ied Maihematical Finance 2, 243-271 (1995) Risk arbitrage in the Nikkei put warrant market of 1989-1990 J. SHAW,' E,O, THORP? and W.T. ZIEMBA?* Woe dv Corporation, Toronte, Canada, 7Edward © Tharp and Associa Catitornic. USA, ‘Alumni Professor of Management Seéence at the University uporr Beach f British Cohimbie Received 2 March 1995 This paper diseusses the Nikkei put warrant market in Toronto and New York during (989-1990. Three classes of long term American puts were traded which when evaluated are product and asset puts, respectively. Type I do not involve exchange rates fo! yen investors. Type IL. in advance the exchange rate to be used on expiry in the home currency. Type II] evalu sirike and spot prices of the Nikxei Stock Average in the home currency rather thin in yen. For typically observed parameters, type I are theoretically more valuable than type H which in tucn ate niore valuable than type IIL. In late 1989 and early 1990 there were significant departures from foir values in various markets. Tais ‘was u market with a set of complex financial instruments that even sophisticated investors needed time to leara about to price properly. Investors in Canada were willing to buy puts at far more than feir value bused on historical volaulity. In addition, US investors overpriced type Il puts fixed in dollars rather than the (5 T's in yen. This led to cross border and US traded (on the same exchenge) low risk hedges The market's convergence to efficiency (that is, all puts priced within transaction cost bands) took about one month afer the introduction of the US puts in early 1990 leading to significant profits for the he quantos, Keywords: option mtispricing, cross-border trading, Nikkei stock exchang 1, Introduction ‘The Japanese stock market rivals that of the US in size and importance. Its growth in trading volume and capitalization has been large. The markets in Tokyo, Osaka, Nagoya and five oihe regional exchanges that now trade were closed during World War If and then reopened in May 1949, Post war construction and aid by the US helped the Japanese economy grow quickly. By 1950, 9% of the world’s equity capitalization was Japanese compared to 58% in the US, 27% in Europe and 6 % in the rest of the world. By 1980 Japan had increased its share to 15% mainly at the expense of Europe whose share fell to 20 %. The rest of the world doubled its capitalization to 12% and the US still had the majority with 53 %. The 1980s were a period of economic excesses in the US that led to a weakening of the strong economic base held in 1980. The Reagan economic policies led to large deficits in both the overall budget and in trade, as well as larg military spending and debt payments (see Modigliani (1988) and Hatsopoulos, Kr Poterba (1989) for analyses). Meanwhile Japan maintained a policy of high investment in plant increases in ‘This paper sas wcltten in 1061 for the presentation of Ziemba (199La). Rete sash aritne Habe WOK wurtatee te 160K! ct references and data occuring inand equipment and R&D, financed through policies that emphasized and rewarded hi npany expansion proceeded more through debt and retained earnings than oquit ‘lable end was at low nominal and real interest rates pa trial grouping (keiretsu) 980s were very fina icularly through bi ncially favourable for Japanese firms. The combination of low interest wo funds, strong export marketing expertise, emphasis on quality. access 1 foreign markets while maintaining structures and rules making imports co Japan dificul lative wealth transfer from the US to Japan. By December 1988. Japan's ewuity cupiulization was 44%, Europe's 21 %, the rest of the world’s 6%, and te US fell 10 29%, Ae: Une 1990-92 stock market decline, Japan’s share fell 1a 25% as of Sepiember 1992. The US was 40%, Europe 25% and the rest of the world 6%. French and Poterba (1991) and Ziemba and Schwartz (1991) have argued that to properly meusure capitalization one must adjust for company cross holdings. If company A holds much of company B's stock and vice versa, then the true market capitalization of A plus B is less than the sum of the individual capitalizations. Japan and many of the Euro; ne nh as Germany and Maly have extensive cross holdings. In 1990 about 71 % of Japan's equity was cross-held and rarely, if ever, traded. Calculations show that the capitalization of Japan in the la 1980s was overstated by about 25 % (See for example, McDonald, 1989 and Ziemba and Sehwartz 1951), After such adjustment, the shares in 1988-89 were about 39 % for Japan. 33% for the US. 22% for Europe and 6% for the rest of the world. Historically, Japanese stock markets have been much more influenced by foreign markets than everse as shown by Becker, Finnerty and Gupte (1990), Hamao, Masulis and Ng (199U nd Ziemba and Schwartz (1991). The transmission of mean returas and tility was mos! nal until the October 1987 world wide stock market crash. price movements in Japan have had more impact on those in New York and Ls the reverse effect is much stronger (see for example, Hameo, Masulis and Ne (1 wegued that this integration of capital markets was small because the insulaied to large extent from foreign influence and not deregula:ed Deregulation of Japan's financial markets began in earnest in 1987 with the introduction of the rsi equity index futures contract, the Kabusaki $0, which traded in Osaka. Futures on the Nikkei Stock Average had begun trading on the Singapore Monetary Exchange (SIMEX) in 1986. Bailey (1989) discusses the eurly history of these two futures contracts, In September 1988. in Iutures contracts on the more popular Nikkei 225 stock average and the Tokyo stock price (Topix) index which were traded in Osaka and Tokyo, respectively. These contracts allowed foreign investors and institutions to easily hedge positions in Japanese equities and 1 engage re fully in a variety of types of progremmed trading including index arbitrage and portiolio insurance. During 1988 and 1989, the Japanese equity markets increased dramatically in wading volume and market capitalization. This was a period of cheap and casily available money fot corporate and individual investors and speculators The equity warrant bonds issued in Luxembourg with the warrants tradir such example, By adding the warrants which were suipped off Paley, easy wcce: ed to an enurmous, jomies J the 2 crash. stock ndon. However. 91}). Ihmay be markets were nee g in London was on ind waded separately: the bonds offer low coupons. Hedging he proceeds of the bond sales, which were mainly in dollar. back into yen with its considerably lower interest rates, provided net costs at borrowing at close to zero percent, The equity warrants, when exercised several years later if they were in-the-moneyBis bivrage in Nikkei put warrant market provided an additional source of funds for the firm to redeem the bonds for a slight dilution, This market was in excess of $200 billion. See Mikami (1990), Takahashi (1990) Marsh (1992, [994) for analyses of the pricing of these warrants. A lar warrants expired worthless because of the 1990-95 decline in stock prices Over-the-counter long-dated typically three year puts were marketed in 1988 by major non. antese brokerage houses to corporate clients who wished to hedge against lon, hara and ze pe of these anlese equ exposure oF to speculate that the high priced Japanese stocks would eventually decline sharply. The sellers of these puts, which typically had premium value of $100000 plus and were prived to trade at volatilities around 16-20 % versus the historical 13 %, were mosily large Japanese corporations ‘The corporations displayed a collective arrogance about the strength of the Japanese stock market and economy by generally not hedging. The high price earnings ratios in the 79 plus range and the asironomically high land prices typified by facts such as the Imperial Palace sn Tokyo being worth" as much as all the land in California led professional and amateur investors and econo- musts to believe that these high prices could not be sustained (see Aroa (1981, 1989), French and Poterba (1991), Friedberg (1990), Ueda (1990), Wood (1990) and Stone and Ziemba (1993) for a variety of analyses with this conclusion). The first Japanese put warrants available to individual investors on an easily purchasable basis were the three year American-lype Nikkei pur warrants that traded on the Toronto Stock Exchange in February 1989. These puts were aot wue warrants as they were cash settled based on the price of the Nikkei Stock Average and were not exercisable into stock. Also