Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market

FINANA-00887; No of Pages 12 International Review of Financial Analysis xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

International Review of Financial Analysis

Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market Min Bai, Yafeng Qin ⁎ School of Economics and Finance, Massey University, New Zealand

a r t i c l e

i n f o

Article history: Received 30 October 2014 Received in revised form 24 July 2015 Accepted 4 August 2015 Available online xxxx JEL Classification: G14 G18 Keywords: Short sales constraints Earnings announcement Post-event analysis

a b s t r a c t This study examines how short sales constraints affect the stock price adjustment to the release of public information in the Hong Kong Stock Exchange. Using a unique feature of this market that allows us to directly investigate the impact of short sales restriction, we find the following. First, non-shortable stocks react more strongly to the publication of negative information than shortable stocks do. Second, non-shortable stocks are overpriced before negative earnings announcements. Hence, part of the strong market reaction of non-shortable stocks on announcement day could be due to the correction of such overpricing. Third, the prices of non-shortable stocks reverse following the announcement of negative information, suggesting that investors overreact to negative information on announcement day. Fourth, it takes longer for the prices of non-shortable stocks to adjust to negative earnings information. On the whole, our results support the research that finds short sales restrictions reduce the efficiency of stock markets. © 2015 Elsevier Inc. All rights reserved.

1. Introduction Generally, during financial market crises, regulators around the world attempt to constrain short selling activities. Looking for an escape route due to political pressure or their genuine beliefs, regulators argue that short sales destabilize the markets and lead prices to deviate far from their fundamental values. They also argue that through banning short sale activities, they could dampen sudden and significant drops in stock markets. Beber and Pagano (2013) report that, among the 30 markets in their sample, 21 imposed a short selling ban during the 2008–2009 crisis, either on whole stock markets or specific sectors. However, research questions the effectiveness of such a drastic practice of imposing a short selling ban. For example, Diamond and Verrecchia (1987) develop a rational expectations model where investors take into account short sales constraints when making their investment decisions. According to their model, since investors with private information are prevented from short selling, it takes longer for information to be incorporated into prices, especially negative information. Hence, when the information is made public, stocks with short sales constraints will react more strongly to the information than stocks without such constraints will. Hence, the model of Diamond and Verrecchia (1987) has two empirically testable predictions: 1) stocks ⁎ Corresponding author at: School of Economics and Finance, Massey University, Private Bag 102 904, North Shore Mail Centre, Auckland 0745, New Zealand. E-mail addresses: [email protected], [email protected] (M. Bai), [email protected] (Y. Qin).

with short sales constraints react to private information more slowly than stocks without such constraints, particularly to negative information; and 2) stocks with short sales constraints react more strongly to public information than stocks without such constraints. Our study empirically examines Diamond and Verrecchia (1987) by looking at how stocks with different levels of short sales constraints react to earnings announcements. Few studies have empirically tested these predictions. For example, Chen and Rhee (2010) use trade-by-trade data to compare the speeds of price adjustment for stocks before and after they are allowed to be sold short. The authors find that short selling increases the speed of price adjustment to both firm-specific and market-wide information in both up and down markets, suggesting that short selling increases market efficiency. Reed (2007), using loan prices determined in the equity market to proxy for the difficulty of short selling, finds that, when short selling is costly, stock prices are slow to incorporate private information and react strongly to information announcements. Fung and Draper (1999), also employing transaction data, provide evidence that lifting short selling constraints speeds up market adjustment and reduces the mispricing of index futures contracts. Bris, Goetzmann, and Zhu (2007) conduct a cross-sectional and time-series study on 46 equity markets around the world. They find that prices incorporate negative information faster in countries where short sales are allowed and implemented. Mashruwala and Mashruwala (2014), though focusing on examining the torpedo effect to earnings news, do provide evidence on the asymmetric reaction of stocks to good and bad earnings announcements when short selling constraints are binding.

http://dx.doi.org/10.1016/j.irfa.2015.08.006 1057-5219/© 2015 Elsevier Inc. All rights reserved.

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

2

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

In this paper, we examine Diamond and Verrecchia's (1987) prediction regarding short selling restrictions and stock price efficiency more directly in two aspects: first, unlike most previous studies that use proxies for short selling constraints, we use a more direct measure. We conduct our study on stocks from the Hong Kong Stock Exchange (HKSE), where only designated stocks that meet certain requirements are allowed to be sold short while the other stocks are strictly forbidden from being sold short. The list of designated shortable stocks is revised regularly, with stocks that become newly eligible being added to the list while no longer eligible stocks are removed. Such a practice enables us to compare the reactions of stocks with and without short sales constraints to the same type of information. More importantly, the direct measure of short sales constraints enables us to rule out other interpretations caused by using imperfect proxies. For example, Mashruwala and Mashruwala (2014), who use short interest data to proxy for short sales constraints, acknowledge that short interest is a measure of short selling activity rather than short selling constrains, hence they cannot definitively rule out the possibility that their findings could be caused by “short-sellers shorting stocks that are overpriced for reasons other than short-sales constraints (p. 540)”. Our direct measure can best prevent this problem. Several studies explore the effects of short selling constraints using HKSE data (e.g., Bai & Qin, 2014; Chang, Cheng, & Yu, 2007; Chen & Rhee, 2010), but their focus differs from ours. Chang et al. (2007) examine Miller's (1977) overpricing theory and Bai and Qin (2014) focus on the impact of short sales constraints on stock liquidity. The most closely related study is that of Chen and Rhee (2010), who also examine Diamond and Verrecchia's (1987) predictions. But their paper focuses on the asymmetric speed of price adjustment to new information between shortable and non-shortable stocks, while ours looks more at the asymmetric magnitude of price reactions to public information, complementing the work of Chen and Rhee (2010). Second, we examine stock reactions to a specific and special event: corporate earnings announcements, which also distinguishes our study from Chen and Rhee (2010). The benefits of selecting such an important event are threefold: 1) earnings announcements are scheduled events when companies release not only current earnings information but also substantial details to significantly reduce uncertainty about earnings (Berkman, Dimitrov, Jain, Koch, & Tice, 2009). Before the public announcements, investors with private information can establish long or short positions in the stocks that they expect to report positive or negative earnings surprises (Christophe, Ferri, & Angel, 2004). Uninformed traders, however, although they may anticipate the information, will observe the information only when it is made public via earnings announcements by the companies. Such a setting and information dissemination structure are consistent with Diamond and Verrecchia's (1987) framework.1 Hence, by looking at the pricing adjustments of stocks with different levels of short selling constraints to earnings announcements, we can provide more direct evidence on how short selling restrictions affect the efficiency of price adjustments to public information. 2) For earnings announcements, we can deliberately and relatively precisely separate negative earnings announcements from positive ones. This is important because Diamond and Verrecchia (1987) predict asymmetric effects not only between shortable and non-shortable stocks in adjusting to private/public information, but also between their reactions to good news and bad news. Chen and Rhee (2010) provide indirect evidence on the latter effect by showing that the difference in the speed of price adjustment for shortable and non-shortable stocks is more significant in a down market than in an up market. However, stocks can receive bad news even in an up market or receive bad news in a down market, particularly firm-specific news. Hence, short selling restrictions could also be binding in an up market for some stocks. Our study separates good news from bad news at the 1 Diamond and Verrecchia (1987, p.298) specifically state that “one can measure efficiency by the average of the absolute value of the returns on the announcement of a piece of regularly released private information: for example, corporate earnings.”

firm event level, which helps us to provide more direct evidence on the asymmetric effect of short selling constraints on price adjustments to good and bad news. 3) We can use an event study methodology to examine our research question at the firm event level. Such a methodology enables us to perfectly time the publication of private information and thus precisely measure the adjustments of stock prices. In addition, by focusing on the short event window around earnings announcements, we can largely rule out other market-wide factors that could contaminate our results. Besides, the literature has long debated whether analyst forecasts reflect all publicly available information and whether they are prone to behavioral bias (e.g., Jagadeesh & Kim, 2010; Jagadeesh & Livnat, 2006). In this study, by identifying the publication of private news, we shall be able to see differences in price changes before and after earnings announcement dates between shortable and non-shortable firms, which helps us to gain more insights into the informational efficiency of stock markets. Consistent with Diamond and Verrecchia's (1987) predictions, we find that, on the announcement day, prices change more significantly for stocks with short sales constraints than for stocks without constraints, for both positive and negative information. On average, abnormal returns of shortable stocks amount to −5.87% upon the announcement of negative information, while those of non-shortable stocks amount to −8.38%. Upon the announcement of positive earnings news, shortable and non-shortable stocks have average abnormal returns of 8.41% and 13.31%, respectively. To gain more insight into the reactions of shortable versus nonshortable stocks, we further look at the price behavior of stocks around earnings announcements, that is, 10 days before and 60 days after the announcements. We find that during the 10 days before the announcement of negative earnings news, the prices of non-shortable stocks increase substantially more than those of shortable stocks (3.47% vs. 1.05%). Such a phenomenon is consistent with Miller's (1977) overpricing theory and Diamond and Verrecchia's (1987) prediction of slow adjustment to bad private information. Before the information is publicly available, there are heterogeneous expectations about the information; however, due to short sales constraints, while optimistic investors are able to incorporate their future expectations of firm performance into prices by simply buying the stocks, many pessimistic investors cannot sell the stocks because they do not own them. Hence, stocks with short sales constraints are overpriced until the information is made public and part of the greater price reaction of non-shortable stocks on announcement day consists of a correction of the previously formed overpricing. For good earnings information, over the 10 days before announcements, the abnormal returns for shortable and nonshortable stocks are − 0.50% and −0.79%, respectively, but the difference is statistically insignificant. We further investigate the pricing of stocks following announcements to see how efficiently prices adjust to public information, applying Savor's (2012) regression framework. Our results indicate that while the prices of shortable stocks react efficiently to negative information on announcement day, the prices of non-shortable stocks seem to overreact to the information, with prices reversing following the announcements. Therefore, part of the greater price reaction of non-shortable stocks on announcement day is caused by investors' overreaction. Further tests show that it takes, on average, 12 days for non-shortable stocks to fully correct for the mispricing. However, when we look at good earnings announcements, both shortable and non-shortable stocks react to the information efficiently. The result confirms Diamond and Verrecchia's (1987) prediction that short sales constraints affect stock price adjustment to negative information. Our study contributes to the debate on the impact of short sales constraints on price efficiency by explicitly examining the price adjustments to earnings announcements of both stocks with short sales constraints and stocks without constraints. Besides providing direct empirical evidence to support the theoretical prediction that short sales constraints enlarge stocks' reactions to negative public information

