Econ Change Restruct (2009) 42:229–262 DOI 10.1007/s10644-009-9069-5
Information-based trade in Russia and the effects of listing abroad Marco Wo¨lfle
Received: 30 July 2008 / Accepted: 26 January 2009 / Published online: 6 March 2009 Ó Springer Science+Business Media, LLC. 2009
Abstract This paper offers empirical evidence about transaction cost in Russia. After relating empirical measures of information and liquidity to corporate characteristics, competitive theories about cross-listings are tested. Since cross-listings generate competitive effects on transaction costs, potential to affect price discovery exists. The results reveal a lower share of private information for cross-listed firms since more transparent accounting standards reduce the incentives to collect superior information. Robustness for the evidence presented in favor of the legal bonding hypothesis is provided by those firms which list with the highest Russian standards. Measures of information-based trade are lower and the likelihood of listing abroad is significantly higher. Keywords
Information-based trade Cross-listing Bonding hypothesis
JEL Classification
D53 D82 G14
1 Introduction It is desirable to possess superior private information about firms since it provides profitable trading strategies at the exchange. While the agency conflict inherent in inside information is straightforward to understand, it is easy to be confounded with information-based trade. Obviously information collection is costly and has to be compensated with sufficiently large expected gains from trade with other market participants before individuals are provided with an incentive. In opaque financial markets as Russia, which is analyzed in this paper, expected gains from information M. Wo¨lfle (&) Institute for Research in Economic Evolution, University of Freiburg, Platz der Alten Synagoge (KGII), 79085 Freiburg, Germany e-mail:
[email protected]
123
230
Econ Change Restruct (2009) 42:229–262
collection are likely to be high. For this reason, stocks traded at Russian exchanges provide an ideal set to describe the empirical properties of measures of price informativeness and trade activity. In this sense, one contribution of this paper is to provide contrasting evidence to what has been found for more transparent markets as the U.S. for example. Since order flow of market participants is revealing information about their individual valuations of goods, the bid-ask spread, which is the intermediaries’ compensation, may contain information rents beyond inventory holding and order processing costs. These components are crucial for the derivation of information measures and reveal the link to liquidity. However, intermediaries may engage in competition on the bid-ask spread in order to attract a larger share in order flow. A comparison of monopolistic and competitive intermediation in a natural experiment can be observed in financial markets, when firms start to list their shares abroad. Hence, the second contribution of the paper is to learn more about the way information and liquidity measures are affected by this structural break in the information environment of the firm. However, the motivations for these cross-listings are actively debated in the literature. Corporate management does not only aim at stimulating liquidity of order flow or signal quality to the stock holder. Managers may be interested in learning private information about specific growth opportunities as well. Listing abroad may induce more market participants to collect information revealed to the managers by market transactions. However, potential growth opportunities for the firm are costly and the funds needed may only be raised by binding to higher disclosure standards. More information is made public, and the incentives for information collection are reduced for the reason that marginal information is more costly to acquire. A reduction in price informativeness as predicted by the legal bonding hypothesis is supported by the Russian data as well. Employing disclosure requirements of different Russian listing standards as a control variable, robustness of the results is confirmed. Firms that list their stocks with the highest Russian standards account for lower information measures and are more likely to list abroad. To a certain extent, domestic listing standards are substitute for cross-listing. Two other results are worth noting: First, the probability of information-based trade and the Roll88 information measure do not quantify the same characteristics of information. It is recommended to use them as complements rather than substitutes. Second, cross-listings are a main determinant of information and liquidity and explain up to 50% of the variation in these measures. This figure is higher than what can be achieved upon the implementation of corporate characteristics in a U.S. set of data. The paper is organized as follows. Section 2 justifies the choice of and then defines relevant measures of information. In Sect. 3, characteristics of the data are discussed before the estimates are related to corporate characteristics. Section 4 begins with a brief discussion on the most important hypotheses about cross-listings and continues with their analysis based on the data with a special note on the reliability of Russian accounting data. Finally, Sect. 5 summarizes the empirical results and concludes.
123
Econ Change Restruct (2009) 42:229–262
231
2 Measures of information This section offers an introduction to the information measures used in this paper. First, the probability of information-based trade (PIN) and price nonsynchronicity (the Roll 1988 measure (further Roll88)), which are principally used in the empirical literature, are discussed. The PIN measure is well-rooted in the literature about market microstructure and based on transaction volumes, while Roll88 is defined in a comparatively pragmatic way. However, its advantage is its ease of calculation since it is only based on stock and index returns. Llorente et al. (2002) propose a newer, more recent measure of information, which was rarely used in the literature hitherto.1 Based on a stylized model of price formation, it combines price with volume data, and complements the measures in a narrow sense. For the reason that the bid-ask spread, which is unfortunately not available for this paper, is composed of information asymmetries as well as order processing and inventory holding cost, three related concepts are presented for completeness and comparability in the second part of this section. Measures of illiquidity are discussed in Amihud (2002) and Lesmond (2005). However, both authors miss to correct information measures by this order processing component as it is done within this paper. Moreover, an alternative estimate of effective bid-ask spreads discussed by Roll (1984) (further Roll84) is used as well. Since information in stock prices cannot be observed but only estimated, the literature on information measures is extensive and it is easy to find other measures (or even classes) of information measures. However, the three measures in a narrow sense used in this paper are the most popular or commonly agreed-upon variants according to their citation records. 2.1 PIN The probability of information-based trading (PIN) has been derived and applied over a range of studies by Easley and coauthors since the basic idea was provided in Easley and O’Hara (1987).2 Moreover, it has been used as a measure in empirical corporate finance by Bessembinder (2003), Brown et al. (2004), and Vega (2006). All of these derivations and applications borrow from the same set of microstructure assumptions as in Glosten and Milgrom (1985) and Kyle (1985), which postulate that market makers clear orders between buyers and sellers. A simple idea of asymmetric information3 assumes two types of traders: informed and uninformed.
1
It is only applied in Grishchenko et al. (2002).
2
See Easley and O’Hara (1992, 2004) and Easley et al. (1996a, b, 1997a, b, 1998, 2002, 2005).
3
Damodaran (1985) defines ‘‘three dimensions of the information structure—the frequency and accuracy of, and the bias in information releases.’’ For the ease of mathematical tractability and in line with the whole market microstructure literature, this paper does not follow this more elaborate idea of information asymmetry.
123
232
Econ Change Restruct (2009) 42:229–262
While order flow from uninformed liquidity traders generates a daily random walk of buy (eb) and sell (es) orders, perfectly informed traders only trade on days when information events occur with probability a. ‘‘Bad news’’ occurs with probability d and ‘‘good news’’ with the corresponding probability 1 - d. Assuming strong-form market efficiency, days are independent and the arrival of information-based trade follows independent Poisson processes. Then, the signal to noise ratio between l (number of informed traders) and e (arrival of uninformed traders) is used to construct a statistic about the probability of information-based trade. Due to the fact that market makers only know the probabilities a and d but not their true realization, they can only update their beliefs about the occurrence of information events based on the arrival rates of trades. In other words, Bayesian updating is the only means by which market makers minimize losses from trade with informed traders. After some algebra, the conditional probabilities of days with no information event (1 - a), days with bad news (ad) and days with good news (a(1 - d)) transmit into the following function for the probability of informationbased trade: al PIN ¼ ð1Þ al þ es þ eb Estimating the parameter vector h = (a, d, l, es, eb) from the arrival of buy (Bt) and sell (St) orders is challenging since a and d cannot be observed. However, the assumption of independent Poisson processes identifies the likelihood function for the three potential information scenarios for each day t. Bb t s Ss t e Bt ! St ! Bt ðl þ s ÞSt þ adeb b eðlþs Þ Bt ! St ! ðl þ b ÞBt s Ss t e þ að1 dÞeðlþb Þ Bt ! St !
LðhjBt ; St Þ ¼ð1 aÞeb
ð2Þ
Maximizing equation (2) for the whole sample according to V ¼ ðhjMÞ ¼
t¼T Y
ðhjBt ; St Þ
ð3Þ
t¼1
requires extensive computations and good initial values for the parameters since the convergence characteristics of the function are not well behaved. In particular, one must take the functions to large powers when estimating the parameters for stocks with high trading volume. It is noteworthy that most of the empirical articles cited above complain about a substantial number of observations for which the likelihood does not converge. Consequently, the more recent papers of Easley et al. (2002, 2005) apply the following factorization in order to facilitate numerical maximization of the joint likelihood function:
123
Econ Change Restruct (2009) 42:229–262
LððBt ; St ÞTt¼1 jhÞ ¼
233
T X ½b s þ Mt ðlnxb þ lnxs Þ t¼1
þ Bt lnðl þ b Þ þ St lnðl þ s Þ þ
T X
ln½ð1
aÞxSs t Mt xBb t Mt
þ
ð4Þ t adl xBb t Mt xM s
t¼1 t þ að1 dÞl xSs t Mt xM b
s b , and xb ¼ lþ : with Mt = min(Bt,St) ? max(Bt,St)/2, xs ¼ lþ s b All estimates presented in this paper are results from this second joint likelihood function in Eq. (4) based on the improved estimation method proposed by Yan and Zhang (2006). Due to the fact that maximizing equations (2) and (4) generates overflow or is at least very time consuming for most statistical software packages, the authors propose a set of initial values in order to facilitate the computation. Specifically, the initial parameters are generated by means of the marginal expected values of S and B for which a and d as well as a multiplier for eb are set to the five fractions of one (0.1, 0.3, 0.5, 0.7, 0.9) yielding at most 125 combinations. Note that this procedure also reduces the frequency of boundary solutions which are more likely for stocks with larger market capitalization.
