Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
Financial Market Design and Equity Premium: Electronic versus Floor Trading
Pankaj Jain*
January 2004
JEL classifications: G14, G15 Keywords: stock exchanges, computerized trading ______________________
* Fogelman College of Business and Economics, The University of Memphis, Memphis, TN 38152, Phone: (901) 678-3810, (901) 327-1404 Fax:(901) 678-0839, Email: [email protected]. Full paper available on Web: http://www.people.memphis.edu/~pjain The paper is abstracted from my doctoral dissertation at Indiana University. I would like to thank Utpal Bhattacharya, Ian Domowitz, Craig Holden, Robert Jennings, an anonymous referee, seminar participants at Indiana University, University of South Carolina, Midwest Financial Association Meeting 2002, Financial Management Association Meeting 2002, Eastern Finance Association Meeting 2003, and American Finance Association Meeting 2004 for their comments and suggestions. Financial support from the Center of International Business Education and Research at Indiana University is gratefully acknowledged. All errors are my responsibility.
1
Financial Market Design and Equity Premium:
Electronic versus Floor Trading
Abstract
We assemble the dates of announcement and actual introduction of electronic trading by the
leading exchange of 120 countries to examine the long term and medium term impact of automation.
Estimates from dividend growth model as well as international CAPM suggest a significant decline
in expected returns after the introduction of electronic trading in the world’s equity markets,
especially in the developing nations. Consistent with this reduction in equity premium in the long
run, there is a positive short-term price reaction to the switch. These findings are sustained even after
controlling for risk factors, economic growth, financial integration, and economic liberalization.
Further analysis of trading turnover supports the notion that electronic trading enhances the liquidity
and informativeness of stock markets leading to a reduction in cost of capital.
2
Rapid technological advancements in telecommunications and the Internet are transforming
the basic business model of a stock exchange. In an increasingly competitive world with low barriers
to entry, exchange-owners are rapidly recognizing that efficient market design and trading
mechanism are keys to winning higher market shares for both trading volume and number of listings.
Scores of stock exchanges around the world have abolished their trading floors on which brokers
manually matched orders using an open-outcry system. Fully automated and transparent electronic
systems have replaced those outcry mechanisms.1 This paper empirically examines whether this
major change in market-microstructure has helped the listed firms lower their cost of equity because
of improvements in liquidity and the informational environment in the secondary market.
Computerization of trading can improve liquidity in secondary markets through several
means. Electronic trading systems can significantly lower investors’ trading costs (spreads, fees,
brokerage, and commissions) and increase the amount of publicly available information about a
stock’s demand and supply. In addition, electronic systems are capable of attracting new pools of
liquidity by providing affordable remote access to investors and by retaining unexecuted orders in a
consolidated order book for possible matching with future orders. Liquidity begets liquidity and
creates network externalities. It certainly reduces barriers to market-making activity, allowing
individual investors to compete with brokers having exchange seats. On automated electronic trading
systems, profit-seeking value traders can closely monitor the market and become suppliers of
liquidity even without their presence on the trading floor. This phenomenon is further facilitated by
the manifestly higher speed of execution and settlement of trades on electronic systems.
Electronic systems are also more transparent than trading floors in displaying detailed order-
flow information such as quotes, depths, and recent transactions from the limit order book to the
market participants in real time. Higher ex-ante transparency reduces the adverse selection problem
1 During the last 5 years most new exchanges started as fully electronic because the costs of developing, operating and
3
(Pagano and Roell (1996)). This information can also be archived more efficiently in electronic
formats and then used ex-post by regulators in audit trails to penalize abusive practices such as
insider trading and front running customers’ orders. Chung and Postelnicu (2003) attribute a lack of
automation as the key reason for the suspicions of front-running by specialists recently investigated
by the NYSE. Exchange officials in both developed and emerging markets such as Germany, Italy,
and Pakistan have commonly cited cost reduction and investor protection as the main reasons cited
for switching from floor to electronic trading.
Several articles in the academic literature (See Domowitz and Steil (1999) and Jain (2001)
for a detailed review and international evidence) and the financial press have documented these
merits of electronic systems in day-to-day trading.2 However, there is relatively little research on the
long-term effects of electronic trading on the equity premium. Improvements in liquidity and
informativeness of stock markets can have far reaching effects on the equity premium. Amihud and
Mendelson (1986) show that investors expect lower returns for stocks with higher liquidity.3 This is
possible because a lower gross return can still yield them the same net return if they face lower
transaction costs. O’Hara and Easley (2002) show that investor demand a higher return to hold
stocks with greater private information. Although their focus is on private versus public information
about cash flows of the stock, greater public information about the order-flow is likely to have
similar effects too. For example, Franke and Hess (2000) examine the role of order-flow
transparency in the information diffusion process. Better order-flow information can enhance the
traders’ ability to react quickly to any new cash-flow information and protect them against adverse
trading with insiders. Ohara (2003) argues that particular trading systems may provide more
information or better information (pp 1342). This can facilitate price discovery, lowers traders’ risks
maintaining electronic systems are typically far lower vis-à-vis floor trading. 2 See Appendix 1 for a comparison of the two types of systems. 3 They predict that observed gross return is an increasing and concave function of relative spread.
4
and therefore affect asset returns. Therefore, one can expect lower equity premium on the electronic
exchanges if they offer better liquidity, lower trading costs and better information to traders than do
the floor based exchanges.
Domowitz and Stiel (2001) document the economic significance of this effect through an
examination of US and some European trading and capital cost data from the period 1996-98.
However, none of the exchanges explicitly switched from floor to electronic trading in the time
period of their study. A related set of papers by Amihud, Mendelson, and Lauterbach (1997), Kalay,
Wei and Wohl (2002), and Muscarella and Piwowar (2001) find that the move from call-auction
trading to continuous trading by the Israeli and the French exchanges resulted in positive price
reactions.
However, there is no dearth of arguments that challenge this standpoint. The world’s largest
stock exchange, NYSE, still relies heavily on floor trading. Beneviste, Marcus, and Wilhelm (1992)
model a situation where repeated trading on the floor between market participants such as the
specialist and floor brokers builds reputation and reduces information asymmetry. Thus, they predict
that bid-ask spread should be lower on the floor although we argue that electronic systems have the
potential to replicate the reputation effects by disclosing the identity of counterparties ex-ante on the
trading screen as is done on yahoo.com, amazon.com, and ebay.com auctions. Schmidt et al. (1993)
report that the spreads on German regional floor exchanges were lower than those on their interbank
electronic trading system. Venkataraman (2001) finds that spreads for similar firms are indeed wider
on the electronic Paris Bourse compared to the floor based NYSE although the study acknowledges
the difficulties in controlling for differences in insider trading laws, competition for order flow, and
trading volume between United States and France. Even if we were to discount the results in these
studies as sample specific and believe that electronic markets have better liquidity, Constantinides
(1986), Vayanos (1998), Kadlec and McConnell (1994), and Barclay, Kandel, and Marx (1998)
5
argue that bid–ask spreads have only a second-order effect on expected returns and such effect is
negligible and insignificant. Thus, it appears meaningful to empirically address these issues with
seemingly contradictory views in a multi-country setting that can provide a bird’s eye view as well
as detailed insights into the benefits of automation.
Our study contributes to the growing literature that examines the impact of market
microstructure on asset pricing. We empirically investigate the changes in market design of stock
exchanges around the world and the effect of these changes on the cost of equity for listed firms. The
main hypothesis tested in this paper is that automation of the trading process leads to a reduction in
the equity premium that investors demand. We gather new information on stock exchanges in 120
countries and find that the leading stock exchange in 101 of those countries has introduced
automated screen-based electronic trading4 within the last 25 years. Of these 101 exchanges 85 are
now fully electronic with no floor-trading. These events provide natural experiments for testing the
impact of this major aspect of market design on the equity premium. The advantages of using such a
comprehensive sample are manifold. First, it enhances our understanding as to how many exchanges
around the World perceive each system to be relatively stronger than the other; is the reliance on
trading floor by the NYSE an exception or the rule? Second, it lets us revisit the accumulated
evidence from the few single or dual country studies to examine if their findings about the positive
stock price reactions are pervasive or limited in their generality. Previous studies, cited above,
examined a move from discrete call trading to continuous trading. Is a similar effect obtained when
exchanges switch from continuous floor trading to continuous electronic trading? Third, by looking
at the long term patterns in cost of equity, we are able to test whether the positive price reaction to
the switch is a sign of temporary optimism by the investors who were newly afforded remote access
4 In this paper the term ‘electronic trading’ is defined as automated matching and execution of trades by a computer algorithm. If the orders on any exchange are only routed electronically and brokers’ intervention is required for final execution of trades, then such an exchange is not classified as an electronic exchange.