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

3

Table 1 Changes in the official short selling list. This table provides information on changes in the official short selling list of the HKSE from January 1994 to December 2011, including the effective date on which a change took place (“Change Date”), the number of stocks added to (“Addition”) and deleted from (“Deletion”) the list, and the total number of stocks appearing on the list (“No. of on-list stocks”). Change date

Addition

Deletion

No. of on-list stocks

Change date

Addition

Deletion

No. of on-list stocks

Change date

Addition

Deletion

No. of on-list stocks

3/01/1994 25/03/1996 1/05/1997 12/01/1998 16/03/1998 9/11/1998 1/03/1999 20/09/1999 12/11/1999 28/02/2000 31/05/2000 28/08/2000 12/02/2001 14/05/2001 20/08/2001 3/12/2001 25/02/2002 21/05/2002 29/07/2002 29/11/2002 27/01/2003 19/05/2003 21/07/2003 4/08/2003 3/11/2003 6/01/2004 10/02/2004 7/04/2004 27/04/2004 1/07/2004 9/07/2004 2/08/2004 8/11/2004 7/02/2005 1/03/2005

17 96 129 69 15 19 7 3 1 24 7 32 15 6 9 17 7 11 24 6 5 18 1 0 36 1 29 1 26 1 1 8 9 15 2

0 0 1 0 0 149 7 17 0 12 0 16 11 0 11 85 14 6 5 15 7 7 16 1 5 0 3 0 4 0 0 21 11 7 0

17 113 241 310 325 195 195 181 182 194 201 217 221 227 225 157 150 155 174 165 163 174 159 158 189 190 216 217 239 240 241 228 226 234 236

17/05/2005 8/07/2005 15/07/2005 15/08/2005 5/09/2005 28/10/2005 18/11/2005 20/02/2006 1/03/2006 29/05/2006 2/06/2006 2/06/2006 25/08/2006 1/09/2006 23/10/2006 27/10/2006 1/12/2006 5/03/2007 14/03/2007 19/04/2007 26/04/2007 21/05/2007 21/05/2007 29/05/2007 4/07/2007 17/07/2007 13/08/2007 27/08/2007 26/11/2007 14/12/2007 14/12/2007 18/02/2008 13/03/2008 13/05/2008 15/05/2008

37 1 1 14 1 1 11 10 2 23 1 1 38 1 1 1 55 30 1 5 4 29 1 1 1 1 137 1 64 2 1 33 1 22 1

9 0 0 12 0 0 7 8 0 17 0 0 10 0 0 0 9 24 0 0 0 14 0 0 0 0 9 0 23 0 0 41 0 47 0

264 265 266 268 269 270 274 276 278 284 285 286 314 315 316 317 363 369 370 375 379 394 395 396 397 398 526 527 568 570 571 563 564 539 540

3/06/2008 7/08/2008 14/11/2008 12/02/2009 14/05/2009 10/07/2009 5/08/2009 5/11/2009 18/11/2009 3/12/2009 15/12/2009 24/12/2009 1/02/2010 1/03/2010 10/03/2010 25/03/2010 10/05/2010 16/07/2010 4/08/2010 30/08/2010 29/10/2010 15/11/2010 22/11/2010 20/12/2010 30/12/2010 28/01/2011 1/02/2011 25/02/2011 24/05/2011 9/06/2011 12/07/2011 12/08/2011 6/09/2011 3/11/2011 14/11/2011

5 10 6 25 13 1 49 58 1 1 1 1 65 1 1 1 59 1 40 1 47 1 2 1 1 1 1 70 65 1 2 24 1 18 1

0 51 144 27 22 0 16 11 0 0 0 0 8 0 0 0 12 0 19 0 18 0 0 0 0 0 0 17 18 0 0 50 0 97 0

545 504 366 364 355 356 389 436 437 438 439 440 497 498 499 500 547 548 569 570 599 600 602 603 604 605 606 659 706 707 709 683 684 605 606

and prolong their price adjustments, our findings suggest that the big price drops of non-shortable stocks upon the announcement of negative news come from three sources: 1) convergence to new fundamentals, 2) correction of previous overpricing, and 3) overreaction to new information. Such insight adds to the literature on short sales constraints and market efficiency, with important implications for regulators and market participants. The rest of this paper proceeds as follows. Section 2 introduces the background of the Hong Kong stock market and the selection of our sample. Section 3 describes our data and methodology. Section 4 presents the empirical results and Section 5 concludes the paper. 2. Short sales in HKSE and sample selection Short sales were prohibited on the HKSE until January 1994, when a pilot scheme was introduced under which 17 stocks became eligible for short selling. From January 1994, the HKSE also adopted the “uptick rule,” which mandated that a short sale could not be made below the best current ask price. The rule was repealed in March 1996 but reinstated on September 7, 1998, following the 1997 Asian financial crisis. The designated list of stocks that can be sold short is revised quarterly or less frequently since 1996. By December 2011, the end of our sample period, the list had been revised 104 times. Table 1 presents the information on historical revisions of the short selling list on the HKSE. The first, fifth, and ninth columns (labeled “Change Date”) indicate the effective dates on which a new version of the list of designated stocks for short selling took effect. The second, sixth, and 10th columns (labeled “Addition”) report the number of

stocks that are newly added to the list during each revision. These stocks will be able to be sold short from the indicated effective date onward. The third, seventh, and 11th columns (labeled “Deletion”) report the number of stocks that are removed from the list during each revision. These stocks will no longer be allowed to be sold short from the effective date onward. The fourth, eighth, and 12th columns (labeled “No. of onlist stocks”) report the total number of stocks that can be sold short after the effective date of each revision. By December 2011,2 there were 1777 additions and 1171 deletions. A closer look at all the stocks involved during these revisions shows that some stocks never appeared on the list, some that were added were later removed, and some that were added were never removed. The unique feature of the HKSE regarding short selling enables us to investigate how stocks with different short sales constraints (i.e., stocks on the list and that can thus be sold short versus stocks not on the list and which are thus constrained from being sold short) react to earnings announcements. Following Diamond and Verrecchia's (1987) prediction, we investigate two types of news, good and bad, and examine how short sales constraints affect stocks' reactions to the news.

2 We are grateful to Eric C. Chang for sharing the dates of the designated short-selling list from 1994 to 2003. The dates of the list from 2004 to 2011 were manually retrieved from the HKSE website. The selection criteria were obtained from the HKSE's official website (see Appendix A). The addition and deletion decisions may result in a certain degree of endogeneity in our tests because additions and deletions in the list may be due to excellent/poor past performance and/or large market capitalization. Chang et al. (2007) adopt two methods to differentiate the pure effect of changes in short sales constraints on stock returns from other possible explanations. They find the endogeneity problem of additions/deletions becomes irrelevant.

