Evidence of market manipulation in the financial crisis
EEvidence of market manipulation in the financial crisis ∗ Vedant Misra, Marco Lagi, and Yaneer Bar-Yam † New England Complex Systems Institute238 Main Street Suite 319, Cambridge, Massachusetts 02142, US (Dated: October 29, 2018)
Abstract
We provide direct evidence of market manipulation at the beginning of the financial crisis inNovember 2007. The type of market manipulation, a “bear raid,” would have been prevented bya regulation that was repealed by the Securities and Exchange Commission in July 2007. Theregulation, the uptick rule, was designed to prevent market manipulation and promote stabilityand was in force from 1938 as a key part of the government response to the 1929 market crash andits aftermath. On November 1, 2007, Citigroup experienced an unusual increase in trading volumeand decrease in price. Our analysis of financial industry data shows that this decline coincidedwith an anomalous increase in borrowed shares, the selling of which would be a large fraction of thetotal trading volume. The selling of borrowed shares cannot be explained by news events as thereis no corresponding increase in selling by share owners. A similar number of shares were returnedon a single day six days later. The magnitude and coincidence of borrowing and returning of sharesis evidence of a concerted effort to drive down Citigroup’s stock price and achieve a profit, i.e., abear raid. Interpretations and analyses of financial markets should consider the possibility that theintentional actions of individual actors or coordinated groups can impact market behavior. Marketsare not sufficiently transparent to reveal or prevent even major market manipulation events. Ourresults point to the need for regulations that prevent intentional actions that cause markets todeviate from equilibrium value and contribute to market crashes. Enforcement actions, even ifthey take place, cannot reverse severe damage to the economic system. The current “alternative”uptick rule which is only in effect for stocks dropping by over 10% in a single day is insufficient.Prevention may be achieved through a combination of improved transparency through availabilityof market data and the original uptick rule or other transaction process limitations. ∗ A report on preliminary results from this work was transmitted to the House Financial Services Committeeand sent by Congressman Barney Frank and Congressman Ed Perlmutter to the SEC on May 25, 2010. † Corresponding author: [email protected] a r X i v : . [ q -f i n . GN ] J a n . INTRODUCTION TO BEAR RAIDS AND MARKET MANIPULATION On July 6, 2007, the Securities and Exchange Commission (SEC) repealed the uptick rule,a regulation that was specifically designed to prevent market manipulations that can triggermarket crashes. While it is widely accepted that the causes of the crash that began laterthat year were weaknesses in the mortgage market and financial sector, the close proximityof the repeal to the market crash suggests that market manipulation may have played a role.Here we present quantitative evidence of a major market manipulation, a “bear raid,”that would not have been possible if the uptick rule were still in force. The timing of thebear raid, in autumn 2007, suggests that it may have contributed to the financial crisis. Bearraids are an illegal market strategy in which investors manipulate stock prices by collectivelyselling borrowed shares. They profit by buying shares to cover their borrowed positions at alower price. While bear raids are often blamed for market events, including financial crises[1, 2], this paper is the first to demonstrate the existence of a specific bear raid.The sale of borrowed shares, called short selling, is a standard form of market trading.Short sellers sell borrowed shares, then buy them back later and return them to their owners.This practice yields profits when prices decline. In a bear raid, investors engage in shortselling with the addition of market manipulation. Instead of profiting from a natural declinein the fundamental value of a company stock, the executors of a bear raid themselves causethe price to decline. Large traders combine to sell shares in high volume, “driving” the pricedown [3, 4].A bear raid is profitable if other investors are induced to sell their shares at the lowerprice. This may happen for two reasons: margin calls and panic. Margin calls occur whenbrokerages force investors to liquidate their positions. Investors who are confident in therising price of a stock may buy shares on borrowed funds, called “buying on margin,” usingthe value of the shares themselves as collateral. When prices decline, so does the value ofthe collateral and at some point brokerages issue “margin calls,” requiring shares to be soldeven though the owners would prefer not to. Panics occur when investors, fearing furtherlosses, sell