6
and greater control in equity trading or if the price reaction is permanent in nature which would
imply a decline in the equity premium in the long run? Fourth, the wide cross-sectional variety in the
financial markets helps us address the possibility that electronic trading may be preferable in certain
financial and legal environments whereas floor trading might be better in others. Franke and Hess
(2000) envisage that the relative importance of information value provided by an insight into the
limit order book in electronic systems, compared to information value of observing traders on the
floor, is less in economies where the intensity of private and public information arrival is very high.
Thus, the decline in the equity premium due to trade automation might be much lower for such
economies. Similarly, Madhavan and Sofianos (1998) show that the value of human intermediation
is inversely related to inherent liquidity in a stock.5 Theissen (2001) reports that electronic trading is
more attractive (lowers spreads) for high volume stocks but not necessarily for low volume stocks in
Germany. Therefore, it is possible that switching to electronic trading might actually hurt the less
liquid markets. The cross-sectional variation in our sample lets us document the economic conditions
and market characteristics that determine the magnitude of cost reduction (or increase) resulting
from trade automation.
We analyze the time series of monthly returns on stock exchanges of 56 countries and annual
returns for 15 additional countries from January 1973 to August 2001 – a period that spans both a
floor trading regime and an electronic trading regime for these countries.6 Several tests are
conducted to verify whether the introduction of electronic trading is associated with a lower cost of
equity for listed firms. We use two different methods to estimate the equity premium – the dividends
growth model as used by Fama and French (2002) and an international asset-pricing model
suggested by Bekaert and Harvey (1995) to control for worldwide and local risk factors. We also
5 They show that the median specialist participation rate at the NYSE drops from 54.1% for illiquid stocks to 15.4% for highly liquid stocks. 6 In many cases the data starts at a later date. We stop at August 2001 to avoid the impact of the 9/11 shock.
7
look at immediate price reaction to exchange automation. A positive price reaction around the
introduction of electronic trading will be consistent with a reduction in the equity premium in the
long run. The main results of the study are discussed below.
The paper finds that electronic trading is associated with a lower cost of equity in the long
term and a positive price reaction in the short term. Dividend yields fell by 0.07% per month or 90
basis points per annum after introduction of electronic trading. The international capital asset pricing
model, which is based on realized equity returns, suggests a much sharper decline. We use a
GARCH model in which we control for variables traditionally found to be related to stock market
returns. These include a systematic world-market risk factor, an idiosyncratic country-specific risk
factor, liberalization, financial integration, development of the economy, and a time trend in returns
and also allow for time variation in betas. The regression model has a negative and statistically
significant coefficient for electronic trading. The reduction is more pronounced in emerging markets
(0.81% per month) than in developed markets (0.15% per month).7 Rather than emphasizing on
these average point estimates, this study looks at 71 exchanges individually as there are significant
cross-county differences. This country-by-country analysis also enables the use of same listed firms
on an exchange before and after the introduction of electronic trading and thus avoids the problem of
imperfect matching of stocks in electronic and floor regimes. The paper finds that, depending on the
measure used, between 62 and 83% of the regime shifts are associated with a reduction in the cost of
equity. The results can partly be explained by improvements in the liquidity measures. Consistent
with several other studies we find that the relative monthly trading turnover increases by 3.04% of
market capitalization after the introduction of electronic trading.
Consistent with the reduction in the equity premium, we find that there is a positive price
jump around the dates of announcement and implementation of electronic trading. Average excess
8
over world (AR) return in the announcement month is 7.87%. AR in the month of actual
implementation is 1.20%. The cumulative excess over world return (CAR) is 10.66% and 2.73%
from six months before to one month after the date of announcement and implementation
respectively.
The remainder of the paper is organized as follows. The next section contains this study’s
hypotheses. Data sources are outlined in section II. Empirical methodology and results are presented
in the following three sections. Section VI concludes.
I. Testable Hypotheses
Several single or dual country studies on floor trading versus electronic trading find that the
latter is associated with higher liquidity in secondary markets. Domowitz and Steil (1999) give an
excellent summary and a somewhat skeptical review of this literature for specific markets such as
the U.S.A., Britain, France, Germany, Singapore, New Zealand, Australia, and India. Moreover,
Venkataraman’s (2001) finding that floor trading produces narrower spreads casts serious doubts on
the efficacy of electronic trading. Instead of focusing on the traditional liquidity measures, we
attempt to address this debate by investigating the impact of automation on a much more important
phenomenon in finance – the equity premium. The primary hypotheses tested in this paper is as
follows:
H10 : Improvement in stock market liquidity induced by introduction of electronic trading lowers the
equity premium and, thus, reduces the cost of equity for listed firms in the long-run.
In our hypothesis, the inverse relationship between liquidity and the equity premium is based
on the unequivocal predictions and empirical findings in many studies including Amihud and
Mendelson (1986), Brennan and Subrahmanyam (1996), Datar, Naik and Radcliffe (1998), Dimson
7 The reduction is more pronounced in both absolute terms and proportional terms. The average dollar returns before introduction of electronic trading are 1.39% in developed markets and 2.14% in emerging.
9
and Hanke (2000), Jones (2001), Pastor and Stambaugh (2001), Swan and Westerholm (2002), and
O’hara (2003).
An immediate implication of the first hypothesis is that when securities move from a high
equity premium regime to a lower equity premium regime, they should experience a positive price
response around the introduction of electronic trading. This is the crux of our next hypothesis:
H20 : When stocks move from floor to electronic trading they will have a positive price reaction in
the short-run.
Tests similar to the second hypothesis have been performed on a stand-alone basis by Amihud,
Mendelson, and Lauterbach (1997), Kalay, Wei and Wohl (2002), and Muscarella and Piwowar
(2001) for moves from discrete call-auction trading to continuous trading by the Tel Aviv stock
exchange and the Paris Bourse. However, we believe that it is important to the first and the second
hypothesis together to rule out some alternative explanations for either one. An alternative
explanation for the evidence supporting the first hypothesis can be a reversal in long-term stock
market returns co-incidental with the introduction of electronic trading.8 However, such a
phenomenon will reject the second hypothesis. Similarly, an alternative explanation for evidence
supporting the second hypothesis can be a coincidence of positive fundamental news and automation
of trading. However, this type of situation would not prevent the rejection of the first hypothesis.
Thus, a failure to reject the first and the second hypothesis will be strong indication that electronic
trading leads to lower cost of equity.
In out third and final hypothesis, we use the cross-sectional variation in the sample to identify
the economic and market specific determinants of the magnitude of reduction (or increase) in the cost of
equity resulting from trade automation:
8 More specifically, if floor trading overlaps a bullish period and electronic trading overlaps a bearish period in each country then we will find spurious support for the first hypothesis. We thank the referee for pointing this out.
10
H30 : The magnitude of the changes in the equity premium after introduction of electronic trading
are a function of a country’s economic and financial environment and the characteristics of their
stocks.
Academic research and anecdotal evidence suggests that not all stocks or exchanges benefit from
electronic trading. Intensity of information arrival (Franke and Hess (2000)), inherent liquidity of a
stock or the stock market (Madhavan and Sofianos (1998) and Theissen (2001)), level of economic
development, enforcement of insider trading laws and other aspects of a country’s economic and
financial environment can affect the relative importance of information value provided by an insight
into the limit order book in electronic systems compared to information value of observing traders
on the floor. Thus, the magnitude of decline in the equity premium due to trade automation might
depend on these characteristics. Particularly in the emerging markets, if formal laws are ineffective
because enforcement is difficult then full automation and transparency of order flow can be an
alternative way of reducing information asymmetry. We test these three hypotheses using the
international data described in the next section.
II. Data
Our sample starts with a set of 120 countries around the world. We first gather the dates of
introduction of electronic trading by the largest exchange in each country.9 This information is
obtained from 12 annual volumes of Handbook of Worlds Stock, Commodities and Derivatives
Markets from 1990 to 2001, and is cross-checked with information given on exchanges’ home pages
on the Internet. We sent emails to the exchange officials for confirmation and clarification. We also
collect announcement dates for automation from Lexis Nexis news retrieval service.
9 In most cases floor and electronic co-exist for some time and the complete switch takes place only after the abolition of trading floor. We assume that the improvement in liquidity is brought about by the introduction of electronic trading and not by disabling the choice between floor and electronic from the traders at the time of floor abolition. Therefore, we concentrate on the date on which the option to trade electronically became available to the investors and present floor-abolition information in Figure 1 only for completeness.
11
In order to compare the equity premium before and after introduction of electronic trading,
we are able to obtain dividend yields (53 countries) from Datastream International (DSI) and
monthly (56 countries) or annual (71 countries) stock market indices from Morgan Stanley Capital
International (MSCI), International Finance Corporation (IFC), DSI or directly from the exchanges.