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

4

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

Table 2 Mean abnormal returns around earnings announcements for shortable and non-shortable stocks. This table reports abnormal returns (in percentage) around the negative (Panel A) and positive (Panel B) earnings announcement dates for shortable and non-shortable stocks in the Hong Kong stock market. The earnings announcement date is defined as day 0. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. A: Bad news

B: Good news

Shortable stocks

Non−shortable stocks

Shortable stocks

Non−shortable stocks

(N = 1415)

(N = 3100)

(N = 647)

(N = 1780)

Day

Mean AR

Median AR

t

Mean AR

Median AR

t

Mean AR

Median AR

t

Mean AR

Median AR

t

−10 −9 −8 −7 −6 −5 −4 −3 −2 −1 0 1 2 3 4 5 6 7 8 9 10

−0.116 −0.113 0.127 0.100 −0.262 0.182 0.130 0.173 0.441 0.390 −5.872 −0.453 0.238 0.207 0.192 0.473 −0.093 −0.166 −0.001 0.315 −0.127

−0.086 −0.072 0.005 −0.178 −0.308 −0.073 0.027 −0.001 0.072 0.048 −4.956 −0.485 0.117 0.004 −0.032 −0.013 −0.093 −0.065 0.020 0.051 −0.065

−0.85 −0.72 0.89 0.60 −1.66* 1.16 0.88 1.02 2.64** 2.24** −33.68*** −1.91* 1.36 1.25 1.15 2.90** −0.65 −1.17 0.00 2.20** −0.92

0.070 0.073 0.150 0.742 0.202 0.215 0.123 0.372 0.513 1.013 −8.375 0.169 −0.002 0.128 −0.109 −0.062 −0.111 0.069 −0.189 0.011 −0.110

−0.015 0.008 −0.004 0.070 0.008 0.003 0.011 0.054 0.063 0.181 −6.545 0.003 −0.006 0.005 −0.038 −0.040 −0.014 −0.023 −0.033 0.002 0.017

0.43 0.44 0.88 3.87*** 1.28 1.41 0.79 2.23** 3.03** 5.18*** −50.39*** 0.78 −0.01 0.84 −0.64 −0.38 −0.75 0.44 −1.23 0.06 −0.76

0.708 −0.454 −0.222 0.131 −0.293 −0.244 0.704 −0.225 −0.353 −0.253 8.410 −0.186 −0.160 −0.166 0.927 −0.016 0.054 0.269 0.194 −0.292 −0.028

−0.311 −0.387 −0.138 −0.090 −0.143 −0.150 −0.062 −0.096 −0.096 0.061 7.081 −0.006 −0.118 −0.089 0.225 −0.047 −0.067 −0.091 0.013 −0.309 −0.059

0.7 −1.75* −0.77 0.39 −1.24 −0.87 1.99** −0.71 −1.52 −0.96 22.01*** −0.5 −0.62 −0.62 3.02*** −0.06 0.2 1.08 0.77 −1.39 −0.11

0.098 0.247 0.015 −0.028 −0.173 −0.124 0.198 0.157 −0.447 −0.733 13.305 −0.067 −0.595 −0.367 −0.282 0.141 −0.441 0.040 0.049 0.158 0.215

0.008 0.014 −0.024 0.045 0.006 −0.021 0.048 0.045 −0.105 −0.067 9.840 −0.153 −0.097 −0.063 −0.083 0.000 −0.032 −0.045 0.008 −0.006 0.038

0.42 0.99 0.06 −0.11 −0.69 −0.58 0.96 0.69 −1.76* −2.36** 33.49*** −0.17 −2.49** −1.74 −1.14 0.56 −2.16** 0.19 0.23 0.63 0.93

out to be smaller than the analyst forecasts. However, if the market were expecting even worse earnings, negative SUE could be associated with a positive market response to the earnings announcement and such an announcement should be classified as good news. Hence, following Engelberg, Reed, and Ringgenberg (2012), we use announcement day returns to identify positive and negative earnings announcements. In particular, we use both the sign and size of the return to classify the contents of the announcements: We define a negative (positive) news event as one where a firm's announcement day return is in the bottom (top) quintile of all returns on a given earnings announcement day and a neutral news event when a firm's announcement day return is in neither the top nor the bottom quintile. As robustness checks, we also use SUE or the cumulative abnormal returns of the stocks from the announcements until 5 or 10 days afterward as an indicator to sign the contents of the news; our results are qualitatively unchanged. We use a standard event study to examine the price reactions of stocks to earnings announcements. In comparison with the

3. Data and methodology We obtain the earnings announcement dates through Thomson Reuters via SIRCA. Daily closing prices, market capitalization, book-tomarket ratios, trading volumes, and numbers of shares outstanding for all Hong Kong stocks traded on the HKSE, including both shortable and non-shortable stocks, and the Hang Seng Price Index are sourced from Datastream. Our sample period is from January 1, 1994, to December 31, 2011. It is important to first classify the types of news. There are various ways in the literature to define the signs of news (positive or negative), such as standardized unexpected earnings (SUE; Michaely, Rubin, & Vedrashko, 2014). While SUE measures provide information about analyst forecasts on the firms' earnings, they do not contain information on the broader market participants' prior expectations. In particular, for some small stocks with very few analysts following, the analyst forecasts could deviate greatly from overall market expectations. Hence, a firm could have negative SUE if the actual earnings announced turn

Table 3 Cumulative abnormal returns around earnings announcements for shortable and non-shortable stocks. This table reports cumulative abnormal returns (CAR) (in percentage) around the earnings announcement dates for shortable and non-shortable stocks when bad/good news is released. The earnings announcement date is defined as day 0. One-tailed test results are obtained by calculating the percentage of the mean cumulative abnormal returns and the difference in cumulative abnormal returns between the non-shortable and shortable stocks from the data sample. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. A: Bad news

B: Good news

Shortable stocks

Non−shortable stocks

(N = 1415)

Shortable stocks

(N = 3100)

Non−shortable stocks

(N = 647)

(N = 1780)

Event windows

Mean CAR

Median CAR

t

Mean CAR

Median CAR

t

Mean CAR

Median CAR

t

Mean CAR

Median CAR

t

(−10, −2) (−10, −1) (−1, 0) (−1, 1) (−5, 5) (0,5) (0, 10) (1, 5) (1, 10)

0.662 1.052 −5.872 −3.899 −6.087 −5.286 −0.453 0.586 1.021

0.177 0.409 −4.956 −4.355 −5.352 −4.971 −0.485 0.554 0.541

1.35 2.00** −33.68*** −6.69*** −16.98*** −9.21*** −1.91* 1.08 1.31

2.460 3.473 −7.362 −7.193 −6.015 −8.251 −8.581 0.124 −0.206

1.085 2.332 −6.673 −7.025 −6.042 −7.583 −8.533 −0.344 −0.621

5.41*** 7.06*** −29.95*** −23.07*** −11.41*** −21.14*** −16.85*** 0.34 −0.41

−0.248 −0.501 8.158 7.972 8.438 8.809 9.007 0.398 0.596

−0.792 −0.576 7.639 7.927 7.091 7.552 7.157 0.366 0.083

−0.22 −0.44 18.05*** 14.47*** 9.24*** 11.96*** 9.58*** 0.57 0.66

−0.057 −0.791 12.571 12.505 11.186 12.135 12.156 −1.17 −1.149

−0.551 −0.851 9.538 8.967 7.677 9.454 9.587 −2.310 −2.174

−0.1 −1.17 24.18*** 19.2*** 12.24*** 17.32*** 15.18*** −2.00** −1.60

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

A) CAR of shortable and non-shortable stocks

To obtain the statistical inferences on the differences between shortable and non-shortable stocks in reaction to earnings announcements, besides the univariate test we also run the following regression on both the bad news and good news samples:

Cumulative average abnormal return

around negative earnings announcements 0.04 0.02

CARðt 1 ;t 2 Þ ¼ α þ β1 DM þ γ0 X þ δ0 DM  X þ μ

ð3Þ

0 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 -0.02 -0.04 -0.06 Bad_NSS

Bad_SS

B) CAR of shortable and non-shortable stocks around positive earnings annoucements Cumulative average abnormal return

5

0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 -0.02 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 Good_NSS

Good_SS

Fig. 1. Cumulative abnormal returns of shortable and non-shortable stocks around earnings announcements. These panels present the cumulative average daily abnormal returns for shortable stocks (SS) and non-shortable stocks (NSS) over a 21-day period around negative and positive earnings announcements, respectively.

cross-sectional analysis adopted by most empirical studies on short sales constraints, where it is hard to control for many exogenous and endogenous factors, an event study is relatively free of this problem. The abnormal return (AR) and cumulative abnormal returns (CAR) are respectively defined as
 ARi ðt Þ ¼ Rit −R f −βi  Rmt −R f CARiðt 1 ;t 2 Þ ¼

Xt 2 
 Rit −R f −βi  Rmt −R f t¼t 1

ð1Þ ð2Þ

where ARi(t) indicates the abnormal return of stock i on event day t, with event day 0 defined as the announcement day; CARiðt 1 ;t 2 Þ represents the cumulative abnormal return between t1 and t2 in different event windows; βi represents the estimated slope coefficient for stock i based on the stock's historical return series from day − 31 to day − 280; and Rf and Rm are respectively the risk-free rate (the 3-month Hong Kong interbank offered rate) and the HKSE market return calculated based on the Hang Seng Index. For each group of stocks (shortable and non-shortable), we rank all the stocks based on their abnormal returns on the announcement days, AR0, and select the stocks whose AR0 values fall in the bottom (top) quintile as our bad (good) news sample. Among the bad news sample, we have 1415 announcements on shortable stocks and 3100 announcements from non-shortable stocks. Among the good news sample, we have 647 and 1780 announcements from shortable and nonshortable stocks, respectively. We then look at the abnormal returns, AR, as well as the cumulative abnormal returns, CAR, of the shortable and non-shortable stocks around the publication of good (bad) earnings news and see if they are different.

