The impact of daily return limit and segmented clientele on stock returns in China

International Review of Financial Analysis 19 (2010) 223–236

Contents lists available at ScienceDirect

International Review of Financial Analysis

The impact of daily return limit and segmented clientele on stock returns in China Haim Kedar-Levy a,⁎, Xiaoyan Yu b, Akiko Kamesaka c, Uri Ben-Zion d a

School of Management, Ben Gurion University of the Negev, and Ono Academic College, Israel Graduate School of Economics, Ryukoku University, Japan c School of Business Administration, Aoyama Gakuin University, Japan d Department of Economics, Ben Gurion University of the Negev, Israel b

a r t i c l e

i n f o

Article history: Received 7 January 2010 Received in revised form 11 May 2010 Accepted 14 June 2010 Available online 25 June 2010 JEL classification: G14 G15

a b s t r a c t Mean and variance of daily type A and B stock returns in Shanghai and Shenzhen exchanges are studied before and after these stocks were subject to a ± 10% daily return limit, and when investors' clientele were segmented, vs. merged. We find that imposing the ± 10% return limit significantly reduced the variance of type A stocks, but increased the variance of type B stocks. This puzzle appears to be related to different liquidity effects. Merging clienteles across stock types reduced their risk, increased mean return, and improved efficiency. Returns were generated primarily at the opening (type A) or trading day (type B) before the clienteles merged, but in a mixed format thereafter. © 2010 Elsevier Inc. All rights reserved.

Keywords: China Anomalies Momentum Reversal Clientele

1. Introduction Chinese capital markets have attracted great research attention over the recent years for a number of reasons. First, they expand rapidly and are gradually opening to international investors, whose trading preferences and information may affect local rates of return and their risks. Second, the significant share of Chinese equity in the investment portfolio of global investors motivates a thorough exploration of anomalous return patterns in the Chinese stock exchanges.1 Third, two major types of stocks are traded in Chinese markets, type A and type B: type A stocks were allowed only to mainland investors until December 1st 2002, but also to Qualified Foreign Institutional Investors (QFII) thereafter. Conversely, type B stocks were permitted only to international investors until February 19 2001, but thereafter were allowed to domestic investors as well. Finally, starting on December 16,

⁎ Corresponding author. POB 653, Beer Sheva, 84105, Israel. Tel.: + 972 8 6472569; fax: + 972 8 6477697. E-mail addresses: [email protected] (H. Kedar-Levy), [email protected] (X. Yu), [email protected] (A. Kamesaka), [email protected] (U. Ben-Zion). 1 The literature is vast and rapidly expanding. A non-exhaustive list of contributions may include Mookerjee and Yu (1999), Sun and Tong (2000), Kim and Shin (2000), Chen, Kwok, and Rui (2001), Wang and Firth (2004), Mitchell and Ong (2006), and Sun, Tong, and Yan (2009). 1057-5219/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.irfa.2010.06.002

1996 a ±10% daily return limit was administratively imposed in both exchanges, following a period where no limit applied.2 The impact of this latter administrative rule was not examined in the literature. Hence, this paper is aimed at exploring whether the ±10% rule had a notable impact on the level and dispersion of daily returns, and whether eliminating the segmentation of clientele between type A and B stocks affected average returns and their risks. In particular, we compare patterns in the mean and standard deviation of daily rates of return before the ±10% limit was imposed (sub-period 1) with a subsequent sub-period (sub-period 2), immediately following the rule. In both sub-periods the clientele of investors who were allowed to invest in either stock type A or B was segmented. Hence, this comparison isolates the impact of the ±10% rule. Further, during sub-period 3 both stock types were allowed to both local and foreign investors, hence by comparing sub-periods 2 and 3 we explore whether allowing all investors to trade both stock types eliminated possible differences that resulted from the segmented clientele. Our dataset includes the value-weighted indexes of type A and B stocks in the Shanghai and Shenzhen exchanges over the period May 22, 1992 (October 6, 1992 for type B stocks at Shenzhen) to December 31, 2007. The first sub-period ends on December 13, 1996 for both stock types, before the ±10% restriction was imposed. The second sub-period starts on December 16, 1996. It ends on November 29, 2002 for type A

2

Until May 20, 1992 a 1% limit applied.

224

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

Table 1 Sample sub-periods by stock type and characteristics. Sub-period

Type A stocks

Type B stocks

Sub-period characteristic

1

May 22, 1992–Dec. 13, 1996

SH: May 22, 1992 SZ: Oct. 6, 1992–Dec. 13, 1996

2

Dec. 16, 1996–Nov. 29, 2002

Dec. 16, 1996–Feb. 19, 2001

3

Dec. 2, 2002–Dec. 31, 2007

Feb. 28, 2001–Dec. 31, 2007

No limit apply to daily rates of return; Type A allowed to mainland investors only Type B allowed to foreign investors only Daily return limit of ± 10% applies Type A allowed to mainland investors only Type B allowed to foreign investors only Daily return limit of ± 10% applies Both stock types allowed to all investors

SH: Shanghai Stock Exchange; SZ: Shenzhen Stock Exchange.

stocks, and on February 19, 2001 for type B stocks. Sub-period 3 makes the reminder until the end of our dataset in December 31, 2007. Because we analyze daily returns, we also discuss the anomalous Monday effect,3 and shed light on its evolution as well. Following Rogalski (1984), we split total daily rates of return (i.e., close-to-close) into “opening return”, defined as the rate of return from the prior day closing price to the opening price of the specific day, and to “intra day” return, measured from the opening to the close. In all tests we analyze each sample based on unconditional returns, and returns that are conditional on the prior day's closing sign (e.g., Abraham & Ikenberry, 1994; Tong, 2000). Our tests are aimed at exploring differences between the mean values and the variances of the return measure. Differences in mean rates of returns reveal plausible profitable trading strategies either by holding the relevant stocks directly or by trading an ETF or a derivative asset. Differences in the second moment strongly affect the value of derivative assets, and hence profitable trading strategies can be implemented by trading options written on the relevant index (Low & Zhang, 2005). The following summarizes our major findings: the ±10% limit indeed reduced the variance of type A daily stock returns, but the variance of type B stock returns increased by about 50%, from 1.9% to 2.7%. We provide evidence suggesting that the explanation involves changes in liquidity. A number of effects are associated with allowing all investors to hold both stock types: it significantly reduced the variance of daily returns of type B stocks, but not of type A stocks; it increased average daily returns of type A stocks by a factor greater than 10, and by a factor of 3–4 in type B stocks. Once foreign investors were allowed to hold type A stocks, a significant Monday effect was found. This finding is puzzling because one would naturally relate it to foreign investors' trading patterns, but no significant Monday effect was found in type B stocks during the first two sub-periods. During the second sub-period, the panels conditional on prior returns of type A stocks show that opening returns obtain the sign of the prior closing return. In the panel conditional on prior positive return the trading day return exhibits a reversal in all days of the week, including Monday, yet the panel conditional on negative return shows reversal in all days but a momentum on Mondays only. During the third sub-period Monday's return became significantly positive and highest across all days, generating a significant Monday effect. In this sub-period Monday's reversal changes signs vs. sub-period 2: rather than a positive opening and a negative trading day return, we find a negative opening and a positive trading day return.

We find that absolute conditional returns on type B stocks were primarily generated throughout the trading day rather than the opening while the clienteles were segmented, apparently since foreigners were absent from the opening session. At the same time, returns on type A stocks were generated primarily throughout the opening, and less throughout the day. This structure changed once clienteles were not segmented during the third sub-period, and returns were generated both at the opening and during the day. This change improved the Chinese markets efficiency, as the predictable return generating patterns were no longer valid. Lastly, we explored the homogeneity of variances across days of the week, revealing that over the entire period Monday's variance was significantly higher than all or most other days of the week, in both stock types and in both exchanges. Still, this pattern is primarily due to the first sub-period in type A stocks, but due to the third sub-period in type B stocks, after they were allowed to mainland investors. The paper continues with a description of the data in Section 2, the methodology in Section 3, analysis of type A stock returns and variances in Section 4, and an analysis of type B stocks in Section 5. Section 6 summarizes.

3 Two alternative null hypotheses exist concerning the day-of-the-week effect: one may expect Monday's return to be either three times higher than other days of the week, if returns are generated on a calendar basis, or equal to other days of the week, if returns are generated through trade. The vast body of literature on the day-of-theweek effect started when French (1980) found that neither assumption holds, but rather Monday's mean return was negative on S&P500 stocks between 1953 and 1977. This finding warranted the findings an “anomaly”, which was found thereafter in the US and many other countries (e.g., Jaffe & Westerfield, 1985; Jaffe, Westerfield, & Ma, 1989; Tong, 2000, and others). However, the anomaly appears to fade in time, at least based on US data, with the specific timing being undecided: according to Kamara (1997) it started fading in the early 1960's, Schwert (2003) points at the late 1970's and Connolly (1989) at the mid-1980's. Such anomalies should not persist in efficient markets once investors become aware of their existence (Fama, 1970).

4 It should be noted that investors could trade the indexes reported in this study either by holding directly the stocks, or through ETFs. Five ETFs were available for investors in Chinese markets in 10/2007: Shanghai 50 (since Dec/30/2004), Shanghai 180 (since Apr/13/2006), Shanghai Dividend (since 2006/11/20), Shenzhen100 (since Mar/24/2006) and the SME (Small & Medium Enterprise) Board (since Jun/8/2006). 5 We excluded earlier data for a number of reasons: a) daily rates of return were administratively bounded to small changes prior to this date. Such restrictions limit the informational content of daily rates of return and prohibit an efficient price discovery; b) data of the Shenzhen stock exchange included Saturday returns; and c) the number of securities traded, and their volume, were very small. 6 In 2004 there were 56 type B stocks and 522 type A stocks in the Shenzhen exchange, and 54 vs. 827 in the Shanghai exchange (Shanghai Stock Exchange Fact Book, 2004; Shenzhen Stock Exchange Fact Book, 2004).