The indices do not start on the same date for every country. The earliest starting point is December
1969. The number of months for which returns data is available ranges from a low of 49 months
(adding both floor and electronic months) for Croatia and Latvia to a high of 383 months each for 18
developed nations. We filter-out data for 6 outlier exchange-months that represent greater than 50%
devaluation of the country’s currency due to unusual circumstances such as a currency Crisis.10 This
results in a final sample size of 12,223 monthly returns. To eliminate possible data errors we also get
rid of observations where values more than double or less than half in successive years. Inclusion of
these outlier months and potentially erroneous data slightly magnifies the magnitude of the drop in
cost of equity, which lends an even stronger support to the primary hypothesis in the paper.
We gather monthly dollar denominated market capitalization on the stock market indices on
these exchanges and dollar trading volumes from DSI. Missing observations are replaced with
annual market capitalization from The LGT Guide of World Equity Markets 1997-2001 and The
Handbook of World Stock, Commodity and Derivative Exchanges 1990-2001 in order to retain these
exchange-months in our sample. Next, we gather data on trading turnover before and after
automation for 63 exchanges and spreads before and after automation on a few exchanges from DSI
and the online archives of the World Federation of Exchanges (formerly known as the International
Federation of Stock Exchanges – FIBV) at www.fibv.com or from IFC. Missing data on turnover are
replaced with information from the Salomon Smith Barney (SSB) Guide to World Equity Markets
1997- 2001.
10 Three months of data from Argentina, two from Venezuela, and one from Indonesia are filtered out.
12
Finally, several papers such as Bakaert and Harvey (2000), Henry (2000), and Bhattacharya
and Daouk (2002) show that the liberalization of markets and the first enforcement of insider trading
laws are important variables that enhance the liquidity of stock markets and reduce the cost of equity
for firms. Therefore, we use these dates provided in the respective papers as control variables. Two
more control variables are included. Quarterly GDP data is obtained from DSI for each country to
compute the rate of economic growth which is one of the key drivers of stock markets. Additionally,
quarterly export and import data are acquired from DSI to compute the level of financial integration
because Bekaert and Harvey (2000) suggest that it affects the sensitivity of returns world market returns.
III. Does electronic trading reduce cost of equity for listed firm?
A. Global shift from floor to electronic exchanges
The pattern of global shift from floor trading to automatic screen-based trading is graphed in
Figure 1 for the leading exchanges 120 countries. The first exchange to introduce electronic trading
is the Toronto Stock Exchange in 1977. The last exchange to do so, in our sample, is Macedonia in
2001. The technology was first introduced by a US brokerage firm, Instinet, in 1969. However, the
NYSE, the leading US exchange, introduced the facility of fully automated trading, known as Direct
+, only in Dec 2000 although electronic routing of orders on SuperDOT has been in place since
1985.11 Today, the leading exchange in 101 of these 120 countries has electronic trading. Of these
101 exchanges, 85 are fully electronic with no floor trading.
[Insert Figure 1 here]
B. Measuring the equity premium and liquidity
Expected returns are estimated with six alternative measures namely monthly dividend
yields, dividend growth model, total returns including dividend and capital gains in local currency,
13
dollar-denominated total returns, excess-over-world returns, and excess-over-T-Bill returns. It was
necessary to conduct the analysis using both local currency indices and US dollar indices to rule out
the possibility that dollar appreciation would drive all the results.
The first measure is simply the dividend yield, A(Dt /Pt-1), obtained by dividing the dividend
for a period with the opening stock price for that period. The dividend yield for an index in
Datastream is the total dividend amount for the index, expressed as a percentage of the total market
value for the constituents of that index.
In a recent article, Fama and French (2002) suggest that although the dividend growth model
and average realized returns have produced similar estimates of expected US equity premium
historically, the two measures have diverged significantly in the more recent periods. They argue
that the dividend model produces estimates closer to the true expected equity premium because
average realized returns are contaminated by price jumps associated with declining discount rates.
Giving heed to their suggestion we include a dividend growth model in our analysis of the equity
premium before and after introduction of electronic trading. According to this model, the average
stock return, A(Rt), is the average dividend yield, A(Dt /Pt-1), plus average rate of capital gain,
A(GPt):
A(Rt) = A(Dt /Pt-1) + A(GPt) … (1a)
We estimate this equation from stock market indices including dividends in local currency as
well as U.S. dollars to get our second and third measure of the equity premium in Table 1.
Fama and French (2002) assume that the dividend-price ratio, Dt /Pt , is stationary (mean
reverting). Stationarity implies that if sample period is long, the compound rate of dividend growth
approaches the compound rate of capital gain. Thus an alternative estimate of expected stock return,
A(RDt), is given by the following dividend growth model:
11 This paper considers the switch to fully automated events as the relevant move. Exceptions such as SuperDOT are
14
A(RDt) = A(Dt /Pt-1) + A(GDt) … (1b)
where GDt = A(Dt – Dt-1)/Dt-1 is the growth rate of dividends. In order to arrive at the monthly
growth in dividends, we calculate the absolute amount of dollar dividends by multiplying the
percentage dividend yield for a country’s index in Datastream with the total market value for the
constituents of that index for that month.
We use equation 1(b) to obtain the fourth measure of cost of equity in Table 1. The last two
measures of the equity premium are based on excess returns. Excess-over-world return for a month
is defined as the difference between dollar-denominated return from stock market i in month t and
the return from the MSCI World Market Index in that month as follows:
Excess-over-world $ returnit = Gross $ returnit – World $ returnt (1c)
For our sixth and last measure, we compute the excess returns by subtracting the risk free
rate from the equity returns as follows:
Excess-over-TBill $ returnit = Gross $ returnit – (One-month US$ T-bill yieldt) (1d)
Finally we measure liquidity using relative turnover. Trading turnover is defined as the
monthly dollar trading volume divided by market capitalization at the end of the month.
C. Average returns and liquidity before and after introduction of electronic trading
In this section, we examine the impact of electronic trading on equity returns in 71 countries
for which returns data are available. In order to sharpen the tests and avoid confounding events such
as liberalization, we throw out the periods that are more than 10 years before or after automation.
This results in 9,052 exchange-months comprised of 4,070 floor-months and 4,982 electronic
months. In Figure 2, we compare the equity premium and turnover in floor and electronic markets.
We observe that all six alternative ways of measuring cost of equity yield the same result -
electronic trading has lower expected returns compared to the floor-trading regime. The differences
analyzed in robustness checks.
15
range from a drop of 0.04% per month for dividend measures to a drop of 1.82% for realized
dividends plus capital gains (in local currency).
Trading turnover increases from 6.14% of market capitalization per month to 9.24%. This
represents a 3.10% gain in liquidity. All changes in expected return and liquidity measures are
statistically significant at 1% level except change in excess-over-world return which is significant at
5% level.
[Insert Figure 2 here]
D. Country by country analysis
Next, we look at individual stock exchanges separately and compare the average excess
returns and liquidity before and after the introduction of electronic trading. This approach offers two
benefits. First, it ensures a more controlled experiment. Almost the same set of listed firms on an
exchange is used before and after the introduction of electronic trading. This avoids the problem of
imperfect matching of stocks in electronic and floor regimes. The differences in legal environment
and other country specific factors are also not an issue with this type of analysis. Second, it also
ensures that the results are not being driven by one or two outlier countries but is more general.
The results are presented in Table 1. The six measures of the equity premium are now placed
in six columns and turnover is in the seventh column. We find that between 62 to 83% of the regime
shifts are associated with a reduction in the cost of equity. For instance, the expected returns measure
based on dividends plus capital gains in dollars indicates that 59 of the 71 countries see a decline in
the equity premium. Automation also results in a increase (decrease) in liquidity in 75% (25%) or 47
of the 63 countries as measured by the trading turnover.
[Insert Table 1 here]
We employ a non-parametric test of statistical significance of these changes. The Wilcoxon
Mann-Whitney rank sum test makes no assumptions about the distribution of the underlying series.
16
The first step involves ordering all floor and electronic averages in a combined series and assigning
ranks to each country-regime. These ranks are then are summed separately for floor and electronic
samples. The test statistic U is the higher of the two sums. Under the null hypothesis of no change,
the expected value is E(U) = n*(n+1)/4, its standard deviation σU is the square root of
n*(n+1)*(2n+1)/24 and {U – E(U)}/ σU is distributed approximately normally N(0,1). Z values, thus
obtained, indicate that changes in four of the six expected return measures and changes in turnover
are statistically significant at 1% level and changes in dividend are significant at nearly 10% level.
Among the countries that experience reduction in dividend yield, 81% also saw an increase in
turnover. Specifically, dividend yields dropped in 38 countries of the 53 countries. Turnover data is
missing for two of these 38 countries. Eighty-one percent or 29 of the remaining 36 countries had in
increase in trading turnover. The correlation between the two variables is a negative -8%. Similarly,
among the countries that experience reduction in dividends plus capital gains, 74% saw an increase in
turnover. Thus, it appears that cost-reduction and liquidity improvement go hand in hand.