where DM is a dummy variable, with DM = 1 indicating shortable stocks and DM = 0 indicating non-shortable stocks. The selection criteria of the HKSE on designated securities for short selling (see Appendix A) indicate that market capitalization, trading volume, and free float, among other requirements, are important factors that determine the inclusion and exclusion of stocks from the short selling list. Therefore, shortable and non-shortable stocks could differ greatly in these firm characteristics, which could lead to the difference in the AR and CAR values we observe. Hence, we include a vector of control variables X as follows: log(MV) and log(BM) are respectively the logarithm transformation of the market capitalization and book-to-market ratio of the stocks; VOL is the trading volume, measured as the average turnover ratio of the stocks from day −60 to day −31 before the earnings announcements; MOM is the past 11-month excess return with 1 month lag to control for the momentum effect; and ILLIQ is the Amihud illiquidity ratio measured from 3 months before the earnings announcements to control for the liquidity effect. The definition and descriptive statistics of the control variables are summarized in Appendix B. The dependent variables CARðt1 ;t 2 Þ are the cumulative abnormal returns before the earnings announcements, CAR(−10,− 1); on the announcements, CAR(0,0); or after the announcements, CAR(1,10). The above analysis focuses on the cross-sectional comparison of shortable and non-shortable stocks in their pricing behavior around earnings announcements. To gain a better understanding of the stocks' response to public information, we also examine the time-series relation between the post-event returns and announcement day returns of stocks with different short selling constraints, applying the following regression framework, proposed by Savor (2012): CARðt 1 ;t 2 Þ ¼ α þ βAR0 þ γX þ μ

ð4Þ

where CARðt 1 ;t 2 Þ is the cumulative abnormal return over a period starting from day 1 to day 40 after the event date and AR0 is the event day abnormal return, with X are a number of control variables we used in the previous regressions. We use the weighted least squares (WLS) regression to estimate the parameters, where the weights are set to be equal cross-sectionally. The t-statistics are calculated using clustered standard errors. The main focus of our analysis is the impact of the event day abnormal return AR0 on post-event cumulative returns, captured by the coefficient β. A positive β indicates a return continuation following the announcements, suggesting that investors underreact to the information on the event day, while a negative β suggests that prices reverse following the announcement, suggesting that investors overreact to the information on the event day. In addition, by varying the event window (t1, t2), we can determine how long it will take for the prices to fully incorporate all the information. 4. Results and analysis 4.1. Empirical tests We first compare the event day abnormal returns of non-shortable stocks with those of shortable stocks. The results are presented in Table 2, where we report the cross-sectional mean, median, and tstatistics of AR0 among our sub-sample stocks on each of the 21 days around the earning announcements. Diamond and Verrecchia (1987) imply that short sales constraints make the excess returns larger in absolute value on public information announcement days. Our results, from both the bad news and good news samples, support such an

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

6

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

Table 4 Regression analysis of cumulative abnormal returns around earnings announcements. This table reports the coefficient estimates of the following regression among stocks with bad (Panel A) or good (Panel B) earnings announcements: CARðt1 ;t2 Þ ¼ α þ β1 DM þ γ0 X þ

δ0 DM  X þ μ, where CARðt 1 ;t2 Þ is the cumulative abnormal return over a period starting t1 trading days and ending t2 trading days around the event day; DM is a dummy variable, with DM = 1 indicating shortable stocks and DM = 0 indicating non-shortable stocks; and X is a vector of control variables, including the logarithm of size (log(MV)), the logarithm of the book-to-market ratio (log(BM)), trading volumes (VOL) measured as the turnover ratio, momentum (MOM), and the Amihud illiquidity ratio (ILLIQ). Parameter estimates are computed using the generalized least squares approach. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Intercept A: Bad news Pre-announcement CAR−10, −1 0.036 7.56*** CAR−10,−1 0.026 1.11 Upon announcement CAR0,0 −0.083 −41.61*** CAR0,0 −0.124 −12.57*** Post-announcement CAR1,10 −0.001 −0.11 0.127 CAR1,10 4.85*** B: Good news Pre-announcement CAR−10, −1 −0.011 −1.60 CAR−10, −1 −0.081 −2.46** Upon announcement CAR0,0 0.133 35.96*** CAR0,0 0.220 12.54** Post-announcement CAR1,10 −0.010 −1.38 CAR1,10 0.087 2.53**

DM

−0.024 −3.15*** −0.018 −2.96*** 0.023 6.77*** 0.030 1.98** 0.008 0.84 −0.028 −0.63

0.008 0.56 0.041 0.65 −0.049 −6.47*** −0.028 −0.84 0.020 1.36 −0.105 −1.63

Log(MV)

Log(BM)

VOL

MOM

ILLIQ

DM * Log(MV)

DM*Log(BM)

DM * VOL

DM * MOM

DM *ILLIQ

R2

4.90% 0.001 0.15

−0.002 −0.47

−0.337 −3.29***

0.007 0.35

0.543 0.85

−0.007 −1.19

0.000 0.03

0.492 0.90

0.002 0.34

−1.012 −2.04**

5.22%

3.50% 0.008 4.38***

0.001 0.49

−0.366 −4.74***

0.005 0.27

1.233 0.96

−0.004 −1.55

0.000 −0.02

−0.057 −0.25

−0.083 −0.47

0.097 1.45

4.29%

−1.09% −0.024 −4.93***

0.011 2.35**

0.246 1.19

−0.003 −1.01

−0.972 −0.55

0.014 2.20**

0.001 0.10

−0.092 −0.15

−0.075 −0.04

0.006 0.23

3.81%

−1.02% 0.011 1.79*

−0.001 −0.26

1.492 5.78***

−0.045 −0.58

0.076 0.44

−0.009 −1.02

0.027 1.72*

1.013 1.3

0.075 1.02

−1.005 −1.12

5.46%

4.36% −0.018 −5.38***

−0.006 −2.29**

0.633 4.63***

0.009 1.14

0.054 0.09

0.005 1.00

−0.001 −0.10

−0.663 −1.61

0.024 1.13

0.862 0.78

6.76%

−0.96% −0.019 −2.98**

0.005 0.85

0.688 2.56**

0.010 1.02

implication. On the day when negative earnings information is released, the prices of shortable stocks drop, on average, by 5.87% (median 4.96%), while those of non-shortable stocks drop by 8.38%, on average (median 6.55%), which is 43% more than for shortable stocks. When positive earnings information is released, the prices of shortable stocks, on average, increase by 8.41% (median 7.08%), while those of nonshortable stocks increase by 13.31%, on average (median 9.84%). Second, for bad news announcements, both types of stocks, on average, exhibit positive abnormal returns around 1 week before the announcement. However, when we compare the magnitudes of positive returns, we can see that the absolute values of abnormal returns from nonshortable stocks are again greater than those from shortable stocks. The finding on positive returns before earnings announcements is consistent with Miller's (1977) overpricing prediction that, before private information is made public, there exists a divergence of opinions among investors regarding the information. When short sales are not allowed, only optimistic opinions are incorporated into prices, causing the assets to be overvalued. The finding on the difference in returns between shortable and non-shortable stocks is consistent with the empirical evidence of informed trading in pre-announcement short selling (Berkman & McKenzie, 2012; and Christophe et al., 2004) and with Diamond and Verrecchia's (1987) prediction that short sales constraints hinder the speed of price adjustment to private information, especially to bad news. The second point above is further confirmed in Table 3, where we present the CAR values with various event windows for both types of stocks. We note that from day − 10 to day − 2 before the announcements of negative earnings news, shortable stocks increase in price by

−0.025 −1.03

0.024 2.57**

0.014 0.83

−0.746 −0.92

0.043 0.89

0.765 0.43

4.18%

0.66%, while non-shortable stocks increase in price by 2.46%, which is almost four times as much as shortable stocks. From day − 10 to day − 1, shortable stocks have a cumulative abnormal return of 1.05%, which is again significantly below the non-shortable stocks' cumulative abnormal return of 3.47%. Such pattern is vividly illustrated in Panel A of Fig. 1; however, from Panel B, we can see that shortable and nonshortable stocks do not seem to have significantly different returns before the publication of good earnings information.

4.2. Regression analysis with a short selling constraint dummy To obtain statistical inferences on the differences between shortable and non-shortable stocks observed from Tables 2 and 3 and Fig. 1, we perform regression analysis with a dummy variable to indicate the shortability of stocks and present the results in Table 4. Table 4 shows that, before the announcement of bad news, shortable stocks have significantly lower returns than non-shortable stocks, even after we control for firm characteristics, while for the good news sample shortable and non-shortable stocks do not show any significant differences in cumulative abnormal returns before the earnings announcements. Such a finding suggests that before bad news is made public, shortable stocks, relative to non-shortable stocks, have already started incorporating the negative information in their prices. However, this phenomenon is absent in the publication of good news. This finding is consistent with Diamond and Verrecchia's (1987) prediction that short sales constraints hinder the speed of price adjustment to private information, especially to bad news.