their issuers were investment banks not individual firms. These warrants provided individual investors with the opportunity to bet against the high siock prices in Japan with 2 minimal investment of capital. The “warrants” were thus long dated put options. US investors were not allowed to purchase those warrants for three months and these warranis ¥ widely advertised and known outside Canada These warrants were purchased in such demand that their price in implied volatility was well above the historical for the NSA index. Most Canadian investors probably had no idea what fair value was. However, despite paying prices up to four times the Black-Scholes (1973) fair value investors who held these put warrants to expiry made very large profits because of the large decline in the Nikkei index. The Toronto warrants were of three types: ordinary puts velued im yen, puts where the final exchange rate for yen is fixed in advance and puts where the NSA was evaluated in Canadian dollars. This latter type allowed investors to profit from declines in the NSA or the Japanese yen or both. Although in principle straightforward to professionals as discussed in Sections 3 and 4 below, besides being unable to evaluate the fair values of these warrants investors were unable to evaluate the relative differences between the various types of warrants. Thus, with complex ‘nsuuments, even the most sophisticated in the market needed time to unders:and the products price them fairly and invest in them to eliminate mispricings.’ The warrants also had d credit characteristics and when exereised were evaluated on the next day’s closing price NSA in Japan if there was trading that day otherwise on the next trading day. Investo warrants could also put in the proviso that the warrant not be exercised if the NS. Te nol " Indeed Reiner's (1992) acticle in Risk is summarized as follows: “Investors and dealers inked forex options ~ then finding they don’t know ow to value taem. Er tnd its variant” einer explains246 Shaw et al more points on the requested exercise day. In late 1989 the type I and type III Canadian NSA put rants were greatly over-priced in comparison to fair values based on histori luding that for the 1987 world wide stock market crash period. There was no way for a small nvestor to hedge these instruments. Large investors or institutions could, or course, hedge in the futares markets on the SIMEX or in Japan, Indeed this was the way that the issuers who sold the puts and were responsible for their exercise payments hedged their investment. Grossman (1988) among others has argued that this is not a fully suitable approach because the futures synthetic does not have the same information requirements as the underlying detivative, Henee the seme types of difficulties associated with the breakdown of portfolio insurance in the 1987 erash could possibly occur in this market as well. See Rubinstein (988) for an analysis of the effect of portfolio insurance on the crash. A better hedge for investors was thus a negotiated over-the-counter put on the NSA that essentially matched the Toronto stock exchange waded warrants. uch instruments were available in late 1989 from investment firms such es the Salomon ind Bankers Trust. The authors and others were aware of the potential of shorting expen- dian NSA puts and hedging with a fair priced puts of similar characteristics and duration in another market. To short the Canadian put warrants these warrants needed to be borrowed since there was a fixed number of them issued, They also had to be shorted according 10 the uptick rule. This was more difficult than shorting an ordinary exchange traded put or call or an tidex option which is essentially from an infinite supply and does not have these restrictions, However, it was possible to short Canadian NSA put warrants in large numbers at the high implied volatilities. It was expected that the market price of these puts would drop to their fair value once a fairly priced product was easily available. Bankers Trust, the Salomon Brothers and the Kingdom of Denmark issued such warrants in January and February 1990 which traded on the American Stock Exchange. These warrants were all fixed exchange rate securities (of type Ll, called quantos) except for the Bankers Trust January put which was a type I with a floating exchange rate 1 volatilities Broth« sive Can Bernard and Thomas (1989, 1990), Afileck-Graves and Mendenhall (1992) and Aberbanell and Bernard (1992) have shown that frequently there are considerable delays in the market price adjust- ment to new earnings announcements, Indeed some of these adjustments take several months to be fully refiected in market prices. Jacobs and Levy (1988) found that lagged earnings surprises are a declining but significant factor in US security prices for one, two and three months after their announcement. The convergence of the NSA puts to efficiency were similar and the proces took over one month from the time the first NSA put warrant was traded on the American Stock Exchange in January 1990. Large profits were made by hedgers, including the authors although they took several risks that are difficult to quantify, Besides the credit and exchange rate risks (Which could be hedged) there was a tisk of forced buy-ins of the shorts at unfavourable prices because it was no longer possible to borrow the puts. All three of the authors had forced y-ias for a small amount of their position. There was also a profitable hedge be:ween fixed end non-fixed exchange rate puts that was affected by the absolute price of the puts. Developing models to account for these imperfections is difficult; see Figlewski (1989) for some results obtained by simulation. In all cases for the Nikkei puts and the Nikkei calls which are discussed in Sections 4 to 6, fixed exchange rate options, the quantos, traded for prices above non-fixed exchange rate options when the theoretical price was less, Investors w re willing toRisk arbitrage in Nikkei put warrant market 247 pay a premium to eliminate this exchange rate risk even though it could have been hedged much cheaper in the foreign exchange futures markets. Also investors paid a premium for low nominal priced warrants that is analogous to that of low priced stocks, see for example, Blume and Stam- baugh (1983). This led to a hedge that was close to arbitrage where investors could buy the high nominal value but low implied volatility Bankers Trust warrants and sell the lower priced Kingdom of Denmark and Salomon type A and type B warrants that had higher implied volati- lities on the same exchange (the American Stock Exchange). These warrants were mispriced for the month of February 1990. Except for slightly different credit risk and strike prices these warrants were virtually identical. One of the authors used this risk arbitrage to win the US stock market championship organized in Barron's in the category of risk adjusted retums for accounts over one million dollars in 1990. The paper is organized as follows. Section 2 contains a brief background to the Japanese stock bracketing the time of this study (mid 1989 to mid 1990). Historical volatility is also cussed there. Additional references on the Japanese stock market include Elton and Gruber (1989), Amihud and Mendelson (1991), Chan, Hamao and Lakonishok (1991), Ziemba and Schwartz (1991), Ziemba, Bailey and Hamao (1991), Ziemba (1989ab, 1991ab), and Stone and Ziemba (1993). Section 3 discusses the various NSA put warrants and call warrants that were trading in (989— 90 on the American and Toronto stock exchanges and categorizes them into the three types which, using the definitions in Rubinstein (1991), are ordinary, product and option to exchange, See also Donnelly (1990), Smith and Dunn (1990), and Tufano (1992) for general discussions of these warrants. Section 4 discusses the fair numerical valuation or the three types of puts using the Cox, Ross and Rubinstein (1979), Boyle (1988) and Boyle, Evnine and Gibbs (1989) two and three dimen- sional binomial lattice models. Other authors such as Derman, Karsinski and Wecker (1990). Babbel and Eisenberg (1991), Rubinstein (1991), Gruca and Ritchken (1991), Clyman (1991, 1992), Chen, Sears and Shabrokhi (1992), Reiner (1992), Wei (1992), Kat and Roozen (1994), and Dravid, Richardson and Sun (1993), have also discussed the pricing of the US Nikkei puts, particularly the type I fixed exchange rate quantos. Section 5 provides a theoretical basis for comparing the three types of puts. Their values depend upon the NSA volatility as well as possibly