E. Regression Analysis
In this section, we conduct a regression analysis that controls for factors that have been
shown in the past studies to account for the differences in equity returns across countries. We use the
three measures of expected returns as dependent variables in three separate regressions. The
regression equation is as follows:
returnit = α + β0 electronicit + β1 worldt + β2 enforceit + β3 liberalit + β4 capit + β5 developi +
+ β6 integrateit + β7GDP_growthit + εit (2)
where returnit is either dividend yield, dividend plus growth, or excess-over-T-bill returns from stock
market i in month t, worldt is the return from the MSCI world market index in month t, electronicit is
an indicator variable that captures the trading mechanism. It takes the value 0 before the introduction
of automated electronic trading markets (this is the floor trading regime) and the value 1 after a stock
17
exchange switches to electronic platform. Enforceit takes the value 1 after the first enforcement of
insider trading laws in a county, liberalit takes the value 1 after the financial markets in a country are
liberalized12, mcapit is the market capitalization of index companies or all listed companies on stock
exchange i in month t expressed in trillions of US dollars, developi takes the value 1 if the country is
classified as a developed economy by MSCI and 0 otherwise, GDP data, available on quarterly basis,
is used to compute GDP_growthit, and integrateit is a measure of integration13 of country i with the
rest of the world at time t.
The results of this analysis are shown in Table 2, which reports the estimates for regression
equation (2). The coefficient on electronic trading is negative and statistically significant at 5% level
for all measures of expected returns, although only three measures are reported for brevity. This
suggests that the advent of electronic trading is associated with a reduction in the cost of equity.
Electronic has a coefficient of –0.0007 for dividend yields, –0.0019 for dividend growth model and
–0.0052 with excess return measure as dependent variable. These estimates support the first
hypothesis in the paper and imply a reduction in cost of equity the estimates for which range from
0.90% to 6.19% per annum. Electronic dummy has a statistically significant positive coefficient of
0.0238 in the turnover regression, which signifies substantial improvement in liquidity.
The coefficients on control variables generally have the expected signs. The coefficient on
World market returns is insignificant for dividend models but positive and highly significant for
excess return regression. Cost of equity is lower in bigger markets as indicated by the negative
coefficients on market capitalization. Enforcement of insider trading laws has negative coefficients
which is also statistically significant for the dividend growth model. Results for market liberalization
12 Liberalization refers to a process by which a government lifts barriers to capital flows and opens its stock markets to foreign investors. Stulz (1999) proposes that liberalization reduces cost of equity because of improved risk-sharing and improved corporate governance. Bekaert and Harvey (2000) and Henry (2000) empirically confirm that liberalization reduces the cost of equity. We obtain official liberalization dates from Table I in Bekaert and Harvey (2000). 13 Integration is defined as follows: Integrateit = (Exportit + Importit ) / GDPit. This measure has been used in several papers such as Bekaert and Harvey (1995) and Bhattacharya and Daouk (2002).
18
are mixed with negative coefficient in dividend equation and positive in dividend growth model
equation.14 We conduct some additional analysis to see if the impact of electronic trading is different
in developed and emerging markets. The level of gross returns and excess returns is lower in
developed markets compared to that in emerging markets. Therefore, we expect that the cost
reduction in developed market will also be of a lower magnitude. This indeed is the case. In Table 3,
we introduce interaction between electronic trading and level of economic development. The
variable “electronic * developed” is set to 1 for all exchange-months when electronic trading is in
place and market is of developed type. It is set to zero if either condition is not met. The other
interaction variable “electronic * emerging” is analogously assigned values for emerging markets.
Many countries in the emerging markets started their capital market within the sample period. This
could mean that the level risk, which is initially high, can decline with maturity. We add a time-trend
variable in Panel B to control for this possibility. Electronic trading is associated with lower cost of
equity especially in emerging markets where both absolute and proportional magnitude of cost
reduction is larger to those in developed markets.
The analysis in this sub-section confirms the findings of the univariate comparisons
discussed in sub-section B. The multivariate regressions attempt to control for differences in the
level of world market returns, insider-trading-law-enforcement, economic liberalization, market size,
economic development, economic integration with world, economic growth rate, and time-trend.
After controlling for all these differences, this section shows that electronic trading is associated
with lower cost of equity, particularly in emerging markets. In the next section we show that this
result holds true even after allowing for a world market risk factor and time-variation in betas.
F. A conditional international asset-pricing model
14 Enforcement and liberalization are not consistently significant in the full sample or the sample centered around electronic trading. However, they do turn out to be significant when samples centered around insider law enforcement dates and liberalization date respectively are used.
19
So far we have shown that the long-term equity premium is different in floor and electronic
regimes. It may be argued that the differences occur due to time-variation in the market risk in those
periods. In this section, we investigate this possibility by using an international capital asset pricing
model (ICAPM) that takes into account the changing exposure to world market risk and domestic
market risk15. The model allows for time variation in both betas. We used a simplified version of the
international asset pricing model of Bekaert and Harvey (1995) shown below:
(returnit–rft)= α0 + φλcovhiwt + (1-φ)λvarhit +eit (3)
where returnit is the monthly dollar return of the stock market index of country i at time t, rft is the
monthly return of the one month US T-Bill at time t, α0 is a constant to be estimated, φ is a measure
of the level of integration of country with the world market, λcov is the price of the covariance (with
world index) risk to be estimated, hiwt is the conditional covariance of the monthly return of the stock
market index of country i with the monthly return of the world index at time t, λvar is the price of
own country variance risk to be estimated (which we are restricting to be the same across all
countries), hit is the conditional variance of monthly return on the stock market index of country i at
time t, and eit is the residual error term. The independent variables in equation (3)– conditional
covariance hiwt and conditional variance hit – are separately estimated pair-wise for each country i
and world pair from the multivariate ARCH model16 specified below:
(returnit–rft) = c1 + εit,
(worldt –rft) = c2 + εwt,
hit = b1 + a1(1/2ε2it-1 + 1/3ε2
it-2 + 1/6ε2it-3), (4)
hiw = b2 + a2(1/2ε2wt-1 + 1/3ε2
wt-2 + 1/6ε2wt-3),
15 We use both a domestic market factor and a world market factor because the past literature has shown that both of these have some merit in explaining the expected returns. See, for example, Harvey and Zhou (1993) and Harvey (1995) 16 The ARCH model described in equation (3) was first introduced by Bollerslev, Engle, and Wooldrige (1988). As in Engle, Lilien, and Robins (1987). the weights of the lagged residual vectors are taken to be ½, 1/3, and 1/6, respectively.
20
hiwt = b3 + a3(1/2εit-1εwt-1 + 1/3εit-2εwt-2 + 1/6εit-3εwt-3),
wtiwt
iwtitwtit hh
hhN ,
00
~,εε
where worldt is the dollar monthly return of the world market index at time t, εit-j is the innovation in monthly
return of the stock market of country i at time t-j, j ∈ {0,1,2,3}, εwt-j is the innovation in monthly return of the
world market index at time t-j, j ∈ {0,1,2,3}, and hwt is the conditional variance of monthly return of the
world market index at time t.
The measure for level of integration of the markets of country i with the world markets at
time t, φit , is defined as follows:
2*5.0expexp1
expexp
−
++
+
=
it
itit
it
itit
it
gdpimportsorts
gdpimportsorts
φ (5)
This measure essentially captures the dependence of an economy (measured by gross
domestic product (gdp)) on exports and imports. φit can take values between 0 and 1. When its value
is zero we assume that the country’s financial markets are completely segmented and when its value
is one we assume that the markets are fully integrated with the rest of the world. Bakaert and Harvey
(1997) find that increases in this ratio are associated with increased importance of world risk factors
relative to local risk factors for the returns generation process. As a robustness check we also restrict
the value of φit to 0.5 thus giving equal importance to the world index and domestic index. This
restriction does not change the results in any significant way.
The results for this international asset-pricing model are given in Table 4. Panel A of the
model shows the estimates for equation (3). Both the covariance risk with the world and own
The constants a2 , b2 , and c2 are constrained to be identical for all country-world pairs. Maximum likelihood is used to estimate equation (3).
21
country variance risk are priced. The price of each risk is positive and significant at 1% level. If the
introduction of electronic trading does not affect the equity returns then the residual eit in equation
(3) should be orthogonal to the electronic trading variable. However, in Panel B we show that this is
not the case. We regress the residuals on a number of variables as follows:
eit =δ0 +δ1electronicit +δ2enforceit +δ3liberalit +δ4mcapit +δ5developi +δ5trendit +ηit (6)
where eit is the residual from international asset pricing equation (3); electronicit, enforceit, liberalit,
mcapit, developi, and trendit retain their definition from sub-section B.