2. Data Our dataset includes open and close prices of value-weighted indexes of type A and type B stocks. The type A stocks sample period starts on May 22, 1992. The type B stocks sample starts on May 22, 1992 in the Shanghai exchange, but on October 6 1992 in the Shenzhen exchange. The ending date is December 31, 2007 for both the Shanghai and Shenzhen stock exchanges and both stock types.4,5 In order to control for the administrative regime changes we are interested in this research, we split the sample into three sub-periods, as presented in Table 1. The first sub-period is characterized by the lack of any limit on daily stock returns, in both exchanges and both stock types. During this period the clienteles of both stock types were segmented. During the second subperiod a ±10% restriction was imposed, and the clienteles were still segmented. However, during the third sub-period both local and foreign investors were allowed to hold both stock types. It should be noted that the Shanghai exchange generally hosts bigger stocks, and the Shenzhen exchange mainly trades smaller stocks.6 Moreover, type A stocks are denominated in RMB, while type B stocks in Shanghai are traded in US

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

Dollar, and those in Shenzhen are traded in HK Dollar. The correlation coefficient between the exchange rates USD/RMB and HKD/RMB was about 0.98 over the examined period, suggesting that our conclusions should be robust to exchange rates effects. All panels of type A and B stocks are further analyzed in two conditional subsets: following Rogalski (1984), Bessembinder and Hertzel (1993), Tong (2000), and others, we split the sample to conditional rates of return, conditional on the sign of the prior day return.

225

1986 and 1998, although non-trading time returns were larger for the entire period. The intra-daily momentum and reversal patterns in both stock types and exchanges were not reported thus far in the literature and we have no explanation for them. They may be related to the particular opening mechanism (collective bidding), or to the identity of market participants at the opening, yet this study is beyond the scope of our paper. 4.2. Days of the week effects

3. Methodology We analyze three measures of daily rates of return. First, “total rate of return” is calculated as the natural  logarithm  of closing prices between PClose;t period t and t − 1: c;t−1 Rc;t = Ln . Second, we break the total   PClose;t−1 POpen;t and rate of return into “opening return” c;t−1 Ro;t = Ln   PClose;t−1 PClose;t . Hence, the sum of the two “trading day” return o;t Rc;t = Ln POpen;t latter returns is equal to the first. This distinction facilitates the analysis of the impact of information arrival on returns, both at the opening, reflecting overnight flow of information, and throughout the trading session. We exclude all observations that are not daily returns, i.e., if the price data are not consecutive for any reason other than weekends, rates of return are excluded for not being daily returns. We compare means and variances of the three return measures between sub-periods 1 and 2, exploring the impact that the ±10% limit. Likewise, we compare between sub-periods 2 and 3 to explore whether the change in clientele had any impact on average returns and their variances. These comparisons are conducted by the Welch Two-sample test for the means and by the Levene test for homogeneity of variances. We further test whether the second moment alone differs across days of the week, by applying the Levene test. All tests are conducted for the entire period and the relevant subperiods for type A and type B stocks in both exchanges, as well as for the conditional panels. 4. Type A stocks Table 2 presents summary statistics of type A stocks in both exchanges, for the entire period and the three sub-periods, as well as the conditional and unconditional panels. We devote separate sections for the major findings. 4.1. Momentum\reversal in conditional daily returns One of the interesting findings is that over the entire period (Panel A of Table 2) opening returns conditional on prior day's signal overreact: prior sign continues into the following-day opening return, but is reversed throughout the trading day in most trading days, except for Mondays, where we find a momentum. During sub-period 2 the panels conditional on prior returns show that opening returns obtain the sign of the prior closing return. In the panel conditional on prior positive return the trading day return exhibits a reversal in all days of the week, including Monday, yet the panel conditional on the negative shows reversal in all days but a momentum on Mondays only. In subperiod 3 Monday's reversal flips signs vs. sub-period 2: rather than a positive opening and a negative trading day return, we find during sub-period 3 a negative opening and a positive trading day return. The fact that returns are now generated throughout the trading day, rather than the opening, is probably attributable to foreigners' trades as they obtain local information in delay. This finding corresponds to the findings of Tsutsui (2003), who found that trading time returns in Japan were larger in absolute values in his three sub-periods between

Generally, unconditional close-to-close returns were not different from zero, and in particular, we report (without a table, for space considerations) that over the entire period and the first two subperiods, there was no significant Monday effect in both exchanges. Only during the third sub-period a strong Monday effect was found, where Monday's average return was higher than all others. During the first sub-period (Panel B) daily total returns were generally insignificantly different from zero across days of the week. The only exceptions were Friday's return, being the highest in both exchanges (significant in SH), and Tuesday's return being the lowest (significant in SH). Monday's return was the most volatile, with standard deviations of 4.159% (SZ) and 4.655% (SH). This pattern continued throughout the second sub-period (Panel C), yet Monday returns were ranked lowest, albeit no average return was significantly different from zero. Not surprisingly, the ±10% restriction significantly reduced standard deviations of total returns by about 50% in both exchanges vs. sub-period 1. During the third sub-period (Panel D), when foreign investors were also allowed to invest in type A stocks, a few changes occurred: 1) standard deviations further declined by about 10% (SH) and 15% (SZ) relative to sub-period 2; 2) average daily return increased from 0.012% to 0.099% (SH), and from −0.014% to 0.095% (SZ); 3) Monday's return became significantly positive, and highest across all days, generating a significant Monday effect; 4) while throughout the first two sub-periods conditional returns were generated primarily at the opening (in absolute terms), most of them were generated through trade during the third sub-period. This may be due to foreign investors' higher activity throughout the trading session, and their absence from the opening session, which is organized by a collective bidding procedure. Throughout the sample period Monday's total return variability was highest across days of the week. One explanation for this phenomenon builds on empirical findings whereby prior day's news which firms release after trading hours of the previous day accumulate and affect the opening of the following day (e.g., Galai & Kedar-Levy, 2005). Over the weekend more such news are released and hence the variance of the opening return on Monday is higher than the rest of the days. We could not obtain a dataset of such news releases for the Chinese markets to assess this hypothesis in this study. 4.3. Comparing average returns and variances across sub-periods The findings of the daily return patterns described above uncover compelling evidence that a number of seasonal daily patterns have changed in response to the administrative rules examined here. Table 3 lists the results of a Welch Two-sample test, comparing differences in returns across sub-periods: Panel A calculates first subperiod minus the second, and Panel B calculates second sub-period minus the third. Panel A reveals a few significant differences in the conditional panels, which do not seem to represent a consistent pattern. However, Panel B clearly shows that Monday's highest return emerged only after type A stocks were allowed to foreign investors. This result is significant in both exchanges.

226

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

Table 2 Summary statistics (%, daily rates of return) — type A stocks. Panel A: Entire period 1992.5.22–2007.12.31 Shanghai type A stocks

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Shenzhen type A Stocks

Mon

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

− 0.050 2.988 0.022 1.588 − 0.072 2.149 725

− 0.139 2.332 − 0.009 1.124 − 0.129 1.949 755

0.176 2.587 0.028 1.373 0.148 2.284 760

− 0.055 2.558 0.090 1.470 − 0.145 2.309 758

0.241⁎ 2.274 0.052 1.234 0.189⁎ 2.058 758

0.036 2.559 0.037 1.366 − 0.001 2.158 3756

0.006 2.815 0.032 0.759 − 0.029 2.665 720

− 0.035 2.140 − 0.010 0.718 − 0.027 1.977 753

0.164⁎ 2.138 − 0.004 0.581 0.167⁎ 2.171 758

− 0.070 2.387 − 0.001 0.898 − 0.070 2.177 756

− 0.002 2.693 0.318⁎ 1.790 − 0.319⁎

0.191 2.332 0.272⁎ 1.106 − 0.082 2.124 356

− 0.120 2.211 0.115⁎ 0.676 − 0.233⁎

0.109 1.886 0.075⁎ 0.573 0.034 1.870 415

0.230 2.409 − 0.100⁎ 0.578 0.328⁎

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.325⁎ − 0.183 0.086 St. deviation 2.719 2.341 2.322 Average 0.297⁎ 0.199⁎ 0.210⁎ c,t−1Ro,t St. deviation 1.531 1.022 0.978 Average 0.029 − 0.382⁎ − 0.123 o,tRc,t St. deviation 1.952 2.030 1.846 Observations 371 376 405 c,t − 1Rc,t

2.430 396

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.443⁎ − 0.094 0.279 − 0.113 c,t − 1Rc,t St. deviation 3.207 2.324 2.860 2.404 Average − 0.269⁎ − 0.216⁎ − 0.179⁎ − 0.159⁎ c,t − 1Ro,t St. deviation 1.599 1.182 1.694 0.952 Average − 0.174 0.122 0.459⁎ 0.046 o,tRc,t St. deviation 2.338 1.833 2.667 2.156 Observations 353 379 355 362