the exchange rate volatility between the yen and the home currency (US or Canadian) and their interactions. Using typical parameter values it is shown that type I puts should be priced higher than type I] and in turn more than type III. This is in contrast to the actual pricing where the type II fixed exchange rate options traded for more then the ordinary type I puts during the study period. However, even though they were both over- priced in relation to historical volatility the type I and type III Canadian put warrants were correctly relatively priced by the market. Section 6 discusses the put warrant risk arbitrage and the convergence to efficiency of the two mispriced markets: the Canadian versus the US and the US fixed versus non-fixed exchange rate puts. Relative option costs for fixed NSA volatility and exchange rate volatility as well as implied volatility comparisons are made. The preference for fixed exchange rate options which applied to puts also applied to calls which began trading in April 1990. The mispriced securities we discuss are referred to as hedge candidates although in most cases they are close to arbitrage. Classical index arbitrage is actively pursued in Japan especially by the foreign firmas, sce for example Miller (1993).248 Shaw et al Discussions of the potential profitability of such arbitrage appear in Brenner, Subrehman} Uno (1989, 1990), Brooks and Yamada (1990), Lim (1992), Chung, Kang and Rhee (1992 Section 7 briefly discusses the relationship between the NSA put warrant prices in North America and the next day's cash market in Japan, With deep in-the-money options, the put discount or premium signalled the up or down direction of the NSA on the next day in Tokyo. For a small data set the conclusion is that the signal was cortect 68% of the time, This is consistent with the conclusion that futures hedging of these instruments had a strong effect on the Japanese stock market. Gruca and Ritchken (1991) also noted similar behaviour on the opening prices in Japan. A discussion of implications of the findings and concluding remarks appears in Section 8 m and 2. The Nikkei stock average 1949-1995 and its historical volatility The NSA is a price weighted average of 225 large capitalized stocks traded on the Tokyo Stock Exchange, It is defined as NSA, = = Be where D, = divisor at time 1. The original divisor was Diggy = 225 and Diyoo, pe = 9.967. Figure L shows the NSA from July 1983 to June 1995. The NSA was 109.9 when it began trading in May 1949. It peaked at 38916 at the end of December 1989. There were twenty declines” of ten percent or more during 1949 to 1989. The index rose 220.84 times in yen and 553.04 in dollars from 1949 to 1989. There were nine declines of ten percent or more during 1990-92. The index fell to 16925 at the end of 1992 a decline of 6.5% from the December 1989 high. Investors from 1949 still had 96.21 for cach yen invested and 277.53 for cach dollar invested. The 1990-92 decline had its minimum at 14309, a decline of 63.2% from the December 1989 peak, on 17 August 1992. ‘There is a very active index arbitrage market in the NSA which has been studied by Brenner, Subrahmanyam and Uno (1989, 1990), Chung, Kang and Rhee (1992), Miller (1993), and others. The value of the futures volume on the NSA trading in Singapore, Osaka and Chicago is the highest of any equity index in the world. ‘The press called the stock market decline during 1990-92 the bursting of a ‘speculative bubble’. See also Tachi (1993) for a similar conclusion by the Ministry of Finance. Ueda (1990) and French and Poterba (1991) among others pointed to the very high Japanese stock prices with price earnings ratios of seventy or higher in 1988-89. Stone and Ziemba (1993) have analysed the steep rise in the stock and land markets during the 1980s in the era of cheap and readily available joney and the subsequent steep deciine largely caused by the Bank of Japan's tight money policy of raising interest rates and decreasing the supply of money. They concluded that the decline in the stock and ‘essential use’ land markets can be explained as an adjustment to changing fundamen- tals. Speculative land such as the membership prices of golf courses and condominiums, on the 2 4 decline is defined as the peak to velley when the fll exceeds ten percent end any subsequent rise would invalidae the ten percent falRisk arbitrage in Nikkel put warrant market 249 40,000 35,000 A 30,000, H 10,000 5,000 0 sna Ln LLL 8 84 85 86 87 88 89 % MM 92 93 94 Jol Jul Aug Avg Aug Aug Avg Avg Aug Aug Aug Jul 31-29 4 03007, 0G— Sh Fig. 1. The NSA weekly close July 1983—June 1995, other hand, appears more likely to have been a speculative bubble, Table 4 (see forward) points to the high stock prices relative to past levels at the end of 1989 This paper studies the period mid-1989 to mid—1990. During 1939 the historical volatility was in the 10% range or slightly below its 1949-1989 average of 13%. Volatility has not beer constant. Figure 2 shows the monthly averages of daily volatility from May 1949 to April 1989. While volatility peaked at 73.5% in October 1987 most of the time the annualized standard ation was less than 20%. Volatility tends to rise in declining markets (Schwert, 1989; Turner and Weigel, 1992), The 1990-91 period in Japan had historical and implied volatilities 60% range for much of this period; sec Fig. 3. 250.2 2 os (Bdu-n ) NSA, NSA,—, where n= re01 for the last 20 trading days of the month, and NSA, is the closing value of the NSA on day i assuming there are 250 trading days per year.250 Shave et al " | Fh ll jl hy wh eae ‘| Fig. 2. NSA historical volatility, monthly averages of daily data annualized in percent, May 1949-April 1989. Source: Jun Uno, Nihon Keizai Shinbun, Ine. 3. NSA put warrants on the Toronto and American Stock Exchanges 1989-1990 ‘The three year American style NSA pul warrants are of three basic types. Let ? puice and NSA, the expiry price of the Nikkei stock average in yen, Let Hy be the current exchange te and EB, be the exchange rate on expiry for Canadian or US dollars into yen. The symbol (x) the greater of X or zero. Then in yen we have using Rubinstein’s (1991) classification of exotic options, where @, 6, and ¢ are constants. Ag be the strike Value on expiry ‘Type of put Currency tisk in U.S./Canadi dollars? L ordinary yes 1 INSAy — product no NSA Ie «2 (0S & oo option to exchange yes, in index value and in this difference with the strike price converted to the home currency It is convenient to velue the puts in yen as discussed in the next section, In their home currency (U.S. or Canadian dollars) using the symbols defined in Table 1, the puts (inchiding the PaineRisk arbitrage in Nikkei pue warrant market 251 0.8 199 1990 Fig. 3. Historical and implied NSA volatility 1989-1991. (—) Trailing 21-day historical volatility, (=) Osaka implied volatility. Implied volatility is the average of closest to money NSA puts and for the nearest month using the Gensaki as the interest rate, Source: Baring Securities. Webber and Sglomon ealls) are: Puts Calls ISH. — + I o( SAA) BT-I, SEK, BTB, London OTC PWA * SAAT wee) BY-III, BT-IV, TFC, DXA, EXW, SXA, We SEL - SXO, PXB Sal Me =) BTA put warrant payoff function can be written as Linx (Sitike-underlying, 0) Curzeney Units That is Z put options oa the named underlying denominated in the specified currency unit. The actual payoff may be in different currency units converted at exercise or expiry at the exchange rate prevailing then, The warrants may be traded in different currency units. Neither of these aifect theShaw et al Table 1, NSA classification of put and call warrants trading ir Canada end the US in 1989-90. Warrant Payot Canadtan puts Bankers Tnwst-I(RP-T) Cdn equivalent at rate then prevailing of yen 0.1168 G2 174—NS4j* Bankers Trst-IE (STAI) Cn 0.1031 (70.54¥754,/8,)" where Ef the number of yen por Canaan dollar at exercise Bonkers Trust-Ill (BT-t01) 1.25% (37 4603: Bankers Trust-1V (BT-1V) Tallon (TFC) ‘Can 82.7547 % ($7416.32 ~ NSA)" 37 416.32 = 00010987 (57416.32— 1 SEK Can equivalent at rate then prevailing of yen 0.1168 (35963.74 — NSay" US pute Kingdom of Denmati (DXA) USS 0.2 (7516.77 —NSA)*/145 33 = 0.0013962 (375156 Selomon-1 (8X) USS 02 (682 14— NSA)*/14582 = 0.0013744 (35821.14 —W: Bankers Trust (BTB) USS equivaleat at rate then prevailing of yen 0.5 (3720642 Sclomton-II (SXO) US$ 02 G7 471.99 NSA)" {144.55 = 00913836 (7471.99 Paine Webber (PXB) USS 0.2 (19246.06-— NS.) /159.80 = 0.001232 (29286.06— 5.4)” Sélomon Warrant OTC, yen 1.0 (32805 NSAP Cordon A'S Eskportfinens (EXW) USS 0.2 (2942458 NSay*/158.84 = 0.0012591 (29424 58 —NS)* US calls Salomon (8X2) 1/15 (vA ~ 28442.94)" 1/1388 = 0.90041982 (S.A ~ 28 Paine Webber OWA) USS equivalent at rat then prevailing of 1/10 (SA ~ 2249.06) underlying put option, but its pricing changes. Thus the fixed characteristics of each warrant namely: the leverage factor (L), the strike price (X), the name of the underlying (Under), and the currency unit of the underlying (CU) places each of the warrants in the standard form shown in Table 1 4, Fair valuation of NSA put and call warrants All of the warrants involve the NSA index and are American type and are valued in yen and may involve an exchange rate. Boyle's (1988) generalization of the Cox-Ross-Rubinsiein (1979) binomial lattice model was used 10 create 3-dimensional lattices to model the evolution of the NSA and the exchange rate and their interaction over time. In terms of calculation steps for time steps, the CRR is of order ” and Boyle's is of order n°. The value of the option is the expected present value of the option payoff in an economy in which the drift of a risky asset is the risk-free te minus its dividend yield. The discount factor used to calculate the present value of the payoff is the risk-free rate. For the dividend yield, we use the foreign interest rate. The NSA dividend yield during [989 was about 0.5 %. While most of the dividends are paid in March and September, sez ‘Ziemba and Schwartz (1991), the continuous approximation used is accurate because the yield is small. We ignore the typical one day or longer time lag between giving notice of exercise and actually being cashed out and other special provisions of the warrants including the credit r of the issuer. We also ignore the possibility of predicting the mean return and use standard analysis; see Grundy (1991) and Lo and Wang (1995) for models based on return predictability.Risk arbitrage in Nikkei put warrant market 253 4.1. Type I put warrants The exercise value is Yen Z Max{(X — NSd,),0). The currency plays no role. At expiry one converts yen immediately into dollars. Hence one has a standard put option which may be valued on a CRR lattice in yen, 4.2. Type I put warrants ‘The exercise value is Yen ZL Max [(X — NS4,), |, Ifthe Cov ( 0 then the put is an American style on the NSA which may be valued on. a CRR lattice." Since the puts are American, lattice methods are required for accurate evaluation. The interest rate used is that in the foreign pay currency. The dividend yield is replaced by the actual dividend yield on the index plus the interest rate differential. The curency tisk may be hedged away by paying an extra dividend yield equal to the differential. By hedging into the foreign currency, the appropriate discount factor is the foreign risk free rate pe Il warrants are generalizations of type I warrants in that a type I warrant is a type TT ant for which the payout currency is yea. r war 4.3, Type IIT put warrants The exercise value is Yen L Max [(XE, ~ NSA, 0], These warrants are fundamentally different and only the Canadian BT-II is of this type, They have fair values above intrinsic even if the NS has zero volatility and they may be valued as an option to exchange with a minor modification of Margrabe's (1978) formula. Two risky assets a, 6 with values S,, S, have payoff = Max (0, 53 — Sa) = S, Max (0,1 — S./Ss) . = 8, Mex (0, 54/So— 1) Thus aa option to exchange may be regarded as a put on the value of S, denominated in units of Sp ora call on the value of S, denominated in units of $,. Let 0 and o,, be the volatility of a and 6, espectively, and p be the correlation of the log of the price relatives. Margrabe’s formula requires the dividend yield to be zero in the Black-Scholes put and call pricing. Using the notation: pricing formula (option type)(asset, X, T, ¢, Div 2, Div A) asset, BS put (Sa/Sp, 1, T, 0; 0, 0) S, or equivalently BS call (Sy/S,, 1, T, 0, 0, 0) Sy, where » Data shows that Cov (NSA, SCdni¥) & 0 and Cov (NS4, SUS) & 0. If the Cov (NS4, £) £0, then one may value liwse puss on a CRR lattice by adjusting oys4 tO Ouse +P cyssce Wilh the dividend yield equal t0 dyse + Fosascay~ ‘upen + 9 2ns4 045 880 68, Reiner (1992) for proof. Thus a dimensional laitice is not needed. “A closed form solution exists for this product option in the European ease assuming log normal NS4 and log normal currency changes since the product of Iognormals is lognormal; se e.g. Merton (1975). This is develo cally in Grucs and Ritehken (1991) and Clyman (1991, 1992). The later author also develops arbitrage selatioaships updating the Merton (1973) analysis to this case, See also Reiner (1992), Dravid, Rishardson and Sua (1992), end Kat and Roceen (1934),Shaw et al eas oes Aat Sosa erst Ke ores vers Ad — 16URF 9 st9s. wsis Adt wv ye 90RZE sa seuz 00°08 sezs Ad oe1sz sas ‘an'0s oss Ad 06 reote sn 0007 E0928 ues Ad HPI IC —OSLSE sn osse srs ies uso ce-ty9 epyse sn ose wuts uss Adi G6dYR GHZET v0 sn oes oo 0s spis asn co-dy3 orzor gtsc1000 sn scre Eels orsis Adl — CEMP OT — HOLE so sn 00c02 wis Bus sn f693 91 CUyLS “ERETOOO sa Sis6 298. Boe F698 GSN CE MPE SI TBO HLETOOO sa 0006 Lis wee BLES ash eeuere 9Ises LELOOU sa 00001 ress Al 16S Adt 26 AONOT p9gSe SOITO. = VD. O00 yess HIE LESS dv s69° CM OSL eBrOTO"D ©=— VO. (0008 zrss “9 ST'SS VD S6IPNST —OSLESDTEDOGD «= VDD oo'0s 6 LCS dyo tuys! eysee GSTIO0 ©=— dvO.—oes't seos he OLS Adf %6Qdi1 pee DIO. VD_oase’z nyo 10 ud Aqdxy py Morar ss fo aay, pupLay opereanng ome tot Ginny By Mat J2qUNDD-21-40A0 254 puoy] pue UEDUaWLY “oWUOLe,|, 24) UO S| 2 pur sind WSN Peper Afeatox Jo ser ‘O66T Aig €z “soBuEypxA Joos aJ0a poud PUL S901 °Z AQUI,Risk arbitrage in Nikkei put warrant market 255 In a risk neutral economy the drift of S, and S, is r, the risk free rate, thus S/S, has drift of zero. Margrabe’s formula can be extended to dividend paying assets with payoff X(S)~ S,), where ¥ is a constant as follows. Since S/S, driits at rate Div ~ Div a where Div is dividend yield. The discount rate is Divé singe the value is measured in units of S,. The value of an American style option to exchange 5, for 5, is then S,CRR (opt type, S,/Sp, X, T, 8, Div B, Div A). Hence the pricing is L EX,CCR (Put, NSA,/E,,¥, 7, 0,i¢, Div NSA) The various trading prices, currency of issue, strike prices, leverage values, expiry dates, implied volatility, 18% volatility price, relative cost (% above or below 18%, volatility price), $% delta and intrinsic values for the various NSA put warrants arc illustrated in Table 2 for 23 July 1990. The basic data is NS = 31895, Cdn$ =¥ 148.13, with interest rates of 11.75% Canadian, 7.361% US and 7.319 % Japanese 5. Numerical comparison of warrant types 1, II and III The three types of warrant puts may be compared as follows, Assume that the American puts have 4 two year exercise period, the home curreney is normalized at 1, the NSA is 100, the Japanese interest rate is 6%, the foreign (Canadian or US) interest rate is 10%, the NSA has « continuous yearly dividend of 0.5% and the standard deviation of the NSA is 20%. The relative values of the three types of puts vary with different assumptions on the volatility of the exchange rate and the covarianee between the NSA and the exchange rate, Assume that the volatility of the exchange rate is 5, 10, or 20 % and that the covariance of the NSA and the exchange rate is 0.5, 0.0 or 0.5 Table 3 contains fair values for these warrants in terms of percentage of the NSA. Table 3°, Comparison of fair values of NSA put warrants, Type ‘Volatility of the exchange rate 5% 10% 20% 05 744 144 7144 Cov (Ns, £) oo 744 7.44 144 os 744 144 744 b. Type 11 ‘Volatility of the exchange rate 15% 10% 20% 05 736 137 80) Cov (NSH, B) 00 713 Ti 23 0s 695 677 6a ©. Type IL ‘Volatility of the exchange rate 5% 10% 20% ~05 7103 8.65 12.31 Cov (NSA, B) 00 602 678 931 05 430 4.59 575 ‘The accuracy of these values depeads upon computer implementation including the number of nodes in the binomial iner in ¢ private comespondence in 1993 found the following similar values using @ 1 e wit, continuous rates of 0.06, 0.10, and 0.005 corresponding to annualized rates of 0.0161366, 01081709, and 0.0059125 all 7.432 “Type Ill: 7.091 ant 1367 1915 aol 6.061 eas 1.167 767 7.167 4933 4932 693 6.786 6431Shaw et al When priced in yen, type I do not involve exchange rates, hence all warrant values are equal With zero covariance between thc NSA and exchange rates, the values of type II warrants are the same regardless of the volatility of the exchange rates, This value is less than nat for type I warrants because of the