Panel A of Table 4(B) shows the results of regression equation (6). All variables except
electronic trading are orthogonal to the residuals from the asset-pricing model. The coefficient on
electronic trading is a negative –0.0046 and statistically significant. Thus, after controlling for time
variation in betas, introduction of electronic trading seems to be sharply lowering the cost of equity
for listed firms. Panels B and Panel C show regressions with interaction terms. The results indicate
that the drop in the equity premium is much sharper for emerging markets (0.81%) than for
developed markets (0.15%). This is true even on proportional terms. The average dollar returns
before introduction of electronic trading were 1.39% in developed markets and 2.14% in emerging
markets.
G. Robustness of results
We have used both gross and excess returns and both full sample and sub-sample for all
analyses through out the paper and find consistent results for all combinations. Nevertheless, we
carry out additional robustness checks in this section. The summary of this additional analysis is
provided below.
First we drop the biggest Internet boom and bust period from January 1999 to August 2001
from the sample and re-run all tests. The results are even stronger for the remaining sample of
10,118 exchange months. For instance, unconditional annualized reduction in expected returns for
22
this truncated sample is 0.08% using dividend yield compared to 0.04% per month for full sample in
Figure 2. Similarly, in the regression with excess returns (in dollars) as dependent variable the
coefficient on electronic becomes more negative from –0.0052 to –0.0059 and also increases in
statistical significance. In the ICAPM framework, the coefficient on electronic trading in residuals
regression reduces in absolute value from -0.0046 in Table 4 to –0.0032 with the reduced sample;
but it remains negative and statistically significant.
Second, we allow for some transition period required for popularization of electronic trading.
This is accomplished by ignoring a period of 1 year (and then 2 years) in the sample immediately
following the introduction of electronic trading. Once again, deleting one year from the sample to
allow for popularity of electronic system results in an unconditional reduction in cost of equity by
0.11% per month using dividend growth model; regression coefficient on electronic trading dummy
in excess return regression remains negative at –0.0039; and ICAPM coefficient is -0.0032 and is
significant at 5.69% level. Results are similar if we allow 2 years for electronic systems to gain
popularity.
Third, we have already excluded the periods of excessive currency devaluation17 from the
sample. Inclusion of such periods in the analysis only strengthens the result; dividend growth
model’s expected returns drop by –0.11%, regression coefficient is –0.0062, and I-CAPM coefficient
is –0.0056 and highly significant.
Fourth, we include a trend variable as an independent variable in the regression equation (2)
with gross/ excess returns as dependent variable to account for the possibility that returns might be
undergoing a downward trend over the years. The coefficient for electronic is still significant and the
coefficient on the trend variable is not statistically significant, which rules out the possibility of
downward trend driving the results.
23
Fifth, it is conceivable that stock exchanges introduce electronic trading after stock
market booms and lower returns observed in electronic trading periods reflect a long term reversal of
returns. We dated the stock market peaks in each country to shed light on this alternative
explanation. We test this and find that stock markets in 42 out of the 71 countries experienced their in-
sample peaks well after the introduction of automated trading. On average the market peak was achieved
6 years after automation. This means that automation was not followed by long term reversals in stock
prices. Rather, the markets were still growing, albeit, at a slower pace because the equity premium
declined, as shown earlier. Similarly, for the remaining 29 countries which had their stock market peaks
in the floor trading regime, this peak was attained almost 3 years before automation on average and more
than 13 years ago in some cases. Thus for these remaining markets, the long term reversals, if any,
affected both the floor trading period returns and the electronic trading period returns.
H. Liquidity
These liquidity improvement results are consistent with the vast market-microstructure
literature that provides evidence of reduction in spreads following automation of stock trading. De
Jong, et al. (1995), Frino and McCorry (1995), Blennerhasset and Bowman (1998), and Jain (2001)
are a few examples. Domowitz and Steil (1999) give an excellent summary of this literature for
specific markets such as the U.S.A., Britain, France, Germany, Singapore, New Zealand, Australia,
and India. Time series of closing spreads data for 6 exchanges are also available from Datastream
International. These are Portugal, Spain, Italy, Switzerland, UK, and France. We examined the
average quoted bid-ask spread two years before and two years after automation for the largest stocks
on these exchanges. For all 6 exchanges, spreads drop significantly after the introduction of
electronic trading. For instance spreads in Spain fall from 0.33% in floor trading to 0.23% after
automation. In Switzerland spreads of 0.12% in electronic regime are half of those in floor trading.
17 Those months when a country’s currency devalued by over 50% were initially dropped from the analysis and included here as a robustness check. There were 6 such months out of 12,229 exchange-months.
24
On average the spreads fall by 39% compared to their floor trading level. Apart from retail trading
costs, the institutional trading costs have also declined over time around the World as documented in
Chiyachantana, Jain, Jiang, and Wood (2003).
IV. Is there a positive price reaction when stocks move from floor to electronic?
In this section we examine the short-term returns around the introduction of electronic
trading to test the hypothesis that when stocks move from floor to electronic trading, they will have a
positive short-run price reaction. We have assembled the dates of announcement of the plan to
switch to electronic trading and dates of abolition of floor trading in addition to the actual
introduction of electronic trading on 69 exchanges. The price response to automation might start
from the date of announcement of such a switch. The average gap between announcement of the
switch and actual introduction of electronic trading is 21 months. In Figure 3 we chart the month-by-
month cumulative excess-over-world returns for a period from 24 months before the introduction of
electronic exchange to 24 months after this date. We do the same thing for announcement dates. The
price reaction in the implementation month is positive 1.20% but is milder than the announcement
month reaction of 7.87%. The cumulative excess over world return (CAR) is 10.66% and 2.73%
from six months before to one month after the date of announcement and implementation
respectively. Benchmarked against their levels 24-months before announcement, cumulative returns
are an impressive 24%. The stock market indices decline from their top levels after announcement of
automation but the cumulative excess returns even after two years of announcements are over 15%.
The results are consistent with the second hypothesis in the paper that automation produces positive
price reaction. It is possible to study the short-term announcement effect of the switch more
precisely by gathering additional daily data on returns. However, we leave this task for future
research.
25
The magnitude of price reaction (7.87%) to announcement is comparable to the results in
several papers that study price discounts due to illiquidity. Amihud, Mendelson, and Lauterbach
(1997) find a 5.5% average cumulative abnormal return on 120 stocks that transfer from a call
market to a continuous market on the Tel Aviv Stock Exchange. Muscarella and Piwowar, (2001)
find that firms switching from single price-fixing system to continuous trading system on the Paris
Bourse experience a positive cumulative abnormal return of 5.4%. In their sample, the exchange
switched some firms from the more liquid continuous trading back to the less liquid single price
fixing and such firm had a negative cumulative abnormal return of -5%. Thus, the market recognizes
the benefits of higher liquidity and stock market valuations improve when electronic trading is
introduced.
V. Conclusion
This study finds that automation of trading on a stock exchange has a long-term impact on
listed firms’ cost of equity. Previous studies had shown that electronic trading improves a stock’s
liquidity and reduces investors’ trading costs. This paper confirms that finding and integrates it with
another stream of literature that examines the effect of liquidity and information on the equity
premium.
After controlling for differences in market size, financial liberalization, enforcement of
insider trading laws, level of economic development, and level of world market returns, time trend in
returns, and time-varying betas and risk factors; this study finds that the introduction of electronic
trading lowers the cost of equity for listed firms. The decline in cost of equity is 0.07% in dividend
yields, 0.19% per month according to dividend growth model and 0.46% per month according to
international capital asset pricing model. The reduction in excess returns is more pronounced in
emerging markets (0.81%) than in developed markets (0.15%). The cost-reduction result holds
26
qualitatively even when the analysis is restricted to certain sub-periods, Internet boom and bust
period is excluded, or a transition period is allowed for popularity of electronic exchanges.
Nevertheless, we do not intend to over-emphasize the point estimates of reductions in cost of capital
due to significant cross country differences. Instead, the most striking take-home result is that
eighty-three percent of the regime shifts (59 of 71 exchanges) are associated with a reduction in the
cost of equity and seventy-five percent of the switches are associated with an improvement in
liquidity. By performing an exchange-by-exchange analysis using the same stocks for electronic and
floor trading regimes, this study also avoids the problems associated with imperfect matching of
stocks.
In addition to the reduced cost of equity, the introduction of electronic trading is associated
with a positive price reaction of 7.87% in listed stocks around the date of the switch. Liquidity also
improves dramatically after automation. These patterns support the notion that electronic markets
improve the liquidity, informativeness, and valuation of listed stocks, which collectively help reduce
the cost of equity.