0.081 2.492 0.259⁎ 1.327 − 0.178⁎ 2.090 1904

0.460⁎ 2.475 0.231⁎ 0.693 0.224⁎ 2.352 382

1.993 386

Thu

Fri

All day

0.130 2.052 − 0.072⁎ 0.743 0.200⁎ 1.995 753

0.039 2.319 − 0.011 0.747 0.049 2.208 3740

0.020 2.605 0.143⁎ 1.071 − 0.122 2.307 397

0.164 2.037 0.084⁎ 0.678 0.082 1.951 354

0.124⁎ 2.264 0.129⁎ 0.760 − 0.005 2.108 1934

− 0.170 2.120 − 0.160⁎ 0.619 − 0.012 2.024 359

0.100 2.068 − 0.210⁎

− 0.052 2.373 − 0.162⁎ 0.702 0.107⁎

0.284⁎ 2.226 − 0.149⁎

− 0.011 2.627 − 0.193⁎

− 0.508⁎ 3.080 − 0.193⁎

0.054 2.061 − 0.140⁎

1.305 0.432⁎ 1.971 401

1.367 0.183⁎ 2.212 1850

0.768 − 0.316⁎ 2.957 338

0.738 0.191 1.939 367

2.481 343

Fri

All day

Mon

Tue

Wed

− 0.002 3.951 0.093 2.255 − 0.094 3.181 1136

− 0.158 4.159 0.080 0.558 − 0.238 4.084 214

− 0.237 2.927 − 0.057 0.816 − 0.180 2.671 224

0.195 2.982 − 0.012 0.499 0.207 3.051 227

0.115 3.375 0.094 1.214 0.021 2.996 228

0.441 0.376⁎ 2.827 226

− 0.581 3.430 − 0.049 1.193 − 0.533 2.934 97

0.075 2.970 0.062 0.664 0.014 3.101 101

0.319 3.958 0.254 1.657 0.065 3.301 116

0.275 2.874 − 0.005 0.486 0.280 2.865 104

0.120 3.400 0.100⁎ 1.036 0.021 3.157 533

− 0.095 2.642 − 0.070⁎ 0.339 − 0.025 2.657 112

0.324 2.793 − 0.134⁎ 0.391 0.458 2.803 122

− 0.021 3.170 − 0.080⁎ 0.356 0.059 3.160 586

0.770 0.304⁎ 2.030 399

2.310 1806

Fri

All day

0.301 2.824 − 0.075⁎

0.046 3.281 0.006 0.765 0.041 3.157 1119

Panel B: Sub-period 1 1992.5.22–1996.12.13

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Mon

Tue

Wed

Thu

− 0.329 4.655 − 0.030 2.623 − 0.299 3.029 219

− 0.548⁎ 3.511 − 0.089 1.778 − 0.459⁎

0.167 4.049 − 0.025 2.363 0.192 3.468 229

0.056 3.906 0.342⁎ 2.450 − 0.286 3.422 230

2.788 226

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.217 − 0.566 0.156 St. deviation 4.244 3.873 4.091 Average 0.405 0.305 0.543⁎ c,t − 1Ro,t St. deviation 2.573 1.958 1.700 Average − 0.188 − 0.871⁎ − 0.387 o,tRc,t St. deviation 2.726 3.093 3.085 Observations 114 95 101 c,t − 1Rc,t

0.057 4.371 0.875⁎ 3.127 − 0.819⁎ 3.699 111

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.922⁎ − 0.535 0.176 0.055 c,t − 1Rc,t St. deviation 5.017 3.238 4.032 3.435 Average − 0.503⁎ − 0.374⁎ − 0.473⁎ − 0.155 c,t − 1Ro,t St. deviation 2.607 1.583 2.700 1.419 Average − 0.420 − 0.161 0.649⁎ 0.210 o,tRc,t St. deviation 3.336 2.514 3.690 3.074 Observations 105 131 128 119

0.616⁎ 3.481 0.255⁎ 1.952 0.361 3.085 232

0.636 3.699 0.640⁎

0.113 4.076 0.560⁎

1.655 − 0.004 3.374 104

2.302 − 0.447⁎ 3.219 525

0.413 3.532 0.199⁎ 0.616 0.214 3.470 115

0.594⁎ 3.321 − 0.077 2.116 0.671⁎ 2.814 127

− 0.102 3.843 − 0.312⁎ 2.136 0.210 3.121 610

− 0.822⁎ 4.718 − 0.058 0.448 − 0.764⁎ 4.661 99

0.026 2.457 − 0.063⁎ 0.305 0.089 2.429 127

0.291 3.001 − 0.071⁎ 0.300 0.363 3.014 126

Thu

Panel C: Sub-period 2 1996.12.16–2002.11.29

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Mon

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

Thu

Fri

All day

− 0.113 1.945 0.109⁎ 0.948 − 0.222⁎

− 0.063 1.688 0.028 0.856 − 0.090 1.446 285

0.160 1.625 0.085⁎ 0.512 0.076 1.581 286

− 0.072 1.743 − 0.011 0.872 − 0.061 1.717 285

0.143 1.463 0.000 0.872 0.143 1.404 281

0.012 1.700 0.041 0.825 − 0.029 1.554 1408

− 0.130 2.172 0.055 0.988 − 0.186 1.799 271

− 0.077 1.910 0.008 0.846 − 0.085 1.710 285

0.12 1.669 0.022 0.689 0.099 1.754 286

2−0.126 1.877 − 0.047 0.937 − 0.078 1.786 285

0.135 1.711 − 0.053 1.089 0.188⁎

− 0.014 1.874 − 0.003 0.918 − 0.011 1.724 1409

1.589 271

1.548 282

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

227

Table 2 (continued) Panel C: Sub-period 2 1996.12.16–2002.11.29 Shanghai type A stocks Mon

Tue

Shenzhen type A Stocks Wed

Conditional panel: (Closet − 1–Closet) return up at t − 1 Average 0.170 − 0.189 c,t − 1Rc,t St. deviation 1.763 1.471 Average 0.332⁎ 0.197⁎ c,t − 1Ro,t o,tRc,t

St. deviation Average St. deviation Observations

0.802 − 0.162 1.527 144

0.402 − 0.385⁎ 1.452 143

panel 0.028 1.243 0.185⁎ 0.447 − 0.157 1.160 156

Thu

Fri

0.035 1.760 0.138 0.952 − 0.103 1.897 147

0.150 1.448 0.182⁎ 0.937 − 0.032 1.347 131

0.037 1.549 0.207⁎ 0.741 − 0.170⁎ 1.496 721

0.826 − 0.001 1.806 146

0.426 − 0.317⁎ 1.586 146

0.137 1.481 − 0.159⁎ 0.780 0.296⁎ 1.440 150

− 0.014 1.847 − 0.134⁎ 0.871 0.120⁎ 1.601 686

− 0.686⁎ 2.155 − 0.284⁎ 1.055 − 0.401⁎ 1.775 125

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.437⁎ 0.064 0.320 − 0.187 c,t − 1Rc,t St. deviation 2.102 1.879 1.983 1.723 Average − 0.156⁎ − 0.143 − 0.036 − 0.169⁎ c,t − 1Ro,t St. deviation 1.033 1.120 0.558 0.749 Average − 0.280⁎ 0.207 0.355⁎ − 0.017 o,tRc,t St. deviation 1.665 1.383 1.939 1.507 Observations 126 142 130 138

All day

Mon 0.345⁎ 2.078 0.346⁎

Tue

Thu

Fri

0.462 − 0.131 1.207 160

− 0.090 1.915 0.118 0.939 − 0.208 1.930 150

0.202 1.651 0.185⁎ 0.998 0.017 1.375 128

0.069 1.735 0.201⁎ 0.760 − 0.132⁎ 1.604 730

− 0.034 2.178 − 0.192⁎ 1.097 0.158 1.805 139

0.238 2.035 − 0.154⁎ 0.869 0.392⁎ 2.237 126

− 0.165 1.840 − 0.231⁎ 0.902 0.066 1.607 135

0.079 1.762 − 0.252⁎ 1.125 0.330⁎ 1.670 154

− 0.104 2.010 − 0.223⁎ 1.018 0.119 1.837 679

− 0.118 1.622 0.199⁎

Wed 0.030 1.309 0.161⁎

All day

Panel D: Sub-period 3 2002.12.2–2007.12.31

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Mon

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

Thu

Fri

All day

0.283⁎ 1.696 − 0.028 0.587 0.313⁎ 1.612 235

0.152 1.370 0.021 0.357 0.133 1.403 244

0.204⁎ 1.555 0.011 0.577 0.192⁎ 1.435 245

− 0.139 1.564 − 0.029 0.321 − 0.109 1.452 243

− 0.001 1.382 − 0.080⁎ 0.432 0.079 1.353 245

0.099⁎ 1.522 − 0.021 0.468 0.120⁎ 1.457 1212

.312⁎ 1.739 − 0.037 0.592 0.341⁎ 1.644 235

00.198⁎ 1.394 0.013 0.380 0.182⁎ 1.436 244

0.184 1.639 − 0.028 0.510 0.208⁎ 1.551 245

− 0.180 1.706 − 0.036 0.305 − 0.146 1.614 243

− 0.032 1.484 − 0.091⁎ 0.392 0.050 1.443 245

0.095⁎ 1.604 − 0.036⁎ 0.447 0.126⁎ 1.545 1212

− 0.089 1.501 0.062⁎ 0.261 − 0.149 1.385 138

− 0.149 1.333 0.054 0.372 − 0.203 1.262 121

0.103 1.453 0.076⁎ 0.403 0.028 1.352 658

0.644⁎ 1.527 0.124⁎ 0.560 0.505⁎

0.214 1.436 − 0.007 0.602 0.219⁎ 1.266 154

− 0.120 1.580 0.075⁎ 0.203 − 0.191 1.504 131

0.030 1.476 0.053⁎ 0.276 − 0.018 1.443 122

0.188⁎ 1.526 0.075⁎ 0.421 0.112⁎

1.432 121

0.189 1.539 0.139⁎ 0.255 0.054 1.494 143

1.440 671

0.143 1.419 − 0.211⁎ 0.448 0.353⁎ 1.386 124

0.094 1.602 − 0.136⁎ 0.512 0.230⁎ 1.565 554

− 0.040 1.882 − 0.208⁎ 0.580 0.167 1.834 114

0.210 1.166 − 0.166⁎

0.133 1.943 − 0.064⁎

0.451 0.364⁎ 1.335 101

0.297 0.190 1.947 91

− 0.251 1.847 − 0.166⁎ 0.351 − 0.093 1.740 112

− 0.095 1.495 − 0.234⁎ 0.436 0.118 1.447 123

− 0.020 1.690 − 0.173⁎ 0.441 0.143⁎ 1.668 541

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.631⁎ 0.085 0.100 c,t − 1Rc,t St. deviation 1.466 1.516 1.342 ⁎ ⁎ Average 0.142 0.129 0.008 c,t − 1Ro,t St. deviation 0.455 0.223 0.577 Average 0.492⁎ − 0.041 0.092 o,tRc,t St. deviation 1.344 1.498 1.171 Observations 113 138 148

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.038 0.238⁎ 0.362 − 0.205 c,t − 1Rc,t St. deviation 1.832 1.153 1.830 1.648 Average − 0.185⁎ − 0.120⁎ 0.016 − 0.149⁎ c,t − 1Ro,t St. deviation 0.651 0.440 0.580 0.353 Average 0.147 0.358⁎ 0.345 − 0.057 o,tRc,t St. deviation 1.816 1.239 1.758 1.542 Observations 122 106 97 105 ⁎ Asterisks indicate 5% significance.