their shares. The executors of a bear raid profit from the price decline by buyingback the shares they borrowed—“covering” their short positions—at the lower market price.In the aftermath of the 1929 market crash, Congress created the Security and ExchangeCommission (SEC). Recognizing the dangers of short selling, Congress specifically required2he SEC to regulate short selling [5]. The regulation that was instituted in 1938, the uptickrule, states that borrowed shares may only be sold on an “uptick”—at a price that is higherthan the immediately preceding price. The rule was designed to limit the intentional orunintentional impact of short selling in driving prices down, and specifically to prevent bearraids. The uptick rule was repealed in July, 2007 by the SEC on the basis of arguments thatmarkets were transparent and no longer needed the protection of the uptick rule [6]. SECclaims that the uptick rule had no significant effect on market stability, even in absenceof specific manipulation, have been refuted [7–9]. Our results implying a bear raid inNovember 2007 contradict the assertion of market transparency.Our evidence points to a bear raid on the large financial services company Citigroup. OnNovember 1, 2007, Citigroup’s stock experienced an unusual increase in trading volume anddecrease in price. To analyze this event, we studied financial industry short trading data(see Appendix A), which reveal the total number of borrowed shares (short interest) at theend of each trading day. Using these data, we show that the increase in trading volume onNovember 1 coincides with an increase in borrowed shares. Six days later, a comparablenumber of short positions were closed during a single trading day. News events to whichthese events might normally be attributed cannot account for the difference between tradingin borrowed shares and trading by owners of shares. The magnitude and coincidence of shortactivity is evidence of a concerted effort to drive down Citigroup’s stock price and achievea profit, i.e., a bear raid.
II. CITIGROUP ON NOVEMBER 1 AND 7, 2007
On November 1, 2007, Citigroup experienced large spikes in short selling and tradingvolume. The number of borrowed shares—short interest—increased by approximately 130million shares to 3.8 times the 3-month moving average. The total trading volume jumpedfrom 73 million shares on the previous day to 171 million shares, 3.7 times the 3-monthmoving average. The ratio of the increase in short positions to volume was 0.77. This is thefraction of the total trading that day that may be attributed to short positions held untilmarket closing. The total value of shares borrowed on November 1 was approximately $6.07billion. Adjusted for the dividend issued on November 1, 2007, Citigroup stock closed onNovember 1 down $2.85 from the previous day, a drop of 6.9%.3he number of positions closed on November 7, 202 million, was 53% larger than thenumber opened on November 1. The short interest before the increase on November 1 andafter November 7 are virtually identical, the larger decrease corresponding to an additionalincrease in short interest between these dates. The mirror image one-day anomalies in shortinterest change suggest that the two are linked. We can conservatively estimate the totalgain from short selling by multiplying the number of short positions opened on November 1by the difference between the closing price on November 1 and closing price on November 7($4.82), which yields an estimated gain for the short sellers of $640 million.The total decrease in short interest on November 7 exceeds the total trading volumeon that day, 121 million, by 82 million shares. This indicates that the reported decreasein borrowed shares is not fully accounted for by recorded trading on the markets. Thedifference may result from off-market transfers, which may be advantageous to short sellersin not causing the price to increase. Alternatively, despite the usual coincidence of borrowingand selling, this may be due to shares that were borrowed and returned without being soldshort. Further investigation of transaction data is necessary to explain the difference inreturned shares and trading volume.Figure 1 shows daily stock price, volume, and short sale data for Citigroup over a two-year period starting January 1, 2007. Short sale data includes short interest—the numberof shares borrowed at the end of each day—and the daily change in short interest. Duringmuch of 2007-2009, the daily change in short interest did not exceed a small fraction of thetotal trading volume. The largest single-day increase in short interest occurred on November1 and is marked with arrows in Figure 1. Figure 2 shows an enlarged view of the periodaround that date.In Appendix B we analyze quantitatively the probability of the events on November 1 andNovember 7. Often probabilities are estimated using normal (Gaussian) distributions