As usual, the findings have to be interpreted keeping in mind the limitations of any large
international study with such a broad scope. The main focus of this study is fully automatic
execution of trading. So the exchange design after the event is clearly defined. However, the term
floor-trading is used to represent the market design before the switch and is somewhat symbolic in
nature as it includes different versions and levels of trading with at least some manual interference.
Future research can explore whether some manual systems are associated with lower capital costs
than other manual systems.
The perceptible positive aspects of electronic trading entreat a move away from floor based
trading to automated screen based electronic trading. This push for a change should come from the
listed firms and investors particularly if the stock-exchange owners’ business incentives are not
27
aligned with theirs. This study also serves to establish an important linkage between financial market
design and asset pricing. The results implore inclusion of market-design factors and liquidity factors
into asset pricing models.
28
Appendix 1. Key institutional differences between floor trading and electronic trading
Institutional Feature Floor Electronic
Identity of Counter-Party as
discussed in Beneviste,
Marcus and Wilhelm
(1992), Rock (1988)
The identity of the counter-
party broker is known
before the trade.
The identity of counter-
party is usually revealed
post-trade although a few
exchange display broker
identity pre-trade.
Ex-ante Transparency of
Available Liquidity as
discussed in Madhavan
(1996)
Usually only the best bid
and offer is known to
traders
Usually top 5 best buy and
best sell prices and
sometimes the whole book
displayed on traders’ screen
Speed of Matching trades or
immediacy as discussed in
Grossman and Miller
(1998)
Trades are matched
manually based on open-
outcry system and can take
10 sec to 1 minute
Trades are matched
automatically by a
computer algorithm within
a second
Operating cost and order
processing costs
Higher Lower (Domowitz and Stiel
(1999))
Speed and cost of
Settlement
Settlement is often paper-
based and with a lag
Usually faster
dematerialized settlement
and lower costs
Order Book Order book often does not
exist and quotes are valid
only as long as “breath is
warm”
Order book matches ir
accumulates customers’
limit order and quotes are
valid until withdrawn
References
Amihud, Y. and H. Mendelson, 1986, Asset pricing and the bid-ask spread, Journal of Financial Economics, 17, 223-249.
Amihud, Yakov, H. Mendelson, and B. Lauterbach, 1997, Market microstructure and securities values: Evidence from the Tel Aviv stock exchange, Journal of Financial Economics 45, 365-390.
Barclay, Michael, Eugene Kandel, and Leslie M. Marx, 1998, The effects of transaction costs on stock prices and trading volume, Journal of Financial Intermediation 7, 130–150.
Bekaert, Geert, and Campbell R. Harvey, 1995, Time-varying world market integration, Journal of Finance, 50, 403-444.
Beneviste, Lawerence M., Alan J. Marcus, and William J. Wilhelm, 1992, Whats special about the specialist? Journal of Financial Economics 32, 61–86.
Bhattacharya, Utpal, and Hazem Daouk, 2002, The world price of insider trading,, Journal of Finance, 57, 75–108.
Blennerhasset, M. and R.G. Bowman, 1998, A change in market microstructure: The switch to electronic screen trading on the New Zealand Stock Exchange, Journal of International Financial Markets, Institutions and Money, 8, 261-276.
Bollerslev, Tim, Robert Engle, and Jeffrey Wooldrige, 1988, A capital asset pricing model with time-varying covariances, Journal of Political Economy 96, 116-131.
Brennan, M.J. and Subrahmanyam, A., 1996. Market microstructure and asset pricing: on the compensation for illiquidity in stock returns. Journal of Financial Economics 41, 441¯464.
Chiyachantana, Chiraphol N., Pankaj K. Jain, Christine Jiang, and Robert A. Wood, 2003, International evidence on institutional trading behavior and determinants of price impact, forthcoming, Journal of Finance.
Chung, Mary, and Andrei Postelnicu, 2003, Specialists' at NYSE face probes of abuses: A lack of automation raises fresh suspicions of manipulation of trades, The Financial Times, April 21, 15.
Compaq Handbook of World Stock, Derivative and Commodity Exchanges, 2001, Financial Publications, UK.
Constantinides, G. M. (1986). Capital market equilibrium with transaction costs, J. Political. Economy 94, 842–862.
Datar, V. T., N. Y. Naik, and R. Radcliffe, 1998, Liquidity and stock returns: An alternative test, Journal of Financial Markets, 1, 203-219.
De Jong, F., Th. Nijman and A. Röell, 1995, A Comparison of the Cost of Trading French Shares on the Paris Bourse and on SEAQ International, European Economic Review, 39,1277-1301.
Dimson, Elroy, and Bernd Hanke, 2000, The expected Illiquidity Premium: Evidence from equity Index-Linked Bonds, Working Paper, London Business School.
Domowitz, I. and B. Steil, 1999, Automation, trading costs, and the structure of the trading services industry, Brookings-Wharton Papers on Financial Services.
Domowitz, I. and B. Steil, 2001, Innovation in equity trading systems: The Impact on transactions costs and cost of capital, Technological Innovation and Economic Performance Richard Nelson, David Victor, and Benn Steil (editors) Princeton University Press.
30
Engle, Robert, David Lilien, and Russell Robins, 1987, Estimating time varying risk premia in the term structure: The ARCH-M model, Econometrica 55, 391-407.
Fama, Eugene F., and Kenneth R. French, 2002, The equity premium, Journal of Finance, 57, 637–660.
Franke, Gunter, and Dieter Hess, 2000, Information Diffusion in Electronic and Floor Trading, Journal of Empirical Finance, 7, 455-78.
French, Kenneth R., and Richard Roll, 1986, Stock Return Variances: The arrival of information and reaction of traders, Journal of Financial Economics, 17, 5-26.
Frino, A. and M. McCorry, 1995, Why are spreads tighter on the Australian Stock Exchange than the NYSE? An electronic open limit order book versus the specialist structure, Working paper, University of Sydney.
Harvey, Campbell, 1995, Predictable risk and returns in emerging markets, Review of Financial Studies 8, 773-816.
Harvey, Campbell, and Guofu Zhou, 1993, International asset pricing with alternative distribution assumptions, Journal of Empirical Finance 1, 107-131.
Haugen, Robert A., 1999, Beast on Wall Street: How Stock Volatility Devours our Wealth, Prentice Hall.
Henry, P.B., 2000, Stock market liberalization, economic reform, and emerging market equity prices, Journal of Finance, 55, 529-564.
Huang, R and H. Stoll, 1997, The components of bid-ask spreads: A general approach, Review of Financial Studies, 10, 995-1034.
Jain P., 2001, Institutional design and liquidity on stock exchanges, Working Paper, Indiana University.
Jones, Charles, 2001, A century of stock market liquidity and trading cost, Working Paper, Columbia University.
Kadlec, G. B., and McConnell, J. J. (1994). The effect of market segmentation and illiquidity on asset prices: Evidence from exchange listings, Journal of Finance, 69, 611–636.
Kalay, Avner, Lei Wei and Avi Wohl, 2002, Continuous trading or call auctions: Revealed preferences of investors at Tel Aviv Stock Exchange, Journal of Finance, 57, 523-542.
Levine, Ross and Sara Zervos, 1998, Stock markets, banks, and economic growth, American Economic Review, 88, 537–58.
LGT Guide to World Equity Markets, 1997. Euromoney Publications. Loderer, Claudio F, and Lukas Roth, 2001, The pricing discount for limited liquidity:
Evidence from switzerland, SSRN Working Paper 288965. Muscarella, Chris, and Michael S. Piwowar, 2001, Market microstructure and security values:
Evidence from the Paris Bourse, Journal of Financial Markets, Vol. 4, No. 3, 209-229.
MSCI Handbook of World Stock, Derivative and Commodity Exchanges, 1998, Financial Publications, UK.
O’hara, Maureen, 2003, Liquidity and Price Discovery, Journal of Finance, 58, 1335-1354.
31
O’hara, Maureen and David Easley, 2002, Information and the cost of capital, SSRN Working Paper No. 300715.
Pagano, M., and A. Roell, 1996, Transparency and liquidity: a comparison of auction and dealer markets with informed trading, Journal of Finance 51, 579–611
Pastor, Lubos and Robert Stambaugh, 2001, Liquidity risk and expected stock returns, Working Paper, Wharton School, University of Pennsylvania.
Salomon Smith Barney Guide to World Equity Markets 1999, 2000 Euromoney Books and Salomon Smith Barney.
Schmidt, H., P. Iversen, and K. Treske, 1993, Parkett oder computer?, Zeitschrift fur Bankrecht und Bankwirtschaft 5, 209–221.
Stoll, Hans R., and Robert E. Whaley, Stock market structure and volatility, Review of Financial Studies, 3, 37–71.
Stulz, Rene, 1999, Globalization, corporate finance, and the cost of capital, Journal of Applied Corporate Finance, Fall 1999, 8-25.