The comparison across sub-periods is repeated in Table 4 for the variances of total daily returns. The decline of variances in the second sub-period vs. the first is highly significant across all days in the Shanghai exchange, and in the trading day and total returns in the Shenzhen exchange. Yet, the opening return variances in the Shenzhen exchange insignificantly increased in some of the days or declined in others. The All days opening return significantly increased in the second sub-period vs. the first, in spite of the ±10% return limit. We explore the reasons for this increase in Section 5.2. 4.4. Comparing the variance of returns across days of the week Last, we test whether the second moment of daily returns differs across days of the week. Predictable seasonal patterns in the variance are important for investors for a number of reasons, among them volatility trading (e.g., Low & Zhang, 2005), and risk assessment. Using Levene's

test for homogeneity of variances we compare all pairs of days of the week, and indicate the resulting p-value in a matrix format. Significant differences (p-value of 5% or lower) are printed in boldface, allowing the observer to discern quickly which pairs of days of the week differ in their second moment. For space consideration we present tables for unconditional returns only (conditional panels are available upon request). The tests are conducted for both exchanges, for total, opening, and trading day returns, as well as over the entire period and sub-periods. One can readily notice from Panel A of Table 5 (entire sample) that Monday's standard deviation of unconditional returns is significantly different (higher) from all other days of the week. This difference stems from the opening return in Shanghai, but from both the opening and trading day return in Shenzhen. The conditional panels (not presented) generally yield a similar conclusion. Panel B of Table 5 shows that during the first sub-period Monday's highest variance stems primarily from opening returns in Shanghai,

228

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

Table 3 Welch two sample t-test — type A stocks. Panel A: sub-period 1 minus sub-period 2 Shanghai type A stocks Mon

Shenzhen type A stocks

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

− 0.485 − 0.116 − 0.369

0.007 − 0.110 0.116

0.128 0.353⁎ − 0.225

0.473 0.255 0.218

− 0.014 0.051 − 0.065

− 0.028 0.025 − 0.053

− 0.160 − 0.065 − 0.095

0.073 − 0.035 0.108

0.241 0.142 0.099

0.167 − 0.021 0.188

0.061 0.009 0.052

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel 0.048 − 0.377 0.128 0.021 0.073 0.109 0.358⁎ 0.737⁎ c,t − 1Ro,t − 0.025 − 0.486 − 0.230 − 0.716 o,tRc,t

0.486 0.458⁎ 0.028

0.076 0.353⁎ − 0.277

0.068 − 0.147 0.215

− 0.464 − 0.248 − 0.216

0.045 − 0.099 0.145

0.408 0.136 0.273

0.073 − 0.190 0.263

0.051 − 0.101 0.153

Conditional panel: (Closet − 1–Closet) return down at t − 1 − 0.486 − 0.600 − 0.144 − 0.346 − 0.232 − 0.438 c,t − 1Ro,t − 0.139 − 0.368 0.294 o,tRc,t

0.457 0.082 0.375

− 0.089 − 0.179 0.090

− 0.136 0.226⁎ − 0.362

0.060 0.129 − 0.069

0.053 0.082 − 0.030

0.070 0.160 − 0.090

0.245 0.118 0.127

Fri

All day

Mon

Tue

Wed

Thu

Unconditional panel − 0.216 c,t − 1Rc,t − 0.139 c,t − 1Ro,t − 0.077 o,tRc,t

c,t − 1Rc,t

c,t − 1Rc,t

panel 0.242 0.014 0.227

Thu

Fri

All day

0.083 0.143⁎ − 0.060

Panel B: Sub-period 2 minus sub-period 3 Mon Unconditional panel − 0.397⁎ c,t − 1Rc,t 0.136⁎ c,t − 1Ro,t − 0.534⁎ o,tRc,t

Tue

Wed

− 0.214 0.007 − 0.223

− 0.043 0.073 − 0.117

Thu

Fri

All day

0.067 0.018 0.048

0.144 0.080 0.064

− 0.087 0.062⁎ − 0.150⁎

− 0.443⁎ 0.092 − 0.527⁎

− 0.275 − 0.004 − 0.268

− 0.062 0.050 − 0.109

0.055 − 0.011 0.068

0.167 0.038 0.138

− 0.109 0.033 − 0.137⁎

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel − 0.461⁎ − 0.274 − 0.073 0.124 0.190⁎ 0.068 0.177⁎ 0.076 c,t − 1Ro,t − 0.654⁎ − 0.345 − 0.249 0.047 o,tRc,t

0.299 0.128 0.171

− 0.066 0.130⁎ − 0.198⁎

− 0.299 0.222⁎ − 0.506⁎

− 0.307⁎ 0.060 − 0.371⁎

− 0.184 0.168⁎ − 0.350⁎

0.030 0.043 − 0.017

0.171 0.132 0.035

− 0.119 0.126⁎ − 0.244⁎

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel − 0.398 − 0.174 − 0.042 0.019 0.029 − 0.023 − 0.052 − 0.021 c,t − 1Ro,t − 0.427 − 0.151 0.010 0.039 o,tRc,t

− 0.006 0.052 − 0.057

− 0.108 0.002 − 0.110

− 0.646⁎ − 0.076 − 0.569⁎

− 0.244 − 0.026 − 0.206

0.106 − 0.090 0.202

0.086 − 0.065 0.159

0.174 − 0.017 0.212

c,t − 1Rc,t

c,t − 1Rc,t

− 0.083 − 0.050 − 0.023

⁎ Asterisks indicate 5% significance.

Friday in Shanghai, but from most other days in Shenzhen. The last sub-period, analyzed in Panel D, shows convergence of both exchanges once foreigners were allowed to hold type A stocks: in

but it stems during trade (primarily conditional on negative returns; not presented) in Shenzhen. Panel C reveals that during the second sub-period Monday's variance was higher only from Tuesday and Table 4 Levene's test for Homogeneity of variances between sub-periods — type A stocks. Panel A: Comparing sub-periods 1 and 2 Unconditional return Mon

Tue

Wed

Thu

Fri

All days

Shanghai type A stocks c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

Shenzhen type A stocks c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

0.00 0.20 0.00

0.00 0.61 0.00

0.00 0.32 0.00

0.00 0.83 0.00

0.00 0.06 0.00

0.00 0.01 0.00

Panel B: Comparing sub-periods 2 and 3 Unconditional return Mon

Tue

Wed

Thu

Fri

All days

Shanghai type A stocks c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

0.90 0.31 0.34

0.15 0.06 0.60

0.78 0.82 0.79

0.99 0.07 0.76

0.80 0.39 0.83

0.39 0.01 0.98

Shenzhen type A stocks c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

0.21 0.21 0.76

0.04 0.02 0.16

0.92 0.25 0.67

0.98 0.03 0.98

0.90 0.04 0.83

0.11 0.00 0.40

P-value b 5% indicates significant dissimilarity of return variance.

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

229

Table 5 Levene's test for homogeneity of variances between pairs of days — Type A stocks. Panel A: Entire period Unconditional returns 1992.5.22–2007.12.31 Shanghai type A stocks

Shenzhen type A stocks

P-value

P-value Tue

Wed

Thu

Fri

Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.000 X

0.001 0.417 X

0.004 0.201 0.671 X

0.000 0.732 0.254 0.105

Mon Tue Wed Thu

0.000 X

0.000 0.138 X

0.004 0.029 0.399 X

0.000 0.429 0.455 0.124

Mon Tue Wed Thu

0.000 X

0.000 0.988 X

0.001 0.768 0.778 X

0.001 0.485 0.525 0.756

Mon Tue Wed Thu

0.007 Xx

0.000 0.112 X

0.007 0.704 0.350 X

0.034 0.585 0.031 0.395

Tue 0.012 X

Wed 0.379 0.155 X

Thu 0.862 0.030 0.501 X

Fri 0.056 0.597 0.364 0.104

Mon Tue Wed Thu

Tue 0.000 X

Wed 0.006 0.043 X

Thu 0.026 0.007 0.533 X

Fri 0.000 0.256 0.336 0.103

Tue

Wed

Thu

Fri

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

Mon Tue Wed Thu Panel B: Sub-period 1 Unconditional returns 1992.5.22–1996.12.13

Shanghai type A stocks

Shenzhen type A stocks

P-value

P-value Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.004 X

0.015 0.796 X

0.021 0.611 0.828 X

0.002 0.769 0.601 0.432

Mon Tue Wed Thu

0.000 X

0.005 0.208 X

0.020 0.138 0.717 X

0.001 0.364 0.681 0.461

Mon Tue Wed Thu

0.001 X

0.005 0.846 X

0.011 0.662 0.829 X

0.003 0.648 0.849 0.957

Mon Tue Wed Thu

0.462 X

0.003 0.145 X

0.955 0.671 0.131 X

0.017 0.376 0.356 0.272

Mon Tue Wed Thu

0.363 X

0.942 0.382 X

0.630 0.184 0.711 X

0.669 0.660 0.649 0.384

Mon Tue Wed Thu

0.000 X

0.019 0.103 X

0.021 0.076 0.924 X

0.005 0.228 0.624 0.546

Tue

Wed

Thu

Fri

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

Panel C: Sub-period 2 Unconditional returns 1996.12.16–2002.11.29 Shanghai type A stocks

Shenzhen type A stocks

P-value

P-value Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.040 X

0.108 0.552 X

0.202 0.401 0.772 X

0.007 0.619 0.234 0.156

Mon Tue Wed Thu

0.043 X

0.015 0.842 X

0.090 0.682 0.507 X

0.006 0.527 0.619 0.274

Mon Tue Wed Thu

0.181 X

0.007 0.250 X

0.136 0.859 0.362 X

0.197 0.974 0.241 0.836

Mon Tue Wed Thu

0.179 X

0.011 0.235 X

0.145 0.837 0.379 X

0.562 0.528 0.095 0.435

Mon Tue Wed Thu

0.111 X

0.648 0.263 X

0.938 0.115 0.611 X

0.123 0.921 0.292 0.129

Mon Tue Wed Thu

0.165 X

0.397 0.587 X

0.682 0.326 0.661 X

0.055 0.665 0.311 0.139

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

(continued on next page)

230

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

Table 5 (continued) Panel D: Sub-period 3 Unconditional returns 2002.12.2–2007.12.31 Shanghai type A stocks