positive differential between the foreign interes! rate and the Japanese rate. However if the covariance is non-zero, then the value of type IT warrants depends both on hat covariance and the volatility of the exchange rate. In general, the value of the warrant nnereases with the volatility of the exchange rate and for put warrants, decreases as the correlation increases. Positive correlation means that negative returns on the NSA are assaciated with a strengthening of the yen. The investor receives returns if the NSA declines so if this is accompanied by a stronger yen, the payoff is less at exercise than otherwise would be received Type ITT warrants have the most interesting behaviour. Their value depends on the volatility the exchange rate even when the correlation is zero. In general, the higher the correlation, the lower the value of the warrant, With positive correlations, the value of the warrant for both low and hi values of exchange rate volatility is higher than that for the intermediate. For typical observed parameters — covariance zero, exchange volatility about 10% and foreign interest rate above Japan’s — type | warrants are generally worth more then type II warrants which are in tum more than type III warrants, all other parameters (leverage, strike price, time to expiration, ete) being equal. There is a similar relationship between the type I (Paine Webber) and type [I (Salomon) calls (see forward to Figs 9 and 10). 6. Constructing the put warrant hedges and the convergence to efficiency The various exchange traded puts in Canada and the US and the over-the-counter puts traded in London had many common and several different characteristics that led to significant price differ- ences. Reasons for the price differences from fair values include currency and cross border risks, different credit cisks, difficulties with borrowing for short sales, price effects due to the differing size of the warrants, differing strike values, inability to value the warrants properly, differing exercise provisions, market sentiment and volatility differences. The London over-the-counter market was active in 1988 and 1989 for large institutional investors, Prices were quoted by the market makers based on historical volatility (in the 15% range) plus a profit margin, Selomon Brothers lesser extent Bankers Trust, made the market with large bid ask spreads as shown below. On November 1989, the NSA was 36484 and three of the Salomon Brothers over-the-counter put warrants were priced as follows, a Price in dollars ike Expiry date Price in yet 24 Nov 1989 Oct 1990 806 24 April 1992 ¥ 934-970 $6.50-6.75 $65.00 31033 7 Aug 1990 503-575 3.25-3.75 Expired 28,139 19 June 1991 75.5-76.5 0.40-0.45 $39.50 By 23 October 1990, the puts had increased at least ten times and the third issue nearly one hundred times.Risk arbitrage in Nikkeé put warrant market 237 ‘The Canadian put warrants BT-I and BT~II were the first opportunity for non-institutional Canadian and US investors (three months after issue) to profit from a fall in the Japanese stock market. BT=I was issued in February 1989 and BT-II in June 1989. These warrants were very popular with investors and waded for very high premiums and implied volatilities. There was considerable good reason to believe that the Tokyo market was overvalued. See for example French and Poterba (1991) for one analysis based on adjusted price earnings ratios and Ziemba and Schwartz (1991) for a synthesis of various studies. One way to evaluate this is via Paul Aron’s adjusted price earnings ratios which are comparable to French and Poterba’s adjustments although somewhat lower. Aron (1981, 1989) computed these ratios to the end of August 1989. His values are shown in Table 4. His adjustments reflect different accounting and business practices, cross holding effects and different capitalization rates. Ziemba and Schwartz (1991) updated Aron’s adjusted values after August 1989 with assumptions concerning the earnings change of the NSA and capitalization rates. The values are shown in Table 4 up to 22 February 1991. The 31 December 1989 value of 23.9 was the highest adjusted price earnings ratio at any time since 1949 and pointed to extreme risk in the stack market. Despite its decline during 1990, it was not until the steep decline on | October that these values decame cheap celative to historical price earnings ratios, Other stock mazket valuation models such bond and stock yield differences, see Ziemba and Schwartz (1991), also were at historical high values at the end of 1989%, All of these models are driven by two factors: earnings forecasts and interest rates. The extreme inerease in interest rates in 1989 from a 2.5% discount rate to 5.25% and later to 6.0% was at the heart of the estimated overvaluation In a multivariate factor model regression study for the period 1979-1989, Zicmbe (1989) found that future earnings forecasts were by far the most important variable for predicting the rates of return of Japanese stocks. Hence there was considerable reason for investors to believe that the Japanese market would crash or at Icast decline sharply. Since the Canadian puts were the only product available to invest in this belief, their prices were understandably very high, particularly given that it was difficult for nearly all of the purchasers of these puts to fairly value thera, Seasonality observers also noted that the decline in January 1990, while only 4.5 %, was a key negative signal since January has histori- cally provided the highest returns in the Japanese markeis, see Ziemba (19910). Moreover, the conditional probability of a decline in the rest of the year following a decline in January is quite righ; see Hensel and Ziemba (1995). Table 3 shows that the fair values of type I warrants generally exceed those of type II which in turn exceed those of type I]. The BT-1 is a type I and the BT-II a type III. In terms of premium, see Table 5, the BT~I was priced higher than the BT-I1. This was the case in the entire trading period from September 1989 to the eventual collapse in February 1990. Table 5 provides insight into pricing differences for Canadian—US hedges (A) and US hedges (B), but those provided from theoretical option pricing models, as we now discuss, were used in our analysis. For exemple, shorting of BT-I or BT-II and buying BT-US looks like a potentially profitable cross border Jhenever the difference was more than two standard deviations above the mean, the market zone”. This model was applied throughout the 1949-1989 period. During this time there were 1 cery time the market was in the danger zone during this 40 year period, x decline of 10 % or more occurred. There wore declines of 10% or more without the difference being the “danger zone”Shaw et al Table 4. Paul Aron’s adjusted PERs for Japan compared with those in the US, 26 April L981 co 3 August 1989, with adjustments for later periods to 22 February 1991 by Ziemba and Schwartz (1991) Date US PER Sapa, adj PER NSal 28 Apr 1981 8 3 Oct 1984 929 Apr 1986 1582 2b May L987 24538 1 Sept 198 2489 31 Dee i987 21333 31 May 1988 v6 A Aug [488 679 I Aug 1989 34808 31 Dee 1989 38915 Mar 1099 20080 Jone 1990 31604 = expense opt 1999 1 Oct 1999 2 Oat 1990 22396 31 Dee 1990 23899 22 Feb 199) 25903 ulves after August 1989 assume: fraerest Eavsings ga Date vs rer Aug 89 Dee 1989 82 64 5 Mar 1990 84 74 ae 1990 82 10 Sept 1990) 82 115 Ger 1990 BL 18 et 1990 BL 13 31 Des 1999 16 6s b 1991 16 60 hedge wade. Similarly shorting Kingdom of Denmark or Salomon I and buying BT-I looks like 2 potentially profitable hedge uade on the Ameri heor stack exchange. Figures 4 and 5 show the etical pricing in (wo ways. Implied volatilities appear in Fig. 4. They illustrate the point However, implied volatilities did not exist at many dates in 1990 when the puts were Uading at discounts (as discussed later in the paper); see the vertical lines in Fig. 4. Hence, a preferable way 10 pare the warrants prices is by their relative cost. That is actual cost minus theoretical value as &Risk arbitrage in Nilekei put warrant market 259 ‘Table 5. Comparison of prices and premium values for four Canadian and (ree US NSA put warrants on 1 February 1990. F ¥ % of ig actions Nsa Expiry Years t@ Premium —Premiwn ~ Wearrane Price anit date expity —% per year A 8 BT-t $2.70edn 11.68% 205 20.1 98 SELL Br S193edn 10.31% 237 164 59 SELL BT-IIl S2s0edn 14.29% 216/93 3.05 20 23 Teilon Fial S27Sedn 13.7% 2/2203 3.05 24 Kof Denmark $5.63us 20% 1593-293 to 34 SELL Salomon Sodus 20% 1993-297 toa 34 SELL BT-US S9.lTas 50% 3 300 0 26 BUY BLY