Anatomy of a Government Interventon n Index Stocks Prce Pressure or Informaton Effects? by Karan Bhanot and Palan-Rajan Kadapakkam 1 Forthcomng n Journal of Busness Ths Verson: May 17, 2004 Keywords: Government Interventon, Sgnal JEL classfcaton: G1, G3 1 Department of Fnance, College of Busness Admnstraton, The Unversty of Texas at San Antono. Emal kbhanot@utsa.edu Tel. (210)-458-7429 Fax (210)-458-6320. We are thankful to Utpal Bhattacharya, Davd Brown, Venkat Eleswarapu, Lalatendu Msra, M. P. Narayanan, Chrs Neely, Sarab Seth, Lla Truett and the referee for helpful suggestons and comments.
Anatomy of a Government Interventon n Index Stocks Prce Pressure or Informaton Effects? Abstract In a massve nterventon desgned to deter speculators, the Hong Kong Monetary Authorty (HKMA) bought Hang Seng ndex stocks over the perod August 14 to August 28, 1998. Our objectve s to document the mpact of nterventon and determne whether the relatve prce changes n Hang Seng stocks are a result of nformaton effects or due to lqudty-based prce pressure effects from nterventon. An equally weghted portfolo of Hang Seng stocks provdes an abnormal return of approxmately twenty-four percent durng the nterventon perod. The abnormal returns are not reversed over the next eght weeks, refutng the hypothess that returns are due to temporary lqudty effects. Usng daly data on Hang Seng stocks and ndvdual stock nterventon amounts, we fnd evdence consstent wth nformaton effects. Evdence from dual class shares also suggests that abnormal returns are due to a credble sgnal of an mplct put opton granted by the government to holders of Hang Seng stocks. 2
Anatomy of a Government Interventon n Index Stocks Prce Pressure or Informaton Effects? Offcal government nterventon n the foregn exchange market and the money market s common practce, but nterventon n the stock market s relatvely rare. Whle there s no reported nstance n Western economes, several Asan governments have ntervened n the stock market n the last few years wth the Japanese government beng the most recent to announce such plans. 1 To our knowledge there s no emprcal study that has documented the mpact of these stock market nterventons. 2 Among the Asan nstances, the nterventon n Hong Kong durng August 1998 stands out by ts sheer magntude. The government purchases accounted for more than seventy-fve percent of market tradng volume durng the nterventon perod. In ths paper, our objectve s to document the mpact of the Hong Kong nterventon on the returns of targeted stocks, and assess whether ths mpact s due to temporary lqudty-based prce pressure effects or nformaton effects assocated wth a credble sgnal from the government. The Hong Kong Monetary Authorty (hereafter HKMA) ntervened n the stock market over the perod August 14 to August 28, 1998. The nterventon operaton nvolved purchases of 33 large captalzaton stocks that comprsed the Hang Seng Index. The stated objectve of HKMA was to counteract the coordnated actvty of speculators 1 The Japanese government announced a plan on February 9, 2002, to provde two trllon yen to a fund that wll buy shares from ts banks. Thaland announced plans to ntervene n 2002, followng an nterventon n 1999. South Korea and Tawan have also ntervened n ther stock markets n the last three years. Tawan has a corpus of funds called the Natonal Stablzaton Fund used to shore up stock prces. Other governments have resorted to nterventon through ntermedares lke government run funds and fnancal nsttutons. 2 Mller, Weller and Zhang (2002) explore a theory of nterventon n the context of the U.S. stock market.
who had taken large short postons n Hang Seng stocks and ndex futures as part of ther attacks aganst the Hong Kong dollar as explaned n Secton I. The Hang Seng stock market ndex declned by 30% n the month precedng nterventon. In response, over ten tradng days, the monetary authorty purchased HK$118 bllon worth of shares n the 33 Hang Seng stocks. Ths amount was very large relatve to the typcal total daly tradng volume n the Hong Kong market of HK$6 bllon durng 1998. At the end of the nterventon perod, the HKMA was a major shareholder n each of the Hang Seng Stocks. In some cases, the ndvdual stock holdng exceeded ten percent of the total outstandng shares n the corporaton. The government dd not lqudate any shares before the end of 1998. Gven the selectve nterventon n Hang Seng stocks, the frst ssue we consder s whether there was a prce mpact on these stocks relatve to other Hong Kong stocks that dd not experence nterventon. 3 We fnd that an equally weghted portfolo of Hang Seng stocks yelded an excess return of 24% over the ten-day nterventon perod relatve to an equally weghted ndex of all stocks. The assessment s robust to the use of alternatve control portfolos. Ths result ndcates that stockholders n targeted frms experenced a szable wealth gan. The next queston s the reason for these szable gans. We consder three possble explanatons: temporary prce pressure effects, permanent prce pressure effects, and nformaton effects arsng from a credble government sgnal to defend a floor for 3 The prce effect of sudden changes n demand s an ssue that s prevously analyzed n the context of ndex addtons. The nterventon sze here s much larger than changes n supply/demand trggered by ndex addtons. Studes on ndex addtons provde evdence about the slope of the demand curve that n turn s relevant for understandng the mpact of corporate events such as stock repurchases and equty ssuances (see, e.g., Bagwell (1992) and Loderer, Cooney and Drunen (1991)). 2