Shenzhen type A stocks

P-value

P-value Tue

Wed

Thu

Fri

Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.000 X

0.058 0.058 X

0.172 0.011 0.563 X

0.002 0.402 0.252 0.074

Mon Tue Wed Thu

0.000 X

0.175 0.036 X

0.588 0.003 0.414 X

0.043 0.094 0.572 0.150

Mon Tue Wed Thu

0.000 X

0.027 0.447 X

0.000 0.732 0.313 X

0.045 0.059 0.527 0.022

Mon Tue Wed Thu

0.000 X

0.001 0.849 X

0.000 0.431 0.691 X

0.006 0.270 0.284 0.042

Mon Tue Wed Thu

0.003 X

0.082 0.166 X

0.211 0.056 0.590 X

0.018 0.400 0.539 0.239

Mon Tue Wed Thu

0.007 X

0.287 0.096 X

0.881 0.009 0.350 X

0.123 0.183 0.667 0.157

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

P-value b 5% indicates significant dissimilarity of return variance.

both markets Monday's variance was higher only from Tuesday and Friday, and Tuesday's variance was different (lower) than Wednesday and Thursday variances (marginal significance of 0.058 in SH). 5. Type B stocks Type B stocks were allowed to foreign investors throughout the entire sample period, but only during the third sub-period they were allowed to local investors as well. The summary statistics for type B stocks, presented in Table 6, reveal a number of similarities and differences between type A and type B stocks. 5.1. Momentum\reversal in conditional daily returns Opening returns for the entire period (Panel A) conditional on prior positive (negative) returns are positive (negative) as they were in the type A stocks panel, but they differ in that these returns do not reverse throughout the trading day, but rather their momentum further intensifies by a factor of 2–3. Hence, while type A stocks overreact at the opening, type B stocks under-react. During the first subperiod, conditional total returns were significantly different from zero, in the direction of the prior day sign. Most of that return was generated throughout the trading day, rather than the opening. This attribute is probably because foreign investors are active throughout the trading day in China, and less so during the opening. Yet, during the second sub-period, the opening return conditional on prior positive return was significantly negative in Shanghai, but significantly positive in Shenzhen (and at higher absolute values, −0.008% in SH vs. 0.153% in SZ). The negative opening return was reversed throughout the trading day in Shanghai, but had a momentum in Shenzhen. This pattern is further puzzling, as the returns conditional on negative prior returns are “well behaved” — the opening and the trading day returns are all negative, i.e., representing an intra-daily momentum. Moreover, it is unclear why conditional on positive closing, type B shares open negative, and reverse, while at the same time type A stocks open positive, and reverse. During the third sub-period (Panel D of Table 6) almost all total returns are insignificantly different from zero, while most conditional opening returns are significant, obtaining the sign of the prior closing. There was a significant momentum on Mondays in three of those conditional opening returns. In most of the other days of the week momentum generally persisted, albeit mostly insignificant.

Because there are no clientele effects during the third sub-period, it is interesting to compare those patterns with type A stocks. Type A stocks had a significant opening return in the direction of the prior closing, just like type B stocks. Like type B stocks, they had a momentum in the panels conditional on prior positive returns and a significant momentum on Mondays. However, unlike type B stocks, they had a reversal in the panels conditional on negative prior (and opening) return, in both exchanges, and this reversal was significant on Tuesdays. It turns out that conditional on a negative prior return, type A and B stocks had a negative opening return, but it reversed on Tuesdays for type A stocks and had a momentum on Mondays for type B stocks. These findings, whereby the momentum\reversal patterns are mixed with the Monday\Tuesday effects are puzzling and call for additional future research. Taking a bird's-eye view on the return generating process in the Chinese markets an interesting and simple structure emerges:7 while investors' clienteles were segmented (first two sub-periods), absolute conditional returns on type A stocks were primarily generated at the opening, rather than during the trading session. At the same time, type B stock returns were generated primarily during the trading session, rather than the opening. The simple explanation for that is the collective bidding procedure at the opening, where local investors participate (type A, but not type B stocks), while during the trading session foreign investors are most active, and generate returns in type B stocks. During the third sub-period, when clienteles are not segmented any longer, absolute returns on type A and B stocks are generated in a mixed format, some during the opening and some during trade. 5.2. Days of the week effects Unconditional returns for the entire period (Panel A of Table 6) show that average daily returns are at level with type A stocks, ranging 0.030%–0.039%. Standard deviations are about 10% lower than type A stocks, ranging 2.21%–2.53%, where variability in the opening is about 1/3 the level of variability throughout the trading day. We note that there was no Monday or any other day effect over the entire period.

7

We thank an anonymous referee for this observation.

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

231

Table 6 Summary statistics (%, daily returns) — type B stocks. Panel A: Entire period Shanghai type B stocks (1992.5.22–2007.12.31)

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Shenzhen type B stocks(1992.10.6–2007.12.31)

Mon

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

Thu

Fri

All day

0.034 2.535 0.009 0.736 0.024 2.400 724

− 0.065 2.135 0.020 0.683 − 0.084 2.104 753

0.078 2.167 − 0.033 0.794 0.111 2.079 759

− 0.044 2.315 − 0.019 0.592 − 0.025 2.206 757

0.144 2.124 − 0.002 0.752 0.146⁎

0.030 2.258 − 0.005 0.714 0.035 2.169 3749

0.154 2.512 0.021 0.753 0.133 2.303 698

− 0.097 2.118 0.106⁎

0.024 2.094 0.020 0.757 0.004 2.068 737

− 0.013 2.165 0.010 0.525 − 0.023 2.173 736

0.131 2.145 0.006 0.504 0.125 2.106 731

0.039 2.210 0.033⁎ 0.699 0.006 2.135 3633

0.100 2.054 0.120⁎ 0.865 − 0.020 1.948 351

0.169 2.256 0.084⁎ 0.664 0.085 2.349 348

0.515⁎ 2.377 0.099⁎

0.312⁎ 2.273 0.128⁎

0.628 0.416⁎ 2.356 349

0.764 0.184⁎ 2.208 1732

− 0.176 2.072 − 0.056⁎ 0.346 − 0.119 2.002 387

− 0.220⁎ 1.842 − 0.079⁎ 0.334 − 0.142 1.811 382

− 0.224⁎ 2.102 − 0.056⁎ 0.620 − 0.168⁎ 2.040 1891

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.709⁎ 0.081 0.208⁎ c,t − 1Rc,t St. deviation 2.466 2.189 2.018 Average 0.089⁎ 0.133⁎ 0.006 c,t − 1Ro,t St. deviation 0.930 0.707 0.764 Average 0.620⁎ − 0.052 0.202⁎ o,tRc,t St. deviation 2.290 2.206 1.985 Observations 356 341 374 Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.625⁎ − 0.186 − 0.072 c,t − 1Rc,t St. deviation 2.435 2.085 2.250 Average − 0.070⁎ − 0.075⁎ − 0.076 c,t − 1Ro,t St. deviation 0.468 0.648 0.817 Average − 0.556⁎ − 0.111 0.004 o,tRc,t St. deviation 2.373 2.018 2.131 Observations 366 412 384

0.273⁎ 2.471 0.081⁎

2.042 756

0.882 − 0.203⁎ 2.008 731

0.524⁎ 2.143 0.145⁎

0.359⁎ 2.272 0.090⁎

0.728⁎ 2.538 0.119⁎

0.389 0.191 2.357 357

0.896 0.378⁎ 1.968 353

0.763 0.270⁎ 2.174 1781

0.836 0.609⁎ 2.260 341

0.797 − 0.164 2.013 343

− 0.327⁎ 2.129 − 0.109⁎ 0.715 − 0.217⁎

− 0.275⁎ 2.195 − 0.092⁎ 0.655 − 0.182⁎

− 0.421⁎ 2.318 − 0.079⁎ 0.640 − 0.342⁎

− 0.247⁎ 2.176 0.010 0.943 − 0.257⁎

2.045 400

− 0.188 2.053 − 0.130⁎ 0.568 − 0.058 2.086 403

2.136 1965

2.228 356

2.004 382

− 0.070 2.083 − 0.074⁎ 0.629 0.005 2.136 384

0.055 2.041 0.219⁎

Panel B: Sub-period 1 Mon

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

Fri

Thu

All day

0.061 1.940 − 0.004 0.674 0.064 1.844 219

− 0.038 1.929 − 0.039 0.696 0.001 1.758 226

− 0.214 1.742 − 0.070 1.025 − 0.144 1.649 229

− 0.195 1.966 − 0.045 0.865 − 0.150 1.747 230

0.182 2.128 0.040 1.090 0.142 1.842 231

− 0.042 1.948 − 0.024 0.888 − 0.018 1.770 1135

0.222 2.239 − 0.028 0.582 0.250 2.020 193

− 0.004 1.559 0.093 1.035 − 0.097 1.374 204

− 0.202 1.774 − 0.020 0.514 − 0.183 1.500 207

− 0.024 1.543 − 0.028 0.295 0.004 1.510 210

0.072 2.258 0.000 0.211 0.072 2.164 206

0.010 1.897 0.004 0.599 0.006 1.741 1020

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.565⁎ 0.544⁎ 0.184 St. deviation 2.214 2.182 1.646 Average 0.069 0.035 − 0.121 c,t − 1Ro,t St. deviation 0.813 0.680 0.977 Average 0.496⁎ 0.509⁎ 0.305 o,tRc,t St. deviation 2.079 2.077 1.668 Observations 105 98 110

0.279 1.771 0.034 0.476 0.245 1.651 95

0.079 1.523 0.051⁎ 0.234 0.028 1.479 87

− 0.149 2.256 − 0.022 0.711 − 0.127 1.781 95

0.216 1.845 − 0.007 0.271 0.222 1.835 84

− 0.110 1.548 0.144 1.380 − 0.254⁎

− 0.249⁎ 1.241 − 0.017 0.250 − 0.232⁎

− 0.446⁎ 1.806 − 0.026 0.218 − 0.420⁎

− 0.222⁎ 1.449 0.006 0.658 − 0.228⁎

1.224 111

1.224 111

− 0.181 1.292 − 0.042 0.311 − 0.139 1.240 125

1.706 119

1.339 576

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

c,t − 1Rc,t

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.412⁎ − 0.483⁎ − 0.581⁎ − 0.529⁎ c,t − 1Rc,t St. deviation 1.518 1.580 1.754 2.033 Average − 0.071 − 0.096 − 0.023 − 0.101 c,t − 1Ro,t St. deviation 0.512 0.705 1.069 1.054 Average − 0.341⁎ − 0.387⁎ − 0.558⁎ − 0.428⁎ o,tRc,t St. deviation 1.504 1.352 1.525 1.765 Observations 113 128 119 135