thatunderestimate the probability of extreme events (“black swans”) that are better representedby long-tailed distributions [11, 12]. We directly fitted the long tails of the distributions andestimated the probability of the events based upon these tails to be p = 2 · − and 8 · − , respectively. Given 250 trading days in a typical year, it would take on average 200 yearsand 500 thousand years, respectively, to witness such events. Moreover, the probability ofthese two events occurring 6 days apart is p = 1 · − , corresponding to 4 billion years,comparable to the age of the Earth. Figure 3 shows that these events are outside the general4 P r i c e ( $ ) Jan 2007 Apr 2007 Jul 2007 Oct 2007 Jan 2008 Apr 2008 Jul 2008 Oct 2008 Jan 20092015105051015202530 D e m a n d Q u a n t i t y ( t e n s o f m illi o n s o f s h a r e s ) Total Short InterestVolumeChange in Short Interest
FIG. 1: Market activity for Citigroup over a two-year period starting January 1, 2007. Top panelshows vertical bars for the daily high and low stock price. Lower panel shows total short interest(yellow), trading volume (gray), and daily change in short interest (red). Arrows indicate November1, 2007 [10]. behavior of the market. We emphasize that our estimates of the probabilities of these eventsreflects the higher probabilities of extreme events in long-tailed distributions.Changes in investor behavior are often explained in terms of specific news items, withoutwhich it is expected that prices have no reason to change significantly [13, 14]. The pressattributed the drop of Citigroup’s stock price on November 1 to an analyst’s report thatmorning [15, 16]. This report, by an analyst of the Canadian Imperial Bank of Commerce(CIBC), downgraded Citigroup to “sector underperform” [17]. Any such news-based expla-nations of investor behavior on November 1 (similarly for November 7) would not account forthe difference in behavior between short sellers and other investors. Under the assumptionsof standard [14] capital asset pricing models, all investors act to maximize expected futurewealth [18], and should therefore respond similarly to news. Furthermore, it has been shownempirically that the ratio of short sales to total volume remains nearly constant, even around5 P r i c e ( $ ) Sep 2007 Oct 2007 Nov 2007 Dec 2007 Jan 20082015105051015202530 D e m a n d Q u a n t i t y ( t e n s o f m illi o n s o f s h a r e s ) Change in Short InterestTotal Short InterestVolume
FIG. 2: Market activity for Citigroup over a five-month period starting on August 15, 2007. Toppanel shows bars for daily high and low stock price (adjusted for dividends). Lower panel showsdaily change in short interest (red bars), total short interest (yellow lines), and trading volume(gray bars). Arrows indicate November 1, 2007 [10]. news events [19]. In the literature, analysis of the residual small differences in the behaviorof short and long investors has been interpreted to indicate that short sellers have an infor-mational advantage or that short sellers are able to anticipate lower future returns [19–23],rather than cause them. Still, these studies do not show that large differences in tradinggenerally occur between short and long sellers. Thus, the existence of such a difference isindicative of specific trader action.Our evidence points to a bear raid during a period of financial stress [24, 25] to whichthe Federal Reserve Bank responded in August 2007 by announcing that they would be“providing liquidity to facilitate the orderly function of markets” because “institutions mayexperience unusual funding needs because of dislocations in money and credit markets” [26].Shortly afterwards, the Dow Jones Industrial Average achieved its historical peak—14,167points on October 9—three weeks prior to November 1, the date our evidence suggests a bear6
IG. 3: Scatter plot of the daily volume of trading divided by the three month prior average (volumeratio), and the increase in number of borrowed shares divided by the volume (short interest changeratio), for Citigroup over a two-year period starting January 1, 2007. Arrows indicate Citigroupon 1 November 2007 and 7 November 2007. These two points are well outside of the behavior ofdaily events even during the period of the financial crisis in late 2007 and throughout 2008. Thetwo measures are described in Appendix A. raid occurred. Bear raids may have long-term price impact if decision makers infer investorconfidence from price movements and act on that basis [27, 28]. Citigroup CEO CharlesPrince’s resignation on November 4 after an emergency board meeting [29] may reflect suchan effect. The months after November 1 saw the beginning of the stock market turmoil of2008-2009 as well as many significant events of the financial crisis, such as the purchase ofBear Stearns by JP Morgan Chase in March 2008 and the bankruptcy of Lehman Brothersin September 2008.