the Hang Seng ndex. Many crtcs contend that nterventon s futle snce t artfcally nflates market prces due to lqudty effects, and prces eventually slde down to fundamental values. 4 Our analyss fals to fnd evdence of such temporary prce pressure effects. The abnormal nterventon perod returns are not reversed over subsequent ntervals. Furthermore, returns over post-nterventon perods are not negatvely correlated wth nterventon perod returns. Another possble explanaton for postve nterventon perod returns s a permanent prce pressure effect (reduced supply coupled wth a downward slopng demand curve) arsng from the Hong Kong government s decson to hold the acqured shares for an extended perod of tme. Under ths explanaton, excess returns are expected to be hgher n stocks experencng a larger nterventon. 5 A fnal explanaton of the excess returns s nformaton effects assocated wth nterventon. The stock market nterventon by the authorty to defend the Hang Seng ndex value provdes a costly and, therefore, credble sgnal of the serousness of stated exchange rate polces and ntentons. Mller, Weller and Zhang (2002) provde a theoretcal model wheren the possblty of nterventon provdes a perceved put opton to market partcpants. The HKMA beleved that coordnated attacks by speculators had drven stock prces below far values, and t ntervened n the stock market to defend stock prces and releve pressure on exchange rates. Ths objectve was communcated wdely and reterated after the nterventon through press statements and ntervews of 4 See text of Mlton Fredman s ntervew on ths subject publshed n the Wall Street Journal (September 3, 1998). 5 Consequently, there s a potental for favortsm when the government selects stocks to emphasze n the nterventon. Meda artcles suggested that the nterventon greatly benefted wealthy ndvduals who controlled a few of the targeted frms. These ndvduals were reported to have a close relatonshp wth government offcals. 3
HKMA executves. For example, n an ntervew n Asa Week (dated October 23, 1998) n the aftermath of the nterventon, the Fnance Secretary stated that f there were to be another speculatve attack, the HKMA would lkely ntervene agan. The avalablty of a large corpus of funds n the Exchange Fund (HK$ 718 bllon) for further nterventons provded sgnfcant credblty to these statements. We do not fnd evdence of a permanent prce pressure effect when we relate total stock-wse nterventon measures to overall nterventon perod returns (from 8/14 to 8/28). Analyss of daly returns and proxes for daly nterventon reveals that the daly return on an ndvdual Hang Seng ndex stock durng the nterventon perod s drven by the average nterventon across all the ndex stocks rather than the amount of nterventon n that specfc stock. Ths evdence provdes support for nformaton effects of the nterventon rather than prce pressure effects. The nformaton hypothess s bolstered by a case study of Swre Pacfc stock. Swre Pacfc has two classes of shares A and B. The two securtes are near perfect substtutes except that the government bought class A shares and not class B shares. Under the nformaton hypothess, the government s decson to defend the value of class A stock provdes a valuable sgnal to the stockholders of class A stock. The evdence shows that class A and class B experenced szeable gans, but, class A shares outperformed class B shares by a total of 6.7% over ten days. We argue that the addtonal return reflects the greater value of the sgnal to Class A shareholders. The remander of the paper s organzed as follows. Secton I of the paper detals the economc background n whch the government made the decson to ntervene. Secton II gves the theoretcal background and outlnes the three hypotheses. Secton III 4
descrbes the data and methodology, whle Secton IV provdes evdence on the effects of nterventon and some polcy mplcatons. Secton V presents concludng comments. I. Background on Speculatve Actvtes and the Economc Envronment Hong Kong follows a monetary system (called a Currency Board system) that provdes for a fxed exchange rate lnked to the Unted States dollar (see Appendx for more detals). Pror to the nterventon perod, large hedge funds had bet that f the HKMA supported the Hong Kong dollar by lettng nterest rates ncrease, then the Hong Kong stock market would fall n reacton. These speculators took large short postons n Hang Seng stocks and futures whle at the same tme sellng the Hong Kong Dollar (referred to as a double play ). The sales of Hong Kong dollars by these funds were desgned to drve up nterest rates. The monetary authorty was aware of the speculatve actvtes taken by hedge funds. The HKMA beleved that the economc fundamentals of Hong Kong were strong and dd not warrant a dramatc realgnment of the exchange rate or an adjustment of equty prces. 6 The HKMA decded that at ths stage an nterventon n the Hang Seng stocks would provde a credble deterrent to speculators n Hang Seng stocks. In a surprse move, the HKMA bought Hang Seng stocks over the perod August 14 to August 28, 1998, and announced these actons after the frst day of nterventon. The market was 6 See text of the speech by HKMA chef on October 26, 1998, avalable on the web ste http://www.nfo.gov.hk/hkma. The HKMA announced after the nterventon that speculators had bet HK $30 bllon aganst Hong markets, and ther postons would have yelded profts of HK$4 bllon for every 1000-pont drop n the Hang Seng ndex. Government offcals clamed that the nterventon had nflcted consderable losses on the speculators. 5
therefore aware of the nterventon n the ndex stocks, but the quantum of stock-wse nterventon was not made publc mmedately. However, gven the magntude of the nterventon, daly volume numbers provded ample clues about the nterventon actvty. Even f the volume of nterventon s not drectly observed by market partcpants, speculators can extract some nformaton about the volume of nterventon and so the target of authortes from the behavor of prces (see for example Bhattacharya and Weller (1997)). Detals on the exact magntude of the stock-wse nterventons were offcally released on October 26, 1998, although the government s holdng n the Hong Kong and Shangha Bankng Corporaton were dsclosed as early as September 1, 1998. The shares were held as a part of the assets of the Exchange Fund. The government ntenton was to sell these shares when the stock market stablzed. In fact, the HKMA holdngs show that these stock purchases were held for well over a year. From an external economc envronment standpont, most Asan economes wtnessed poor macroeconomc performance durng 1998. The pre and post-nterventon wndow does not reveal any unusual returns n Hong Kong relatve to the other markets n general. Even though most of the economes experenced slow growth and deprecaton n exchange rates, there was no general deprecaton n currences durng the event. Durng the event wndow there was an apprecaton n nterest rates n Hong Kong (overnght rates between 8.25% and 19.25%) consstent wth severe pressures on the Hong Kong dollar. The overnght nterest rates declned to around 5% wthn two days after the nterventon. 6
