market returns lose their statistical significance and lowing model: contemporaneous returns have lower coefficients compared to intraday returns. SETS changed whether foreign markets ity of individual stock returns (using its own lagged returns) is Granger-cause the JSE. M. more obscure stocks.0306*** β 0. using the Quotecenter application for the period January 1997 until December 2007. GARCH (1.0258*** 0. For overnight returns. eling daily equity returns.t−2 ) · (1 − SETS) stocks are affected by their own market during the trading day. ket being considered.0003* 0. RCC (ROC) refers to close (open) -to-close returns.0003 −0.4420*** 0. between the pre-SETS and post-SETS samples are. + εt εt ∼N(0.0205*** −0.0001* γ8 0. 5% and 1% level.0007*** 0.0651*** 0.0005*** 0. Levendis / Review of Development Finance 3 (2013) 204–213 211 Table 4 Evaluation of individual stock level returns for the Johannesburg Stock Exchange listed stocks before and after SETS. This table simply shows whether the foreign In summary.0101* 0.0017*** −0.t−1 + β6 Foreigni.t−1 + β8 PRSi. J. It increased the predictability of intraday returns.0651*** 0.0038 0.0468*** β2 −0.0001** γ7 −0. but only its contem.3581*** β 0.0002** 0.0009*** 0.0240*** β 0.0615*** 0. This find- ing contradicts those at the market level. We cannot conclude that the effects on intraday individual stock returns.0001 γ9 −0. “*”.3275*** 0.0002** 0.t + β7 PRSi.0008*** 0.0583*** 0.3582*** 0. JSE has become more efficient after SETS either at the market poraneous effect is statistically significant for overnight returns.0307*** 0.0010** β10 −0.0009*** −0. + (γ0 + γ1 WRt−1 + γ2 WRt−2 + γ3 PRSt + γ4 PRSt−1 + γ5 Foreigni. After controlling for the market’s return.t−1 + γ6 Foreigni. “**” and “***” refer to statistical significance at the 10%.0800*** 0. Similar to the previous tests.0002** −0.0017*** 0.0010*** 0.

755*** 0.125 Singapore SE of Singapore 24.425 0.003 5.066 8.602 1.058** 8.870 3. The JSE is more liquid after SETS.667* 42. 5% and 1% level. W refers to either equally weighted (EW) or value weighted (VW).915 0.074*** 15.366 2. more listed companies.730*** 0.809 6. EW VW Country Exchange ROC RCC ROC RCC Before After Before After Before After Before After Australia National A.451*** 0.098 0.814 5.101 8.247 0.607** 23.325*** 7.037** 0. “*”.879 5.933 2.463 3.239** 1.212 M.777** 0. three of the four measures of market.T.669 Thailand SE of Thailand 1. while only 7 and there are more trades at JSE after SETS.723*** 8.089*** 0.494** 0. returns.318 6.010 0.882* 1.196*** 6. We also do the same for the post- SETS coefficients.918 2.219 0.395 0.190 33. using the Quotecenter application for the period January 1997 until December 2007.866* 3.344*** 15.127** 0.982* 2.920*** 0.960 Austria Vienna SE 7.325 1.729*** 6. 39. and liquidity). trading is cheaper.476* 0.309 2.825 5.663 0.306 Hong Kong Hong Kong SE 1.398*** 6. Using close-to-close efficiency. 23.955*** 6.297*** 166.467 3.449 10.659 2.598** Germany Frankfurt SE 10.962 Indonesia Jakarta SE 4.079 Canada Toronto SE 340. 16 exchanges caused the JSE before SETS.861 Jordan Amman F.109 1.005 Italy Milano SE 10.407 0.556*** 1.946 0.239 0.334 4. Data are through Reuters.964 0.824*** results of a Wald Chi2 test of β5 = β6 = 0 against the alternative 6. 