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

7

Table 5 Regression analysis of cumulative abnormal returns around earnings announcements with SUE. This table reports the coefficient estimates of the following regression among stocks with bad (Panel A) or good (Panel B) earnings announcements: CARðt 1 ;t2 Þ ¼ α þ β 1 DM þ 0

0

β 2 jSUEj þ β 3 DM  jSUEj þ γ X þ δ DM  X þ μ, where CARðt1 ;t2 Þ is the cumulative abnormal return over a period starting t1 trading days and ending t2 trading days around the event day; DM is a dummy variable, with DM = 1 indicating shortable stocks and DM = 0 indicating non-shortable stocks; |SUE| is the absolute value of SUE; and X is a vector of control variables, including the logarithm of size (log(MV)), the logarithm of the book-to-market ratio (log(BM)), trading volumes (VOL) measured as the turnover ratio, momentum (MOM), and the Amihud illiquidity ratio (ILLIQ). Parameter estimates are computed using the generalized least squares approach. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Intercept DM

|SUE|

DM * |SUE|

A: Bad news Pre-announcement CAR−10,−1 0.039 1.22 −0.416 CAR−10,−1 −1.53 Upon announcement CAR0,0 −0.061 −2.74*** CAR0,0 −0.072 −0.39 Post-announcement CAR1,10 0.017 0.37 CAR1,10 −0.609 −1.64 B: Good news Pre-announcement CAR−10,−1 −0.301 −3.53*** CAR−10,−1 −0.280 −3.37*** Upon announcement CAR0,0 0.100 3.06*** CAR0,0 0.106 3.47*** Post-announcement CAR1,10 0.055 0.59 CAR1,10 0.043 0.46

Log (MV)

Log(BM) VOL

−0.032 −0.94 0.027 1.48

0.001 0.8 0.002 1.31

−0.002 −2.91*** −0.511 0.064 −0.005 −1.98** 1.66* −0.11

0.008 0.35 −0.015 −0.07

0.000 −0.02 0.000 0.01

0.007 3.31*** 0.007 3.13***

0.011 0.23 0.650 1.67*

0.000 −0.07 −0.001 −0.70

0.009 1.93* 0.013 2.94**

0.304 1.53 0.432 0.83

−0.004 −1.58 −0.005 −1.84*

0.001 0.23 0.003 0.80

−0.034 −1.03 0.080 1.43

0.001 0.62 0.000 0.49

0.000 −0.07 0.001 0.38

−0.024 −0.25 0.139 0.81

0.002 0.61 0.002 0.71

−0.001 −0.11 0.000 0.04

MOM

ILLIQ

DM *

DM *

Log(MV)

Log(BM)

DM * VOL

DM * MOM

DM * ILLIQ

R2

1.69% 3.690 1.03

0.015 0.27

1.096 0.33

−0.074 −1.86*

−0.007 −0.13

−3.271 0.013 −0.89 0.41

−0.098 5.37% 1.99** 5.85%

0.003 0.005 0.10 0.15

−0.654 1.034 −0.27 0.68

1.562 0.84

0.002 0.08

−0.004 −0.11

−0.599 0.006 −0.24 0.14

0.873 1.57

6.67%

1.12% 0.087 −0.006 1.66* −0.08

5.500 1.12

−0.025 −0.889 −0.086 −0.37 −0.68 −1.58

0.058 0.83

−5.771 −0.063 0.017 −1.15 −0.12 0.33

7.08%

0.44% 0.000 −0.030 0.80 −1.13

1.708 1.92*

−0.102 0.175 −0.66 0.32

−0.014 −0.99

0.050 1.76*

1.708 1.92*

0.066 0.95

−0.864 3.06% −1.53 0.90%

0.000 −0.009 0.00 −0.95

0.159 0.49

0.012 1.01

0.029 0.58

−0.012 −2.29**

0.011 1.04

0.159 0.49

0.074 0.92

0.054 2.38**

3.48%

−0.26% 0.000 0.017 0.57 0.56

−0.655 0.585 −0.65 0.43

On announcement days, the regression results show that for the bad news sample shortable stocks have significantly higher abnormal returns than non-shortable stocks, even after we control for firm characteristics. Since the abnormal returns are negative for the bad news sample, the non-shortable stocks have significantly lower negative returns than shortable stocks or, put another way, the nonshortable stocks react more strongly to the announcement of bad news than shortable stocks. Hence, the results are supportive of Diamond and Verrecchia's (1987) second prediction, that announcement day price reactions will be larger when short selling restrictions are binding. However, for the good news sample, though nonshortable stocks have higher abnormal returns than shortable stocks on announcement days, such a difference is statistically insignificant after we include control variables. After earnings announcements, during the window 1 day to 10 days after the event, shortable and non-shortable stocks do not differ significantly in their returns, for both the bad news sample and the good news sample, with and without control variables. 4.3. Regression with earnings announcement informational content So far we have documented the difference in stock responses to earnings announcements between shortable and non-shortable stocks, particularly to bad news. We further examine whether such a difference is associated with the informational content of the earnings announcements, measured by SUE (Standardized Unanticipated Earnings). In doing so, we merge our sample with data from the

−0.009 −0.018 −0.25 −1.12

0.005 0.17

−0.655 0.088 −0.65 0.52

−0.736 −1.07% −0.79

Institutional Brokers' Estimate System (I/B/E/S), where we can obtain SUE3 scores, and run the following regression4: CARðt 1 ;t 2 Þ ¼ α þ β1 DM þ β2 jSUEj þ β3 DM  jSUEj þ γX þ δDM Xþ μ

ð5Þ

The results, shown in Table 5, are qualitatively consistent with our results in Table 4. For the bad news sample, the coefficients of DM * |SUE| are significantly negative before earnings announcements but significantly positive during and after the announcements. Such results indicate that before bad earnings information is made public, shortable stocks underperform non-shortable stocks, despite the significance of the information, probably due to the incorporation of private bad news into the shortable stocks. Upon the publication of the earnings and thereafter, non-shortable stocks significantly underperform shortable stocks, suggesting that non-shortable stocks react more strongly to the publication of bad news than shortable stocks, and such a difference lasts until after the event period. For the good news sample, however, after we control for the contents of the surprise in earnings news, shortable and nonshortable stocks do not show any difference before, during, or after the 3 I/B/E/S calculates their SUE Score the number of standard deviations the actual (reported) earnings differ from the I/B/E/S Surprise mean estimates for a company, for the fiscal period indicated. 4 At the cost of losing statistical power, however, since such merging greatly reduced our sample size.

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

8

A: Shortable stocks and bad news

B: Non−shortable stocks and bad news

Dependent

Intercept

AR0

Log(MV)

Log(BM)

VOL

MOM

ILLIQ

R2

Dependent

Intercept

AR0

Log(MV)

Log(BM)

VOL

MOM

ILLIQ

R2

CAR1,1

0.061 2.60*** 0.075 2.43** 0.139 3.17*** 0.185 3.20*** 0.219 3.64*** 0.279 3.97***

0.105 1.03 0.146 1.09 0.321 1.68* 0.336 1.33 0.424 1.62 0.530 1.73*

−0.005 −1.97** −0.006 −1.98** −0.013 −2.82*** −0.016 −2.64** −0.018 −2.87*** −0.024 −3.35***

0.006 1.28 0.009 1.41 0.011 1.25 0.047 3.88*** 0.061 4.87*** 0.066 4.50***

−0.618 −1.90* −0.593 −1.39 0.316 0.52 −0.700 −0.88 −0.783 −0.94 −0.309 −0.32

−0.009 −0.21 −0.013 −0.36 0.005 0.29 0.011 \0.44 0.048 0.54 0.036 0.88

−1.425 −2.04** −1.113 −1.74* −1.052 −1.63 0.046 0.88 0.084 0.52 −0.893 −1.02

3.23%

CAR1,1 CAR1,5

2.31%

CAR1,10

5.70%

CAR1,20

7.92%

CAR1,30

7.70%

CAR1,40

−0.162 −4.14*** −0.208 −3.12*** −0.251 −2.83*** −152 −1.25 −0.295 −1.34 −0.367 −1.24

−0.006 −2.86*** −0.015 −4.22*** −0.022 −4.58*** −0.040 −6.11*** −0.053 −6.78*** −0.071 −7.95***

0.004 2.05** 0.007 2.08** 0.011 2.46** 0.032 5.11*** 0.052 7.13*** 0.074 8.87***

−0.255 −2.84*** −0.256 −1.68* 0.154 0.75 −0.193 −0.69 −0.200 −0.60 −0.490 −1.30

0.002 0.86 0.004 1.64 0.015 1.73* 0.019 1.89* 0.240 2.04** 0.048 2.98***

−0.174 −1.42 −0.253 −1.02 −0.185 −0.95 −0.123 −0.64 −0.097 −0.76 −0.144 −0.52

3.55%

1.36%

0.023 1.91* 0.068 3.24*** 0.096 3.45*** 0.206 5.42*** 0.264 5.81*** 0.350 6.78***

CAR1,5 CAR1,10 CAR1,20 CAR1,30 CAR1,40

C: Shortable Stocks and Good News

3.33% 3.94% 6.72% 9.19% 12.30%

D: Non−Shortable Stocks and Good News

Dependent

Intercept

AR0

Log(MV)

Log(BM)

VOL

MOM

ILLIQ

R2

Dependent

Intercept

AR0

Log(MV)

Log(BM)