III. CONCLUSIONS AND POLICY IMPLICATIONS
The 2007–2011 financial crisis resulted in widespread economic damage and introducedquestions about both our understanding of economic markets and about the practical needfor regulations that ensure market stability. The Financial Crisis Inquiry Commission7FCIC) reported that over 26 million Americans were unemployed or underemployed inearly 2011, and that nearly $11 trillion in household wealth evaporated. Moreover, theFCIC concluded that the crisis was avoidable and was caused in part by “widespread fail-ures in financial regulation and supervision [that] proved devastating to the stability of thenation’s financial markets” [30]. Regulatory changes that preceded the financial crisis in-clude the June 2007 repeal of the uptick rule, which was implemented in 1938 to increasemarket stability and inhibit manipulation [5–8, 31].Within the resulting deregulated environment, it is still widely believed that the crisis wascaused by mortgage-related financial instruments and credit conditions, and that individualtraders did not play a role [32–35]. Our analysis demonstrates that manipulation may haveplayed a key role. Methods for detecting manipulation and its effects are necessary to bothinform and enforce policy.When the SEC repealed the uptick rule on July 6, 2007, one of its main claims was thatthe market was transparent, and that such regulations were not needed to prevent marketmanipulation [6]. Our results suggest that, not long after the uptick rule was repealed,a bear raid may have occurred and remained undetected and unprosecuted. Our analysisreinforces claims that lax regulation was an integral part of the financial crisis [30].In response to requests for reinstatement of the uptick rule after the financial crash,the SEC underwent extended deliberations and finally implemented an alternative uptickrule, which allows a stock to fall by 10% in a single day before limitations on short sellingapply [36]. This weaker rule would not have affected trading of Citigroup on November 1,2007, as its minimum price was just 9% lower than the close on October 31. Subsequentday declines until November 7 were also smaller than 10%.The existence of a major market manipulation should motivate changes in market models,analysis, regulation and enforcement. In particular we conclude that: • Large traders may have a significant influence on the market. Scientific analysis andmodels should recognize the role of large traders and consider both past events andpotential future events they may cause. For example, market time series analysis thatdoes not specifically consider the effect of manipulation may be unable to discover it,because manipulation events may not manifest in averages and distributions that areusually considered. 8
Improved access to data can enable the detection of market manipulation. This wouldfoster transparency in the markets, which has been lauded but not realized. Regulatoryagencies should mandate the increased availability of relevant data for the detectionof manipulation. If these data cannot be made available in real-time or for public use,they may be provided with time delays or only for scientific use. Data of importanceinclude not only the opening of short positions but also their closing, as aggregateshort sale activity cannot be determined when only opening trade data are available.These data should be made available at the transaction level. • Current legislation, which focuses on retroactive penalties, is ineffective due to thediscrepancy between the timescale of enforcement response and that of market manip-ulation. Severe failures in the financial system may include cascading global marketcrises and numerous takeovers and bankruptcies, making the disentanglement of indi-vidual events difficult if not impossible. Regulatory agencies should adopt preventivemeasures such as the uptick rule, which would be more effective than punitive ones.The uptick rule was designed to minimally restrict trader’s actions while simultane-ously providing underlying stability for the financial system and inhibiting particularforms of manipulation, including bear raids. • The limitations of our data prevent definitive conclusions about individual events ortheir attribution to individual investors. Enforcement agencies should perform inves-tigations into specific candidate events, including the candidate event we identified onNovember 1, 2007. • Until effective regulations and enforcement are in place, market price changes may notreflect economic news. They may reflect market manipulation.The complexity of financial markets and their rapid dynamics suggest that data analysisand market models are increasingly necessary for guiding decisions about setting marketregulations and their enforcement [37–39]. Independent of the role it may play in financialcrises, understanding market manipulation may be important for characterizing market dy-namics. Recent decades have seen significant advances in financial market theory, includingthe mean-variance portfolio theory [40], the capital asset