0.751⁎ 2.335 0.268 1.414 0.483⁎

0.460⁎ 2.045 0.053 0.934 0.407⁎

0.665⁎ 3.045 − 0.035 0.801 0.700⁎

1.863 98

1.872 506

2.692 83

− 0.238 1.862 − 0.128⁎

− 0.447⁎ 1.768 − 0.086⁎

0.730 − 0.110 1.793 133

0.845 − 0.361⁎ 1.605 628

− 0.112 1.261 − 0.022 0.338 − 0.090 1.212 110

0.780⁎ 2.608 0.036 0.198 0.744⁎ 2.525 87

0.307⁎ 2.330 0.005 0.514 0.303⁎ 2.124 436

Panel C: Sub-period 2 1996.12.16–2001.2.19 Shanghai type B stocks

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Shenzhen type B stocks

Mon

Tue

Wed

Thu

Fri

0.036 2.974 − 0.009 0.087 0.044 2.987 188

− 0.452⁎ 2.516 − 0.007⁎

0.209 2.812 0.000 0.086 0.209 2.805 201

0.077 2.756 − 0.002 0.017 0.079 2.755 199

0.248 2.353 − 0.011⁎

0.022 2.696 − 0.006⁎

0.050 0.259 2.348 195

0.061 0.028 2.696 984

0.030 − 0.445⁎ 2.515 201

All day

Mon

Tue

Wed

Thu

Fri

0.064 2.894 0.007 0.705 0.056 2.672 188

− 0.656⁎ 2.567 0.085 0.843 − 0.741⁎ 2.538 201

0.084 2.534 − 0.002 0.640 0.087 2.610 201

0.023 2.694 0.048 0.835 − 0.025 2.829 198

0.318 2.298 0.030 0.467 0.288 2.238 195

All day − 0.037 2.618 0.034 0.712 − 0.071 2.604 983

(continued on next page)

232

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

Table 6 (continued) Panel C: Sub-period 2 1996.12.16–2001.2.19 Shanghai type B stocks Mon

Shenzhen type B stocks Thu

Fri

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.718⁎ − 0.332 0.425 c,t − 1Rc,t St. deviation 2.684 2.505 2.753 Average − 0.016⁎ − 0.007⁎ − 0.010 c,t − 1Ro,t St. deviation 0.095 0.030 0.081 Average 0.734⁎ − 0.325 0.435 o,tRc,t St. deviation 2.711 2.504 2.759 Observations 94 95 78

Tue

Wed

0.439 2.998 − 0.002 0.023 0.441 2.996 95

0.615⁎ 2.113 − 0.006⁎ 0.023 0.621⁎ 2.109 97

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.658⁎ − 0.560⁎ 0.072 c,t − 1Rc,t St. deviation 3.119 2.533 2.851 Average − 0.009 − 0.008⁎ 0.007 c,t − 1Ro,t St. deviation 0.046 0.030 0.088 Average − 0.650⁎ − 0.553⁎ 0.065 o,tRc,t St. deviation 3.119 2.532 2.835 Observations 93 106 123

− 0.254 2.484 − 0.002 0.010 − 0.252 2.484 104

− 0.115 2.527 − 0.017⁎ 0.067 − 0.098 2.524 98

All day

Mon

Tue

0.372⁎ 2.636 − 0.008⁎

0.666⁎ 2.458 0.079⁎

− 0.123 2.269 0.314⁎

0.058 0.379⁎ 2.641 459

0.263 0.587⁎ 2.319 95

1.048 − 0.438 2.474 87

− 0.285⁎ 2.715 − 0.005⁎ 0.057 − 0.280⁎

− 0.652⁎ 3.043 − 0.091 0.938 − 0.561 2.826 92

− 1.064⁎ 2.713 − 0.091 0.594 − 0.973⁎

2.710 524

Wed

Thu

Fri

All day

0.840⁎ 2.638 0.125⁎

0.415⁎ 2.480 0.153⁎

0.576 0.715⁎ 2.573 89

0.813 0.261⁎ 2.554 441

0.266 2.242 0.117 0.647 0.149 2.044 77

0.376 2.660 0.136 1.169 0.240 3.067 93

2.572 114

− 0.029 2.702 − 0.077 0.627 0.048 2.914 124

− 0.290 2.698 − 0.031 0.313 − 0.259 2.593 105

− 0.121 1.872 − 0.050 0.331 − 0.071 1.849 106

− 0.422⁎ 2.645 − 0.068⁎ 0.594 − 0.354⁎ 2.601 541

Panel D: Sub-period 3 2001.2.28–2007.12.31

Unconditional panel Average c,t − 1Rc,t St. deviation Average c,t − 1Ro,t St. deviation Average o,tRc,t St. deviation Observations

Mon

Tue

Wed

Thu

Fri

All day

Mon

Tue

Wed

Thu

Fri

All day

0.014 2.621 0.029 0.959 − 0.016 2.357 317

0.155 1.984 0.077 0.859 0.078 2.024 326

0.202 1.958 − 0.028 0.848 0.230⁎

− 0.012 2.248 − 0.012 0.535 0.000 2.121 328

0.057 1.977 − 0.025 0.681 0.082 1.983 330

0.084 2.168 0.008 0.789 0.076 2.062 1630

0.165 2.432 0.058 0.864 0.107 2.231 317

0.190 2.055 0.128⁎ 0.799 0.062 1.913 326

0.129 1.977 0.058 0.931 0.071 1.999 329

− 0.027 2.155 0.012 0.378 − 0.039 2.081 328

0.057 1.971 − 0.004 0.638 0.061 1.988 330

0.102 2.121 0.050⁎ 0.748 0.052 2.042 1630

0.034 2.204 0.101⁎ 0.345 − 0.067 2.118 171

− 0.094 2.106 − 0.085⁎ 0.389 − 0.009 2.047 157

1.798 329

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel Average 0.799⁎ 0.040 0.131 c,t − 1Rc,t St. deviation 2.501 1.918 1.852 Average 0.165 0.288⁎ 0.088 c,t − 1Ro,t St. deviation 1.228 0.899 0.771 Average 0.634⁎ − 0.249 0.043 o,tRc,t St. deviation 2.156 2.025 1.757 Observations 157 148 186

0.174 2.490 0.156⁎ 0.431 0.019 2.285 167

0.735 0.165 1.932 158

0.846 0.122 2.047 816

1.039 0.577⁎ 1.982 163

0.827 − 0.123 1.983 169

0.161 1.848 0.197⁎ 1.005 − 0.036 1.996 179

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel Average − 0.757⁎ 0.251 0.229 c,t − 1Rc,t St. deviation 2.512 2.038 1.952 Average − 0.104⁎ − 0.099 − 0.192⁎ c,t − 1Ro,t St. deviation 0.559 0.784 0.910 Average − 0.653⁎ 0.350⁎ 0.421⁎ c,t − 1Ro,t St. deviation 2.378 1.989 1.730 Observations 160 178 142

− 0.204 1.955 − 0.185⁎ 0.576 − 0.019 1.942 161

− 0.191 1.898 − 0.197⁎ 0.577 0.006 2.032 172

− 0.135 2.109 − 0.154⁎ 0.690 0.019 2.060 813

− 0.503⁎ 2.392 − 0.112⁎

0.249 1.949 − 0.012 0.745 0.261 1.821 157

0.029 1.995 − 0.115 0.805 0.145 1.890 149

0.327⁎ 2.031 0.162⁎

0.290⁎ 2.183 0.167⁎

0.796⁎ 2.303 0.220⁎

0.584 − 0.391⁎ 2.374 154

0.135 2.152 0.258⁎

0.215 2.071 0.118⁎

0.262⁎ 2.128 0.178⁎

0.778 0.097 2.110 173

0.836 0.084 2.050 855

− 0.116 1.845 − 0.138⁎

− 0.087 2.075 − 0.092⁎ 0.606 0.006 2.013 774

0.395 0.022 1.849 157

⁎ Asterisks indicate 5% significance.

Table 7 Percentage of zero returns observations. Type A stocks

Type B stocks

SH

SZ

SH

SZ

0.0% 2.5% 0.1%

0.7% 45.7% 0.4%

0.9% 68.8% 1.6%

4.8% 78.1% 5.0%

0.0% 0.0% 0.0%

0.1% 0.5% 0.1%

0.0% 93.4% 0.0%

2.6% 20.2% 3.3%

0.0% 0.0% 0.0%

0.1% 0.4% 0.2%

0.0% 0.7% 0.0%

0.1% 0.2% 0.0%

Sub-period 1 c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

Sub-period 2 c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

Sub-period 3 c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

During the first sub-period (Panel B) daily unconditional returns in both exchanges were lower than those of type A stocks by about 40 basis points, and their standard deviations about half (±1.9%). No significant daily effect was found. Yet, once the ±10% rule was imposed (Panel C), unconditional standard deviations increased to the range of 2.6%–2.7%. This increase is puzzling since the implementation of the new rule is expected to reduce daily variability, as was the case for type A stocks. Moreover, standard deviations in the opening declined by more than 90% in Shanghai (from 0.888% to 0.061%) while increasing by about 15% in Shenzhen (to 0.712%). Trading day standard deviations increased by about 50% (from about 1.7% to 2.6%) in both exchanges. A possible explanation for this puzzle builds on different levels of liquidity due to the stale price problem. This problem reflects the empirical finding where no trade occurs at the opening, and hence the opening price is set equal to the last closing price, generating zero

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

233

Table 8 Welch two sample t-test — Type B stocks. Panel A: Sub-period 1 minus sub-period 2 Shanghai tape-B stocks Mon