2.2 Roll88 The second measure of information that is applied in this paper goes back to Roll (1988). Although it was first developed 20 years ago, this measure and its variants are actively used in the empirical literature as shown by the papers of Lesmond et al. (1999), Morck et al. (2000), Wurgler (2000), Bushman et al. (2004), Durnev et al. (2003, 2004), DeFond and Hung (2004), and Chen et al. (2007). The basic idea of this measure originates from Roll (1988), where the performance of multi-factor models (APT) is tested against the performance of a single factor model (CAPM). Roll’s main result is that regressions of individual monthly stock returns on the market and their corresponding industry index yield an average adjusted R2 of around 35%, which is only marginally improved by means of additional factors. Consequently, he suggests that the remaining variation in the stock returns, i.e. 1 - R2, quantifies firm-specific private information and noise. Given the ability to predict firms’ future earnings as shown in Durnev et al. (2003) or Chen et al. (2007) and to explain corporate investment as shown in the latter paper, empirical literature favors the hypothesis that 1 - R2 measures firmspecific information over the alternative hypothesis of reflecting only noise. Consequently, the second measure of information is 1 - R2 of the following equation: ri;t ¼ b0 þ b1 rm;t þ b2 rb;t þ ei;t
ð5Þ
ri,t is the return on stock i, rm,t represents the market return and rb,t describes the return of the industry index, all at time t. Despite the fact that the PIN and the Roll88 measure are easy to compare since both are bounded between 0 and 1, one
123
234
Econ Change Restruct (2009) 42:229–262
advantage of the Roll88 measure is its ease of calculation. Daily price quotations as well as industry and market indices are easy to acquire for most stocks around the world, while high frequency records on seller- and buyer-initiated orders (St and Bt) are rare for many stocks from emerging markets. 2.3 Llorente Llorente et al. (2002) propose a newer, more recent measure of information, which was rarely used in the literature hitherto.4 It rests on the assumption that ‘‘returns generated by risk-sharing trades tend to reverse themselves, while returns generated by speculative trades tend to continue themselves.’’5 As with PIN, this model is well-rooted in the market microstructure literature. Following the idea of risk-sharing traders, whose trade does not affect long-run returns, is technically equivalent to the role of liquidity traders in Glosten and Milgrom (1985) and Kyle (1985). In contrast to Easley et al. (1996b) and Roll (1988), only informed (speculative) trade shifts the equilibrium price as the pricing kernel of Glosten and Milgrom (1985) postulates. Consequently, one can learn about price information from the persistence in price changes based on historical information as in the following equation: ri;t ¼ c0 þ c1 ri;t1 þ c2 ri;t1 vi;t1 þ ei;t
ð6Þ
As in the previous model, ri,t represents the return on stock i at date t, which is regressed on the constant c0, first-order autocorrelation c1, and the interaction of the lagged price with lagged trading volume (vi,t-1). Its coefficient c2 is the parameter of interest; specifically, this interaction coefficient measures the extent to which insiders trade on their informational advantage. Combining price and volume information, the Llorente measure can be understood as a hybrid between PIN and the Roll88 measure. Note that it combines the strong theoretical foundation of the PIN measure with the ease of computation of the Roll88 measure. Nevertheless, depending on the data set, the Llorente measure is empirically weak as the extensive robustness tests in Llorente et al. (2002) and the small number of empirical applications suggest. 2.4 Complementary measures Given these three measures of information, empirically related concepts are presented. Demsetz (1968) decomposes the bid-ask spread into information asymmetries as well as order processing and inventory holding cost. The first component is evidently the reason to use it as a measure of information, while the two remaining components relate to liquidity. Lesmond et al. (1999) develop a price-based estimate on (effective) transaction costs using the nature of daily security returns. It is straightforward that informed investors will only trade on their information when the value of this information exceeds the boundary of transaction 4
It is only applied in Grishchenko et al. (2002).
5
See Llorente et al. (2002), abstract, and Campbell et al. (1993).
123
Econ Change Restruct (2009) 42:229–262
235
costs yielding positive or negative returns. When daily returns are zero, which is the case for around 40% of the observations of large firms in their sample, the authors suggest that individuals do not possess information valuable enough to trade. It is interesting to note that the authors are able to explain up to 88% in the variation of the bid-ask spread with their measure. Abstracting from theoretical details how these components are connected, this result shows that information measures may be influenced by illiquidity and vice versa. Consequently, it is worthwhile to compare the three measures of information with two established measures of illiquidity from which the second one is used to correct Roll88. The first measure of illiquidity and transaction is developed in an earlier paper by Roll (1984), who shows that first-order serial covariance in returns can be used to construct a proxy for the bid-ask spread. For efficient markets, only unanticipated information may change prices, which means that serial covariation cannot exceed amounts generated by serial covariation in expected returns. In the presence of transaction costs, measured by the bid-ask spread, transaction prices fluctuate randomly around the fundamental value of the underlying. However, when transactions ‘‘too far away’’ from the fundamental value are recorded, incentives are provided to trade against this deviation. This generates negative serial covariation in transaction prices. Based on the likelihood of successive price changes, equation (7) provides a measure for the ‘‘effective bid-ask’’ spread (si) of stock i consisting only of transaction costs (further Roll84). qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð7Þ si ¼ 2 covðpi;t ; pi;t1 Þ Another measure of illiquidity is developed by Amihud (2002) using absolute percentage price change per dollar of daily trading volume. Beyond the correlation found between this measure and the small firm effect, the author shows that these higher required returns of small firms may be due to liquidity premia. Supportive evidence in favor of this measure are provided by Lesmond (2005). For this task the Vuong (1989) test is applied. Based on the residuals of null-models and parameterized models comparable likelihood ratios for several empirical models are computed. It is striking to note the dominance of his own measure of illiquidity and Amihud over other measures as Roll84 for example. Given these empirical results the Amihud measure of illiquidity is applied to the Russian data as well. The daily absolute percentage price change per dollar of daily trading volume (|ri,t |/(pi,t vi,t)) is calculated. First, its quarterly average as shown in the following equation T ri;t X Amihud ¼ 1=T ð8Þ p vi;t t¼1 i;t is a proxy for illiquidity and is used as an additional measure of ‘‘information’’. Second, its daily realizations are used in order to correct the Roll88 measure for illiquidity (further Roll88 corrected) as shown in the following Eq. (9):
123
236
Econ Change Restruct (2009) 42:229–262
ri;t ¼ b0 þ b1 rm;t þ b2 rb;t þ b3
ri;t pi;t vi;t
þ ei;t
ð9Þ
Given that Lesmond (2005) finds that liquidity costs amount to 47% of the bidask spread for Russian stocks, it is not surprising that the results from the regression in Eq. (9) substantially differ from the Roll88 measures generated by means of Eq. (5). A technical note has to be made regarding the numerical properties of the measures. Clearly, the PIN measure as well as both variants of Roll88 range in the interval from zero to one, while the others are unbound from negative to positive infinity. In order to improve comparability, most of the tables present logistically transformed values in addition to the unbound values. More advanced tests, however, are only applied to the transformed values. Observe that beyond these prominent examples, additional or other measures could be used because the literature provides a multitude of variants of these information measures and potentially classes that use variables other than price and volume. Especially regarding variants of the PIN measure, the literature is very active for the reason that its empirical properties are not very attractive.6 Moreover, contributions such as Dennert (1993) show that simpler versions of order imbalance- or volume-based measures can work equally well. However, the preceding subsections show prototypes of classes of information measures that are generally agreed-upon in the literature.
3 Data and descriptive results 3.1 Data Data for the empirical analysis is obtained from the archives of the Russian Trading System (RTS).7 RTS is a trading system consolidating transactions on five regional trading floors8 in order to provide efficient prices on Russian financial instruments. This aggregation is done for the reason that trade in Russia is decentralized to several local exchanges, which results in low trading frequencies and volumes relative to observations on U.S. exchanges. In particular, the trading system allows to register OTC trades as well. It was established in 1995 with the leading Russian composite index, the RTS index, starting September 1. Trade in stocks, bonds and derivatives is feasible in corresponding segments of the market. It is interesting to note that prices are quoted in U.S. dollars for the reason that settlement in foreign currency is feasible for stocks in the ‘‘Classic Market’’ segment. However, only for the most liquid stocks, anonymous, order-driven trade is feasible. Less liquid stocks are traded on a quote-driven basis with the preliminary deposition of collateral. 6
E.g. van Oppens (2004), Venter and de Jongh (2004), or Brown et al. (2006).
7
http://www.rts.ru/en/archive/securityresults.html.
8
Russian Trading System Stock Exchange, OJSC Russian Trading System Stock Exchange, RTS Clearing Center, RTS Settlement Chamber and St.Petersburg Stock Exchange.
123
Econ Change Restruct (2009) 42:229–262
237
On the RTS database a list of all transactions performed on regular Russian exchanges is provided. For each transaction, exact times, prices and the volume of shares traded is provided as well as the corresponding figures for market capitalization and the value of each trade. One important advantage of this database for the calculation of the PIN measure is the registration of the origin of the trade (buyer- or seller-initiated) for most transactions. This means that the algorithm developed by Lee and Ready (1991) in order to generate a proxy for the origination (buyer or seller) of transactions based on observed prices has to be applied only spottily, namely when origin information is missing. Moreover, for some stocks, institutional address and repo trades are registered which will later help as a robustness test when analyzing the Llorente measure. For this paper, data is collected over the period from 01/01/2006 until 12/31/2007 on all stocks that are traded at one of the official quotation levels provided by the RTS.9 For each stock, quarterly time series are generated from the trade records. However, only liquid stocks are analyzed. Time series of stocks with less than two quarters are excluded as well as those for which less than three consecutive months of trade records exist. In particular for the calculation of the PIN, in each quarter, a minimum of 20 observations is set. A set of 130 firms yields 926 valid firm-quarters, which are used in order to compute and analyze the empirical measures described in the previous section. A complete list of the names of all stocks analyzed in this paper as well as their ISIN number and ticker symbols in the RTS is provided in the Appendix. 3.2 Informativeness and liquidity Based on the data from the Russian exchanges, information is quantified by means of the measures described in Sect. 2. Aggregate results on the six measures of information on 926 firm-quarters are presented in the following Table 1. The table gives a descriptive overview on the estimated information measures. It gives the three information measures in a narrow sense (PIN, Roll88, and Llorente) in the first three rows, while the following three rows are filled with the two figures on illiquidity (Roll84 and Amihud), and the corrected Roll88 measure. It has to be noted that the figures on the Llorente measure and illiquidity are logistically transformed in order to compare them to the other measures since the original figures range between negative and positive infinity. However, a probability of information-based trade or R-squared figures can only range from zero to one. Untransformed figures on Llorente, Roll84 and Amihud are presented in the last three rows of the table. Although for each information measure 926 observations could exist, a substantial number are missing for all measures except Roll88. To the contrast, 350 estimates on the PIN suggest that only about one third of all time series can be fitted. However, as already noted before, time series on the PIN are only estimated with a minimum of 20 observations, which is the case for 487 of the 926 Roll88 observations and corresponds to a convergence ratio of 71.87%. Given that the 9
For more information, please see Table 10 and http://www.rts.ru/s718.