pricing model [18], arbitrage pricingtheory [41], and the theory of interest rates [42]. However, the financial crisis and anoma-9ous events such as “flash crashes” [43] demonstrate limitations in existing approaches. Morerecent efforts seek to explain market phenomena via methods such as agent-based model-ing [44–49] and analysis of the long-tailed distributions of price fluctuations [11, 50–53].While these methods have been successful in describing some aspects of market behavior,they generally do not consider the impact of individual traders who have the ability to sig-nificantly impact the market [54–60]. Current approaches, whether analytical or statistical,may not reveal isolated—or even frequent—instances of trader influence.Among the possible forms of individual trader influence, intentional actions—includingmanipulation—are of particular relevance, as they undermine the role of markets in settingprices so as to reflect economic value. Market manipulation is illegal under Section 10 ofthe Securities Exchange Commission Act of 1934 [5]. Some forms of manipulation are welldocumented, including indirect price manipulation through the generation of false news [61].Direct price manipulation through market transactions is also commonly thought to occur [1,2, 54], but methods for its detection that are based on statistical analysis [62, 63] are limitedby their inability to independently account for news events and other anomalies. No directevidence of recent price manipulation has been presented based upon these methods.The timing of the event we identified raises questions about the potential role it may haveplayed in the financial crisis. Understanding the wider impact of such an event requires thatwe consider the vulnerability of the overall market.Whereas a highly stable system is not vulnerable to any but the largest impacts, a vul-nerable system can be destabilized by much smaller shocks [64, 65]. This is a general aspectof the behavior of complex interdependent systems, not just of financial markets. Specificevents can have large effects if the underlying physical, biological or social system is vul-nerable. For example, while mass extinctions have been shown to coincide with meteorstrikes [66], underlying vulnerabilities are thought to contribute to the severity of extinctionevents [67]. Similarly, market manipulation during a period of instability and high intercon-nectedness, such as before the financial crisis [24, 25, 68], may exacerbate or even trigger acollapse. The financial system can be expected to exhibit this general property of complexsystems, in which the coincidence of underlying vulnerability and extreme events can triggercrises.We thank Yves Smith and Matt Levine for helpful comments. This work was supportedby the New England Complex Systems Institute.10 ppendix A: Methodology: Data and Event Detection
It is generally difficult to characterize the investments of individual traders, especiallyfor short positions. Unlike those who own large stakes in companies, those with large shortpositions are not required to report their holdings [69]. Short interest data is publiclyavailable by ticker symbol at two-week intervals for a rolling 12-month period [70]. Thistime resolution is too low to detect the bear raid candidate we will describe, and does notinclude historical data for the period of the financial crisis. The recent availability of off-market transaction systems that enable large volume transactions, such as crossing networks[71, 72], makes it difficult, if not impossible, to trace intentional large short sale transactionsusing market data. A short sale transaction between cohorts on a crossing network mayallow one trader to execute a short sale while the other trader accumulates a long position.This long position can then be sold on the open market without leaving a signature of itsshort sale origins.Our study is based on industry data on daily securities lending. While this data doesnot identify the individuals borrowing the shares, the time resolution proved sufficient toprovide evidence of a bear raid.We obtained price and volume data from Thomson Reuters Datastream. Short interestdata was obtained from Data Explorers and included a daily record of the value and quantityof loaned securities as reported by brokerages. These included separate time series forthe total number of borrowed securities (total demand quantity) and for daily incrementalchanges in the number of borrowed shares. Daily incremental changes were approximatelygiven by day-to-day differences in total demand quantity, with small corrections arising fromthe addition and removal of reporting organizations from the data set. The reconstructionof short selling data from security lending data is an inexact process, because borrowedsecurities may be used for purposes other than short selling, including tax arbitrage, dividendarbitrage, and merger arbitrage. Furthermore, reported data may be incomplete, becausenot all lenders supply data to industry data providers. Nevertheless, because short sellingis the predominant