Shenzhen tape-B stocks

Tue

Wed

Thu

Fri

All day

Mon

0.414 − 0.032 0.446⁎

− 0.423 − 0.071 − 0.352

− 0.272 − 0.043 − 0.229

− 0.067 0.051 − 0.118

− 0.063 − 0.018 − 0.045

0.159 − 0.035 0.194

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel − 0.154 0.876⁎ − 0.241 − 0.160 c,t − 1Rc,t 0.084 0.042 − 0.111 0.037 c,t − 1Ro,t ⁎ − 0.238 0.833 − 0.130 − 0.196 o,tRc,t

0.136 0.273 − 0.137

0.089 0.061 0.028

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel 0.246 0.077 − 0.653⁎ − 0.275 c,t − 1Rc,t − 0.063 − 0.089 − 0.030 − 0.099 c,t − 1Ro,t 0.309 0.166 − 0.624⁎ − 0.176 o,tRc,t

− 0.123 − 0.111 − 0.012

− 0.162 − 0.081⁎ − 0.081

Fri

All day

Mon

0.191 0.014 0.178

− 0.062 − 0.014 − 0.048

Conditional panel: (Closet − 1–Closet) return up at t − 1 panel − 0.080 − 0.371 0.294 0.264 c,t − 1Rc,t − 0.180 − 0.295⁎ − 0.098 − 0.158⁎ c,t − 1Ro,t 0.100 − 0.076 0.393 0.422 o,tRc,t

0.288 − 0.168⁎ 0.456

Conditional panel: (Closet − 1–Closet) return down at t − 1 panel 0.098 − 0.812⁎ − 0.157 − 0.049 c,t − 1Rc,t 0.095⁎ 0.091 0.199⁎ 0.183⁎ c,t − 1Ro,t 0.003 − 0.903⁎ − 0.356 − 0.233 o,tRc,t

0.076 0.180⁎ − 0.104

Unconditional c,t − 1Rc,t c,t − 1Ro,t o,tRc,t

panel 0.025 0.005 0.020

Tue

Wed

Thu

Fri

All day

0.653⁎ 0.008 0.645⁎

− 0.287 − 0.017 − 0.270

− 0.047 − 0.075 0.028

− 0.246 − 0.030 − 0.216

0.047 − 0.030 0.077

− 0.001 − 0.114 0.113

0.202 − 0.264⁎ 0.466

− 0.415 − 0.139 − 0.276

− 0.161 − 0.143 − 0.018

− 0.059 − 0.089 0.029

− 0.107 − 0.149⁎ 0.042

0.540 0.069 0.471

0.954⁎ 0.235 0.719⁎

− 0.221 0.059 − 0.280

0.109 − 0.010 0.120

− 0.326 0.023 − 0.349

0.199 0.074⁎ 0.126

Tue

Wed

Thu

Fri

− 0.101 − 0.051 − 0.050

− 0.846⁎ − 0.043 − 0.803⁎

− 0.045 − 0.060 0.016

0.050 0.036 0.015

0.260 0.034 0.227

− 0.139 − 0.016 − 0.123

0.082 − 0.175⁎ 0.257

− 0.130 − 0.140 0.010

− 0.258 0.057 − 0.315

0.106 − 0.080 0.186

0.342 0.036 0.307

0.624 0.006 0.618

0.152 − 0.025 0.177

− 0.150 0.149⁎ − 0.299⁎

− 0.149 0.021 − 0.170

− 1.312⁎ − 0.078 − 1.234⁎

− 0.058 0.039 − 0.097

− 0.196 0.054 − 0.250

− 0.004 0.089⁎ − 0.093

− 0.335⁎ 0.024 − 0.359⁎

Panel B: Sub-period 2 minus sub-period 3 Mon Unconditional panel 0.022 c,t − 1Rc,t − 0.038 c,t − 1Ro,t 0.060 o,tRc,t

Tue

Wed

− 0.607⁎ − 0.084 − 0.523⁎

0.008 0.028 − 0.021

Thu 0.088 0.010 0.079

All day

⁎ Asterisks indicate 5% significance.

overnight return (Tsutsui, 2003 found a similar issue for Nikkei 225 stocks). The presence of stale prices is related to liquidity: if a stock rarely trades at the opening there will be many zero returns, and the measured standard deviation will be small. However, when such an illiquid security eventually trades, its return volatility is high. This explanation is consistent with our findings, as presented in Table 7: during the first sub-period 69% of all opening returns of type B stocks in Shanghai were zero. During this period, the opening standard deviation was 0.888%. However, during the second sub-period 93% of all opening returns were zero and the associated opening standard deviation declined to 0.061%. Consistent with this finding, there were 78% of zero opening returns in Shenzhen, but unlike the pattern in Shanghai, this rate declined to 20% in the second sub-period. As a result, the opening standard deviation, which was 0.599% in the first sub-period, increased to 0.712% during the second sub-period. Because Chinese markets open when a significant portion of foreign investors are not active, most of the trades take place throughout the trading day, rather than the opening, hence the variance of trading day returns is higher. It turns out that foreign investors traded more actively during the second sub-period than in the first sub-period, and primarily in Shenzhen. This may imply that foreign investors favor a certain degree of regulation over no regulation in a foreign, illiquid market. Table 7 further shows that type A stocks were more liquid during the first two sub-periods. During the first sub-period average unconditional returns in both exchanges were not significantly different from zero. The highest daily returns were on Fridays in Shanghai, but on Mondays in Shenzhen. With the allowance of local investors in type B stocks, being our third sub-period (Panel D), standard deviations decline to an

Table 9 Levene's test for Homogeneity of Variances between sub-periods — type B stocks. Panel A: Comparing sub-periods 1 and 2 Unconditional return Tue

Wed

Thu

Fri

All days

Shanghai type B stocks 0.00 c,t − 1Rc,t 0.00 c,t − 1Ro,t 0.00 o,tRc,t

Mon

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

0.01 0.00 0.00

0.00 0.00 0.00

Shenzhen type B stocks 0.00 c,t − 1Rc,t 0.72 c,t − 1Ro,t 0.00 o,tRc,t

0.00 0.55 0.00

0.00 0.15 0.00

0.00 0.03 0.00

0.03 0.00 0.02

0.00 0.00 0.00

Panel B: Comparing sub-periods 2 and 3 Unconditional return Tue

Wed

Thu

Fri

All days

Shanghai type B stocks 0.06 c,t − 1Rc,t 0.00 c,t − 1Ro,t 0.00 o,tRc,t

Mon

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

0.00 0.00 0.00

Shenzhen type B stocks 0.19 c,t − 1Rc,t 0.00 c,t − 1Ro,t 0.13 o,tRc,t

0.04 0.66 0.01

0.02 0.05 0.01

0.02 0.96 0.01

0.24 0.01 0.39

0.00 0.00 0.00

P-value b 5% indicates significant dissimilarity of return variance.

234

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

average level of 2.1%–2.2%, and average returns increase 3–4 times vs. sub-period 2. Interestingly, average unconditional returns on Monday were not the highest, nor the lowest, contrasting our findings for type

A stocks. Recall that the Monday effect was found in type A stocks once foreign investors were allowed, but the effect was not present in type B stocks, where foreign investors held the stocks throughout.

Table 10 Levene's test for Homogeneity of Variances between Pairs of Days — Type B stocks. Panel A: Entire period Unconditional returns Shanghai type B stocks (1992.5.22–2007.12.31)

Shenzhen type B stocks (1992.10.6–2007.12.31)

P-value

P-value Tue

Wed

Thu

Fri

Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.001 X

0.004 0.634 X

0.072 0.127 0.285 X

0.003 0.675 0.949 0.252

Mon Tue Wed Thu

0.007 X

0.003 0.759 X

0.036 0.509 0.332 X

0.010 0.922 0.687 0.577

Mon Tue Wed Thu

0.373 X

0.974 0.371 X

0.279 0.895 0.283 X

0.646 0.684 0.633 0.576

Mon Tue Wed Thu

0.269 X

0.597 0.108 X

0.082 0.005 0.266 X

0.086 0.005 0.279 0.954

Mon Tue Wed Thu

0.005 X

0.005 0.940 X

0.128 0.174 0.193 X

0.007 0.830 0.889 0.235

Mon Tue Wed Thu

0.021 X

0.019 0.911 X

0.192 0.328 0.288 X

0.054 0.743 0.668 0.527

Tue

Wed

Thu

Fri

Tue

Wed

Thu

Fri

Mon Tue Wed THu

0.277 X

0.382 0.754 X

0.367 0.861 0.904 X

0.935 0.271 0.370 0.353

Mon Tue Wed Thu

0.155 X

0.147 0.882 X

0.085 0.703 0.845 X

0.940 0.184 0.170 0.102

Mon Tue Wed Thu

0.805 X

0.125 0.085 X

0.448 0.332 0.427 X

0.256 0.187 0.783 0.643

Mon Tue Wed Thu

0.270 X

0.473 0.099 X

0.397 0.076 0.974 X

0.053 0.017 0.256 0.062

Mon Tue Wed Thu

0.168 X

0.285 0.685 X

0.277 0.750 0.941 X

0.698 0.321 0.514 0.488

Mon Tue Wed Thu

0.129 X

0.124 0.915 X

0.112 0.863 0.951 X

0.920 0.120 0.114 0.103

Tue

Wed

Thu

Fri

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

Panel B: Sub-period 1 Unconditional returns 1992.5.22–1996.12.13 P-value

P-value

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

Panel C: Sub-period 2 Unconditional returns 1996.12.16–2001.2.19 Shanghai type B stocks

Shenzhen type B stocks

P-value

P-value Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.098 X

0.470 0.357 X

0.545 0.274 0.889 X

0.061 0.864 0.264 0.191

Mon Tue Wed Thu

0.271 X

0.289 0.953 X

0.689 0.459 0.488 X

0.055 0.401 0.360 0.109

Mon Tue Wed Thu

0.126 X

0.121 0.548 X

0.002 0.000 0.359 X

0.776 0.044 0.101 0.000

Mon Tue Wed Thu

0.105 X

0.775 0.143 X

0.378 0.488 0.499 X

0.956 0.052 0.681 0.282

Mon Tue Wed Thu

0.094 X

0.450 0.360 X

0.526 0.276 0.886 X

0.057 0.859 0.263 0.190

Mon Tue Wed Thu

0.493 X

0.530 0.965 X

0.812 0.363 0.395 X

0.075 0.275 0.267 0.049

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

235

Table 10 (continued) Panel D: Sub-period 3 Unconditional returns 2001.2.28–2007.12.31 P-value

P-value Tue

Wed

Thu

Fri

Tue

Wed

Thu

Fri

Mon Tue Wed Thu

0.001 X

0.001 0.962 X

0.102 0.096 0.100 X

0.003 0.764 0.798 0.159

Mon Tue Wed Thu

0.018 X

0.007 0.769 X

0.109 0.425 0.268 X

0.012 0.938 0.822 0.365

Mon Tue Wed Thu

0.521 X

0.290 0.657 X

0.066 0.242 0.531 X

0.190 0.518 0.882 0.577

Mon Tue Wed Thu

0.459 X

0.536 0.952 X

0.005 0.046 0.074 X

0.166 0.556 0.555 0.106

Mon Tue Wed Thu

0.023 X

0.002 0.488 X

0.271 0.206 0.039 X

0.028 0.885 0.384 0.249

Mon Tue Wed Thu

0.024 X

0.035 0.931 X

0.233 0.288 0.345 X

0.079 0.606 0.680 0.580

c,t − 1Rc,t

c,t − 1Ro,t

o,tRc,t

P-value b 5% indicates significant dissimilarity of return variance.