123
238
Econ Change Restruct (2009) 42:229–262
Table 1 Overview: measures of information Measure
Obs
Mean
SD
Skew
Kurt
Min
Max
PIN
350
0.4529
0.1115
-0.0040
3.5568
0.0000
Roll88
926
0.7434
0.2140
-1.1681
3.7880
0.0000
0.7812 0.9955
Llorente
415
0.5205
0.0341
4.3698
23.2387
0.4737
0.7319
Roll84
615
0.5053
0.0053
4.7716
38.6102
0.5000
0.5588
Amihud
499
0.7927
0.3790
-1.4575
3.2497
0.0000
1.0000
Roll88 cor
499
0.6064
0.2037
-0.6611
2.7306
0.0141
0.9544
Llorente unt
415
0.0834
0.1441
4.5162
24.7187
-0.1051
1.0045
Roll84 unt
615
0.0212
0.0212
4.7931
38.9094
0.0000
0.2363
Amihud unt
499
5.7804
6.7612
-0.76537
2.6047
-11.1421
16.6355
This table gives a descriptive overview of the estimated information measures. The first column gives the names of the measures. In the second column, the number of observations (firm-quarters) for which the specific measure can be estimated is counted for the whole sample. The third and fourth columns give means and standard deviations, respectively. Higher moments, i.e. skewness and kurtosis are presented in the following columns. The last two columns give minimum and maximum values for the estimated measures. It has to be noted that the figures on the Llorente measure and illiquidity (Amihud and Roll84) are logistically transformed in order to compare them to the other measures since the original figures range between negative and positive infinity. However, a probability of information-based trade or Rsquared figures can only range between zero and one. Untransformed figures on Llorente, Roll84 and Amihud are presented in the last three rows of the table
evidence in Easley et al. (2002) or Yan and Zhang (2006) already suggests that PIN likelihood functions frequently do not converge. This ratio is only about 20% lower than their estimates for the U.S. and speaks in favor of the quality of this Russian data set. Moreover, it is not surprising to find only 615 Roll84 observations and only 415 significant Llorente coefficients. Autocorrelation may be present for the majority of the time series, while not for the whole sample. Allowing only for first-oder negative autocorrelation as is done for the computation of Roll84 and analyzing only the interaction with volume as is done for the estimation of Llorente, the potential for significant observations is even reduced to the figures presented. For the illiquidity measure according to Amihud, 499 observations can be found, more than half of the sample. Based on these observations, means, standard deviations as well as minimum and maximum values on the estimates of the measures are displayed in the columns of the table. Yielding an average of 0.79, it is easy to observe that the Amihud measure is higher than all information measures. Moreover, it is highly variable as implied by a standard deviation of 0.38 and extreme values at the boundaries of the potential interval. This variation suggests that the measure may be well-suited to correlate with corporate characteristics. The Roll88 measure is second highest, and it is striking to note how it is affected by liquidity premia in spite of Roll’s prediction that the market model inherent in his information measure is only weakly dominated by refinements. Both variants of the measure, however, are fairly volatile with standard deviations ranging around 20%.
123
Econ Change Restruct (2009) 42:229–262
239
Table 2 Measures of information: different industries Industry
PIN
Roll88
Lloren
Roll84
Amihud
Roll88 corr
Cons & Ret
0.4827
0.7909
0.5122
0.5048
0.9792
0.7297
Elec Util
0.4395
0.7901
0.5268
0.5071
0.8784
0.6223
Financial
0.4678
0.6577
0.5199
0.5050
0.7242
0.5200
Industrial
0.3965
0.8469
0.5116
0.5049
0.9442
0.7386
Met & Min
0.4432
0.6670
0.5203
0.5047
0.7335
0.6037
Oil & Gas
0.4239
0.6056
0.5497
0.5047
0.4241
0.4544
Telecom
0.4910
0.6960
0.5139
0.5044
0.8455
0.5693
This table presents mean estimated information measures for the different industry indices of the Russian stock exchanges
Even after correcting for liquidity, Roll88 corrected is on average 10% higher than the PIN and the Llorente measure, which range around 50%. It is interesting to note the strong variation in the samples. Except the estimates on the PIN measure, which are almost normally distributed, all measures have significant skewness coefficients and Roll84 and Llorente have excessive kurtosis coefficients. For these measures, most of the estimated figures are observed below the means around 50%, while the negative skewness coefficients computed on Amihud and Roll88 indicate that observations above the mean are more likely. When the Roll88 measure is corrected for liquidity, however, significance in the deviation from a normal distribution is reduced substantially as implied by higher skewness and lower kurtosis coefficients. The close distribution of the Llorente and Roll84 measures around the means illustrated by an excessive kurtosis coefficient explains why disaggregated figures for the RTS industries10 are comparatively flat for these measures as can be seen in Table 2. In contrast, the second illiquidity measure, Amihud, ranges between averages of 42% for oil and gas producers and 98% for consumer and retail firms. Given active trading and immense public interest in energy markets due to tremendous rises in price in the most recent past, it is not surprising that oil and gas enterprises’ stocks are highly liquid. However, it is still striking that the financial services industry’s Amihud figure amounts to 72% despite that fact that these firms should have a strong interest in openness in order to raise funds on international financial markets. Only two firms (one third of all financial observations) account in IAS. All firms in the consumer and retail, as well as the metals and mining sectors, do not follow internationally agreed accounting standards, which corresponds to higher illiquidity figures.11 In spite of the fact that variation in the PIN estimates is substantially weaker than in illiquidity, it can still be seen in the Table that the average oil and gas, industrial and electric utility firm has a significantly lower probability of information-based 10 The industries used for the presentation in the table are RTS Consumer and Retail, RTS Electric Utilities, RTS Financials, RTS Industrial, RTS Metals and Mining, RTS Oil and Gas, and RTS Telecom. 11
See Appendix for a tabulation of accounting standards in the different industries.
123
240
Econ Change Restruct (2009) 42:229–262
trade than consumer and retail or telecommunication enterprises. Despite its high correlation with the Roll88 measure, the industry specific differences in the averages are not according between the two measures. For example, the average industrial firm has the highest Roll88 score while it generates the lowest PIN measure, even after correcting for liquidity premia. These findings strongly support the procedure in Chen et al. (2007) of using the PIN and the Roll88 measure as complements instead of substitutes for their empirical implementation. Note that the authors find correlation figures around 20%, twice as large as for the data on Russian stocks shown in Table 3. Correlation of Amihud amounts to 16% with the PIN and to 70% with the uncorrected Roll88 measure in spite of the fact that the Roll measure does not incorporate trade volume information. These highly significant coefficients may come from the relationship with the bid-ask spread which is shared by both the information measures and the illiquidity measures. It also supports that both ideas are empirically related concepts and may be used simultaneously to test information discovery and legal bonding by means of Russian data. Another result of the analysis of correlation is the exclusion of the Llorente measure from more advanced tests. The low variation which is shown in the descriptive tables and the finding of significant negative correlation with the other information measures suggests that it is not a sound measure for the Russian data analyzed in this paper. Given that the data allows correction of the volume information applied to estimate the Llorente measure, qualitative differences cannot be found. Based on non-institutional trade, the negative correlation with the Roll88 measure even increases. Roll84 is not correlated with any other measure. In order to relate these figures to data from the U.S., it is straightforward to use the descriptive statistics presented in Yan and Zhang (2006). It is not surprising that the authors quantify significantly lower averages of the PIN in the range between 10% and 20% for their sample of U.S. stocks. Due to the shorter horizon of the Russian sample, it is not possible to support decreasing PIN estimates over time. Moreover, January effects as shown for the American data amount to only 1.5% for Russian data. The t-test reveals that the PIN estimates are not significantly lower in the first quarter than during the rest of the year.
Table 3 Measures of information: correlations PIN PIN Roll88 Llorente
Roll88
Lloren
Roll84
Amihud
1 0.0998* -0.1988***
1 -0.5399***
1
Roll84
0.0170
0.0534
-0.0880
1
Amihud
0.1648***
0.7039***
-0.6514***
0.0496
Roll88 corr
0.1193**
0.9493***
-0.5325***
-0.0406
1 0.6752***
This table presents correlation coefficients among the different information measures Coefficients with *, **, and *** indicate statistical significance on the 10%, 5%, and 1% level
123
Econ Change Restruct (2009) 42:229–262
241
While the Llorente measure does not show significant differences in the t-test for the first quarter, the average Roll88 score decreases by 6% from 0.76 to 0.70. Correcting for liquidity premia, the decrease drops to 3% and the t-test becomes only marginally significant (11.9%). Again this finding supports the correction for liquidity, especially since the illiquidity measures themselves are not significantly different in the first quarter. Moreover, it raises the question whether the PIN and the Roll88 measure are affected by other measures of trade activity as well. The following Table 4 shows the relationship between different proxies of trade activity and the measures of information. Despite the fact that Roll84 is defined as a measure of illiquidity, it is not correlated with any proxy for trade activity. It is straightforward to relate this result to the low variation in this measure and the absence of significant correlation with the other measures, which to the contrast correlate with the proxies for trade activity. While the PIN does not show significant covariation with skewness or kurtosis coefficients of the quarterly volumes of traded shares, Amihud and both Roll88 variants positively correlate with these figures. Correlation with trade activity is reduced by means of the liquidity correction of the Roll88 measure, however, it must be noted that it is not removed completely. For the lower double rows of the table beginning with the total number of trades, qualitative results for the information measures are more in accordance to each other, again except for the Llorente measure. Quarterly total and average volume of traded shares do not significantly correlate with any measure despite the fact that the frequency of trades does. By definition, it is clear that the total and average number of trades per quarter substantially reduce the illiquidity measure by around 61%. However, correlation with the Roll variants is almost as large, while the coefficient for the PIN only amounts to approximately one third. Based on the correlations with Table 4 Measures of information: correlation with trade activity PIN
Roll88
Lloren
Roll84
Amihud
Roll88 co
Skewness (volume shares)
-0.0322
0.3821*
-0.2672*
0.1058
0.2870*
0.2600*
Kurtosis (volume shares)
-0.0377
0.3234*
-0.1734*
0.1011
0.2180*
0.2065*
Quart. total (vol. shares)
-0.1127
0.0147
0.0458
0.0711
-0.0582
-0.0701
Quart. average (vol. shares)
-0.1117
0.0131
0.0469
0.0709
-0.0592
-0.0700
Quart. total (trades)
-0.2197*
-0.5374*
0.7620*
-0.0552
-0.6166*
-0.5329*
Quart. average (trades)
-0.2175*
-0.5332*
0.7298*
-0.0511
-0.6108*
-0.5264*
Quart. total (vol. money)
-0.2227*
-0.4168*
0.6929*
-0.0311
-0.4663*
-0.4397*
Quart. average (vol. money)
-0.2220*
-0.4105*
0.6553*
-0.0301
-0.4595*
-0.4323*
Quart. total (cap)
-0.1804*
-0.4450*
0.6472*
-0.0535
-0.5264*
-0.4619*
Quart. average (cap)
-0.1800*
-0.4489*
0.6430*
-0.0533
-0.0001
-0.4644*
This table presents correlation coefficients between measures of information and trade activity. Asterisks on the correlation coefficients indicate significance on the 1% level. The first double row gives correlation between information measures and skewness (kurtosis) coefficients of the trading volume measured in the number of traded shares. The second double row presents the corresponding figures with quarterly total and average number of traded shares. The following double rows display quarterly total and average numbers for the number of trades, trading volume in monetary units and market capitalization
123
242
Econ Change Restruct (2009) 42:229–262
volume in monetary terms and with market capitalization, the results in Yan and Zhang (2006) are supported by the Russian data as well. Stocks which trade more frequently and which have a higher market capitalization tend to have lower probabilities of information-based trade. Another potential application of these correlations is to use them in order to develop proxies for information risk as is done by Aslan et al. (2007). The authors exploit the relationship between estimates on the PIN measure and corporate characteristics such as age, analyst coverage, total assets, or Tobin’s q to find a valid proxy for information risk since they admit that the empirical properties of the PIN are not very attractive. Finally, they show that their proxy for information risk (PPIN) is able to explain around 45% of the PIN variation and that its in-sample asset-pricing properties are comparable to those of PIN.