reason for securities lending, securities lending is a reasonable proxyfor short selling [73, 74]. We also were able to eliminate the possibility of the most likelyalternative explanation to a bear raid, dividend arbitrage, as described in Appendix C.The signature of a successful bear raid is an anomalous spike in the number of shares11f a company’s stock that are sold short, followed by a price decline, then a correspondinglarge spike in the number of positions that are covered—a decrease in the number of shortpositions. A sufficiently large increase in short selling would also increase the total volumeof trades, so we monitored also the total daily trading volume.We searched data for several prominent companies to identify candidate events, andcalculated two ratios, R and Q , for each trading day. R is the ratio of the change in shortinterest to daily volume, R ( t ) = ∆ S ( t ) V ( t ) , (1)where ∆ S ( t ) = S ( t ) − S ( t −
1) is the change in short interest, V is trading volume, and t is thedate. A large absolute value of R indicates that a high proportion of trading is accountedfor by securities lending activity—that the volume of borrowed shares was a substantialfraction of the total volume, and that short sales might have affected the stock price. Ahigh positive value indicates that shares were borrowed, and a high negative value indicatesshort covering. Note that if a large number of short positions were opened and closed onthe same day (i.e. an intraday bear raid), it would not be revealed by daily short interestdata. We cannot exclude the possibility of intraday bear raids occurring during this period. Q is the ratio of the trading volume to the three month moving average, Q ( t ) = V ( t ) V ( t ) , (2)where V is the prior 3-month (63 trading day) moving average of volume. A value of Q substantially greater than one indicates an anomalously high trading volume. The event weanalyzed was identified by a high absolute value of R and high value of Q , indicating thatthe increase in borrowed shares was large in comparison to trading activity, and that totaltrading activity increased dramatically. Appendix B: R and Q distributions In this appendix we present our analysis of the distributions of R (the ratio of the changein short interest to daily volume, see Eq. 1) and Q (the ratio of the trading volume to thethree month moving average, see Eq. 2) for Citigroup, from January 2007 through December2008. The analysis allows us to obtain a probabilistic estimate of the inherent likelihood of12 and Q values for each day, and in particular for the events on November 1 and 7, 2007.The positive and negative tail cumulative distributions for Citigroup for R are plotted inFig. 4. The two sides of the distribution behave differently: while the positive tail followsa power law distribution (top panel), the negative tail is well described by a Laplaciandistribution (bottom panel). The distribution for Q , shown in Fig. 5, has a power lawtail. November 1 and 7, 2007 are omitted in the plots, but this does not affect the fitteddistributions. From the fitted distributions we extracted the expected probabilities of thetwo events. Appendix C: Tests and Technical Notes
We have tested a number of alternative explanations of the data: • Is it possible that the borrowed shares were used to receive a dividend payment, i.e.dividend arbitrage?Sometimes borrowing shares provides benefits of dividends to the borrower rather thanto the owner. In such cases the borrower may not necessarily sell the shares short,which precludes a bear raid.The date on which shares were borrowed, November 1, was an “ex-dividend” date, i.e.a date on which ownership determines dividend payments. In order for borrowers toreceive the benefit of dividends they are required to hold the shares at the prior day’sclosing. Thus, there was no dividend paid to shares borrowed on November 1. • Is it possible that the reported dates for borrowed shares is delayed so that the actualdate of borrowing is a different date than what is reported (for example, could it bereported on the date of settlement three days after a market transaction)?We verified the agreement of reported borrowing and short selling date by looking atthe period of the short sale ban starting in September 2008. The dates of the startand stop of borrowing coincide with the dates that they should for the ban, whichshows that there is no delay in reporting. • Does commercial market transaction data corroborate the short selling?13 .001 C u m u l a ti v e D i s t r i bu ti on R Citigroup R (positive tail) Power Law C u m u l a ti v e D i s t r i bu ti on - R Citigroup R (negative tail) Laplacian FIG. 4:
Citigroup R distribution - Cumulative distribution functions (CDF) of the short interestchange ratio for Citigroup, for 2007 and 2008. Top panel : Positive tail of the distribution, blueline is the best fit power law (CDF( R ) ∼ R α , with α = − . Bottom panel : Negative tail ofthe distribution, blue line is the best fit Laplacian distribution (CDF( R ) ∼ R − β )(1 − exp( −| R − β | /γ )), with β = 0 .