26 of 30 exchanges Overall.169 4. respectively.823 4. We posit that this is Granger-caused the JSE before SETS.013*** 5. With the new trading platform.066*** 0.640** 7.884*** 7.945** Greece Athens SE 3.452 0.130*** 1.075*** 6.931 2.418 3.990 47.809 1.196*** 1.771 16. however.583 2.152 2. wider investment opportunity set.152 2. Stock Exchange.115*** 7.365 1.469 3.925*** 3. J.910 12.760 10.212 0.230 0.542v 7. We then compare whether the JSE The adoption of the SETS trading platform was supposed to was Granger-caused by more foreign markets before or after represent a watershed moment in the history of the Johannesburg SETS.491** 0. it would open-to-close returns.230 1.900 0.560 24.154*** 15.096 19.242 3. the γs.529*** 73. Levendis / Review of Development Finance 3 (2013) 204–213 Table 5 Evaluation of market level causal relationship of Johannesburg Stock Exchange with foreign equity markets before and after SETS.559*** Korea Korea SE 4.865 Spain Barcelona SE 33. of SETS has actually increased the independence of the Johan.020 12.773*** 9. Dicle.477*** 11.219 2. we interpret that the incorporation lation of market efficiency.226* 2.076* 57.468 39.288*** 33.309 4.492 0.416 15.443** 7.319*** 4.543** India Bombay SE 0. average daily returns are higher. and only 7 did after.343*** China Shanghai SE 1.887*** 130.029 1. 8.218 14.773 4.504*** 6.428** Japan Tokyo SE 13.467 9.595 17.031** Taiwan Taiwan SE 5.292*** 49. the results are less consistent. Using close-to-close returns.704 2.623*** 0. “**” and “***” refer to statistical significance at the 10%.894*** 2. and 13 did after SETS.853* US New York SE 356.F.012*** 0.597 3. all would imply more market efficiency.630** 14.279 10. This is mainly because of .224*** 5.005 29.582 Norway Oslo SE 8.242*** 11.729 8.691* 74.081*** 10.552 8. δ2t ) and δ2t = λ0 + λ1 ε2t−1 + λ2 δ2t−1 .023** 0. exchanges Granger-caused the JSE before SETS.214 0.633** 0.521** 2. likewise.985 0.135** 7.981* 0.126*** 1.228 Sweden Stockholm SE 21. The results provided in this table are obtained using Eq.892 0.373*** 46.865*** 1. Another expected improvement of SETS would be that the nesburg Stock Exchange.050*** 2. R refers to either close-to-close (RCC) or open-to-close (ROC) returns.238 Belgium Euronext Brussels 1.549** 9.557 Denmark Copenhagen SE 2.591 3.985*** 8.488 6.727** 17. In summary.632 9. this is a clear vio- Based on these results.450* 6.441** France Euronext Paris 0.224** 0.772* Switzerland CHE SWX Swiss E.076 8.686 0. The JSE has Using equal-weighted open-to-close returns. Using for the associated risk.273 1.708 0.669** New Zealand New Zealand SE 1.297** 0. market-wide and individual-level of foreign exchanges.706*** 5. and 13 did faster trading and more trade (evidenced with trading activity after SETS.064 0.503 2.099 0.814*** 5.061** 3.751 6. stock returns are still somewhat predictable.103 1.108 24.258 3.979*** 0.780 7.113 2.703*** 0.057*** 11. thereby increasing its diversification JSE would offer unique opportunities to investors with the benefits for international investors.920*** 2.755*** 132.116 Peru Lima SE 3.148*** 1. Higher numbers of investors.391*** 2.682 0.476 16.324** Finland Helsinki SE 6.998 0.731 0.982 25. did after SETS.169 0.118 UK SETS 33.687 60.799** 2.544 3.990 0. (7): WRt = (β0 + β1 WRt−1 + β2 WRt−2 + β3 PRSt + β4 PRSt−1 + β5 Foreignt−1 + β6 Foreignt−2 )•(1 − SETS) + (γ 0 + γ 1 WRt−1 + γ 2 WRt−2 + γ 3 PRSt + γ 4 PRSt−1 + γ 5 Foreignt−1 + γ 6 Foreignt−2 )•(SETS) + εt where εt ∼N(0.640** 3. Contrary level returns indicate that the JSE has become more independent to our expectations. 18 of 30 nearly doubled its trading activity (volume).795*** 4.129 0.482 8.345 1.215** 12. Using mainly because the returns are increased to the levels demanded value-weighted returns.361*** 106.785 Malaysia Kuala Lumpur SE 7.360 7.010 3.002* 2.380** 1.M.977 1.008 Netherlands Euronext Amsterdam 2.127 0.485 11.360 0.477*** 1.486 15.414*** 0.267* 73.742*** 5.030** 1.392*** 9.674 47. Conclusion that either β5 or β6 = / 0.605 10.570** 1. 10 of 30 exchanges Granger-caused the also be expected that there would be improvements in market JSE before SETS.