II. Theoretcal Background and Hypotheses Ths secton provdes a theoretcal background for the emprcal tests. Followng Wurgler and Zhuravskaya (2002), consder a model for the demand and supply of a stock (Fgure 1 (a)). The ntersecton of the total demand curve ( TD ) and total supply curve ( TS ) of a stock gves the equlbrum prce (P) and quantty traded (Q). The dfference between total demand and total supply at a gven prce s termed the excess demand. Therefore, by defnton, excess demand s zero at the equlbrum prce. We consder four types of market partcpants: regular nvestors, arbtrageurs, the government and speculators. The excess demand of a regular nvestor depends on hs belef about the fundamental value of an asset. Ths class of nvestors has heterogeneous belefs about the value of an asset. In contrast, arbtrageurs have correct and homogenous belefs about the value of an asset but are subject to a zero net nvestment constrant. In general, these two classes of nvestors absorb demand and supply shocks. A postve demand shock because of government nterventon causes the demand curve to shft upwards (e.g., Fgure 1(b)), whle a negatve supply shock from speculators shfts the supply curve to the left (e.g., Fgure 1(c)). The demand shock s the demand from government nterventon net of the supply from speculators. Wurgler and Zhuravskaya (2002) show that the return due to a demand shock (D) s approxmated by: R D (1) 1 N + h ka where R s the expected return on a stock, h s a parameter that ndexes the heterogenety of belefs of regular nvestors, k s a parameter of rsk averson of arbtrageurs, A s the 7
mnmzed varance of the arbtrageur s portfolo wth a zero net nvestment constrant, and N s the number of arbtrageurs. They pont out that demand shocks nduce hgher returns, the greater the heterogenety of belefs (larger h). Also, a better possblty for arbtrage (a lower value for the varance parameter A) causes a smaller response to a demand shock. We now dscuss the three prmary sources of nfluence from nterventon: two prce pressure channels (temporary and permanent) and a sgnalng channel (these are collected n Table I for convenence). Hgher nterventon results n a larger net demand shock (parameter D n equaton (1)) and a correspondng larger return. Cross-sectonally the prce-pressure effect wll nduce a postve correlaton between return and nterventon after controllng for the presence of arbtrageurs and heterogenety of belefs among regular nvestors. If the demand shock does not contan any new nformaton about the stock, the excess demand curve for a stock should revert back to ts equlbrum poston and the observed prce pressure effects would be temporary. Kaul, Mehrotra and Morck (2000) argue that reductons n the float of stocks nduce a permanent prce pressure effect (leftward shft n the supply curve (Fgure 1 (c)). 7 The HK government ntenton was to hold the stocks acqured n the nterventon tll the market had stablzed. Due to the uncertanty about the duraton of the government ownershp, traders wll be dscouraged 7 Several other studes have examned prce pressure effects n the context of addtons and deleton from equty ndexes. For example, Harrs and Gurel (1986) fnd sgnfcant postve abnormal returns for addtons and negatve abnormal returns for deletons. They nterpret these results as prce pressure effects. Lynch and Mendenhall (1997) fnd that returns are partally reversed. Other recent studes that examne ths ssue nclude Benesh and Whaley (1996) and Wurgler and Zhuravskaya (2002). 8
from arbtragng away ths prce ncrease. Under the permanent pressure hypothess, the prce ncrease s not reversed mmedately after the nterventon perod. The stated objectve of the government was to deter a speculatve attack. If government nterventon provdes new nformaton (a sgnal) to nvestors, stock prces must adjust as a result of a reassessment of the rsk and return tradeoffs. 8 The HKMA stock nterventon s restrcted to the Hang Seng stocks because most speculators use these assets for ther actvtes. Hence, any sgnalng nfluences are restrcted only to Hang Seng stocks, and manfested n each of them. If ths deterrence s effectve, the post-nterventon perod abnormal returns should be zero rather than negatve as n the case of the temporary pressure hypothess. If the nterventon provdes a sgnal common to all Hang Seng ndex stocks, the return on any Hang Seng stock wll be related to the overall nterventon rather than nterventon n that specfc stock. Thus, n contrast to the permanent prce pressure hypothess we do not expect a sgnfcant relatonshp between abnormal return and stock specfc nterventon actvty. 9 Outlne of tests The purpose of ths paper s to try to sort out the three prce effects for postve nterventon perod returns lsted above (collected n Table I). If the prce pressure effect were transent (temporary prce pressure), t would reverse tself as soon as nterventon stops. Ths reversal mples that abnormal returns after the nterventon wll be negatve. Furthermore, the magntude of the cumulatve abnormal returns on a stock n the post- 8 In effect, nterventon to support Hang Seng stocks provdes a valuable mplct put opton provded to stock holders (as n Mller, Weller and Zhang (2002)). 9 If the sgnal s stock specfc there wll be a postve relatonshp between abnormal returns and nterventon actvty n that stock. Emprcally, ths wll be ndstngushable from prce-pressure effects. 9