4 Determinants of price informativeness It is commonly agreed among economists that one of the most important functions of markets is that they serve as a means to aggregate information about the true value of goods. For example, Fama (1970) is frequently cited for affirming that the equilibrium price ‘‘reflects all the available information in the market’’ and communicates it to the participants. Clearly, this statement takes recourse to Hayek (1945) and his contributions to general equilibrium theory. Equilibrium theory guarantees efficient prices although information is dispersed among market participants as long as their information sets sufficiently overlap. For the aggregation of information, the existence of a Walrasian auctioneer is necessary to perform the clearing of individual residual demand and supply yielding equilibrium prices and to coordinate transactions for individuals willing to trade. Unfortunately, this framework is static in nature and does not provide details about the formation of transaction prices. Transaction prices, however, reveal market participants’ information and are the means by which information is aggregated into equilibrium prices, i.e. true values. In this respect, the contribution of Grossman and Stiglitz (1980) is seminal. The authors show that when market participants are asymmetrically informed, uninformed traders can learn from transaction prices. Since information collection is costly, the potential to learn from market prices reduces the incentives to collect information and a rational expectations equilibrium may fail to exist. This situation constitutes the information paradoxon for the reason that fully revealing prices erase the incentive to collect information for any market participant. In this case, prices cannot embody information at all. The information paradoxon clearly reveals that market efficiency crucially depends on the cost of information collection, the asymmetry of information and the potential of the price mechanism to reveal information. However, real markets are frictional and individuals may prefer intermediation over searching for trading partners themselves. When choosing between the search market and potentially several competing intermediaries, individuals face a related tradeoff between the probability of trading and gains from trade. It depends on individuals’ residual
123
Econ Change Restruct (2009) 42:229–262
243
demand and the spread set by the intermediary as well as the likelihood to find transaction partners on the search market. Regarding international financial markets, firms listing their stocks on several exchanges with different pools of investors constitute such a frictional market. Asymmetrically informed investors look for arbitrage opportunities and switch to the other exchange or to other intermediaries, whenever they offer a better combination of gains from trade, i.e. spreads, and the probability of trading, i.e. liquidity. In this sense, every cross-listing on another foreign exchange generates a structural break in the informational environment of the firm. Given that intermediaries compete on their spread for attracting order flow, it is interesting to learn the way this affects market efficiency. In other words, the empirical question that is raised is whether the change in the expected value from trade, which is due to more attractive spreads for traders at the cost of lower probabilities of trade, is net positive or negative. When the expected value from trade is reduced, it deters individuals to collect information. For this reason, it is possible to show that market efficiency may decrease when switching from monopolistic to competitive intermediation. The following analysis provides evidence on the size and determinants of this net effect based on a short account of more detailed hypotheses discussed in the literature about cross-listings and specific circumstances of the Russian information environment. 4.1 Cross-listings Given that the benefits from a reduction in transaction costs by increased competition between intermediaries after a firms’ cross-listing, it has been shown that only the potential to incorporate smaller pieces of information into prices is not sufficient to improve market efficiency when disregarding the fact that incentives to collect information is reduced. Beyond these basic structures affected by the crosslisting, additional aspects are relevant and motivated a scientific debate during the last 20 years almost as active as foreign listing on the exchanges has been. In the following, the four most important strands in the literature are briefly discussed without claiming to be exhaustive. First, cross-listings may reduce regulatory burdens and transaction costs for foreign investors, particularly when markets are segmented to a high degree. Consequently, the opportunity to acquire an additional pool of investors is often welcomed by the corporate management when it tries to overcome local credit constraints. Contributions by Kaplan and Zingales (1997) or Baker et al. (2003) show that the decision to list abroad is often motivated by the desire to raise funds from foreign investors. Merton (1987), Fuerst (1998), and Domowitz et al. (1998) favor another thesis, which provides a closer link to information. Cross-listing not only increases visibility to foreign investors, it is also often used as a device to signal quality to all investors. Given that only profitable firms may be able to bear the substantial direct and indirect listing costs, corporate management may use cross-listings as an opportunity to convince investors of the quality of the firm.
123
244
Econ Change Restruct (2009) 42:229–262
Third, indirect listing costs may be substantially high when the listing standard a company is applying for requires compliance with different accounting rules. This is one of the basic ideas of the legal bonding hypothesis proposed by Stulz (1999), Coffee (1999, 2002). Operating in a tighter regulatory environment may generate efforts in order to comply with the accounting rules and reduce the management’s ability to extract private benefits. However, it is in the interest of the management to reduce the agency conflict, only when the cross-listing potentially generates the possibility of raising additional funds needed to pursue a specific growth opportunity. The proceeds generated by this growth opportunity may compensate the management’s losses in private benefits. It is straightforward to understand that binding to more transparent accounting rules may reveal information that was private before and should reduce the amount of information-based trade. Fourth, being a means of aggregating and communicating investors’ information, the market mechanism can be attractive for the corporate management as well. By means of official price quotations, the corporate management can learn about the external evaluation of its actions and strategies. For example, Chen et al. (2007) provide empirical evidence that market prices incorporate information which is not at the possession of the management. Given that managers can actively learn from market transactions, they will adjust their behavior to the information inherent in market prices. However, the management’s sensitivity crucially depends on the amount of information revealed by market prices, and consequently this sensitivity is increasing in the likelihood of information-based trade. Based on these findings, it is vital to assume that a second external opinion as provided by a cross-listing may be beneficial to the corporate management which is the core argument of the theoretical model derived by Foucault and Gehrig (2008). This thesis is often discussed in opposition to the legal bonding hypothesis for the reason that it implies a rise in the amount of information-based trade after the cross-listing. 4.2 Accounting standards and the legal bonding hypothesis Before the theory of price informativeness can be tested against the legal bonding hypothesis, it is appropriate to discuss the nature and quality of the Russian accounting and listing standards since both substantially relate to the results generated by the empirical analysis. Despite the fact that Russian legislation and listing standards borrow from IAS to a large extent, de facto accounting standards are relatively opaque. For example, a questionnaire conducted on Russian accountants and auditors by Rozhnova (2000) reveals that 71% of the respondents do not think that financial statements (in Russia) adequately reflect the company’s economic reality. Vorushkin (2001) highlights that IAS generate advantages for both outside and inside investors such as transparency (outsiders) and efficiency, i.e. lower cost of capital when raising funds on financial markets or a better organizational structure within the enterprise. Nevertheless, interviews conducted by Preobragenskaya and McGee (2003a, 2004) on Russian branches of the big four auditing companies reveal that corporate governance standards in Russia are still very weak. In spite of profiting from uniform reporting, a better comparability of their subsidiaries and ease in analyzing financial
123
Econ Change Restruct (2009) 42:229–262
245
statements, large enterprises such as Lukoil publish their quarterly reports with a huge delay, sometimes exceeding one year. Gazprom, the largest, state-owned Russian enterprise does not issue (sound) financial statements at all. However, Preobragenskaya and McGee (2003b) show that foreign capital is attracted by credibly reported financial information, especially since it allows market outsiders to have a comparable decision basis. This thesis is supported by Goncharov and Zimmermann (2007) who show that financial consideration affects reporting incentives of Russian companies. The authors analyze window dressing before credit decisions. Financial statements of firms which have applied for credit always improve in the period before the funds are granted while the quality of the statements substantially declines after the decision. In other words, good reporting is important to acquire money on financial markets but from banks as well. In this sense Russian enterprises’ need for funds may improve accounting standards in the long run. Given the evidence on these internal mechanisms of information production, external mechanisms such as the information produced or revealed by trading at the stock exchange are analyzed as well, for example in Christoffersen and Slok (2000) and Grishchenko et al. (2002). In both papers, it is found that Russian stock markets contain a high degree of private information and that this information is predictive about the future economic performance of firms. Black et al. (2006) find a positive relationship between the quality of corporate governance and firm value measured by Tobins q and other balance sheet ratios. One central motivation for analyzing this relationship between internally and externally generated information is the ‘‘legal bonding’’ hypothesis of cross-listings. In essence, Stulz (1999) and Coffee (1999, 2002) highlight the importance of the firms’ management for the decision to list the firms’ shares abroad. Beginning with a typical agency conflict scenario, it is assumed that the firm’s management is able to extract private benefits from the firm’s economic performance and that consequently its interest in sound corporate governance standards is only moderate except a situation in which the management can pursue a specific project. Given that this marginal project is a growth opportunity for which the management’s private benefits are increasing in the proceeds, managers are willing to ‘‘bond’’ themselves to higher disclosure and legal standards in general. Despite the fact that investors require a return premium for more opaque firms, the above logic helps to explain why stocks of firms that operate with a higher listing standard are traded at comparatively higher prices than stocks with lower standards. Firms which are more transparent are able to attract more funds and can pursue more of their potential projects. Given that asymmetries in the disclosure and listing standards between exchanges in Europe or the U.S. and exchanges in less-developed countries are very strong, the results support Doidge et al. (2004) who find that firms listed in the U.S. are worth more. Depending on whether the stock of a firm is listed on a foreign exchange or not, Table 5 presents averages of Tobin’s q in 2006 for Russian enterprises and the corresponding p-value of the t-tests conducted on the averages. Cross-listings are only collected for Germany, the U.K. and the U.S. for the reason that Russian enterprises use other exchanges only spottily. However, analyzing one representative for Continental Europe and two Anglo-American countries helps to contrast the efficiency of the different legal systems and listing standards.
123
246
Econ Change Restruct (2009) 42:229–262
Table 5 Tobin’s Q and the cross-listing Average q not listed
Average q listed
T-test (p-value)
Obs
Germany
1.6235
1.2138
0.0009
148/165
U.K.
1.3171
1.7524
0.0041
248/65
U.S.