Swan, P. L. and P. J. Westerholm, 2001, Asset prices and liquidity: The impact of endogenous trading, SSRN working paper.
Tiehang, Niu, 1996, Regional stock market integration in China and Hong Kong, Consulting Assistance on Economic Reform (CAER II), Discussion Paper # 16.
Theissen, Erik, 2001, Floor versus screen trading: Evidence from the German Stock Market, Working Paper, University of Bonn.
Vayanos, D. (1998). Transaction costs and asset prices: A dynamic equilibrium model, Review of Financial Studies, 11, 1–58.
Venkataraman, Kumar, 2001, Automated versus floor trading: An analysis of execution costs on the Paris and New York Exchanges, Journal of Finance 56, 1445–1485.
World Federation of Exchanges’ (formerly known as the International Federation of Stock Exchanges – FIBV) archives at www.fibv.com
Table 1. Country-by-country analysis of declining equity premium after Electronic Trading This table reports the changes in equity premium for 71 exchanges after the introduction of electronic trading. The six measures of expected returns are calculated for the floor trading months and the electronic months and the differences are reported here. Changes in trading turnover are in the last column. Excess Returns
Country
Change in
dividend yields
(DY)
Change in dividend
yields plus
capital gains (local
currency, DYCG)
Change in
dividend yields
plus capital
gains (in U.S.
Dollars, DYCG$)
Change in
dividend yields
plus dividend
growth (DYG)
Country return minus world
return (in U.S.
dollars, ERW$)
Country return minus
U.S. dollar t-
bill return (ERT)
Changes in Trading
turnover (Turnover)
Panel A. Developed Countries 1 Australia -0.04% -1.52% -1.13% -1.76% -0.30% -0.82% 0.0253 2 Austria 0.01% -0.46% -1.19% -0.01% -0.84% -1.14% 0.3063 3 Belgium -0.17% 0.06% -0.88% -0.83% -0.38% -0.83% 0.1724 4 Canada -0.08% 0.79% 0.45% -0.46% -0.41% 0.18% 0.0291 5 Denmark -0.10% -0.19% 0.27% 0.75% 0.80% 0.59% 0.0375 6 Finland -0.02% -0.48% -0.37% -0.50% 0.67% -0.26% 0.0212 7 France -0.18% -1.13% -0.60% 0.51% -0.24% -0.31% -0.0170 8 Germany -0.11% -0.21% -0.58% -0.19% -0.10% -0.29% 0.1345 9 Hong kong -0.08% 0.31% 0.76% 0.43% 1.03% 1.05% 0.0105
10 Ireland -0.08% -0.66% -0.60% 0.55% 2.01% -0.61% -0.0171 11 Italy -0.07% -0.48% -0.85% -1.72% -0.39% -0.75% 0.0546 12 Japan -0.10% -0.17% 0.17% -0.40% -0.59% 0.19% 13 Luxembourg -0.14% -1.41% -1.39% 1.14% -1.54% -1.20% 0.2398 14 Netherlands -0.22% 0.20% -0.67% -0.24% -0.07% -0.60% 0.0908 15 New Zealand -0.04% 0.76% 0.80% -0.26% 0.81% 1.02% 0.0001 16 Norway -0.12% -0.82% -0.85% -1.77% -0.22% -0.54% 0.0529 17 Portugal 0.07% 1.48% 1.32% -0.70% 1.29% 1.53% 0.0277 18 Singapore -0.09% -0.68% -0.67% -1.11% -0.01% -0.35% 0.0358 19 Spain 0.03% -0.54% -0.31% -1.08% 0.26% 0.01% 0.0271 20 Sweden -0.04% -1.11% -0.89% -0.64% -0.31% -0.57% -0.0146 21 Switzerland -0.09% 0.16% -0.65% -0.27% -0.41% -0.61% 0.0246 22 UK -0.17% -0.74% -0.98% -1.40% -0.48% -0.93% 0.0543 23 US-NYSE -0.20% -2.94% -2.94% -0.33% -0.28% -2.90% 0.0018 Country DY DYCG DYCG$ DYG ERW$ ERT Turnover Panel B. Emerging Countries 24 Argentina 0.12% -12.82% -4.39% -2.30% -4.44% -4.35% 0.0032 25 Bahrain 0.00% -1.88% -1.88% 0.86% 0.57% -1.90% 26 Bangladesh 0.06% -1.68% -2.03% 1.02% -0.74% -2.02% 0.0366 27 Barbados -1.97% -1.93% 0.21% -1.94% -0.0114 28 Bermuda -0.09% -0.09% -0.82% -0.09% 29 Brazil -0.27% 6.84% -2.88% -4.05% -2.84% -2.66% -0.0545 30 Bulgaria -2.75% -0.98% -0.96% -0.98% 31 Chile -0.46% -1.42% -1.47% -1.15% -1.02% -1.27% 0.0003 32 China -0.04% 0.48% 1.97% -2.08% 0.71% 2.23% -0.0023 33 Colombia 0.04% -2.83% -2.88% -1.06% -2.81% -2.85% 0.0067 34 Croatia -0.03% 2.20% 2.40% 3.15% 4.03% 2.38% 0.0009 35 Cyprus 1.71% 1.54% 3.14% 1.52% 0.2461 36 Czech 0.03% -0.08% 0.02% 2.73% 1.52% 0.04% 0.0201 37 Ecuador -0.15% -2.18% -1.34% -2.19% 38 Egypt 0.22% -3.74% -4.21% -1.74% -3.23% -4.19% 0.0549 39 Greece -0.03% -1.84% -1.87% 1.04% -2.10% -1.71% 0.0492 40 Hungary -0.02% -3.66% -2.80% 2.68% -1.65% -2.82% 0.1212
Continued…..
33
….Table 1 Continued Country DY DYCG DYCG$ DYG ERW$ ERT Turnover
41 India 0.06% -2.55% -1.88% -2.12% -2.18% -1.86% 0.0400 42 Indonesia -0.03% -2.86% -2.59% 0.51% -2.51% -2.55% 0.0195 43 Israel 0.15% -0.20% -0.11% -1.83% 0.94% -0.14% 0.0240 44 Jordan 0.01% -0.23% -0.14% 1.68% 2.28% -0.16% -0.0030 45 Korea -0.21% -1.49% -1.96% 0.76% -1.28% -1.70% 0.1715 46 Kuwait -4.04% -4.07% -3.62% -4.02% 0.0260 47 Latvia -8.68% -8.41% -9.35% -8.40% 0.0148 48 Lebanon -0.05% -1.46% -1.60% -3.17% 1.63% -1.55% 0.0077 49 Malaysia -0.05% -0.93% -1.21% -0.04% -1.53% -1.07% 0.0299 50 Mexico -0.12% -2.11% -1.26% -0.38% -1.13% -1.21% -0.0577 51 Morocco -0.02% -2.09% -2.20% 2.67% -1.43% -2.18% 0.0203 52 Oman 0.24% -3.85% -3.85% 2.82% -3.47% -3.84% -0.0225 53 Pakistan 0.43% -2.23% -2.36% -2.04% -1.99% -2.27% 0.1372 54 Peru 0.13% -4.00% -3.40% -3.57% -2.79% -3.48% -0.0123 55 Philippines -0.03% -1.53% -1.91% 1.21% -2.21% -1.79% 0.0149 56 Poland 0.00% -7.82% -6.73% 0.24% -6.10% -6.78% -0.0340 57 South Africa -0.05% -2.66% -3.09% 0.63% -2.47% -3.14% 0.0287 58 Sri lanka 0.31% -1.77% -2.16% 0.67% -1.31% -2.20% -0.0024 59 Thailand -0.11% -3.38% -3.67% -1.41% -3.73% -3.45% 0.0023 60 Turkey -0.40% -1.26% -2.24% -2.24% -2.33% -2.17% 0.1238 61 Iran -6.47% -4.07% -4.15% -6.47% -0.0789 62 Ivory Coast -2.41% -2.76% -1.75% -2.41% 63 Jamaica 0.16% 1.04% 3.51% 2.55% -0.0756 64 Mauritius -2.04% -2.07% 0.41% -2.04% 65 Mongolia -4.10% -3.20% -3.78% -4.10% -0.2097 66 Namibia -3.47% -3.11% -2.80% -2.67% -0.0011 67 Nigeria -1.56% -0.28% 0.20% -2.30% 0.0504 68 Panama -4.81% -4.81% -3.87% -4.81% 0.0008 69 Taiwan 1.40% 2.20% 1.91% 0.34% 70 Tunisia -1.50% -2.05% -2.23% -1.50% 0.0342 71 Venezuela -5.98% -5.27% -4.77% -5.98% 0.0678
72% 82% 83% 62% 72% 82% 25%
Proportion of countries that experience reduction Wilcoxon Rank Sum Test U = Rank Sum - n*(n+1)/2 1195 1263** 1139** 1279 1694** 1172** 2817**Z = {U - E(U)}/Std(U) -1.62 -5.13 -5.64 -0.47 -3.37 -5.50 2.91
Table 2. Effect of Electronic Trading in a Regression Framework
Dividend, excess return (ERT), and turnover regressions are based on exchange-months from 53, 71, and 63 countries respectively for which all required data was available. Explanatory variables are given in rows. Electronic trading is an indicator variable that signifies the introduction of electronic trading in the country. World market returns are computed from the MSCI world index. Market capitalization are in trillions of US dollars. Integration of the markets is measured by the ratio of exports plus imports to GDP. GDP growth rate is computed quarter to quarter. The remaining variables are indicator variables for date of first enforcement of insider trading laws, official liberalization date, and developed vs. emerging market. Statistical significance of the coefficients is marked ** and * at 1% and 5% levels respectively.