W. M. Harvey. Efficient capital markets: a review of theory and empirical Piesse.. E. 1970. Equity market integration versus segmentation in eral causal relationship? Journal of Comparative Policy Analysis 12 (3). (3)..R.. Journal of Finance 25 (2)...A. 1783–1831. C. C. 131–140. Shleifer. South African Journal of Economics 74 (2). Hearn. South African Journal of Economics 73 (1). K. A. Structural change and the bid-ask spread: evidence from Chuhan. Levendis / Review of Development Finance 3 (2013) 204–213 213 the expectation that more local companies would list with the Griffin. M. 2005.. 1998. 2009.. nal of Financial Economics 56 (1). B. Military spending and GDP growth: is there a gen. N. lessons from emerging markets. securities exchange. 2010. J. 1998..J. Smith. Review of Financial Studies 20 (6). M. tion and causality tests. Law and finance.. African Finance Journal 8 (2).. Research 32 (1). 2001. 3–27.. 2002. R. Intraday stealth trading: which Liu. 2002. Working Paper. As expected. from http://news. 2006. A. J. R. 2006.. Vishny. Piesse. Integration and volatility spillovers in References African equity markets: evidence from Namibia and South Africa. South African Journal of Economics 72 (4). Middle Eastern Finance and Economics. Zimbabwe Appiah-Kusi. L. B... contracts and their underlying spot market indices.M. Emerging Markets Review 2 (3). 59–76.D. global perspective. Negash. 33–63. Journal of Studies in Economics and Econometrics money supply and changes in share prices: the semi-strong form efficiency 19 (3). 2008. 2012.J. Blau. K. R.. T. Commonality in liquidity. 2000. Jefferis.. Shleifer.. 684–707. 1995. market participation. of the Johannesburg Stock Exchange revisited. Kelly. Van Ness. 2010. 2007. E. Michello. Stock market development and market efficiency on China stock trades move prices during periods of high volume? Journal of Financial market... Commonality in liquidity: a Journal for Studies in Economics and Econometrics 29 (2).R.N. P. South African Journal of Economics 76 better diversification opportunities for international investors. Jefferis. 2010. Dicle. Liquidity and expected returns: tection and corporate governance..... South African Journal of Business Man- 171–185.. Journal of Political Economy 106 (6). 61–76. A. Smit.W. South Africa Upgrades Trading System.D. M. and Johannesburg stock exchanges.. Saharan Africa. Hearn. Volatility and volatility spill-overs in emerging markets: the case of the African stock markets. 131–140. Rudolph. 2004. Roll...F. F. 1999. Ferret...stm La Porta. G. 69–90. F. J.. 23–40. Investor pro- Bekaert. The reaction of stock prices to unexpected infla- tion. Fama. It now offers in SADC’s smallest equity markets. Ward.. Chung. three dominant markets of the Southern African Customs Union: cointegra- 311–345. the JSE became more Hearn.F. 2006. R.. B. The relationship between changes in the efficient market: a review. Mbeki. An event study of the Botswana. African Finance Journal 3 (1). La Porta. Pérignon.. Regional integration of equity markets in Sub- work. Jour.. 851–882. F. A. Subrahmanyam.. Lopez-de-Silanes. . A. 1995.F. 3225–3277. P. Floros. 1998. 1113. Lopez-de-Silanes... Presidential Speeches.. In: the Johannesburg stock exchange (JSE). BBC. Journal for Studies in Thompson. Structural changes and dating of stock mar- ket liberalization: evidence from the JSE securities exchange South Africa. Humavindu. Lundblad.. Dicle. Brockman. Address at the Launch of JSE Sets and Infowiz.A. D. 36–53. The relationship between international equity market behaviour and the JSE. R. G. C... 1992. Okeahalam. Studies in Economics and Econometrics 30 (1). Page.. Ekonomia 2 (2). Applied Economics 34 (14). R. Review of Financial Studies 23 (8).D.M. B. Sources of portfolio investment in emerging markets. M. J. R.. 3– B. 2005. B. 383–417.. 2000. Construction of all SADC stock markets indices. Co-integration of South African index futures Samouilhan. Business Management 26. Retrieved December 5. Van Ness.Y. Piesse.M. Do market efficiency measures JSE due to lower trading costs (higher liquidity) and higher yield correct inferences? A comparison of developed and emerging markets... Vishny.. Opportunities and costs of portfolio: diversification independent after the incorporation of SETS... Biekpe. 1–21. 280–291. D. World Bank.. Dicle. Journal of Financial and Quantitative Analysis 44 (4). Makina. Nardari. J. The Johannesburg stock exchange as an Economics and Econometrics 22 (1). G. International Economics Department. N. agement 30 (4). Journal of Financial Economics 58 (1–2). Capitalisation and weak-form efficiency in the JSE Auret. Glass. C. 399–426. T. J. South African Chordia. T. P. 2002. South African Journal of Tyandela.. 2001. 248–260.J. C.. 31–51. 1711–1722.. C. 2009. 7.V. Pescetto..