1.3492
1.4759
0.3071
169/144
Depending on whether the stock of a firm is listed on a foreign exchange or not, this table presents averages of Tobin’s q in 2006 for Russian enterprises and the corresponding p-value of the t-tests conducted on the averages. The last column displays the ratio between non-cross-listed and cross-listed firms for the specific listing destination
On average, firm-quarters of listings in the U.K. reach a q of 1.75, which is larger by around 0.44 than the qs of non-cross-listed peers. It is striking to note that not only the difference between the means of cross-listed and non-cross-listed firms in the U.K. is larger than in the U.S., but also that the average q is substantially higher than for firms listed only in the U.S. Despite the fact that the average q for U.S. cross-listings increases from 1.35 to 1.48, the t-test does not show significant differences. However, firms cross-listed in Germany have the lowest q-values with 1.21 on average. In other words, firms which are listed in Germany have a lower market capitalization than their peers. These results are supported by analyzing the combined effects as well. Regressing q on all cross-listing dummies and their combined effects reveals that only the parameters of the dummies for the U.S. and for Germany remain significant. While the marginal cross-listing in Germany reduces Tobin’s q by 0.81, the marginal cross-listing in the U.S. increases q by 0.88. Cross-tabulating the realizations of the cross-listing reveals that the effect from the U.K. is insignificant in the combined regression. Only a small share of firms listed in the U.K. do not list in the U.S. Large q-values for those firms listed in both countries explain why the t-test reveals such a large difference, which seems to be due to the listing in the U.K. The difference originates from the combined effect since the largest average (1.85) is computed for firms listed in the U.K. and the U.S. simultaneously. Given that legal standards in the U.K. and the U.S. are higher than in Germany, these results offer support for the legal bonding hypothesis. However, lacking reliable accounting data for 2007, the sample had to be reduced to 2006 for this test, which weakens the robustness of the figures. In order to contrast these findings, cross-listings are related to the measures of information, which has the additional advantage that both the price-informativeness and the legal bonding hypothesis can be tested alternatively. For each measure of information, three t-tests are conducted in order to analyze the effects of cross-listings in Germany, the U.K. and the U.S. Table 6 provides the results of these tests. For each measure, the results of the tests are given by blocks of four rows. The first two rows present the average measure of information depending on whether the firm-quarter observed a cross-listing or not. The third row presents the p-value of the t-test. Finally, the last row displays the ratio between non-crosslisted and cross-listed firm-quarters.
123
Econ Change Restruct (2009) 42:229–262
247
Table 6 Cross-listing effects on the measures (t-test) Country
Parameter
Germany
Average not listed
0.4591
0.8015
0.5135
0.5057
0.9208
0.6945
Average listed
0.4486
0.6626
0.5313
0.5048
0.6381
0.5014
t-test (p-value)
0.3814
0.0000
0.0000
0.0371
0.0000
0.0000
Observations
192/155
581/387
249/163
334/268
269/227
269/227
Average not listed
0.4612
0.7880
0.5126
0.5056
0.9099
0.6641
Average listed
0.4319
0.5491
0.5503
0.5042
0.4436
0.4358
t-test (p-value)
0.0377
0.0000
0.0000
0.0126
0.0000
0.0000
Observations
267/80
745/173
325/87
491/111
370/126
370/126
Average not listed
0.4518
0.7994
0.5129
0.5058
0.8814
0.6826
Average listed
0.4579
0.6400
0.5326
0.5045
0.6719
0.5046
t-test (p-value)
0.6094
0.0000
0.0000
0.0050
0.0000
0.0000
Observations
195/152
593/325
252/160
383/219
283/213
283/213
U.K.
U.S.
PIN
Roll88
Lloren
Roll84
Amihud
Roll88 co
Depending on whether the stock of a firm is listed on a foreign exchange or not, this table presents t-tests of the measures of information. For each country (listing destination) in column one a t-test is performed, for which the first row of the block gives the average information measure for firm-quarters not listed in the specific country, while the second row gives the average over firm-quarters listed in the specific country. The following row presents the p-value of the t-test. The last row displays the ratio between noncross-listed and cross-listed firm-quarters for the specific listing destination
The table reveals that all measures decrease by listing abroad except the Llorente measure and except PIN, which is only significantly affected by listings in the U.K. Comparing the differences in the measures of information among the three potential listing destinations, it is clear that the cross-listing effects are strongest for the U.K. For example, the mean Amihud score for non-cross-listed firms amounts to 0.91 and decreases by 0.44 for firms listed in the U.K. Moreover, both variants of the Roll88 measure decrease by around 0.24, which is also slightly larger than the decrease by listing in Germany and in the U.S. Observe that Roll84 decreases only slightly in spite of the fact that all t-tests are highly significant. Evidence on this measure is more straightforward when analyzing untransformed figures. The decrease in the effective spread quantified by this measure amounts to 0.4% for German, and 0.5% for U.K. as well as U.S. cross-listings corresponding to a mean decrease of around 1US$. Comparing firms without cross-listings with those being listed on all three other exchanges reduces the Roll84 by 1%. Although, it does not generate correlation with information or liquidity as shown in the previous tables, it qualifies an adequate measure for transaction costs and supports the prediction made by the frictional market theory that intermediaries compete by reducing the spread to attract order flow. Overall, reductions in information-based trade and illiquidity by listing abroad offer a first support for the legal bonding hypothesis. However, only aggregate statistics have been presented without showing the change in the distribution of the measures conditional on the number of listings. In order to overcome this, Table 7 shows the first four statistical moments disaggregated for the number of listings.
123
248
Econ Change Restruct (2009) 42:229–262
Table 7 Cross-listing effects on the measures (parameters) Listings 1-4 Mean
Standard deviation
Skewness
Kurtosis
Observations
PIN
Roll88
Lloren
Roll84
Amihud
Roll88 co
0.4538
0.8137
0.5115
0.5058
0.9466
0.7085
0.4953
0.8001
0.5124
0.5061
0.9065
0.6750
0.4534
0.6778
0.5222
0.5045
0.6839
0.5257
0.4352
0.5288
0.5571
0.5043
0.4821
0.4203
0.1137
0.1547
0.0068
0.0065
0.1985
0.1355
0.0989
0.1527
0.0042
0.0048
0.2462
0.1467
0.0956
0.2347
0.0267
0.0029
0.4361
0.2055
0.1268
0.2496
0.0739
0.0024
0.4648
0.1988
-0.0480
-1.6438
3.1758
4.4862
-4.2424
-0.7027
0.2539
-0.8495
0.5533
1.4413
-2.6496
-0.4286
-0.2387
-0.7526
3.0439
1.3747
-0.7715
-0.2922
0.3545
0.0500
1.3529
1.0194
0.0600
0.2034
4.1347
6.7410
38.7077
30.5676
19.7178
3.2519
2.8461
2.9285
4.2891
5.0768
8.6288
2.6277
3.8181
2.8309
13.4516
6.5601
1.6880
2.4744
2.4003
1.8902
3.3213
3.3116
1.1080
166
492
221
320
237
2.0716 237
32
79
28
49
37
37
92
235
107
150
137
137
57
112
56
83
85
85
Depending on the number of listings from one to four (one indicates no cross-listing), this table gives a descriptive overview of the estimated information measures. For each measure (columns) the first four statistical moments as well as the number of observations for the respective number of listings are presented below. Means from the first to the fourth listing are given in panels of rows followed by standard deviation, skewness, kurtosis and the number of observations, i.e. firm-quarters
The table presents means from the first to the fourth listing in panels of rows followed by standard deviation, skewness, kurtosis and the number of observations, i.e. firm-quarters. Except for the Llorente measure, all mean figures monotonically decrease in the number of listings. Despite the fact that PIN and Roll84 observe a small increase from no cross-listings to one cross-listing, these parameters are in accordance with the t-tests presented before, supporting the legal bonding hypothesis that price informativeness is reduced by listing abroad. Interestingly enough, listing abroad changes the whole distribution of the measures. From the first to the fourth listing, skewness and kurtosis coefficients approach what is indicated by a normal distribution for the measures Roll88, Roll84 and Amihud. For example, skewness on Roll84 is highly positive for firms without a cross-listing while it amounts to only 1.02 for firms being listed on all foreign exchanges. Observing a higher share of observations below the mean, skewness of 4.49 versus 1.44, may be the reason why the first cross-listing generates a small increase in the mean figure. The evolution of the Roll84 figures from non-crosslisted firms to firms with three foreign listings is of particular interest regarding the frictional markets theory since this measure of transaction cost shows that
123
Econ Change Restruct (2009) 42:229–262
249
monopolistic intermediation may dominate competition to a certain extent. For the first cross-listing the net effect on the tradeoff between gains from trade and probability to find matching partners is negative. In other words, transaction costs increase for the majority of the sample. However, observations on firms with more than one foreign listing indicate a positive net effect. Expected conditions for market participants may improve with a sufficiently large number and potentially a specific set of foreign exchanges. Regarding this statistic, evidence on the Amihud measure supports the results found with Roll84. Illiquidity is decreasing sharply, whereas only a few observations lay at the upper end of the distribution for the third cross-listing. Moreover, kurtosis decreases from excessive 19.72 to only 1.11. Qualitatively, the evolution of skewness and kurtosis on Roll88 is similar when comparing firms without to those with three foreign listings. However, in contrast to the correlations with trade activity, skewness and kurtosis on the corrected Roll88 measure do not change dramatically. Correcting Roll88 already generates values close to the normal distribution, similar to the PIN measure. Given the fact that the distributions of the subsamples are depending on the number of cross-listings, it is vital to analyze the way correlation between the measures is affected. For this purpose, Table 8 is similar to Table 7 and presents correlation coefficients disaggregated for the number of listings. Given that the correlation between PIN and Roll88 is only weakly significant on aggregate and amounts to only 10%, disaggregated figures show why Chen et al. (2007) find coefficients twice as large. Only three cross-listings generate significant correlation between these information measures for Russian firms. For the reason that the aggregate sample of U.S. stocks analyzed in Chen et al. (2007) is more likely to crosslist, it is not surprising that their aggregate figure is larger. Moreover, the economic significance of this result is that PIN and Roll88 do not measure exactly the same object. If possible it is recommended to use them as complements instead of substitutes for interpretation or for use in secondary regressions as is done by the authors. On aggregate, PIN and Roll88 are correlated with the Amihud measure. Disaggregated figures reveal an increase in correlation for both combinations. Especially, the PIN is only significant for firms with two and three cross-listings. In other words, the empirical overlap between information rents, inventory holding and order processing cost increases with more foreign listings. This result suggests that competition between intermediaries has a merging effect on these three components of the bid-ask spread. Alternatively, another evident explanation for this finding is the change in the potential of prices to reveal information generated by listing abroad. However, regarding the hypothesis of legal bonding, another alternative explanation for this result can be found. Russian firms listing their stocks with higher foreign listing standards generate an increase in correlation among measures of information and illiquidity on their stock prices. For the reason that a decrease in transaction costs in the number of cross-listings is shown, which according to the positive correlation coefficients translates into a decrease in information rents, a decrease in price informativeness by listing abroad may be concluded. While the Llorente measure provides similar unreasonable insight as before, corrected Roll88 figures again provide joint evidence from Roll88 and Amihud. Cross-listing subsamples for Roll84 do not show substantial differences to the
123
250
Econ Change Restruct (2009) 42:229–262
Table 8 Cross-listing effects on the measures (correlations) Listings 1–4 PIN
PIN
Roll88
Lloren
Roll84
Amihud
1 1 1 1
Roll88
Llorente
Roll84
-0.0312
1
0.2626
1
0.0424
1
0.2731**
1
-0.0220
-0.3002***
1
-0.1852
-0.1697
1
-0.2959***
-0.4965***
1
-0.3598**
-0.6221***
1
0.0547
Roll88 corr.