Dependent Variable >
Dividend yields
(DY)
Dividend yields plus
growth (DYG)
Excess Returns
(ERT) Turnover Number of Observations 8710 8710 12223 6692 Adjusted R-Square 11.48% 0.56% 13.49% 3.21% Instruments Intercept 0.0029** 0.0212** 0.0061** 0.0268** Electronic Trading -0.0007** -0.0019* -0.0052** 0.0238** World Market Return 0.0005 -0.0038 0.0080** 0.0006 Enforce Insider Laws -0.00003 -0.0026** -0.0008 -0.0200** Liberalized Market -0.0002** 0.0036** 0.0012 -0.0333** Market Capitalization -0.0002** -0.0009* -0.0008 0.0019 Developed Market 0.0004** -0.0052** -0.0021 0.0003 Integration of Market -0.00003 0.0016* -0.0002 -0.0013 GDP Growth -0.00003 -0.0012 0.0021 0.0006
Table 3. Interaction between electronic trading and level of economic development.
This table introduces an interaction between electronic trading and financial market development in the regressions. A time-trend control variable is also added. Time trend gives the relative position of the month from the starting date in data. DY DYG ERT Turnover Intercept 0.0075** 0.0156* 0.0418** -0.0139 Electronic * Developed -0.0003** -0.0010 0.0034 0.0051 Electronic * Emerging -0.0005** -0.0061** -0.0115** 0.0349** World Market Return 0.0006 -0.0038 0.0080** 0.0006* Enforce Insider Laws 0.00004 -0.0027** -0.0005 -0.0206** Liberalized Market -0.0002** 0.0040** 0.0016 -0.0341** Market Capitalization -0.0001** -0.0010* -0.0003 0.0008 Developed Market -0.0001 -0.0078** -0.0128** 0.0237** Integration of Market 0.00002 0.0015* 0.0001 -0.0012 Time Trend -0.0001** 0.0002 -0.0009* 0.0010 GDP Growth -0.00005 -0.0012 0.0019 0.0005
35
Table 4. A. Effect of Electronic Trading: ICAPM with Time Varying Betas
The following regressions is based on equity index returns from January 1973 to June 2001. First, an international capital asset pricing model is estimated as follows:
(returnit–rft)= α0 + φλcovhiwt + (1-φ)λvarhit +eit (E1)
where returnit is the monthly dollar return of the stock market index of country i at time t, rft is the monthly return of the one month US T-Bill at time t, α0 is a constant to be estimated, φ is a measure of the level of integration of country with the world market, λcov is the price of the covariance (with world index) risk to be estimated, hiwt is the conditional covariance of the monthly return of the stock market index of country i with the monthly return of the world index at time t, λvar is the price of own country variance risk to be estimated, hit is the conditional variance of monthly return on the stock market index of country i at time t, and eit is the residual error term. The measure for level of integration of the markets of country i with the world markets at time t, φit , is defined as follows:
2*5.0expexp1
expexp
−
++
+
=
it
itit
it
itit
it
gdpimportsorts
gdpimportsorts
φ
The independent variables in equation (3)– conditional covariance hiwt and conditional variance hit – are separately estimated pair-wise for each country i and world pair from the multivariate ARCH model specified below:
(returnit–rft) = c1 + εit, (worldt –rft) = c2 + εwt,
hit = b1 + a1(1/2ε2it-1 + 1/3ε2
it-2 + 1/6ε2it-3),
hiw = b2 + a2(1/2ε2wt-1 + 1/3ε2
wt-2 + 1/6ε2wt-3),
hiwt = b3 + a3(1/2εit-1εwt-1 + 1/3εit-2εwt-2 + 1/6εit-3εwt-3),
wtiwt
iwtitwtit hh
hhN ,
00
~,εε
where worldt is the dollar monthly return of the world market index at time t, εit-j is the innovation in monthly return of the stock market of country i at time t-j, j ∈ {0,1,2,3}, εwt-j is the innovation in monthly return of the world market index at time t-j, j ∈ {0,1,2,3}, and hwt is the conditional variance of monthly return of the world market index at time t. Dependent Variable -> Excess Returns Parameters Coefficient Standard P-Value Error Alpha 0.0044 0.0011 <.0001 Price of covariance risk with World 4.3927 1.8419 0.0171 Price of own country variance Risk 1.3015 0.1147 <.0001
36
Table 4 B. Effect of electronic trading on residuals from ICAPM model
The residuals from equation E1 in Table (4A.) above form the dependent variable in the following regression equation:
eit =δ0 +δ1electronicit +δ2enforceit +δ3liberalit +δ4mcapit +δ5developi +δ5trendit +ηit
where eit is the residual from international asset pricing equation; Electronic is an indicator variable that signifies the introduction of electronic trading in the country, enforce becomes 1 after the date of first enforcement of insider trading laws, liberal becomes 1 after the official liberalization date, mcap is market capitalization in trillions of US dollars, developed vs. emerging market classification is based on MSCI, and trend gives the relative position of the month from the starting date in data, which is December 1969. Statistical significance of the coefficients is marked ** and * at 1% and 5% levels respectively. Panel A. Effect of Automation of Trading after controlling for risk factors and other events Dependent Variable -> Residuals from ICAPM model Independent Variables Coefficient Standard P-Value Error Intercept 0.0006 0.0132 0.9617Electronic Trading -0.0046 0.0023 0.0411Enforcement of Insider trading laws -0.0025 0.0020 0.1964Liberalization -0.0038 0.0025 0.1280Market Capitalization -0.0010 0.0009 0.2928Developed Markets 0.0015 0.0024 0.5431Time Trend 0.0001 0.0004 0.7150 Panel B. Regression with Interaction between electronic trading and level of economic development. Intercept 0.0025 0.0011 0.0300Electronic * Developed -0.0036 0.0020 0.0762Electronic * Emerging -0.0069 0.0021 0.0008 Panel C. Regression with Interaction between electronic and development and other events. Intercept 0.0045 0.0116 0.6974Electronic * Developed -0.0015 0.0025 0.5454Electronic * Emerging -0.0081 0.0027 0.0025Enforcement of Insider trading laws -0.0027 0.0020 0.1760Liberalization -0.0051 0.0023 0.0262Market Capitalization -0.0011 0.0009 0.2490Time Trend 0.0001 0.0004 0.7963
Figure 1. The Global Shift from Floor Trading to Electronic TradingBased on the leading stock exchange in 120 countries from 1975 to 2001
Fully ElectronicCombined Floor and
Electronic
Floor Trading Only
0
20
40
60
80
100
120
140
1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002
Year
Num
ber o
f Exc
hang
es
38
Figure 2. Declining Cost of Equity after Automation
0.79%
0.02%
0.25%
0.38%
0.79%
2.14%
0.22%
1.05%
1.57%
0.85%
1.46%
2.00%
2.61%
2.36%
0.26%
-0.17%
6.14%
-1.00% 0.00% 1.00% 2.00% 3.00% 4.00% 5.00% 6.00% 7.00%
Trading turnover
Volatility (Variance of total dollarreturns)
Country return minus t-bill return (indollars)
Country return minus world indexreturn (in dollars)
Country return minus world indexreturn (local curr)
Dividend yield plus capital gains (inU.S. dollars)
Dividend yield plus capital gains(local currency)
Dividend yield plus dividend growth
Dividend yield
Floor TradingElectronic Trading
9.24%
39
Figure 3. Cumulative Excess-over-world Returns
-5%
0%
5%
10%
15%
20%
25%
30%-2
4-2
3-2
2-2
1-2
0-1
9-1
8-1
7-1
6-1
5-1
4-1
3-1
2-1
1-1
0 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Month Relative to Announcement/ Implementation of Fully Automated Trading
Exc
ess
Cou
ntry
Inde
x R
etur
n M
inus
Exc
ess
Wor
ld R
etur
ns
ImplentationAnnouncement