VOL

MOM

ILLIQ

R2

CAR1,1

−0.007 −0.30 0.021 0.44 −0.003 −0.05 0.033 0.47 0.052 0.55 0.049 0.50

−0.126 −1.48 −0.195 −1.19 −0.078 −0.37 −0.099 −0.40 −0.046 −0.14 −0.184 −0.54

0.003 1.05 0.000 0.04 0.003 0.53 −0.002 −0.24 −0.002 −0.17 0.002 0.18

0.002 0.45 0.000 0.04 0.018 1.41 0.042 2.75*** 0.051 2.56** 0.087 4.21***

−0.166 −0.66 0.024 0.05 −0.060 −0.1 0.156 0.21 −1.189 −1.21 −1.703 −1.68*

−0.003 −1.02 −0.011 −0.42 0.004 0.21 0.011 1.05 0.022 1.28 0.046 1.48

0.178 1.98** 0.046 1.35 −0.147 −0.98 0.098 1.44 0.076 1.21 −0.25 −0.48

1.19%

CAR1,1 CAR1,5

−0.70%

CAR1,10

4.14%

CAR1,20

4.52%

CAR1,30

8.61%

CAR1,40

0.086 1.32 −0.014 −0.24 −0.139 −1.56 −0.269 −1.73* −0.200 −1.57 −0.138 −0.95

−0.004 −1.19 −0.013 −2.24** −0.022 −3.14*** −0.042 −4.30*** −0.053 −4.24*** −0.070 −4.92***

−0.002 −0.73 −0.003 −0.63 0.004 0.66 0.011 1.33 0.021 2.03** 0.041 3.43***

0.026 0.18 0.473 2.02** 0.776 2.71*** 0.813 2.04** 0.733 1.43 0.818 1.4

0.001 0.22 0.008 0.36 0.102 0.86 0.113 0.87 0.143 1.35 0.201 1.08

0.258 2.04** 0.214 1.79* 0.185 1.75* 0.136 1.68* 0.115 1.23 0.154 1.59

1.65%

−0.97%

0.009 0.45 0.054 1.69* 0.118 2.98*** 0.243 4.41*** 0.316 4.46*** 0.410 5.08***

CAR1,5 CAR1,10 CAR1,20 CAR1,30 CAR1,40

3.66% 3.69% 4.23% 4.10% 6.47%

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

Table 6 Determinants of stock returns following earnings announcements. This table reports the coefficient estimates of the following regression among shortable stocks (Panels A and C) and non-shortable stocks (Panels B and D) with bad (Panels A and B) and good (Panels C and D) earnings announcements: CARðt1 ;t 2 Þ = α + βAR 0 + γ'X + μ, where CARðt1 ;t 2 Þ is the cumulative abnormal return over a period starting t1 trading days and ending t2 trading days around the event day; DM is a dummy variable, with DM = 1 indicating shortable stocks and DM = 0 indicating non-shortable stocks; and X is a vector of control variables, including the logarithm of size (log(MV)), the logarithm of the book-to-market ratio (log(BM)), trading volumes (VOL) measured as the turnover ratio, momentum (MOM), and the Amihud illiquidity ratio (ILLIQ). Parameter estimates are computed using the WLS approach, where the weights are set so that each cross section has equal weight. The t-statistics (in italics) are calculated using clustered standard errors. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively.

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

9

Table 7 Determinants of post-event returns: shortable and Non-shortable stocks. This table reports the coefficient estimates of the following regression among stocks with good earnings announcements: CARðt 1 ;t2 Þ = α + β1AR0 + β2DM*AR0 + γ1'X + γ2'DM*X + μ, where CARðt 1 ;t2 Þ is the cumulative abnormal return over a period starting t1 trading days and ending t2 trading days around the event day; DM is a dummy variable, with DM = 1 indicating shortable stocks and DM = 0 indicating non-shortable stocks; and X is a vector of control variables, including the logarithm of size (log(MV)), the logarithm of the book-to-market ratio (log(BM)), trading volumes (VOL) measured as the turnover ratio, momentum (MOM), and the Amihud illiquidity ratio (ILLIQ). Parameter estimates are computed using the generalized least squares approach. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Dependent

Intercept

AR0

A: Bad News 0.035 CAR1,1 3.13*** CAR1,5 0.070 4.05*** CAR1,10 0.110 4.65*** CAR1,20 0.199 6.27*** CAR1,30 0.249 6.87*** CAR1,40 0.327 7.86***

-0.147 -3.48*** -0.205 -3.16*** -0.234 -2.65*** -0.160 -1.34 -0.312 -1.3 -0.395 -1.53

B: Good News CAR1,1 0.006 0.35 CAR1,5 0.048 1.75 CAR1,10 0.093 2.78*** CAR1,20 0.200 4.37*** CAR1,30 0.262 4.44*** CAR1,40 0.336 5.05***

0.088 1.63 -0.009 -0.16 -0.121 -0.83 -0.239 -1.34 -0.162 -1.38 -0.086 -0.65

DM*AR0

Log(MV)

Log(BM)

0.192 2.26** 0.340 2.61*** 0.487 2.74*** 0.530 2.22** 0.808 2.96*** 1.040 3.32***

-0.008 -3.02*** -0.016 -3.11*** -0.024 -3.83*** -0.039 -3.96*** -0.051 -4.92*** -0.068 -4.13***

0.004 1.68* 0.007 2.07** 0.011 2.32** 0.032 3.13*** 0.053 3.49*** 0.075 4.25***

-0.004 -1.24 -0.012 -2.39** -0.018 -2.96*** -0.035 -4.23*** -0.044 -4.16*** -0.058 -4.85***

-0.002 -0.76 -0.003 -0.62 0.005 0.84 0.012 1.62 0.023 2.36*** 0.043 3.94***

-0.245 -1.32 -0.249 -1.47 -0.182 -0.87 -0.250 -0.88 -0.377 -1.02 -0.771 -1.46

VOL

MOM

ILLIQ

DM*log(MV)

DM*log(BM)

DM*VOL

DM*MOM

DM*ILLIQ

-0.249 -2.51** -0.255 -1.67* 0.161 0.77 -0.197 -0.70 -0.207 -0.65 -0.502 -1.37

0.012 0.76 0.009 0.89 0.047 1.26 0.185 1.57 0.193 1.61 1.043 1.33

-1.267 -3.63*** -1.021 -2.15** -1.774 -3.14*** -0.099 -1.87* -0.085 -1.63 -0.072 -1.41

0.006 2.63** 0.010 3.15*** 0.015 3.54*** 0.022 3.26*** 0.030 4.49*** 0.038 5.31***

0.004 0.85 0.002 0.33 0.002 0.23 0.014 1.09 0.007 0.44 -0.012 -0.68

-0.364 -1.23 -0.337 -0.74 0.162 0.26 -0.507 -0.61 -0.582 -0.61 0.181 0.17

-0.011 -1.68* -0.138 2.01 -0.074 1.24 -0.062 1.46 -0.993 -2.14** -0.105 -1.27

0.035 1.33 0.188 1.47 0.489 1.22 1.482 2.13** 1.014 2.26** -0.752 -1.59

0.001 0.34 0.002 0.95 0.001 0.87 0.021 1.55 0.048 1.43 0.051 1.79*

1.174 2.57*** 0.263 3.24*** 0.158 1.68* 0.177 1.12 0.189 1.47 0.096 1.22

0.005 2.48** 0.009 2.85*** 0.012 2.95*** 0.017 3.15*** 0.023 3.20*** 0.033 4.07***

0.004 0.45 0.002 0.12 0.007 0.47 0.020 0.90 0.016 0.55 0.027 0.84

-0.213 -0.52 -0.490 -0.75 -0.983 -1.22 -0.911 -0.83 -2.243 -1.58 -2.959 -1.84*

-0.004 -1.43 -0.007 -1.81* -0.028 -2.07** -0.049 -2.51** -0.047 -1.87* -0.077 -2.80***

0.069 1.36 -0.153 -1.87* -0.384 -2.71*** 0.041 1.13 -0.023 -0.45 1.279 2.28**

0.025 0.18 0.470 2.13** 0.765 2.81*** 0.794 2.14** 0.709 1.48 0.785 1.45

publication of the news. The results are again consistent with Diamond and Verrecchia's (1987) predictions.