-0.3064***
1
-0.1107
-0.1643
0.4675
1
-0.0381
-0.0786
0.0526
1
-0.1288 Amihud
0.0343
0.0869
-0.1451
0.0730
0.4645***
-0.5999***
-0.0549
1
0.2295
0.4425***
-0.2883
-0.4881*
1
0.1938*
0.6586***
-0.6548***
0.1453
1
0.2308*
0.6851***
-0.6676***
0.1924
0.9268***
-0.2492***
-0.1303
-0.0418
1
1 0.4176***
0.2793
0.9255***
-0.1629
-0.6339***
0.3884**
0.1134
0.8954***
-0.4793***
-0.0878
0.6103***
0.3127**
0.9876***
-0.6134***
0.0267
0.6714*
Depending on the number of listings from one to four (one indicates no cross-listing), this table correlation coefficients among the different information measures. Correlations between two measures are organized in panels of rows increasing from the first to the fourth listing. Coefficients with *, **, and *** indicate statistical significance on the 10%, 5%, and 1%-level
aggregate figures supporting the way to concentrate its use on the interpretation of transaction cost instead of information or illiquidity. As shown in the discussion of the effects on Tobin’s q, it is straightforward to analyze all cross-listing dummies simultaneously since most firm-quarters observe multiple cross-listings. By regressing measures of information on the different dummies, Table 9 allows comparison of their marginal importance. It is not surprising that most significances are very similar to what has been found by conducting t-tests. Again, the effects on the PIN are weak and in contrast to what can be found for the market value based on Tobin’s q, only the U.K. and U.S. dummies are significant. At the exception of Roll84, the U.K. dummy is generally the most effective for other measures as well. However, based on the evidence generated by means of the t-test, it is interesting to observe that in combination, U.K. and U.S. cross-listing dummies are significant. Given that only a small number
123
1.10%
R2
-0.1126*** (-0.0192)
23.40%
0.8140*** (-0.0069) 23.30%
0.5105*** (-0.0007)
0.0190*** (-0.0045)
0.0400*** (-0.0072)
-0.0145*** (-0.0047)
Lloren
1.20%
0.5058*** (-0.0004)
-0.0012** (-0.0005)
-0.0010*** (-0.0003)
0.0003(-0.0006)
Roll84
29.10%
0.9389*** (-0.0133)
-0.0366 (-0.0459)
-0.4146*** (-0.0521)
-0.0578 (-0.0509)
Amihud
31.60%
0.7099*** (-0.0087)
-0.0880*** (-0.0215)
-0.1554*** (-0.0245)
-0.0579** (-0.0249)
Roll88 co
The table presents results of regressing the measures of information on the cross-listing dummies. Each regression is represented by one column in which the coefficients are displayed with stars indicating statistical significance (* for significance on the 10%-level, ** for 5%, and *** for significance on 1*). Standard deviations are printed in parentheses. The last row gives R-squared
0.0260* (-0.0141)
0.4567*** (-0.0085)
U.S.
Con
-0.1996*** (-0.0221)
0.0153 (-0.0185)
-0.0142 (-0.0154)
-0.0319** (-0.0157)
Ger
U.K.
Roll88
PIN
Table 9 Cross-listing effects on the measures (regressions)
Econ Change Restruct (2009) 42:229–262 251
123
252
Econ Change Restruct (2009) 42:229–262
of firms listed in the U.K. do not list in the U.S., the small and weakly significant decrease in PIN may be understood as the additional reduction in price informativeness generated by this listing. In other words, what tended to look like an effect of German and U.S. cross-listings in the analysis of t-tests is to a large extent due to fact that most firms which list on these exchanges list in the U.K. as well. For example, the decrease in Amihud that seemed to be created by all crosslisting destinations is only significant when listing in the U.K. Despite the fact that mean Amihud showed a significant difference of 0.29 for Germany and 0.21 for the U.S., the dummies are not significant in the regression analysis. In effect, these results provide strong evidence for the legal bonding hypothesis. Note that all significant cross-listing dummies generate negative coefficients in the regressions with the measures. Aggregate marginal effects range between -0.001 for Roll84 and -0.41 for Amihud. It is interesting to note that except for the PIN and the Roll84 measure, between 23% and 32% of the cross-sectional variation is explained only by the cross-listing dummies. In particular, the corrected variant of the Roll measure is strongly affected by the cross-listing. In order to test the robustness of these results, evidence supporting the legal bonding hypothesis is generated by means of the different quotation levels on the Russian exchanges. The characteristics of the different listing standards provided by Table 10 in the Appendix show that both kinds of A-level-quoted firms account according to US GAAP or IFRS as opposed to other listing standards. Clearly, the enforcement of these standards is not on the same level of diligence as it is in the cross-listing destinations (Siegel 2005). However, Table 12 in the Appendix shows that firm-quarters from the highest quotation level have significantly lower measures of information and that the effect generated by the cross-listing is only weakly affected. Binding to better reporting standards at home and abroad significantly reduces the probability of information-based trade. Another issue of robustness is correcting for trade intensity since the contribution by Yan and Zhang (2006) and Table 4 reveal correlation between information measures and the number of traded shares. However, this relationship has already been analyzed throughout the whole paper by providing a liquidity corrected version of the Roll88 measure without showing substantially different qualitative results. Moreover, the plots in Table 12 already incorporate the quarterly total number of trades as an explanatory variable for the measures without changing the qualitative statement of the regressions. It has to be noted that cross-listing dummies and these two control variables contain high explanatory power for the information and illiquidity measures based on Russian data. R-squared figures for Roll88 and Amihud yield 41.70% and 50.10%. This corresponds to an increase of 18.30% and 21.00% respectively and especially for the Amihud measure, it is slightly more than PPIN corporate proxies tested by Aslan et al. (2007) are able to explain from the PIN variation. Altogether, the results strongly suggest that foreign listings account for a large share of information and liquidity in the Russian stock market. Due to the fact that listing with higher disclosure standards decreases information rents by disseminating more information to the public, price informativeness is decreasing in the number of cross-listings as predicted by the legal bonding hypothesis.
123
Econ Change Restruct (2009) 42:229–262
253
5 Conclusion Based on an almost ideal set of transaction data, as well as on corporate characteristics and accounting, information-based trade in Russia is analyzed. After presenting different measures of information and liquidity in stock prices, it is shown that oil and gas enterprises are highly liquid, which translates into lower probabilities of information-based trade. Given the correlations between information and trade intensity, the results provide support for the thesis of Yan and Zhang (2006), who suggest that information-based trade is less likely in stocks which are traded more often and which have a higher market capitalization. It is interesting to note that the PIN, the Roll88 and the Amihud measure generate significant, positive correlation as found by Chen et al. (2007). However, distributions of information and liquidity measures are crucially affected by the number of foreign exchanges at which the underlying stock is traded. Comparing firms without cross-listings with those being listed on all three other exchanges reduces the effective spread by 1%. This figure can amount to 2 US$ in transaction prices. The tradeoff inherent in market participants’ choice for a specific intermediary as stated by the frictional markets variant of general equilibrium theory is adequately shown analyzing the evolution of information and liquidity measures over the range of cross-listings. Comparing monopolistic intermediation, i.e. no cross-listing, with competition induced by the first listing abroad, the results indicate that the net effect on the expected value from gains from trade with the probability of trade is negative. Expected conditions for market participants may only improve with a sufficiently large number and potentially a specific set of foreign exchanges. For the Russian data analyzed in this study, two foreign listings on the exchanges in Germany, U.K. or U.S. usually yield a decrease in illiquidity and price informativeness. Moreover, the Roll measures on information and transaction cost as well as the Amihud measure on illiquidity monotonically approach the normal distribution with a higher number of cross-listings. An increase in the number of foreign listings is associated with higher positive correlations between the measures. Given that intermediaries’ competition reduces transaction costs, the positive relationship found with information supports the prediction of the legal bonding hypothesis. Price informativeness decreases by listing abroad. Since both information measures, PIN and Roll, are only weakly correlated, it is recommended to treat these measures as complements rather than substitutes. In order to understand the structure behind information-based trade, impacts generated by different listing destinations are analyzed as well. Measures of information are significantly lower for firms listing with a higher standard. Obeying superior legislation and more transparent accounting standards disseminates more information to the public. However, the data reveals that price informativeness decreases since marginal cost of collecting superior information decreases as well. Robustness for the results on the legal bonding hypothesis is provided by those firms which list with the highest Russian standards. Indeed, higher standards are a central motivation for listing abroad. Except for the PIN measure, listing dummies and trade activity account for 40% to 50% in the variation of the information and liquidity measures. Although a higher domestic listing standard may substitute for
123
254
Econ Change Restruct (2009) 42:229–262
listing abroad, cross-listing dummies remain substantial in size and significance. For cross-listed firms, measures of information-based trade are lower and the likelihood of listing abroad is significantly higher. Having presented empirical evidence for one of the most opaque financial markets, the results provide contrasting evidence to what is found for the U.S. It is not surprising that information-based trade is more likely in Russia. However, it is striking that all technical properties, such as weak empirical properties for the PIN or the relationship with the number of trades, as well as the characteristics of the information measures, such as weak pairwise correlation and the relationship with market capitalization, hold. Since the quality of data on corporate accounting tends to improve over time, it would be interesting for further research to analyze other determinants of measures of information and cross-listings. For example, this would allow replication and development of empirical tests conducted by Aslan et al. (2007), Chen et al. (2007), and Kaplan and Zingales (1997) to a larger extent. Another important issue is the dimension of enforcement of legal standards. For foreign supervisors of accounting and legal standards, it is difficult to verify balance sheet information and if court decisions are carefully respected. Collecting data on compliance with these two standards would be challenging and interesting. From a technical point of view, unreasonable evidence generated by means of the Llorente measure and weak correlations between the PIN and the Roll88 measure point to the need to rethink the quality of these information measures. It is interesting whether these measures quantify different characteristics of information or whether one dominates the others. In this case, different market conditions would be worth analyzing in order to learn more about the measures’ empirical properties.
Appendix See Tables 10, 11, 12, and 13.