nterventon perod should be negatvely related to that stock s cumulatve abnormal returns durng nterventon (row 1 of Table I). Both permanent prce pressure and sgnal effects mply postve returns that are not reversed. To dfferentate between these effects we examne the senstvty of stock returns to nterventon actvty by estmatng the followng relatonshp: AR = β 0 + β1 PSB + β 2 Slope + β 3 Sg + β 4 Indxwt + β 5 ProrRet + ε (2) where AR, s the abnormal return on stock on day t and ε consttutes other frm t specfc shocks. Our prmary metrc of nterventon actvty, PSB (a proxy for D), computes the rato of dollar nterventon n each stock to the market value of the stock at the end of July 1998. A postve coeffcent for PSB provdes support for prce pressure effects. Followng equaton (1), we control for the heterogenety of belefs of regular nvestors (h) and the varance of the arbtrageurs portfolo (A). A natural proxy for h, the slope of the excess demand curve, s obtaned by aggregatng bd and ask quotes n the electronc lmt order book (denoted Slope and descrbed n more detal later). We expect a negatve coeffcent on Slope because a steeper excess demand curve wll result n a larger prce reacton (note that the Slope varable s negatve). Wurgler and Zhuravskaya (2002) argue that the non-systematc rsk of a stock (denoted Sg ) s a convenent proxy for A n equaton (1). We expect a postve coeffcent on Sg because hgher non-systematc rsk dmnshes the ablty to arbtrage away prce pressure effects. 10
The net demand shock D s determned by the demand from nterventon ( PSB ) mnus the supply from speculators. Speculators, wth short postons n ndex futures, are more concerned about stocks wth a greater mpact on the ndex. Thus, the weght of a frm n the Hang Seng ndex ( Indxwt ) should be a good proxy for sellng by speculators. We would expect a negatve sgn on the Indxwt varable, snce greater sellng by speculators wll lead to lower returns. Fnally, t s possble that speculatve actvty may have artfcally depressed a stock s prce pror to the nterventon perod. The onset of government nterventon may correct ths under-valuaton resultng n postve returns durng the nterventon perod. Under ths explanaton, nterventon perod returns should be negatvely related to PorRet, the abnormal return durng the sx weeks before nterventon. We select ths perod because HKMA authortes ndcated that they suspected speculatve attacks at the begnnng of July 1998. III. Data A. Prce data The stock market data for Hong Kong s obtaned from the Pacfc Basn Captal Markets (PACAP) database compled at the Unversty of Rhode Island. The database contans daly stock returns, closng transactons prces, and volume data for all exchange-lsted frms. The HK government made a publc dsclosure of the exact level of nterventon n each stock on October 26, 1998. Ths nformaton s collected from the text of the speech by HKMA chef on October 26, 1998, avalable on the HKMA web ste cted prevously. An electronc search was conducted usng the LEXIS/NEXIS database to dentfy other news events pertanng to the Hang Seng stocks durng the nterventon 11
perod. Ths nformaton s used to screen out daly observatons that mght be subject to other confoundng nfluences such as earnngs announcements. B. Abnormal return computatons The tests are based on an examnaton of Hang Seng stocks relatve to returns on an equally weghted control portfolo of all stocks that are traded on the Hong Kong stock exchange. Because the government ntervened n the Hang Seng stocks, any prce pressure and sgnal effects should be prmarly restrcted to the Hang Seng stocks. We assess the relatve mpact on Hang Seng stocks by computng ther abnormal returns. The control perod s set to the frst sx months of the calendar year (January 1 to June 30, 1998) and the last one and a half months (November 15 to December 31, 1998). The choce was motvated by the fact that there were no nterventon-related announcements n ths perod, and to also allow examnaton of post-nterventon returns over a reasonably long wndow. The followng cross-sectonal regresson s run usng returns over the control perod for each stock: where R r r = α + β R + ε. (3) t, mt, t, R, t s the raw return of stock on day t, R m, t s the return on the equally weghted portfolo of all stocks traded on the Hong Kong stock exchange on day t and α r, β r are constants. The abnormal return for each day on the event perod s then computed as: r r ( mt) AR = R α + β R (4) t, t,, We chose the equal weghted ndex rather than the value weghted ndex to mnmze the mpact of the Hang Seng stocks on the ndex. Although Hang Seng stocks account for roughly 70% of the market captalzaton, they account for less than 10% of 12
the lsted stocks n terms of numbers. In addton to the equally weghted ndex, we employed other ndces to measure the performance of the overall market. We reconstructed the ndex returns by explctly excludng the mpact of the Hang Seng stocks on the ndex. We also used an ndex of the 33 largest frms that are not ncluded n the Hang Seng ndex. Fnally, we constructed an ndex of 33 frms matched by sze and ndustry wth the Hang Seng ndex. The reported conclusons are the same f the equal weghted ndex s replaced by any of these ndces. We report the results usng the equal weghted ndex, as ths ndex s readly avalable n the PACAP database. C. Pattern of government nterventon Table II provdes summary statstcs of the nterventon data. The frst row shows the wde varaton n the amounts spent on the 33 Hang Seng stocks. Row 2 presents detals on normalzed nterventon actvty ( PSB ) measured as the rato of amount of purchase to the frm equty value on July 31, 1998. The nterventon resulted n the government holdng anywhere from 3% to 12% of the outstandng shares n a partcular stock. The volume of nterventon amounted to an average of 75% of the total tradng volume n the nterventon perod (row 3). The last row of Table II presents another measure of nterventon. Ths metrc gves the purchases n each frm relatve to the tradng volume n the frm shares durng the month of July 1998. The data reveals that the nterventon purchases were very large relatve to the average tradng volume n the precedng month of July. The data also ndcates that the government purchased more n bgger frms. 13