4.4. Time-series regression to examine post-event pricing behavior Table 6 reports the time-series regression of cumulative abnormal returns CAR on the announcement day abnormal returns AR0 to examine the stocks' efficiency in incorporating public information. The dependent variable is CAR for various specifications of the event windows—namely, (1,1), (1,5), ( 1,10), (1,20), (1,30), (1,40), (1,50) and (1,60)—and we again classify our samples based on the shortability of the stocks and the signs of the earnings announcements: Panels A to D report the regression results on shortable stocks with bad news, nonshortable stocks with bad news, shortable stocks with good news, and non-shortable stocks with bad news, respectively. Panel A of Table 6 shows that, for all the dependent variables with different event windows, all the coefficients of AR0 are statistically insignificant, with only one exception: when the dependent variable is CAR1,10, the coefficient is marginally significant at the 90% level. Such results suggest that the public earnings information can be fully incorporated into the prices of shortable stocks on the announcement day, with returns that show no continuation or reversal following the announcements. Hence, when bad news, such as disappointing earnings news, is made public, shortable stocks can fully digest such information in one day. In Panel B, however, the coefficients of AR0 are significantly negative in regressions with dependent variables CAR1,1, CAR1,5, and CAR1,10. This result indicates that the prices of non-shortable stocks reverse in the next 1-day, 5-day, and 10-day event windows following the publication of bad earnings news, suggesting an overreaction of non-shortable stocks toward negative public information on the announcement days. The coefficients of AR0 from other specifications are insignificant, implying that the overreaction is corrected within 10 days following an announcement. In Panels C and D, none of the coefficients of AR0 are statistically significant, suggesting that when companies announce

R2 2.41% 2.68% 2.92% 6.15% 9.36% 12.25%

0.53% 0.48% 1.56% 3.02% 2.97% 5.64%

good earnings information, both shortable and non-shortable stocks react quickly and effectively to such news: the prices of all the stocks fully incorporate the information on the event day, showing no price continuation or reversal following the announcement. Given the findings on the differences between shortable and nonshortable stocks in their post-event pricing behavior, particularly after bad earnings announcements, we conduct the following regression analysis to provide more statistical inferences on the differences we observe: CARðt 1 ;t 2 Þ ¼ α þ β1 AR0 þ β2 DM  AR0 þ γX þ δDM  X þ μ

ð6Þ

where we add the dummy variable DM and its interaction terms with the control variables to the regression model (4) and all the other variables are defined as before. In this regression framework, β1 and β1 + β2 capture the post-event reactions of non-shortable and shortable stocks, respectively. Hence, the statistics of β2 provide us with a clear inference on whether shortable and non-shortable stocks have significant differences in their post-event reactions to earnings announcements. The results are presented in Table 7, where, we again separate the results from the good news sample from those from the bad news sample. Panel A of Table 7 shows the regression results for the bad news sample. We focus on the coefficients of the interaction term DM * AR0, that is, β2s. For all six specifications, β2 is significantly positive, suggesting that shortable and non-shortable stocks have significantly different pricing patterns following the publication of bad earnings information. The negative and significant coefficients of AR0 , that is, β1, when the dependent variables are CAR1,1, CAR1,5, and CAR1,10 again indicate that nonshortable stocks experience a return reversal up to about 10 days following the earnings announcements. The significant but positive β2 values suggest that shortable stocks do not show as much of a reversal as nonshortable stocks. Actually, the absolute values of β2 are larger than those of β1, indicating that there is actually no reversal for shortable stocks at all, which is consistent with our previous findings. Combining this with the results from Table 6, we can conclude that when bad news is made public, shortable and non-shortable stocks react significantly differently

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

10

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

Table 8 Speed of price adjustment for Non-shortable stocks following Bad news. This table reports the coefficient estimates of the regression CARðt 1 ;t2 Þ = α + βAR0 + γ'X + μ, where CARðt1 ;t2 Þ is the cumulative abnormal return over a period starting t1 trading days and ending t2 trading days around the event day; DM is a dummy variable, with DM = 1 indicating shortable stocks and DM = 0 indicating non-shortable stocks; and X is a vector of control variables, including the logarithm of size (log(MV)), the logarithm of the book-to-market ratio (log(BM)), trading volumes (VOL) measured as the turnover ratio, momentum (MOM), and the Amihud illiquidity ratio (ILLIQ). Parameter estimates are computed using the WLS approach, where the weights are set so that each cross section has equal weight. The t-statistics (in italics) are calculated using clustered standard errors. The symbols *, **, and *** denote statistical significance at the 10%, 5%, and 1% levels, respectively. Dependent

Intercept

AR0

Log(MV)

Log(BM)

VOL

MOM

ILLIQ

R2

CAR1,1

0.023 1.91* 0.017 1.08 0.041 2.40** 0.041 2.13** 0.068 3.24*** 0.052 2.31** 0.064 2.66*** 0.083 3.29*** 0.094 3.53*** 0.096 3.45*** 0.093 3.31*** 0.113 3.87*** 0.120 3.93*** 0.165 5.12*** 0.182 5.47*** 0.184 5.36*** 0.171 4.76*** 0.201 5.46*** 0.206 5.49*** 0.206 5.42***

−0.162 −4.14*** −0.170 −3.40*** −0.190 −3.49*** −0.187 −3.05*** −0.208 −3.12*** −0.351 −4.90*** −0.322 −4.24*** −0.339 −4.21*** −0.256 −3.02*** −0.251 −2.83*** −0.218 −2.44** −0.195 −2.10** −0.138 −1.42 −0.074 −0.73 −0.068 −0.65 −0.061 −0.56 −0.161 −1.42 −0.088 −0.75 −0.119 −1.00 −0.367 −1.24

−0.006 −2.86*** −0.005 −1.87* −0.010 −3.27*** −0.010 −2.90** −0.015 −3.22*** −0.015 −3.76*** −0.016 −3.95*** −0.021 −4.67*** −0.022 −4.69*** −0.022 −4.58*** −0.021 −4.32*** −0.024 −4.77*** −0.025 −4.67*** −0.032 −4.67*** −0.035 −5.01*** −0.035 −4.91*** −0.034 −4.47*** −0.038 −4.97*** −0.040 −5.11*** −0.040 −5.11***

0.004 2.05** 0.002 0.75 0.003 1.21 0.006 1.96** 0.007 2.08** 0.008 2.21** 0.009 2.39** 0.009 2.29** 0.009 2.10** 0.011 2.46** 0.014 2.97*** 0.016 2.40** 0.020 2.95*** 0.019 2.56** 0.022 3.02*** 0.026 3.71*** 0.029 4.06*** 0.030 4.03*** 0.029 3.84*** 0.032 4.11***

−0.255 −2.84*** −0.177 −1.54 −0.237 −1.90** −0.361 −2.56** −0.256 −1.68* −0.412 −2.50** −0.441 −2.52** −0.685 −3.70*** 0.085 0.44 0.154 0.75 0.004 0.02 −0.089 −0.42 0.008 0.03 0.039 0.16 −0.002 −0.01 0.055 0.22 −0.104 −0.40 −0.024 −0.09 −0.048 −0.18 −0.193 −0.69

0.002 0.86 −0.010 −0.41 0.017 0.25 −0.001 −0.04 0.004 1.64 0.056 0.75 0.054 0.47 0.088 0.92 0.046 0.79 0.015 1.73 −0.176 −0.69 0.062 1.13 −0.311 −1.46 −0.108 −0.78 −0.292 −1.62 −0.184 −1.17 −0.263 −1.67* −0.225 −1.88* −0.246 −1.53 0.019 1.89*

−0.174 −1.42 −0.183 −1.51 −0.279 −1.17 −0.244 −1.59 −0.253 −1.02 −0.221 −1.59 −0.163 −0.87 −0.203 −1.13 −0.199 −1.02 −0.185 −0.95 −0.176 −0.83 −0.177 −1.14 −0.165 −0.59 −0.160 −1.33 −0.152 −1.31 −0.136 −0.97 −0.141 −1.18 −0.139 −0.69 −0.128 −0.73 −0.123 −0.64

3.55%

CAR1,2 CAR1,3 CAR1,4 CAR1,5 CAR1,6 CAR1,7 CAR1,8 CAR1,9 CAR1,10 CAR1,11 CAR1,12 CAR1,13 CAR1,14 CAR1,15 CAR1,16 CAR1,17 CAR1,18 CAR1,19 CAR1,20

to the news. While shortable stocks tend to react to the news relatively more efficiently, without no return continuation or reversal following the announcements, non-shortable stocks seem to overreact to the information on the announcement dates, which is indicated by a significant reversal around 10 days following the announcements. In contrast, Panel B of Table 7 shows that, for the good news sample, all the β2 values are statistically insignificantly different from zero. This result implies that when good news is made public, shortable and nonshortable stocks tend to behave quite similarly up to 40 days following the announcement dates. Combining these findings with the results from Table 6, we can conclude that there is no difference in the postevent reactions of both types of stocks: there is no return continuation or reversal from either type of stock. 4.5. Speed of price adjustment for non-shortable stocks following bad news Our evidence so far indicates the efficiency in both shortable and nonshortable stocks reacting to good earnings announcements and shortable stocks reacting to bad earnings announcements. However, for shortable stocks with bad earnings announcements, since prices reverse following the announcements, this suggests that a market with short selling constraints tends to overreact to the bad news on the announcement day and it takes time for the price to fully adjust to the correct level. To investigate the speed of such a price adjustment, we further refine our test in

1.35% 2.45% 2.45% 3.33% 4.45% 4.22% 5.31% 3.59% 3.94% 3.42% 3.99% 4.06% 4.57% 5.32% 5.83% 5.95% 6.27% 6.31% 7.03%

the above session by changing the event windows of our regressors, CAR. Since the previous results from Table 6 indicate that the prices of non-shortable stocks reverse around about 10 days following the announcement day, we further define our dependent variables as CAR(1, t2), where t2 changes from one to 20. The regression results are presented in Table 8. We can see that while all the coefficients of AR0 are negative, they become statistically insignificant after day 12. Before that day, all the coefficients are negative and statistically significant. For example, when CAR(1, 1) is regressed on AR0, the coefficient is −0.162, with t-statistics of −4.14. This finding implies that, following significant drops on the event day, the prices of non-shortable stocks reverse the next day and the higher the price drop on day 0, the higher the price rises on day 1. Such price reversal continues until day 12, though the statistical significance gradually declines. After day 12, the CAR values are no longer related to AR0. Hence, it takes about 12 days for the information (bad news) to be correctly and fully incorporated into the prices of non-shortable stocks. 4.6. Robustness checks We sign the contents of the earnings announcements based on the sign and magnitude of the abnormal returns of the stocks on the announcement days. However, it could be problematic if the abnormal returns on the announcement days do not reflect the “true” response