5.3. Comparing average returns and variances across sub-periods Panel A of Table 8 shows that except for Tuesday's return, there were no material differences in average returns between the first and second sub-periods. Tuesday's return declined in both exchanges from nearly zero (−0.038% in SH and −0.004% in SZ) to the lowest and only negative return among weekdays in the second sub-period ( −0.452% in SH and −0.656% in SZ). Negative Tuesday's returns were found in other Asian and Australian markets (e.g., Aggarwal & Rivoli, 1989; Chen et al., 2001 and Jaffe & Westerfield, 1985). One of the explanations to this finding is the negative Monday returns in NYSE, being the leading market. Yet, Panel B of Table 8 shows that during the third sub-period Tuesday's total return was insignificantly positive (0.155% in SH, 0.190% in SZ), and higher than the average in both exchanges. Tuesday's positive return was generated in Shanghai at the opening following a positive prior return, and throughout the trading day following a negative prior return. The same pattern was evidenced in the Shenzhen exchange, although the trading day return following a negative prior return was not significant. To examine the effect of changing administrative rules on the variances of daily returns we conducted a Levene test for the homogeneity of variances between sub-periods. Panel A of Table 9 compares the variances of unconditional returns between sub-periods 1 and 2 (the comparison of conditional return variances is available upon request). Evidently, the variances of all three return measures, opening, trading day, and total return, were significantly lower in Shanghai, for both stock types. The case of Shenzhen was not as clearcut, primarily due to the variances of opening returns. This finding is probably related to the notion of liquidity at the opening, as explained above and described in Table 7. 5.4. Comparing the variance of returns across days of the week Panel A of Table 10 shows results of the Levene test for the homogeneity of variances across days of the week, over the entire period. Generally, Monday's variance is significantly higher than the variance of all or most other days of the week, in both exchanges. This higher variance stems primarily from trading day returns, rather than the opening, and almost exclusively from those returns conditional on prior positive closing. Exploring this finding over time reveals that there were no significant differences in the Shanghai exchange during the first sub-

period (Panel B, Table 10), both for the unconditional, and the conditional panels. The picture is somewhat more complicated in Shenzhen: while the unconditional panel shows no significant pattern, the panel conditional on prior positive return reveals that Monday's variance differs from all days of the week, except for Friday. Because this pattern was not strong enough and because the conditional on prior negative return contained no pattern, Monday's uniqueness did not show on the unconditional panel. The absence of a significant pattern in daily variances generally continued into the second sub-period (Panel C, Table 10), when the ± 10% rule applied. Yet, during the third sub-period (Panel D, Table 10), when local investors were allowed to invest in type B stocks the Monday effect in the variance of returns emerges. In both exchanges Monday's variance is higher than all other days of the week, except for Thursday. As before, it primarily stems from the trading day return, and less from the opening return. Here we find another distinction between type A and B stock returns: the variance of Mondays differs from that of other days due to trading day returns in type B stocks, but due to the opening return in type A stocks. Although both clienteles hold both stock types during the last sub-period of our study, foreign investors are less active during the opening, and more during the trading session. This unique property was not sufficient to make a difference when foreigners were the only clientele allowed to hold type B stocks, possibly due to lack of liquidity. 6. Summary This paper explores the impact that administrative regime changes in the Chinese financial markets had on daily rates of returns, and their variances. We focus on three distinct sub-periods: before a daily return limit applied, and after a ±10% limit was administratively imposed. In both sub-periods type A and B stocks were allowed to segmented clienteles, locals and foreigners, respectively. During the third sub-period both type A and B stocks were allowed to local and foreign investors, and the ±10% limit applied. We find that the ±10% limit indeed reduced the variance of type A stock returns, but the variance of type B stock returns increased by about 50%, from 1.9% to 2.7%. We provide evidence suggesting that the explanation involves changes in liquidity. Allowing all investors to hold both stock types had a number of effects: it reduced the variance of daily returns; it increased average daily returns of type A stocks by a

236

H. Kedar-Levy et al. / International Review of Financial Analysis 19 (2010) 223–236

factor greater than 10, and by a factor of 3–4 in type B stocks. Once foreign investors were allowed to hold type A stocks, a significant Monday effect emerged. This is a puzzling finding because while one would naturally relate it to foreign investors' trading patterns, no Monday effect was found in type B stocks that were held by foreign investors throughout. The return generating process in the Chinese markets seems to have been greatly affected by the segmentations of clienteles across stock types. Because foreign investors are generally absent from the collective bidding process at the opening, but rather trade throughout the trading day, absolute conditional returns are primarily generated in the following pattern: as long as the clienteles were segmented (sub-periods 1 and 2) absolute conditional returns on type A stocks were primarily generated at the opening, rather than throughout the trading day. At the same time absolute conditional returns on type B stocks were generated primarily throughout the trading session, rather than at the opening. Yet, once the clienteles were not segmented, during the third sub-period, returns were generated in a mixed format, both during the opening and during trade. This change has important implications for market efficiency, since the predictable pattern could have been utilized by sophisticated investors to earn excess returns while the clienteles were segmented. We reveal many more findings, including intra-daily momentum and reversal dynamics, systematic distinctions between both exchanges, between the two types of stocks, between opening- and intra-daily returns, and across different days of the week. The paper discusses the findings and analyzes the ways those differences changed following the administrative changes. References Abraham, A., & Ikenberry, D. (1994). The individual investor and the weekend effect. Journal of Financial and Quantitative Analysis, 29, 263−277. Aggarwal, R., & Rivoli, P. (1989). Seasonal and day-of-the-week effects in four emerging stock markets. Financial Review, 24, 541−550. Bessembinder, H., & Hertzel, M. G. (1993). Return autocorrelations around nontrading days. Review of Financial Studies, 6, 155−189.

Chen, G., Kwok, C. C. Y., & Rui, O. M. (2001). The day-of-the-week regularity in the stock markets of China. Journal of Multinational Financial Management, 11, 139−163. Connolly, R. A. (1989). An examination of the robustness of the weekend effect. Journal of Financial and Quantitative Analysis, 24, 133−171. Fama, E. F. (1970). Efficient capital markets: A review of theory and empirical work. Journal of Finance, 25, 383−417. French, K. R. (1980). Stock returns and the weekend effect. Journal of Financial Economics, 7, 579−596. Galai, D., & Kedar-Levy, H. (2005). Day-of-the-week effect in high moments. Financial Markets, Institutions and Instruments, 14, 169−186. Jaffe, J., & Westerfield, R. (1985). The weekend returns in common stock prices: The international evidence. Journal of Finance, 40, 433−454. Jaffe, F., Westerfield, R., & Ma, C. (1989). A twist on the Monday effect in stock prices: Evidence from the U.S. and foreign stock markets. Journal of Banking and Finance, 13, 641−650. Kamara, A. (1997). New evidence on the monday seasonal in stock returns. Journal of Business, 70, 63−84. Kim, Y., & Shin, J. (2000). Interactions among China-related stocks. Asia Pacific Financial Markets, 7, 97−115. Low, B. S., & Zhang, S. (2005). The volatility risk premium embedded in currency options. Journal of Financial and Quantitative Analysis, 40. Mitchell, J. D., & Ong, L. L. (2006). Seasonalities in China's stock markets: Cultural or structural?. International Monetary Fund: IMF Working Paper. Mookerjee, R., & Yu, Q. (1999). Seasonality in returns on the Chinese stock markets: The case of Shanghai and Shenzhen. Global Finance Journal, 10, 93−105. Rogalski, R. J. (1984). New findings regarding the day-of-the-week returns over trading and non-trading periods: A note. Journal of Finance, 39, 1603−1614. Schwert, W. G. (2003). Anomalies and market efficiency. In G. Constantinides, M. Harris, & R. Stulz (Eds.), Handbook of the economics of finance (pp. 937−972). Amsterdam: Elsevier Science. Shanghai stock exchange fact book, 2004, (http://www.sse.com.cn/sseportal /en_us/ps/ about/fact.shtml). Shenzhen stock exchange fact book, 2004, (http://www.szse.cn/main/en /Catalog_1389. aspx). Sun, Q., & Tong, W. H. S. (2000). The effect of market segmentation on stock prices: The China syndrome. Journal of Banking and Finance, 24, 1875−1902. Sun, Q., Tong, W. H. S., & Yan, Y. (2009). Market liberalization within a country. Journal of Empirical Finance, 16, 18−41. Tong, W. (2000). International evidence on weekend anomalies. Journal of Financial Research, 23, 4. Tsutsui, Y. (2003). Stock prices in Japan rise at night. Japan and the World Economy, 15, 391−406. Wang, S. S., & Firth, M. (2004). Do bears and bulls swim across oceans? Market information transmission between greater China and the rest of the world. Journal of International Financial Markets, Institutions and Money, 14, 235−254.