123
Econ Change Restruct (2009) 42:229–262
255
Table 10 Listing standards at the Russian Trading System (RTS) Level Market Cap. Ord. (Pref.) Stock
Existence No Losses in years for years
Transactions volume last 3 months
Accounting Free float
A1
10 (3) bill RUR
min. 3
2 out of last 3 min 25 mill RUR
US GAAP and/or IFRS
1 Shareholder or affiliates not hold more than 75%
A2
3 (1) bill RUR min. 3
2 out of last 3 min 2.5 mill RUR
US GAAP and/or IFRS
1 Shareholder or affiliates not hold more than 75%
B
1.5 (0.5) bill RUR
min. 1
-
1.5 bill RUR -
1 Shareholder or affiliates not hold more than 90%
V
–
–
min. 3
2 out of last 3 –
–
I
60 (25) mill RUR
–
–
–
Placement of at least 10%
IPOs –
This table gives a stylized overview of listing standards at the Russian Trading System (RTS). The first column presents the different quotation levels with a special segment of listing standards for initial public offerings (IPOs). The second column gives the minimum amount of market capitalization that is required in order to list ordinary or preferred stock in the corresponding listing standard. Columns three and four give the minimum number of years for which a firm should exist and for which a firm should not generate accounting losses. Special accounting rules and restrictions on concentrated ownership are given in the last two columns
Table 11 Listing standards: different industries Industry Consumer and retail
A1
A2
B
I
V
0
0
74
3
0
16
24
120
0
0
Financial
8
8
33
0
3
Industrial
0
24
56
0
0 0
Electric utilities
Metals and mining
0
0
76
0
Oil and Gas
16
8
38
0
0
Telecom
16
88
31
0
0
This table presents the number of observations of the different listing standards according to the different industry indices at the Russian stock exchange
123
123
-0.2152*** (-0.0332)
5.10%
0.4677*** (-0.0088)
41.70%
0.8273*** (-0.0069)
-0.0002*** (0.0000) 60.80%
0.5054*** (-0.0021)
0.0001** (0.0000)
0.0157 (-0.0150)
0.0084** (-0.0033)
0.0162* (-0.0088)
-0.0116** (-0.0049)
Lloren
0.90%
0.5058*** (-0.0004)
0.0000 (0.0000)
-0.0006 (-0.0005)
-0.0012** (-0.0005)
-0.0010** (-0.0004)
0.0004 (-0.0006)
Roll84
50.10%
0.9752*** (-0.0149)
-0.0004*** (-0.0001)
-0.2350*** (-0.0785)
0.0325 (-0.0422)
-0.2901*** (-0.0578)
-0.0425 (-0.0449)
Amihud
46.10%
0.7255*** (-0.0088)
-0.0001*** (0.0000)
-0.1453*** (-0.0291)
-0.0641*** (-0.0201)
-0.1112*** (-0.0237)
-0.0373 (-0.0230)
Roll88 co
The table presents results of regressing the measures on the cross-listing dummies, a dummy for being listed with the highest Russian listing standard (A1) and the total amount of the quarterly number of trades (# Trades). Each regression is represented by one column in which the coefficients are displayed with stars indicating statistical significance (* for significance on the 10%-level, ** for 5%, and *** for 1%). Standard deviations are printed in parentheses. The last row gives R-squared
R2
Constant
-0.0001*** (0.0000)
0.0156 (-0.0179)
A1
# Trades
-0.1344*** (-0.0206)
-0.0776*** (-0.0167)
-0.0041 (-0.0181)
0.0371*** (-0.0143)
U.K.
U.S.
0.0188 (-0.0163)
-0.0254 (-0.0165)
Roll88
Germany
PIN
Table 12 Cross-listing effects on the measures (regressions(robustness))
256 Econ Change Restruct (2009) 42:229–262
Econ Change Restruct (2009) 42:229–262
257
Table 13 Overview: Ticker symbol, ISIN, and Names Ticker
ISIN
Name and type
AFKS
RU000A0DQZE3
Sistema JSFC, Common
AFLT
RU0009062285
JSC ‘‘Aeroflot’’, Common
AGAMIT
RU000A0JNG89
CJSC ‘‘AG Capital Asset Management’’, FundUnits
AKRN
RU0009028674
JSC Acron, Common
AMEZ
RU000A0B88G6
Ashinskiy metallurgical works, Common
APTK
RU0008081765
‘‘Pharmacy Chain 36,6’’, Common
ARMD
RU000A0JP4J4
PJSC ‘‘ARMADA’’, Common
AVAZ
RU0009071187
JSC ‘‘AVTOVAZ’’, Common
BANE
RU0007976957
Bashneft, Common
BANEP
RU0007976965
Bashneft, Pref
BEGY
RU0009044242
Bashkirenergo, Common
BISV
RU0009059216
BashInformSvyaz, Common
CHEP
RU0009066807
JSC ‘‘hTRP’’, Common
CHMF
RU0009046510
JSC ‘‘Severstal’’, Common
CHZN
RU0009093918
JSC ‘‘CZP’’, Common
DGEN
RU0006752649
JSC ‘‘Dagenergo’’, Common
DIXY
RU000A0JP7H1
OJSC ‘‘DIXY GROUP’’, Common
DVEC
RU000A0JP2W1
JSC ‘‘FEEC’’, Common
EESR
RU0008959655
RAO UESR, Common
EESRP
RU0009029532
RAO UESR, Pref
ENCO
RU0009087456
OJSC Sibirtelekom, Common
ENCOP
RU0009088280
OJSC Sibirtelekom, Pref
ESMO
RU0009075840
JSC CenterTelecom, Common
ESMOP
RU0009075857
JSC CenterTelecom, Pref
ESPK
RU0009101158
OJSC ‘‘Far East Telecom’’, Common
FESH
RU0008992318
FESCO, Common
GAZA
RU0009034268
GAZ, Common
GAZP
RU0007661625
Gazprom, Common
GCHE
RU000A0JL4R1
OJSC ‘‘Cherkizovo Group’’, Common
GMKN
RU0007288411
OJSC ‘‘MMC’’ ‘‘NORILSK NICKEL’’, Common
GRAZ
RU000A0HG4P4
RAZGULAY Group OJSC, Common
GUMM
RU0008913751
GUM, Common
IRGZ
RU0008960828
JSC ‘‘Irkutskenergo’’, Common
IRKT
RU0006752979
Irkut Corporation, Common
KHEL
RU0009102404
Kazansky Helicopter Plant, Common
KIRZ
RU0009084263
Kirovsky Zavod, Common
KLNA
RU0007247243
OJSC Concern ‘‘KALINA’’, Common
KMAZ
RU0008959580
KAMAZ Inc., Common
KUBN
RU0009043426
‘‘UTK’’ PJSC, Common
KUBNP
RU0009091920
‘‘UTK’’ PJSC, Pref
KZBE
RU0009045652
SC ‘‘KUSBASSENERGO’’, Common
KZOS
RU0009089825
OJSC ‘‘Kazanorgsintez’’, Common
123
258
Econ Change Restruct (2009) 42:229–262
Table 13 continued Ticker
ISIN
Name and type
LEKZ
RU000A0D8G13
OJSC Lebedyansky, Common
LKOH
RU0009024277
‘‘LUKOIL’’, Common
LSNG
RU0009034490
JSC ‘‘LENENERGO’’, Common
LSNGP
RU0009092134
JSC ‘‘LENENERGO’’, Pref
MAGN
RU0009084396
OJSC ‘‘MMK’’, Common
MASZ
RU000A0B8366
MASHINOSTROITELNY ZAVOD, Common
MGNT
RU000A0JKQU8
OJSC ‘‘Magnit’’, Common
MGRS
RU000A0ET7W1
Moscow City Electricity Distribution Company, Common
MGTS
RU0009036461
MGTS, Common
MGTSP
RU0009036479
MGTS, Pref
MSNG
RU0008958863
AO MOSENERGO, Common
MSRS
RU000A0ET7Y7
Moscow Integrated Electricity Distribution Company, Common
MSSB
RU000A0ET7Z4
Mosenergosbyt, Common
MTLR
RU000A0DKXV5
Mechel, Common
MTSS
RU0007775219
MTS OJSC, Common
NKNC
RU0009100507
‘‘Nizhhnekamskneftekhim’’ INC., Common
NLMK
RU0009046452
NLMK, Common
NMTP
RU0009084446
PJSC ‘‘NCSP’’, Common
NNSI
RU0009058234
OJSC ‘‘VolgaTelecom’’, Common
NNSIP
RU0009058242
OJSC ‘‘VolgaTelecom’’, Pref
NOMP
RU0009084453
JSC ‘‘Novoship’’, Common
NTRI
RU000A0JP3B3
PJSC ‘‘NUTRINVESTHOLDING’’, Common
NVTK
RU000A0DKVS5
JSC ‘‘NOVATEK’’, Common
OGKB
RU000A0JNG55
JSC ‘‘OGK-2’’, Common
OGKC
RU000A0HMML6
JSC ‘‘WGC-3’’, Common
OGKD
RU000A0JNGA5
JSC ‘‘OGK-4’’, Common
OGKE
RU000A0F5UN3
OJSC ‘‘OGK-5’’, Common
OMZZ
RU0009090542
OMZ, Common
OPIN
RU000A0DJ9B4
JSC ‘‘OPIN’’, Common
PHST
RU000A0JP7F5
JSC ‘‘Pharmstandard’’, Common
PIKK
RU000A0JP7J7
PIK Group, Common
PKBAP
RU0009107692
Baltika Breweries, Pref
PLZL
RU000A0JNAA8
OJSC ‘‘Polyus Gold’’, Common
PMTL
RU000A0JP195
JSC ‘‘Polymetal’’, Common
RASP
RU000A0B90N8
OAO Raspadskaya, Common
RBCI
RU0005707834
RBC Information Systems, Common
RENRTS
RU000A0JNWU0
‘‘RCAM’’ Ltd., FundUnits
RITK
RU0006935947
RITEK, Common
ROSB
RU000A0HHK26
‘‘ROSBANK’’ (OJSC JSCB), Common
ROSN
RU000A0J2Q06
OJSC ‘‘OC’’ ‘‘Rosneft’’, Common
RTKM
RU0008943394
OJSC ‘‘Rostelecom’’, Common
RTKMP
RU0009046700
OJSC ‘‘Rostelecom’’, Pref
123
Econ Change Restruct (2009) 42:229–262
259
Table 13 continued Ticker
ISIN
Name and type
RTMC
RU000A0JP7P4
OJSC ‘‘RTM’’, Common
RU0009029540
Sberbank, Common
SBERP
RU0009029557
Sberbank, Pref
SCON
RU000A0DM8R7
JSC ‘‘The Seventh Continent’’, Common
SIBN
RU0009062467
JSC Gazprom Neft, Common
SILM
RU0005928307
OJSC ‘‘Power machines’’, Common
SILV
RU0009018469
Silvinit, Common
SITR
RU000A0JP187
JSC SITRONICS, Common
SNGS
RU0008926258
Surgutneftegas, Common
SNGSP
RU0009029524
Surgutneftegas, Pref
SNTZ
RU0009152144
Sinarsky Tube Works, Common
SPTL
RU0009046585
OJSC N. W. Telecom, Common
SPTLP
RU0009107338
OJSC N. W. Telecom, Pref
STBK
RU0009100945
‘‘Bank’’ ‘‘Saint-Petersburg’’ OJSC, Common
SVAV
RU0006914488
SOLLERS OJSC, Common
TATN
RU0009033591
Tatneft, Common
TATNP
RU0006944147
Tatneft, Pref
TGKA
RU000A0JNUD0
JSC ‘‘TGC-1’’, Common
TGKB
RU000A0JNGS7
JSC ‘‘TGC-2’’, Common
TGKD
RU000A0JNMZ0
JGC ‘‘TGC-4’’, Common
TGKE
RU000A0JKZF0
JSC ‘‘TGC-5’’, Common
TGKF
RU000A0JNG06
JSC ‘‘TGK-6’’, Common
TGKG
RU000A0HML36
OJSC ‘‘Volga TGC’’, Common
TGKH
RU000A0JNG48
JSC ‘‘SGC TGC-8’’, Common
TGKI
RU000A0JNAC4
OJSC ‘‘TGC-9’’, Common
TGKJ
RU000A0F61T7
JSC ‘‘TGK-10’’, Common
RUSI SBER
OJSC ‘‘IC RUSS-INVEST, Common
TGKM
RU000A0F6SZ9
JSC ‘‘Yenisei TGC (TGC-13)’’, Common
TRMK
RU000A0B6NK6
TMK, Common
TRNFP
RU0009091573
Transneft, Pref
UFMO
RU0009095244
JSC ‘‘Ufa Engine Industrial Association’’, Common
UFNC
RU0007665048
Ufaneftekhim, Common
URKA
RU0007661302
OJSC Uralkali, Common
URSAP
RU000A0JNHW7
URSA Bank, Pref
URSI
RU0009048805
O.J.S.C. ‘‘Uralsvyazinform’’, Common
URSIP
RU0008013438
O.J.S.C. ‘‘Uralsvyazinform’’, Pref