percentage of theoretical value. This is shown in Fig. 5 assuming a volatility for the NSA of 20% and an exchange rate volatility of 10%. There were no NSA put warrants trading in the United States until the Kingdom of Denmark type 11) put warrant began trading on the American Stock Exchange on 3 January 1990. The Salomon A (type Il) and Bankers Trust (type 1) put warrants began trading two weeks later With the availability of these three warrants investors in the Canadian put warrants could replace these warrants with the much cheaper US instruments. Figures 4 and 5 show that it took more than a month for the Canadian puts to converge to efficiency. A gradual decline began with the introduction and market knowledge of the three cheaper US instrament and then there was a sudden collapse in late February 1990 just after the second Salomon pu ‘warrant (@ type Il) began trading. The slowness of the market to react to new information gous 10 that of stock prices which are frequently slow to react to new earnings information, see Affieck-Graves and Mendenhall (1992), Bernard and Thomas (1990, 1992) and Aberbanell and Bernard (1992). ‘The market needed time to understand, evaluate and then fairly price these complex instruments, e collapse occurred at a time of minor decline in the NSA in February 1990 well before the steep declines in March and April 1990. Hedge investors who were able to short the Canadian put warrants and buy cheaper US warrants particulerly the Bankers Trust BTB could have made considerable profits’ A second advantageous hedge is illustrated in Figure 6, Despite the fact that the theoretical fair values of type I warrants was larger than type Il, US investors had a preference for type I inst ments. Apparently they preferred a fixed exchange rate in dollars upon expiry rather than to value the puts in yen. To eliminate the currency risk, they paid more for type II warrants than if they had bought type I warrants and hedged the currency risk in the futures market, Hence type [I warrants traded for prices which were much larger than those of the type ] warrants. There was also a price fis hedge had relatively low risk but was not a true arbitrage given that there were different credit risks and othe: characteristics of che various put warrants, There was also the difficulty of securing end holding borrowed warrant short positions, The threat of forced buy-in wes also present. All thrce authors did have foxced buy-ius of shott positions, but the amount was small so that the overal] hedge was very suecessiul,260 Shaw et al NKP vs NKP.A vs BTB 4s Implied Volatility (@Val hh 108, 30 30 1S-uu plicd volatility of BT-I, BT-IL, and BTB NSA put warrants assut xchange rate voltility of *%, 17 February 1989 to 21 September 1990, (Q) BT-L, type 1, Canadian, (+) BT-Il, type 3, Canadian, and ) BTB, type 1, US. effect. The BTB warrant represented 0.5 of an NSA unit and the DXA, SXA and SXOs were worth only 0.2 ofan NSA. Hence the BTB should trade, other things being equal, at about 2.5 times plus or minus a transactions cost band around the other warrants. In fact the BTB usually traded at prices much lower. This is analogous to the low priced stock effect that captures much of the January small firm effect, see Blume and Stambaugh (1983). These two factors yielded the profit- able hedge from January to March 1990. After convergence to efficiency these markets have since generally traded within transactions costs bands.® Figure 7 shows the relative costs of various put warrants in Canada and the US with similar sirike prices. These warrants were all issued in early 1990 and had NSA strikes between 36822 and © Additional analysis of the post March 1990 period for vasious US NSA put warrants , particular (Clyman (1991, 1992), Chen, Sears and Shabrokhi (1992) and Dravid, Richarison and Sun (1993) an speshing, after transactions costs ere considered, the market was efficien of type IL, appears in ug others. Generallyarbitrage in Ni kei put warrant market 261 Rel Cost @ Nik Vol 20%, Exch Vel 10% -89 13. 07-Dec—89 04-May-90 | 01-May-83 26-Sep-88 22-Feb-90 18-Jul-90 Fig. Relative costs of BT-I, BT-II and BTB NSA put warrants with NSA volatility of 20% and & rate volatility of 10%, 17 February 1989 to 21 September 1990. Relative deviation from model price (actual cost — theoretical value)/(theoretical value). (+) BT-I, type 1, Canadian, (©) BT-II, wpe 3, Canadian and (4) BTB, type 1, US. 37472. Here are shown the higher prices paid for type If warra: January to the end of February 1990. Figure 8 shows the relative costs of Canadian type I, Il and IT NSA put warrants. Investors, relative to fair prices paid more for type I (BT-I) than for type TH (BT-II) until the market converged 10 efficiency in late February 1990. From March to September 1990, all three 1 of put warrants had relative costs within transaction cost bands. Figures 9 and 10 give the implied volatilities and relative costs of the two NSA call warrants taded on the American Stock Exchange. The Paine Webber call is a type I and the Salomon is a type II. The fair value of a type I should be higher than a type II. However, US investors preferred the type IT with its fixed exchange rate of dollars into yen and bid its price higher during most periods from April to October 1990. in comparison to type I from262 Shaw et al Ral Cost @ Nik Vol 20% raul ny, cal yey one } “snl “ -o.2 | 9g = Lass (03-Jan~-30 14-Mar-90 25-May—G0 08—Aug—30 Fig. 6, Relative cost of US type I (BTB) versus US type Il (DXA, SXA, SXO) NSA put warrants, Ja September 1990, assuming NSA volatility of 20%, (Cl) BTB, type 1, 0.5 NSA, (+) 2 0.2 NSA, and (—) normalized Nikkei ry to g DXA, SKA, SXO, type Table 6, Percent of days the intrinsic value exceeded the market value (Source: Cly: DXA SYA SxO PXB ExW 0.0 00 na 00 00 00 2 Mar 132, 27 273 Apr 65.) 70.0 88.0 May 409 40.9 499) Tune 50 00 00 July 0.0 00 90 Aug 435 435 as Sep 519 52.6 52.6 Det 522 478 32.2 Average 2.1 29.5 358 Exercise Pri 37817 36821 31472Risk arbitrage in Nikkei put warrant market 263 (yen Zeno vol 10%) Of | = x 4 1 1 @3~Jan—20 14—Mer=90 25-Moy-30 08-Aug-30 Fig, 7. Relative costs puts with similar strike prices, assuming 20% NSA volatility, Yen/Canadi ol of 10%, Janus 9) BTB dollar ry to Septernber 1990. (O) BT-II1, (+) TFC, (0) DXA, (4) SXA, (x) SXO, anc 7. The relationship between NSA put warrant prices in North America and the cash market in Tokyo During much of 1990 the Toronto and New York NSA put warrants were trading deep in the money. Frequently the puts traded for less than their intrinsic value, see Table 6. Tokyo’s next trading session was the following day. Since much futures hedging was required to protect the ssuers’ positions, and that trading would lead to index arbitrage if the futures prices deviated much from fair value, the prices in North America provided a forecast of the likely prices in Tokyo. If the put was trading at a discount one would expect the Tokyo market to rise. Similarly the forecast was for a fall in the Tokyo market if the put was trading at a premium. An indication of the size of the market is that during 1990 the NSA puts averaged 13% of the trading volume (and a similar fraction of the trading value) on the American Stock Exchange. Informal estimates by the authors of the size of the market in the US and Canada suggests that it was possible that264 Shaw et al Cdn Type | vs Type Ill vs Type tls 9 |= ~04 : 03-Jan-90 14—-Mor—90 25-Moy—90 08-Aug-30 Fig. 8. Relative costs of Canadian type I, I] and III put warrants based on 20 % NSA volatility and 10% Yen, ‘adian dollar volatility, January 10 September 1990. (+) type II, avg BT-IIl, TFC, SEK no Volume, (©) type 1, BT-I, (4) type IN, BT-Ul and (%) normalized Nikkei upwards of 20% of the NSA futures uades in Osaka and Singapore were related to NSA put edging. Consider for example, the SXA Salomon January 1993 NSA put. It had a strike price of 36821.14, a currency conversion rate of 145.52 yen per US dollar, and is worth 0.2 of an NSA unit. The intrinsic value of the put was 2.