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

of the market. For example, a large negative return on announcement days may not indicate negative news if it reverses later. Similarly, a zero abnormal return may not indicate neutral news if there are subsequent positive or negative returns. As we determined earlier, nonshortable stocks could overreact to bad news on announcement days. Hence, using announcement day returns to classify news could introduce bias. We therefore vary our ways of signing the news. We first classify the news based on the cumulative abnormal returns of the stocks in the event windows (0, 5), (0, 10), (0, 30), and (0, 60). We also use the SUE score obtained from the I/B/E/S database to sign the news. The results are qualitatively unchanged and sometimes even stronger. We also alter our ways of defining the abnormal returns by changing the event windows in calculating stocks' β coefficients. This basically does not change our results. Our sample period runs from 1994 until the end of 2011, which covers two financial crises, namely, the 1997 Asian financial crisis and the 2007–2008 financial crisis. To look at how our study is affected by these two turbulent periods, we remove 1997, 1998, 2007, and 2008 from our sample. The results are, again, qualitatively the same. Therefore, our conclusions are very robust. 5. Conclusions This study looks at how short sales constraints impact price adjustment to information. In particular, we look at how stocks with short sales constraints react differently to earnings announcements from stocks without such constraints. Our empirical tests show that, consistent with the prediction from a theoretical framework of Diamond and Verrecchia (1987), shortable stocks adjust to private bad information faster than non-shortable stocks, causing the non-shortable stocks to be relatively overpriced before the publication of the private information. However, non-shortable stocks react more strongly to the publication of bad information than shortable stocks. Hence, part of the stronger reaction from non-shortable stocks could be correction of the overpricing. We further find that the prices of non-shortable stocks reverse following the announcement of negative information, suggesting that investors may overreact to negative information on announcement days. However, such evidence is absent among shortable stocks, showing that stocks without short sales constraints react to information quickly and efficiently. Lastly, our test indicates that it takes about 12 days for the prices of non-shortable stocks to fully adjust to negative earnings information. Furthermore, our results are robust when we vary our ways of signing earnings announcements, when we alter our calculation of abnormal returns, and when we remove the crisis period from our sample. Our findings contribute to a better understanding of the effect of short sales constraints on market informational efficiencies and have important implications for regulators, practitioners, and academic theorists.

January 1994. Under the pilot scheme, 17 securities could be short sold and a short sale could not be made below the best current ask price (the so-called “tick rule”). The scheme was revised in March 1996 with the number of designated securities for short selling increased and the tick rule abolished. The tick rule was reinstated on 7 September 1998 upon changes in market conditions. The number of Designated Securities for short selling is revised on a quarterly basis. (A full list of Designated Securities is posted under “Designated Securities Eligible for Short Selling”). The selection criteria for Designated Securities are as follows: 1. all constituent stocks of indices which are the underlying indices of equity index products traded on the Exchange; 2. all constituent stocks of indices which are the underlying indices of equity index products traded on HKFE; 2. all underlying stocks of stock options traded on the Exchange; 4. all underlying stocks of Stock Futures Contracts (as defined in the rules, regulations and procedures of HKFE) traded on HKFE; 5. stocks eligible for structured product issuance pursuant to Rule 15A.35 of the Main Board Listing Rules or underlying stocks of Structured Product traded on the Exchange; 6. stocks with market capitalization of not less than HK$3 billion and an aggregate turnover during the preceding 12 months to market capitalization ratio of not less than 50%; 7. Exchange Traded Funds approved by the Board in consultation with the Commission; 8. all securities traded under the Pilot Program; 9. stocks that have been listed on the Exchange for not more than 60 trading days, with a public float capitalization of not less than HK$10 billion for a period of 20 consecutive trading days commencing from the date of their listing on the Exchange and an aggregate turnover of not less than HK$200 million during such period; 10. all underlying stocks of Structured Product which is based on one single class of shares traded on the Exchange; and 11. applicable Market Making Securities (other than the securities described in categories 7. and 8. above) approved by the Board in consultation with the Commission. Appendix B. Definition and descriptive statistics of control variables

Variable

Definition

Mean

Median

Min.

Max.

Std. Dev

Log(MV)

Logarithm transformation of the market capitalization Logarithm transformation of the book-tomarket ratio Trading volume, measured as the average turnover ratio of the stocks from day −60 to day −31 before the earnings announcements The past 11 month average return with one month lag to control for the momentum effect Average Amihud illiquidity ratio measured from 3 months before the earnings announcements

6.325

5.780

1.649

14.297

2.108

−0.201

−0.068

−8.372

3.219

1.136

0.007

0.002

0.000

0.581

0.021

−0.034

0.000

−3.217

4.156

0.812

4.837

0.172

−4.940

1970.960

42.470

Log(BM)

Acknowledgment We would like to thank Eric C. Chang for kindly providing us with the data on the short selling regulations of the Hong Kong Stock Exchange. We are grateful to the valuable comments from two anonymous referees. We thank Alireza Tourani-Rad and other seminar participants at Auckland University of Technology, and appreciate the helpful comments from the discussant and participants at the 26th AFBC annual conference and 2014 New Zealand Finance Colloquium. Appendix A. Regulation on the implementation of short selling in the HKSE5 In line with reform of the securities borrowing and lending regime, the Exchange introduced a pilot scheme for regulated short selling in 5

From https://www.hkex.com.hk/eng/market/sec_tradinfo/regshortsell.htm.

11

VOL

MOM

ILLIQ

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006

12

M. Bai, Y. Qin / International Review of Financial Analysis xxx (2015) xxx–xxx

References Bai, M., & Qin, Y. (2014). Short-sales constraints and liquidity change: Cross-sectional evidence from the Hong Kong market. Pacific-Basin Finance Journal, 26, 98–122. Beber, A., & Pagano, M. (2013). Short-selling bans around the world: Evidence from the 2007–09 crisis. Journal of Finance, 68(1), 343–381. Berkman, H., Dimitrov, V., Jain, P. C., Koch, P. D., & Tice, S. (2009). Sell on the news: Differences of opinion, short-sales constraints, and returns around earnings announcements. Journal of Financial Economics, 92, 376–399. Berkman, H., & McKenzie, M. D. (2012). Earnings announcements: Good news for institutional investors and short sellers. Financial Review, 47(1), 91–113. Bris, A., Goetzmann, W. N., & Zhu, N. (2007). Efficiency and the bear: Short sales and markets around the world. Journal of Finance, 62(3), 1029–1079. Chang, E. C., Cheng, J. W., & Yu, Y. (2007). Short-sales constraints and price discovery: Evidence from the Hong Kong market. Journal of Finance, 62(5), 2097–2121. Chen, C. X., & Rhee, S. G. (2010). Short sales and speed of price adjustment: Evidence from the Hong Kong stock market. Journal of Banking & Finance, 34, 471–483. Christophe, S. E., Ferri, M. G., & Angel, J. J. (2004). Short-selling prior to earnings announcements. Journal of Finance, 59(4), 1845–1876. Diamond, D. W., & Verrecchia, R. E. (1987). Constraints on short-selling and asset price adjustment to private information. Journal of Financial Economics, 18(2), 277–311.

Engelberg, J. E., Reed, A. V., & Ringgenberg, M. C. (2012). How are shorts informed? Journal of Financial Economics, 105(2), 260–278. Fung, J., & Draper, P. (1999). Mispricing of index futures contracts and short sales constraints. Journal of Futures Market, 19, 695–715. Jagadeesh, N., & Kim, W. (2010). Do analysts herd? An analysis of recommendations and market reactions. Review of Financial Studies, 23(2), 901–937. Jagadeesh, N., & Livnat, J. (2006). Post-earnings-announcement drift: The role of revenue surprises. Financial Analyst Journal, 62(2), 22–34. Mashruwala, C., & Mashruwala, S. (2014). Is there a “torpedo effect” in earnings announcement returns? The role of short-sales constraints and investor disagreement. Journal of Accounting, Auditing & Finance, 29(4), 519–546. Michaely, R., Rubin, A., & Vedrashko, A. (2014). Corporate governance and the timing of earnings announcements. Review of Finance, 18, 2003–2044. Miller, E. M. (1977). Risk, uncertainty, and divergence of opinion. Journal of Finance, 32, 1151–1168. Reed, A. V. (2007). Costly short selling and stock price adjustment to earnings announcements. Unpublished working paper. Savor, P. G. (2012). Stock return after major price shocks: The impact of information. Journal of Financial Economics, 106(3), 447–684.

Please cite this article as: Bai, M., & Qin, Y., Short sales constraints and price adjustments to earnings announcements: Evidence from the Hong Kong market, International Review of Financial Analysis (2015), http://dx.doi.org/10.1016/j.irfa.2015.08.006