In order to better understand the pattern of nterventon, we estmate the relatonshp between the prmary metrc of nterventon actvty, PSB (percentage of outstandng shares bought), and varables such as the average frm weght n the Hang Seng ndex over the nterventon perod ( Indxwt ), the return on a stock durng the 30- day wndow before nterventon ( PorRet ) and the slope of the excess demand curve ( Slope ) 10. Note that the slope of the excess demand curve s always negatve. We obtan the followng estmates for the regresson parameters (fgures n brackets ndcate t-statstcs based on Whte s correcton for heteroscedastcty): PSB = 0.06 + 0.17 Indxwt 0.03 PorRet + 31.28 (7.34) (4.65) (-1.40) (3.66) 2 Slope + ε (5) R = 0.52 The regresson shows that the government ntervened more n frms wth a hgher ndex weght consstent wth greater vulnerablty of these stocks to speculators who had short postons n Hang Seng futures. The nterventon s not related to poor performance n the prevous month. A more lqud frm wll have a flatter excess demand curve, and hence a less negatve slope. Consequently a large amount of nterventon s requred to nduce a small change n prce. The regresson, consstent wth ths noton, reveals that there s larger nterventon n frms wth a flatter excess demand curve. 10 The slope of the bd (ask) curve s computed as the rato of the percentage change n prce (for the best fve quotes) to the change n the percentage of shares bd (offered). The dfference between slope of the bd curve and the ask curve s the excess demand and the average of these slopes over the last one hour of the tradng day gves the average slope for that day. The slopes of the curves are computed as the averages of the slope over the frst two weeks of August. 14
D. Generatng measures of daly nterventon In our analyss of the mpact of nterventon, we are hampered by the lack of data on daly nterventon actvty. Fortunately, the nterventon s on such a massve scale that t s feasble to obtan a reasonably good proxy for the daly nterventon actvty by analyzng daly volume data. For each stock, the followng cross-sectonal regresson s run usng dollar tradng volume over the control perod: where V v v = α + β V + ε (6) t, mt, t, V, t s the dollar volume of stock on day t, V m, t s the total dollar volume on all stocks traded on the Hong Kong stock exchange excludng Hang Seng stocks on day t, and α v, β v are constants. The abnormal volume for the event perod s then computed as: v v ( mt) AV = V α + β V (7) t, t,, The sum of the dollar abnormal volumes over the ten-day nterventon perod has a crosssectonal correlaton of 1.0 wth the dollar nterventon n each stock. Ths correlaton may reflect a frm sze effect captured by both varables. When both the volume varables are normalzed by July market value to remove the frm sze effect, the correlaton s 0.94 and s very hghly sgnfcant. Thus, the procedure descrbed produces a very good estmate of the daly nterventon actvty, and enables us to analyze the relatonshp between returns and nterventon actvty usng ten daly observatons on the 33 stocks for a total of 330 observatons compared to 33 observatons on reported nterventon amounts. Note that the volume of government nterventon s not drectly observable by the market partcpants, whch rases the queston of whether there s stll nformaton content 15
n the sgnal of nterventon? Bhattacharya and Weller (1997) analyze the mpact of nterventon when the monetary authorty has a target for the exchange rate, and trades wth speculators who have prvate knowledge about fundamentals. In ths settng, the equlbrum prce conveys nformaton about the fundamentals and the target prce. Then, speculators can extract nformaton about the volume of nterventon and so the target of authortes from the behavor of prces. A smlar argument apples n ths case too. IV. Emprcal Analyss Ths secton frst detals the returns and volume around government nterventon. The next two sectons provde evdence on the temporary prce pressure and the permanent prce pressure hypotheses. A. Returns and volume around government nterventon Table III gves a day-by-day overvew of returns around nterventon (columns 1 and 2). Raw returns were largely negatve n the perod pror to nterventon. However, nterventon was accompaned by a sharp ncrease n both raw and abnormal returns over the perod August 14 to August 28. The average abnormal return s 24% durng the nterventon perod. The medan abnormal nterventon perod return s 23.3%, and the standard devaton s 7.57%. The daly abnormal returns are statstcally sgnfcant on each day of the nterventon perod except August 20 and August 21. The largest abnormal return of 4.8% occurred on the frst day of nterventon. On the last day of the nterventon, the abnormal return was 3.6%, although the raw return was negatve. Ths observaton ponts to the mportance of relance on abnormal returns rather than raw 16
returns. The huge ncrease n the sze of the nterventon on August 28 helped to hold the raw returns on the ndex stocks to 1.29%, compared to a drop of 4.99% on the control portfolo. Table III (columns 3, 4 and 5) also gves an overvew of tradng volume around nterventon. The table reports both the dollar volume and the relatve volume on Hang Seng stocks and an ndustry-matched portfolo. Relatve volume for a stock s computed as a percentage of the raw volume n the 22 tradng days of July 1998. Table III shows that total volume n Hang Seng stocks had a sharp ncrease durng the nterventon perod from August 14 to August 28 (columns 3 and 4). In partcular, the average relatve volumes on August 27 and August 28, the last two days of the nterventon perod, were 7.7 tmes and 27.3 tmes the normal volume, respectvely. A large part of the ncreased volume on August 28 was a result of futures and optons expraton. Speculators who were short futures contracts attempted to drve down equty prces whle the government ntervened to support the Hang Seng stocks. Of the HK$118 bllon spent by HKMA durng the nterventon, newspapers reported that HK$70 bllon was spent on August 28. The massve and unprecedented nterventon on August 28 was unable to prevent a declne n stock prces, although t reduced the downward slde n stock prces. The control portfolo (ndustry matched portfolo) had a relatvely modest up tck n volume on a few days n the nterventon perod. The tradng volume n the Hang Seng stocks tapered off followng the cessaton of government nterventon. However, volume remaned at hgher levels compared to the average July daly volume. In summary, the data shows both ncreased volumes and large postve abnormal returns durng the nterventon perod. 17