(36821.14 — NSA) 14552 Hence the implied NSA is 36 821.14 — 5 (145.52)P. Table 7 shows that on the seventeen of the twenty-five trading days from 1 August to 6 September 1990, the forecast was correct (six of the incorrect predictions were expected rises). There are many control aspects to a full study of this relationship such as open versus closed prices, futures effects, are futures and warrants giving the same estimate of cash prices, etc? * For some progress on this see Dravid, Richardson and Craig (1993) end Yuen (1994).Risk arbitrage in Nikkei put warrant market 265 hnplied Yolatitt wl i if 0 | spunermemuneonasusteuectansath——th—h-— ees 17—Feb-€9 19-Apr-89 19-Jun-89 22-mug-eg | a3-net-Rg 20-Mar-89 18-May-89 20-Jul-89 2 Fig. 9. Implied voletility of the Paine Webber and Salomon NSA call warrants, April to October 1990. ( PXA, and (+) SXZ. However, there seems to have been a strong relationship between the price of the NSA put warrants in North America and cash prices in Tokyo on the next trading day 8. Implications of the findings and concluding remarks The paper has described two Favourable hedges involving Nikkei put warrants during the period November 1989 to February 1990. The cross border hedge involved shorting overpriced Canadian Nikkei put warrants which traded on the Toronto Stock Exchange and purchasing either Nikkei puts with negotiated terms over the counter in London or exchange traded puis on the American Stock Exchange. Since the Canadian puts were unavailable to US investors for three months from their issue in February and April 1989 they were not heavily advertised or known in the United States. US residents and citizens could have traded them at the time of the hedge, however. The266 Shaw et al. Rel Cost @ Nk Vol 20% -o.4 | =08 + + — o3-son—90 | 02-Mor-s0 | 02-Mey-30 | 03—sul-90 04-Sep-99 O1-Fev—so ‘02-Apr-20 01-Jun-30 02-Aug-20 Fig. 10. Relative costs of the Paine Webber and Salomon NSA call warrants using 2: volatility, April to October 1990. (O)) PXA, (+) SXZ, and (—) normalized Nikkei NSA historical easons for the mispricing are several. The puts were complex for most ordinary investors and all duit experienced option traders in Canada likely evaluated them incorrectly. Evidence of this is found in the literature on them from various Canadian brokerage houses. Many investors in Canada and academics ~ see for example, French and Poterba (1991) and Ueda (1990) ~ were quite convinced that the Japanese market was overpriced. Even the Canadian investors bidding up of the price did not prevent them from making considerable profits later when the Nikkei fell sharply". The Canadian puts finally declined into their theoretically correct pricing about a month after the US puts were trading on the American Stock Exchange. ‘The studies of Bernard and Thomas (1989, 1990), Affleck-Graves and Mendenhall (1992) and Aberbanell and Bernard (1992) show the slowness of individual stocks to react to new earnings information. Hence, it is not surprising that this convergence to efficiency of more complex cross border investments would take about a month to occur. '® According to Slocum (1993) investors in the four Bankers Trust warrants made a total profit of about CdnS500 million,Risk arbitrage in Nikkei put warrant market 267 Table 7. The Selomon Nikkei January 1993 put warrants record at predicting the following day's change in the NSA in Tokyo, 1 August to 6 September 1990. Date SHA implied Nikkei Nikkei close Predicsion* 9.375 30000 29181 28.908 27635 28600 28099 27653 28126 28509 x 2827 27616 x 27453 27330 x 26817 27453 27817 26 125 14.125 26544 B30) 26998 x 1475 26089 1.00 23179 18.125 23633 x 16.625 24725 14.625 26180 Iss 25543 16.00 25179 160 5179 x 15.625 25432 closed na 16.25 24997 rise X i700 24482 re X 1735 23905 23812 * The prediotion is for « fal (rise) in the Nikkei on day ¢ + 1 ifthe implied Nikkei on day «is below (above) the close on & Souree: Mocified from The Wall Streer Jeurnai (1990), Interestingly, Bankers Trust also issued Canadian dollar against the US dollar put warrants in June 1990. These traded oa the Toronto Stock Exchange al a time when many Canadians expected « sharp decline in the Canadian dollar while US exchange traded options on the Canadian dollar were actively traded. These puts were also overpriced and they stayed overpriced for the entire year until they and the US puts expired worthless in June 1991. This latter case has some parallels with une Nikkei put hedge but important differences, With the Canadian dollar puts the difference in price could be explained by the fac: that it was exiremely difficult to short these puts'". For those that did, including two of the authors, there were considerable profits in a percentage but not absolute basis, In contrast it was not difficult to short the Canadian Nikkei puts in large numbers. Salomon Brothers and other issuets w Uon to participate in the mispricing hedge. Presumably for business reasons concerned with such products, they did not converge the mispricing to efficiency sooner than February 1990. The market did price the relative values of the type II and type III Canadian puts correctly. Other '! Another example ia Holland, with similar mispricings related to the inability to shor! the overpriced warrants, is diseusseo by Veld and Verboven (1992),Shaw et al seasos for the lemporary mispricing of the Canadian puts as discussed in the text are the risks of buy-ins, cross border risks, small relative currency risks and differing credit risks, The latter is mi ted somew! in the hedge: sell Canadian Bankers Trust puts and buy US Bankers Trust puts. Still if an extraordinary event occurred and there was no tading in the Nikkei index the liquidity of the two types of puts could have been different. The second hedge involved securities of fixed versus floating exchange rate on the American Stock Exchange. The explanation for this mispricing which also lasted about one month seems to be the price effect and the different ways one can view the currency risk and pricing, The price effect ‘where the Bankers Trust US warrants had NSA sizes two and half times as large as the Kingdom of Jenmark and the Salomon puts is totally analogous to the effect of low priced stocks in January died by among others Blume and Stambaugh (1983). It is known that much of the January small firm effect can be equally viewed as a low price effect. Hence, it is not surprising that in the very beginning of their trading the much higher nominally priced BT warrants traded for somewhat lower actual prices. Another possible reason for the discrepancy involves the currency risk, The theoretical models assume that curreney prices are based on their forward rates, Hence, if investors were assuming that the lower yielding yen would not appreciate against the higher yielding US or Canadian dollars as evidence summarized by Froot and Thaler (1990) suggests for such cusrenei: then higher prices were warranted for the fixed exchange rate puts”. Since even this explanation, that is, assuming that the Forward rate equals the spot is not enough to explain the full extent of the gs it appears that a combination of the two effects and the premium that is warranted for t k is the logical explanation, Acknowledgements Without implicating them, the authors thank Stephen Figlewski, Stewart Hodges, Eric Reiner, Mark Rubinstein, Michacl Selby, Andy Tumer aud seminar participants at the Berkeley Program in Finance, Monterey, Harvard Business School, Simon Fraser University, the ORSA, TIMS San Francisco, University of Helsinki, Society of European Financial Modelers, Memtove, Italy and the Financial Options Research Center, University of Warwick for helpfull discussions and comments on an earlier version of the paper, References Abarbanell, JS. and Bernard, VL. (1992) Tests of analysts’ overreactionjunderreaction to earnings information as aa explanation for anomalous stock prive behavior, J. Finance, 47, 1181-2 AMeck-Graves, J. aud Mendenhall, R. (1992) The relation between the value Lin ouncement drift, J. Financial Economics, 31, 75, Amihud, ¥, and Mendelson, H. (1991) Volatility, efficiency, end trading: evidence fiom the Japanese market, J. Finance, 46, 1765-89 P.H. (1981) Are Japanese P/E multiples too high? Daiwa Secu enigma and post earnings ck, re ies of Ameriva, NY. * In retrospect, the yen did rise against the US and Canadian dollars so the fuiures prices were more accurate predictors of the prices in the future than the martingale constant price model.Risk arbitrage in Nikkei put warrant market Aron, PAL (1989) Japanese P/E ratios revisited, Duiwa Securities of America, NY Bubbel, D, and Eisenberg, L. (1991) Quantity-adjusting options and forward contracts, Working Paper, Rodney L. White Center for Financial Research, The Wharton School. Bailey, W. (1989) The market for Japanese stock index futures: some preliminary evidence, J. Futures Markets, 9, 283-95. Barron's (1990) Kamikaze instruments: index futures wreak havoc on Japanese stocks (16 April) Becker, K.G,, Finnerty, LE, and Gupta, M, (1990) The interternporal relation between the US and