USBN
RU0006929536
URALSIB, Common
VRPH
RU000A0JL475
‘‘ VEROPHARM’’, Common
VSMO
RU0009100291
‘‘VSMPO-AVISMA Corporation’’, Common
VTBR
RU000A0JP5V6
JSC VTB Bank, Common
VZRZ
RU0009084214
V.Bank, Common
WBDF
RU0005344356
WBD Foods, Common
123
260
Econ Change Restruct (2009) 42:229–262
Table 13 continued Ticker
ISIN
Name and type
WTCM
RU0008137070
JSC ‘‘WTC Moscow’’, Common
WTCMP
RU0008137088
JSC ‘‘WTC Moscow’’, Pref
YARE
RU0007796892
Yarenergo, Common
YAREP
RU0007796884
Yarenergo, Pref
This table gives RTS ticker symbols, ISIN numbers, names and types (common or preferred stock) of the stock time series analyzed in this paper
References Amihud Y (2002) Illiquidity and stock returns: cross-section and time-series effects. J Financ Mark 5(1):31–56 Aslan H, Easley D, Hvidkjaer S, O’Hara M (2007) Firm characteristics and informed trading: implications for asset pricing. Working paper, available at SSRN. http://ssrn.com/abstract=971311 Baker M, Stein JC, Wurgler J (2003) When does the market matter? Stock prices and the investment of equity-dependent firms. Q J Econ 118(3):969–1005 Bessembinder H (2003) Trade execution costs and market quality after decimalization. J Financ Quant Anal 38(4):747–77 Black B, Love I, Rachinsky A (2006) Corporate governance and firms’ market values: time series evidence from Russia. Emerg Mark Rev 7:361–379 Brown S, Hillegeist S, Lo K (2004) Conference Calls and information asymmetry. J Account Econ 37(3):343–366 Brown S, Hillegeist S, Lo K (2006) The effect of meeting or missing earnings expectations on information asymmetry. Available at SSRN. http://ssrn.com/abstract=922128 Bushman R, Piotroski J, Smith A (2004) What determines corporate transparency?. J Account Res 42(2):207–52 Campbell J, Grossman S, Wang J (1993) Trading volume and serial correlation in stock returns. Q J Econ 108:905–939 Chen Q, Goldstein I, Jiang W (2007) Price informativeness and investment sensitivity to stock price. Rev Financ Stud 20(3):619–50 Christoffersen P, Slok T (2000) Do asset prices in transition countries contain information about future economic activity? IMF Working Paper, WP/00/103 Coffee J (1999) The future as history: the prospects for global convergence in corporate governance and its implications. Northwest Univ Law Rev 93(3):641–708 Coffee J (2002) Racing towards the top? The impact of cross-listings and stock market competition on international corporate governance. Columbia Law Rev 102(7):1757–1831 Damodaran A (1985) Economic events, information structure, and the return-generating process. J Financ Quant Anal 20(4):423–434 DeFond M, Hung M (2004) Investore protection and corporate governance: evidence from worldwide CEO turnover. J Account Res 42(2):269–312 Demsetz H (1968) The cost of transacting. Q J Econ 82(1):33–53 Dennert J (1993) Price competition between market makers. Rev Econ Stud 60(3):735–751 Doidge C, Karolyi GA, Stulz RM (2004) Why are foreign firms listed in the U.S. worth more? J Financ Econ 71(2):205–238 Domowitz I, Glen J, Madhavan A (1998) International cross-listing and order flow migration: evidence from an emerging market. J Finance 53(6):2001–2027 Durnev A, Morck R, Yeung B, Zarowin P (2003) Does greater firm-specific return variation mean more or less informed stock pricing? J Account Res 41(5):797–836 Durnev A, Morck R, Yeung B (2004) Value-enhancing capital budgeting and firm-specific stock return variation. J Finance 59(1):65–105 Easley D, O’Hara M (1987) Price, trade size and information in securities markets. J Financ Econ 19:69–90 Easley D, O’Hara M (1992) Time and the process of security price adjustment. J Finance 47(2):576–605
123
Econ Change Restruct (2009) 42:229–262
261
Easley D, O’Hara M (2004) Information and the cost of capital. J Finance 59(4):1553–83 Easley D, Kiefer N, O’Hara M (1996a) Cream-skimming or profit sharing? The curious role of purchased order flow. J Finance 51(3):811–833 Easley D, Kiefer N, O’Hara M, Paperman J (1996b) Liquidity, information, and infrequently traded stocks. J Finance 51(4):1405–1436 Easley D, Kiefer N, O’Hara M (1997a) The information content of the trading process. J Empir Finance 4:811–833 Easley D, Kiefer N, O’Hara M (1997b) One day in the life of a very common stock. Rev Financ Stud 10(3):805–835 Easley D, O’Hara M, Papermann J (1998) Financial analysts and information-based trade. J Financ Mark 1:175–201 Easley D, Hvidkjaer S, O’Hara M (2002) Is information risk a determinant of asset returns? J Finance 57:2185–2221 Easley D, Hvidkjaer S, O’Hara M (2005) Factoring information into returns. Working paper, Cornell University Fama E (1970) Efficient capital markets: a review of theory and empirical work. J Finance 25(2):383–417 Foucault T, Gehrig T (2008) Stock price informativeness, cross-listings and investment decisions. J Financ Econ 88(1):146–168 Fuerst O (1998) A theoretical analysis of the investor protection regulations argument for global listing of stocks. Working paper, Yale University, New Haven, CT Glosten LR, Milgrom PR (1985) Bid, ask, and transaction prices in a specialist market with heterogeneously informed traders. J Financ Econ 14(1):71–100 Goncharov I, Zimmermann J (2007) Supply and demand of accounting information: the case of bank financing in Russia. Econ Transit 15(2):257–283 Grishchenko O, Litov L, Mei J (2002) Measuring private information trading in emerging markets. NYU Department of Finance Working Paper; EFA 2002 Berlin meetings presented paper available at SSRN. http://ssrn.com/abstract=301098 Grossman S, Stiglitz J (1980) On the impossibility of informationally efficient markets. Am Econ Rev 70(3):393–408 Hayek FA (1945) The use of knowledge in society. Am Econ Rev 35(4):519–530 Kaplan SN, Zingales L (1997) Do investment-cash flow sensitivities provide useful measures of financing constraints? Q J Econ 112(1):169–215 Kyle A (1985) Continuous auctions and insider trading. Econometrica 53(6):1315–1335 Lee C, Ready M (1991) Inferring trade direction from intraday data. J Finance 46(2):733–746 Lesmond D (2005) Liquidity of emerging markets. J Financ Econ 77(2):411–452 Lesmond D, Ogden J, Trzcinka C (1999) A new estimate of transaction costs. Rev Financ Stud 12(5):1113–1141 Llorente G, Michaely R, Saar G, Wang J (2002) Dynamic volume-return relation of individual stocks. Rev Financ Stud 15(4):1005–1047 Merton RC (1987) Presidential address: a simple model of capital market equilibrium with incomplete information. J Finance 42(3):483–510 Morck R, Yeung B, Yu W (2000) The information content of stock markets: why do emerging markets have synchronous stock price movements? J Financ Econ 58(1–2):215–260 Preobragenskaya G, McGee R (2003a) Corporate governance in a transition economy: a case study of Russia. Working paper, available at SSRN: http://ssrn.com/abstract=459366 Preobragenskaya G, McGee R (2003b) The role of international accounting standards in foreign direct investment: a case study of Russia. Working paper, International Trade and Finance Association, forthcoming. Available at SSRN: http://ssrn.com/abstract=409020 Preobragenskaya G, McGee R (2004) Recent developments in corporate governance in Russia. Working paper, available at SSRN: http://ssrn.com/abstract=480702 Roll R (1984) A simple implicit measure of the effective bid-ask spread in an efficient market. J Finance 39(4):1127–1139 Roll R (1988) R2. J Finance 43(2):541–566 Rozhnova O (2000) The problem of perception of the New Russian Accounting Standards. International Center for Accounting Reform Newsletter. November/December Siegel J (2005) Can foreign firms bond themselves effectively by renting to U.S. securities law? J Financ Econ 75(2):319–359 Stulz RM (1999) Globalization, corporate finance, and the cost of capital. J Appl Corp Finance 12:8–25
123
262
Econ Change Restruct (2009) 42:229–262
van Oppens H (2004) Information asymmetry: a new measure. Working paper, University de Strasbourg, France Vega C (2006) Stock price reaction to public and private information. J Financ Econ 82(1):103–133 Venter J, de Jongh D (2004) Extending the EKOP model to estimate the probability of informed trading. Available at SSRN: http://ssrn.com/abstract=547062 Vorushkin V (2001) IAS benefits for Russian enterprises: managerial issues. International Center for Accounting Reform Newsletter, March/April Vuong Q (1989) Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 57(2):307–333 Wurgler J (2000) Financial markets and the allocation of capital. J Financ Econ 58:187–214 Yan Y, Zhang S (2006) An improved estimation method and empirical properties of PIN. Available at SSRN: http://ssrn.com/abstract=890486
123