B. Rejectng the temporary prce pressure hypothess If nterventon effects were due to temporary lqudty pressures, then there should be a negatve reacton n prces as soon as nterventon stops. Accordngly, we examne abnormal returns n the eght weeks followng the nterventon wndow. As Fgure 2 ndcates, the Hang Seng ndex contnued to apprecate n the weeks followng the nterventon. Table IV dsplays the abnormal returns and cumulatve abnormal returns on the Hang Seng stocks after the nterventon perod. Recall that the cumulatve excess return for all Hang Seng stocks s 24% for the event wndow. The followng three weeks (weeks 1, 2 and 3) dsplay negatve returns for a cumulatve retracement of 9.3%. However, abnormal returns are once agan postve for weeks 4 to 7. In other words, there s no evdence that the abnormal returns generated durng the event wndow reversed n the weeks followng nterventon. We also test a more powerful predcton of the temporary prce pressure hypothess: namely, that the prce ncreases for each stock durng the event perod s reversed over the subsequent weeks. In other words, for every Hang Seng stock n the test sample, we run the followng regresson: CAR, 1 T = + θ IPAR, t = IP + ε, t T α (8) where CAR, 1 T, s the cumulatve abnormal stock return after nterventon begnnng n week +1 through week +T nclusve, and IPAR =, t IP s the cumulatve abnormal return n the nterventon perod. Under a complete reversal of nterventon returns, the ntercept s zero and the slope s 1. Under a partal reversal, the slope coeffcent wll be negatve. Table IV presents the results. The reported t-statstc for the coeffcent of the nterventon perod cumulatve abnormal return s based on Whte s correcton for 18
heteroscedastcty. The table (columns 5 and 6) shows that none of the coeffcent s sgnfcantly dfferent from zero. Thus the evdence does not support temporary prce pressure effects. 11 C. Evdence on the permanent prce pressure hypothess Gven that the evdence ponts aganst temporary prce pressure, we seek to dstngush between the permanent prce pressure and nformaton hypotheses. In other words, we want to dscern whether the mere fact that the government was buyng and ts promse to support stocks led to a change n equty prces (sgnalng) or was t because of the demand for stocks from nterventon (prce pressure). The prce pressure hypothess posts a postve cross-sectonal relatonshp between abnormal returns and nterventon amounts. In ths secton, we present evdence on ths cross-sectonal relatonshp. C.1. Overall nterventon perod results To nvestgate the presence of permanent prce pressure effects, we estmate equaton (2) usng data for the entre nterventon perod, 33 observatons, and obtan the 11 The mean abnormal return s 6.54% on August 31, 1998, the frst day after the nterventon. The szable average negatve return rases the ssue of a partal reversal of nterventon perod returns. However, there s no sgnfcant cross-sectonal relatonshp of these returns wth the ether the nterventon perod abnormal returns or the nterventon amounts. Thus, we need to be cautous about nterpretng the abnormal returns on August 31 as a partal reversal of postve nterventon perod returns. The negatve returns on August 31 for ndex stocks could possbly be a delayed adjustment to factors that affected the rest of the market on August 28. In ths case, the abnormal returns on these two days are expected to have a negatve correlaton, snce stocks that were propped up more by nterventon on August 28 should fall more on August 31 when nterventon stops. However, they are postvely correlated. 19
followng estmates for the regresson parameters (fgures n brackets ndcate t-statstcs based on Whte s correcton for heteroscedastcty): 12,13 IPAR = PSB 3.1 Slope + 1.92 Sg 0.19 Indxwt 0.004 PorRet + ε (2.95) (-0.69) (-0.05) (0.72) (-1.08) (-0.05), t IP = 0.21 0.33 (9) Under the prce pressure hypothess, larger nterventon amounts lead to larger nterventon returns. Contrary to ths predcton, we fnd that the abnormal return on a stock s not postvely related to nterventon n that stock. The reported results are not affected by the excluson of the control varables n the regresson. Another possble reason for the reported results s that the sgnal effect and prce pressure effects are cumulated over the ten days, leadng to a potental loss of nformaton. C.2. Results based on daly abnormal volume measures The next test uses measures of daly abnormal volume as descrbed by equaton (7) n Secton III.D as a proxy for nterventon. Unlke the pervous regresson (equaton (9)) that used 33 observatons, ths regresson uses 330 daly observatons. We then exclude observatons concdng wth major corporate announcements that mght contamnate abnormal returns. Durng the ten-day nterventon perod, sx frms reported earnngs. We excluded the earnngs announcement day and the next day, snce t was 12 Kaul, Mehrotra and Morck (2000) use pror prce run up to account for captal gans tax nduced frctons on tradng. There are no such taxes n Hong Kong. We nclude pror returns to nvestgate whether postve nterventon perod returns are a correcton of pror abnormal negatve returns. 13 We expect a postve relatonshp wth Sg snce t proxes for the dffculty of arbtrage. However, arbtrageurs, who beleve that the government nterventon wll fal, are more lkely to trade the ndex rather than ndvdual stocks. The nsgnfcance of the term Sg lends support to ths vew. 20
unclear whether the announcements were made before or after close of tradng on the announcement day. Ths procedure led to the excluson of 12 observatons (retenton of these observatons does not materally alter the results). The daly abnormal volume measure s standardzed by market value at the end of July 1998 to yeld comparable measures across frms ( AR AV, t ). Panel A of Table V reports regresson estmates of the followng equaton:, t β 0 + β1 AV, t + β2 Slope + β3 Sg + β4 Indxwt, t + β5 ProrRet + ε, t = (10) Here, abnormal return, AR, t, consttutes the gross effect of nterventon that ncludes both the sgnal and prce pressure components. The slope coeffcent for abnormal volume s sgnfcant (t-stat of 4.197) 14. Ths result can be nterpreted as evdence consstent wth prce pressure effects. The low R 2 s consstent wth the poor explanatory power of the varable measurng these effects. The abnormal returns on August 14, 1998, provde a unque set of observatons. The government dsclosed ts nterventon only after close of tradng on that day. Although news reports ndcate that market partcpants speculated about government actvty, the sgnalng content of the nterventon should have been mlder, f not nonexstent, snce t was unannounced. Thus, the postve abnormal returns on August 14 can be nterpreted as evdence consstent prce pressure effects. Furthermore, consstent wth such effects, Panel B of Table V reveals a sgnfcant postve correlaton between abnormal returns and abnormal volume on August 14. However, t could well be that the 14 It s possble that the results are drven n part by speculatve actvty over the last two days of nterventon. However, we get a smlar result even f we drop the observatons from the last two days. 21