Japanese stock markets, J. Finance, 45, 1297-306. aud Thomas, KK, (1989) Post-earnings-announcement drift: delayed price x premium?, J. Accounting Research, 27 (Supplement), 1-36. ard, V.L, and Thomas, IK. (1990) Evidence that stock prices do not fully refiect the implications of current earnings for future earnings, J. Aecaumiing and Economics, 13, 305~40. Black. F. and Scholes, M. (1973) The pricing of options and corporate liabililies, J. Poitical Economy, 81, 637-59. Blume, ME. and Stambaugh, R. (1983) Biases in computed returns: an applicatio Financial Economics, 12, 387-404. le, P, (1988) A lattice framework for options pricing with two state variables, J. Financial and Quantitative Analysis, 23, 1-12. P., Eynine, J. and Gibbs, S. (1989) Numerical evaluation of multivariate conti Financial Studies, 2, 241-50. Brenner, M., Subrahmanyam, M. and Unom J, (1989) The behavior of prices in the Nikkei spot and futures markets, J. Financial Economics, 23, 363-83. Brenner, M., Subrahmanyam, M. and Uno, J. (1990) Arbitrage opportunities in the Japanese stock and futures markets, Financial Analysts Journal, 46, 14-24, Brooks, DJ. and Yamada, J. (1990) The relationship between the cash options and the futures, Security Analysts Jounal, 28, 24-34 (in Japanese) Chan, LC, Hamao, ¥. and Lakonishok, J. (1991) Fundamentals and stock returns in Japan. 46, 1139-64 K.C,, Sears, RS. and Shuhrokhi, M. (1992) Pricing Nikkei put warrants: some empirical evidence, J Financial Research, 15, 231~51. Chung, Y.P., Kang, JK. and Rhee, §.G, (1992) An intraday transactions data test of the Nikkei stock average index futures price behavior and index arbitrage profitability, Mineo, University of California. Riverside Clyman, D.R. (1991) Anatomy of an anomaly: or what know about fixed exchange rate Nikkei put warrants, Mimeo, Harvard Business School, January. Clyman, D.R. (1992) Arbitrage and xed exchange rate Nil School, June Cox, JC, Ross, S.A. and Rubinstein, M. (1979) Option pricing: a simplified approach, J. Finaricia' Economics, 7, 229-63 Derman, E., Karsinski, P. and Weeker, J, (1990) Understanding guaranteed exchange- stock investments, Intemational Equity Strategies, Goldman, Sachs & Co. Donnelly, B. (1990) New York bears influence on the Tokyo stack exchange, Wall Stree: Journal Dravid, A., Richardson, M. and Craig, A. (1993) Explaining overnight variation in Japanese sicek retams: the information content of derivative securities, Working Paper, ‘The Wharton School of the University of osyvania , Richardson, M. and Sua, T. (1993) Pricing foreign index contingent claims: an application co Ne put warrants. J. Derivatives, (Fall), 33-51 sponse oF tisk 1 the size effect. J. Bo: Bo: nt claims, Review of J. Finance. Kei put Warrants, Mineo, Harvard Business te contracts in fore’270 Shaw ct al Elon, Ed. and Gruber, MJ. eds (1989) Japanese Capital Markets, Harper and Row, New York Figiowski, 8. (1989) Options arbitrage in imperfect markets. J. Finance, XLIV, 1289-311 Friedberg's Commodity and Currency Comment (1990) Earthquake, 11, 25 French, K. and Poterba, J, (1991) Were Japanese stock prices too hig! Froot, KA, and Thaler, R.H. (1990) Foreign exchange, J. Eoonomic Perspectives, 3, 179-9) Grossmun, S.J. (1988) An analysis of the implications for stack and futures price volatility of program trading ind dynamic hedging strategies, J. Business, 61, 275-98. Grundy, B. (1991) Option prives and the underlying assets” return distrib Gruci, E. and Ritchken, P, (1991) Exchange treded foreign University, Cleveland August Hamaa, Y., Masulis, R.W and Ng, V. (1990) Correlations in price changes x finzncial integration. Review of Financial Studies, 3, 281-307 Haruo, Y., Masulis, RW. and Ng, V. (1991) The effect of the 1987 stock crash on international financial antegration in Ziemba, Bailey and Hemeo (199). Hatsopoulos, G.N, Krugman, P.R. and Poterba, JM. (1989) Overvonsumption: the cha eoonomi cial Eeonon ion. J. Fina cauts, Mineo, Ce nce, 46, 1045-69. Western R 1d volatility across interns enge to US ic policy. Report by the American Business Conference and the Thermo Electron Corporaiton, Waltham, Massachusetts, March, Hensel, C., and Ziemba, W.T, (1995) The January barometer: Swiss, European and globul +. Finanzmarket and Portfolio Management, 9(2), 187-96 Jacobs, BI. and Levy, K.N, (1988) Disentangling equity return regulasities: new insights and inves opportunities, Financial Analysts Journal, 44, 18-43, Kat HLM. and Roozen, H.N.M. (1994) Pricing and hedging international equity derivatives, J. Derivanives, Winter), 7-19. Kuwahara, H. and Marsh, T.A, (1992) The pricing of Japanese equity warrunts, Meniagement S 1610 vara, H. and Marsh, T.A. (1994) Why doesn't the Black-Scholes model vertible bonds? Japanese Journal of Financial Economies, 1, 33-65 Liu, K.G. (1992) Arbitrage and price behavior of the Nikkei stock index futures, J. Fun Lo, A.W, and Wang, J. (1995) Implementing option pricing models when asset reiuras ar Finance, Le, 87-129. McDonald, J. (1989) The Mochiai effect: Japanese corporate cross-hol ults, ment jes Markets, IL, 151— edictable, J Journal of Portfolio Management, 16, 90-4 ‘abe, W. (1978) The value of an aption to exchange one asset for another, J. Finance, 23, 177-86 Merton, R.C. (1973) Theory of rational option pricing, Bell Journal of Econonnics ard Management Seionee, 4, I. (1990) Investment strategy: convertible bonds and equity warrants, paper presented so ihe Be ‘gram in Finance in Asia Seminar, Tokyo, June, Miller, M.H. (1993) The economics and polities of index arbitrage in the US and Japan, Pa Journal, 2, 1-23, Mowigiiani, F, (988) Reagan’s economic policies: a critique, Oxford Economie Papers, 40, 397-426 Reiner, E, (1992) Quanto mechanics. Risk, 5, 59-63 Rubinstein, M, (1988) Portfolio insurance and the market crash, Financial Analysis Journal, 44, 34-47 Rubinstein, M. (1991) Exotic options, Mimeo, University of California, Berkeley. Schwert, G.WW. (1989) Why does stock market volatility change over time? J. Finance, 44, 1115. D, (1993) Profitable Nikkei puts expire. Globe end Mail, April, 13. Basin FinanceRisk arbitrage in Nikket put warrant market m . C. and Duna, K. (1990) Tokyo exchange pushes to halt US sales of Nikkei warrants, Wall Street Journal, April 26, 1 Stone, D. and Ziemba, W.T. (1993) Land and stock prices in Japan. J. Economic Perspectives. 13 (Summer), 149-65. ‘Tachi, R. etal (1993) The Mechanism and Economic Eifects of Asvet Price Fluctuations, Institute of Fiseal and Monetary Policy, Ministry of Finance, April, Mineo. ‘Takahashi, M. (1990) Kizon Warrant Kakaku Hyouka Model: Ichi Kosatsu, Toshi Kogaku, April, S1-80. ‘Tufano, P. (1992) Goldman Sachs and Co Nikkei Put Warrants ~ 1989, Harvard Business School Case N 9— 292-113. Turner, A. and Weigel, E. (1992) Volatility of the S&PS00. 1926-1990, Management Science, 38, 1586-609. Ueda, J. (1990) Are Japanese stock prices too high? J. Japanese and International Economies, 4, 331-70. Veld, F. and Verboren, A. (1992) Anomalies in the pricing of warrants versus the pricing of long term call options, Mimeo. University of Tilburg, Presented to the European Finance Association, August. Wail Street Journal (1990) Tokyo stock exchange pushes to halt further US sales of Nikkei warrants (April 3) Wei, J.Z. (1992) Pricing Nikkei put warrants, J. Multinational Financial Management, 2, 45-75 Wood, C. (1990) Falling apples, The Economist, 3-22. Yuen, R. (1994) The influence of the Nikkei put warrant market in North America on the Tokyo stock exchange 1989-94, MSc Thesis, University of British Columbia. Ziersba, W.T. (19892) Seasonality effects in Japanese futures markets, in S.G. Rhee and R.P. Chang, eds, Research on Pacific Basin Security Markets, Amsterdam, North Holland, W-T. (19895) Cumulative effects of fundamental variables on the Tokyo stock exchange: 1979-89, Working paper, University of British Columbia, Vancouver. Canada. Shortened version in February 1991. pp. 73-9. Full details appear in Ziemba and Schwartz (1991). Zicmba, W.T. (19912) Convergence of efficiency of the Nikkei put warrant market of 1989-90, presentation of this paper to the Berkeley Program in Finance, Montery, California Ziemba, W.T. (1991) Japanese security market regularities: monthly, turn of the month and year, holiday and golden week effects, Japan and the World Economy, 3, 119~46, Ziemba, W-T., Bailey, W. and Hamao, Y. eds (1991) Japanese Financial Market Research, North Holland, Amsterdam. Ziemba, W-T. and Schwartz, S.L. (1991) Invest Japan: The Structure. Performance and Opportu Stock and Bond Markeis, Probus, Chicago. 25 of the