postve returns were due to the market s nterpretaton that the buyng was beng drven by prvate nformaton. The explanatory power remans low n ths case too. To further nvestgate the sgnalng mpacts of the nterventon, we re-estmate equaton (10) wth two proxes for nterventon: AV t, whch s the average abnormal volume on day t, and AdjAV t, calculated as: AdjAVt, = AVt, AV t,. The average abnormal volume across all Hang Seng stocks on a gven day ( AV t ) s a proxy for the strength of the government nterventon on that day. Ths average should not be nfluenced by abnormal volume drven by stock-specfc prvate nformaton, and t should be related to the sgnal content of the nterventon on that day. The adjusted abnormal volume ( AdjAV t, ) captures the cross-sectonal varaton n abnormal volume on day t and should be more drectly related to prce pressure and prvate nformaton effects n ndvdual stocks. Panel C of Table V reports that the coeffcent of AdjAV t, s nsgnfcant, suggestng mnmal prce pressure effects. However, the coeffcent of AV t, a measure of the common sgnal component, s sgnfcant and provdes support for sgnalng effects of the nterventon. In summary, even though we would expect nterventon measures should explan return behavor to a large extent, the overall evdence provdes only mld support consstent wth the prce pressure hypothess and ponts to the alternatve nformaton hypothess, whch argues that the government credbly sgnaled the grantng of an mplct put opton to holders of Hang Seng stocks. 15 15 The average nterest rate durng July 1998 s 10%. The annual volatlty of realzed weekly returns on the Hang Seng ndex durng the year precedng nterventon s 45%. Usng these parameter estmates, the Black-Scholes formula yelds a premum of 13.4% 22
D. The case of Swre Pacfc Swre Pacfc offers a unque case study. Swre Pacfc has two classes of shares A and B. Only class A shares are a part of the Hang Seng ndex. The rsk assessment of both these shares s smlar except for dfferences n votng rghts Class A shares have dfferent votng power than Class B shares. The two securtes are near perfect substtutes except that the government ntervened n one and not n the other. Wurgler and Zhuravskaya (2002) argue that stocks wth close substtutes are more reslent to prce pressure effects. Ths reasonng would suggest that n the case of Swre Pacfc A shares, the avalablty of very close substtute Class B shares should have mmunzed them from prce pressure effects. However, under the sgnalng hypothess, the government s decson to defend the value of Hang Seng stocks provdes a valuable put opton to the stockholders of Swre Pacfc A. The consequent bolsterng of the frm s equty value should also be benefcal to Class B shareholders, although the mpact wll be more drect for Class A shareholders. Table VI gves an overvew of the return behavor of these two shares. The dfferental n returns between Class A and Class B shares s statstcally nsgnfcant durng the control perod as well as perods mmedately before and after the nterventon. Ths evdence confrms that these two securtes are very good substtutes. Durng the nterventon perod, both classes experence hghly sgnfcant average abnormal returns: 2.33% for Class A and 1.67% for Class B. These postve returns for the two close substtutes provde evdence aganst the prce pressure hypothess. The dfference for one-year at-the-money put optons. In comparson, average nterventon abnormal returns are 24%. If the returns of 6.5% on August 31, 1998, are ncluded, the average abnormal return to stockholders drops to 17.5%. The Black-Scholes estmate does not account for perceptons of ncreased volatlty around nterventon. 23
between the average return on Class A and Class B s 0.66% for a cumulatve total of 6.7% over ten days. The addtonal return reflects the greater value of the put opton to the Class A shareholders. These excess returns were not reversed n the weeks followng nterventon. The Swre Pacfc case lends credence to the presence of a sgnal component mplct n the stock prces. E. Addtonal Remarks The consderable reducton n tradng volume on August 31 compared to the prevous day provded clear clues about the government wthdrawng from nterventon on that day. The government ssued statements n whch t asserted that t would contnue to montor the stock market and would ntervene f the stablty of the market were threatened. Consstent wth the nformaton hypothess, the negatve returns on August 31 can be nterpreted as a reacton to clues about the lmts of the government s nvolvement. Under the permanent prce pressure hypothess, stock prces should stay up as long as the government holds the stock. Thus, ths hypothess cannot explan the observed negatve return on ths day. Fnally, we address whether the nterventon was bound to ncrease stock prces gven ts sheer magntude. One vew of the success of an nterventon s the profts garnered by the ntervenng authorty. HKMA reported the average purchase prce for each of the stocks n ts dsclosure on October 26, 1998. At the concluson of the nterventon, the government had a meager proft of only 1.4%. The low profts are consstent wth the bulk of the purchases occurrng at the end of the nterventon. In fact, on the day after the nterventon concluded, the government faced a loss of 6.9% on ts portfolo, whch worsened to a loss of 9.8% on the next day. The government faced a 24
loss on each of the acqured stocks at some pont durng the two months after the nterventon. Nevertheless, by October 23, 1998, these losses had been erased (see Fgure 1), and the government enjoyed a proft of 24.7%. On the next day, the government voluntarly dsclosed detals about the nterventon. 16 The government profts tended to cast the nterventon n a postve lght. However, the varablty of profts suggests that net government profts were not a foregone concluson despte the massve nterventon. F. Polcy Implcatons There are two possbltes n relaton to ths nterventon. One s that the Hang Seng ndex had been drven down below a level consstent wth fundamentals by the concerted actons of speculators, n whch case the nterventon transmtted to the market a wllngness on the part of the authortes to expend resources to protect nvestors aganst the effects of such actons. The other possblty s that speculators were reactng to fundamentals, mplyng that the market was correctly valued before nterventon. Then, the belefs created by the nterventon would have drven the market above the level consstent wth fundamentals. The HKMA publcly stated ther vew that speculators had drven the market below a value warranted by fundamentals. The Hang Seng stocks lost more than a quarter of ther value n the month before nterventon even though other Asan economes wth smlar economc performance dd not experence such declnes n equty prces. As Table IV reveals, the postve returns generated durng the nterventon were 16 The government announced nterventon detals after close of trade on October 26, 1998. The abnormal return on the next day s an nsgnfcant 0.11%. Further, the returns on October 27 are not related to nterventon amounts. 25
not reversed after the nterventon ceased. Although not conclusve, the overall evdence s consstent wth the vew that the government nterventon corrected an under valuaton nduced by speculatve attacks on the currency and the stock market. Under the alternatve vew that the speculators were merely forcng the stock prces to ther correct values based on fundamentals, the government nterventon results n over valuaton of the stocks. Mller, Weller and Zhang (2002) analyze how regulators should unwnd nvestors expectatons n such crcumstances. If the government unwnds such expectatons rapdly, stock prces are lkely to drop sharply and even fall below prces based on fundamentals. Such a sharp correcton could be assocated wth real effects n the economy. In ths context, they develop mportant polcy mplcatons for the government s management of nvestor expectatons. The HKMA unwound a large proporton of ts holdngs n one nstance on November 12, 1999 (n the form of an exchange traded fund). Ths was one of the largest and most successful offerngs ever n Asa wth approxmately HK$33 bllon rased from 180,000 partcpants, the vast majorty of whom were local retal nvestors. The lqudaton by the government of ts stockholdngs dd not have any adverse mpact on the market ether on the offerng day or around the announcement day (October 11,1999). Ths evdence supports the HKMA vew that the nterventon ncreased market effcency by protectng nvestors aganst the effects of market manpulaton rather than the vew that the nterventon artfcally nflated stock prces. V. Concluson Should governments ntervene n the stock market and what are the effects of nterventon? These questons of vtal mportance to polcy makers and monetary 26
authortes have not been explored n the lterature. To our knowledge, ths s the frst study that explores the effects of stock market nterventon. We analyze the case of Hong Kong, where the monetary authorty ntervened n the Hang Seng ndex stocks over the perod August 14 to August 28, 1998. The nterventon was amed at deterrng speculators va a credble sgnal that the Monetary Authorty would use avalable funds to buy ndex stocks n the event of a speculatve attack. The nterventon had a salutary effect mmedately, and an equal-weghted portfolo of Hang Seng stocks experenced sgnfcant abnormal returns to the tune of 24%. These abnormal returns were not reversed n the weeks followng the nterventon and cannot be attrbuted to temporary prce pressure effects. Usng daly data on ndvdual stocks we nvestgate whether relatve prce changes were a result of a credble sgnal or permanent prce pressure effects arsng from the government buyng and holdng large amounts of stock. The overall nterventon perod results do not provde any evdence of prce pressure effects. Analyss of daly data and nterventon proxes reveals that ndvdual stock returns are related to the overall nterventon actvty rather than stock-specfc nterventon. Ths and related evdence on dual class shares ponts to an nformaton effect of nterventon on stock returns. The lessons from the Hong Kong nterventon must be carefully consdered and are not drectly transferable to other stock markets. The Hong Kong government had consderable credblty snce t had large reserves relatve to tradng actvty n the Hong Kong stock market. Some countres (the most recent beng Japan) have taken a cue from the successes n Hong Kong and have proposed a large nterventon program. When and f such nterventons should be orchestrated are questons of crucal mportance that are left to future research. 27
References Bagwell, L. S. 1992. Dutch aucton repurchases: An analyss of shareholder heterogenety. Journal of Fnance 47: 71-105. Benesh, M. D. and R. E. Whaley. 1996. An anatomy of the S&P game: the effects of changng the rules. Journal of Fnance 51: 1909-1930. Bhattacharya, U. and P. Weller. 1997. The advantage to hdng one's hand: Speculaton and central bank nterventon n the foregn exchange market. Journal of Monetary Economcs 39: 251-277. Harrs, L. and E. Gurel. 1986. Prce and volume effects assocated wth changes n the S&P 500 lst: New evdence for exstence of prce pressures. Journal of Fnance 41: 815-829. Kaul, A., V. Mehrotra and R. Morck. 2000. Demand curves for stocks do slope down: New evdence from an ndex weghts adjustment. Journal of Fnance 55: 893-912. Loderer, C., J. Cooney and L. V. Drunen. 1991. The prce elastcty of demand for common stock. Journal of Fnance 46: 621-651. Lynch, A. and R. Mendenhall. 1997. New evdence on stock prce effects assocated wth changes n the S&P500. Journal of Busness 70: 351-384. Mller, M., P. Weller and L. Zhang. 2002. Moral hazard and The U.S. stock market: Analyzng the 'Greenspan Put'. The Economc Journal 112: C171-C186. Wurgler, J. and E. V. Zhuravskaya. 2002. Does arbtrage flatten demand curves for stocks? Journal of Busness 75: 583-608. 28