11 and γ = 0 . .0010.010.11 C u m u l a ti v e D i s t r i bu ti on Q Citigroup Q Power Law
FIG. 5:
Citigroup Q distribution - Cumulative distribution function (CDF) of the volumeratio for Citigroup for 2007 and 2008. Blue line is the best fit power law (CDF( Q ) ∼ Q α , with α = − . We have studied commercially available NYSE short selling data [75] from these dates,and found it to be unreliable because the transactions reported are inconsistent withreported trade and quote data [76] at the transaction level. Despite dialog with theNYSE staff we have not received an explanation of the inconsistency. For the presentanalysis, the inconsistency inhibits our efforts to use this data to cross-validate theresults in this report. More generally, it raises questions about the reliability of marketprovided short sale data. • Is it possible that the analyst report downgrading Citigroup that morning was releasedin collusion with the bear raid?We have no specific evidence, but such collusion would be consistent with strategiesused by those who manipulate stocks [1, 2, 54, 61]. • Is it possible that those who engaged in the bear raid also used trading in options toincrease their profits by buying put or selling call options?15ur estimate of the profits made on the bear raid are conservative. • Is it possible that the large block trades on November 1 and 7 represented tradingbased upon information that was not yet available to the public on November 1?Our evidence suggests that a single individual or group of individuals traded a largevolume of borrowed shares on November 1 and November 7. If this represented po-tentially illegal insider trading, the traders would have avoided attracting attention.Neither the large trading volume nor the abrupt price drop on November 1 at theopening of the market appear to be consistent with a low-profile trading approach.The rapid price drop is also inconsistent with the expected behavior of insider traders,which is to maximize profits by selling gradually to avoid affecting prices until the neg-ative news becomes public. Both the large volume of trading and the rapid drop areconsistent with trading intended to affect prices, i.e. a bear raid. While the intentionsof traders can only be determined from a more detailed inquiry once those traders areidentified, the available information strongly supports a bear raid over the possibilityof insider trading per se. It is possible that traders with insider information chose tohelp matters along by performing a bear raid at the same time as they were tradingon insider information.
Addendum: Additional Tests and Technical Notes
Following the release of this paper, we were contacted by the NYSE with additionalinformation about the NYSE short selling transaction data [75] described in Appendix C.The new information enabled us to reconcile the short sale and trade data [76] by aggregatingand shifting the times of multiple transfers to correspond with market transactions. Thereare residual issues with a small minority of transactions that are being resolved, but theseissues appear to be irrelevant to conclusions about the volume of trading.The additional information enables us to identify with some confidence the reported shortsale volume on the NYSE on November 1 and other dates. The short sale volume is notunusual as a proportion of total volume, constituting about one quarter of the total volumeon this market. NYSE transactions constituted 30% of the total market volume on November1, 2007. This limits the volume of reported short selling on the markets, and diminishes the16ikelihood that the reported increase in borrowed shares was directly reflected in reportedshort sales.Absent an alternative interpretation, if shares were sold in a way that concealed theirorigin as borrowed shares the data sets would be consistent. One method to achieve this,using “short to buy” transactions, was reported in Senate investigations of the PequotCapital hedge fund in 2009 [77]. In this approach a single trader moves shares from oneaccount to another, creating a short position in one and a long position in the other. Sincethere is no change in beneficial ownership, such transactions may be reported in a way thatis not consistent with standard reporting requirements, resulting in share borrowing withouta market record. Long positions created this way may be sold on any market without beingidentified as short sales, even though in doing so a net short position is created.This method appears to have been developed to hide short selling at a time when theuptick rule was in effect. Short to buy transactions require a close relationship with abroker dealer. The necessary access to market trading systems, called “sponsored” or “directmarket” access, needed to perform the short to buy transaction is not available to mosttraders but constitutes a significant fraction of reported trading [78, 79]. Only recently,beginning in 2011, were brokers required to apply standard regulations to transactions oftraders using sponsored access [80, 81]. Previously, non-compulsory self-regulation was ineffect [82]. In the absence of oversight, market data may not properly record the volume ofshort selling.An explanation in these terms for the events in November of 2007 is also consistent withthe observation that there was a larger volume of returned shares on November 7 than thetrading volume. In the “short to buy” scenario, residual positions can be closed through“back office” transactions and may never be recorded on the market.The new information we received implies that the sale of borrowed shares reflected in theincrease in borrowed shares on November 1 and the corresponding decrease on November 7may have been done in a way that would not have been prevented by the uptick rule. Amore detailed inquiry into the means by which such selling could have been done is beyondthe current work. 17e thank Steven Poser and Wayne Jett for helpful discussions. [1] G. Matsumoto, Naked short sales hint fraud in bringing down Lehman,
Bloomberg (March 19,2009) .[2] G. Soros, One way to stop bear raids,
Wall Street Journal (March 24, 2009) .[3] M. K. Brunnermeier, L. H. Pedersen, Predatory trading,
The Journal of Finance , 1825(2005).[4] M. G. Ferri, S. E. Christophe, J. J. Angel, A short look at bear raids: Testing the bid test Georgetown University Working Paper; Financial Management Association Annual Meeting,Fall 2005 (2004).[5] Securities Exchange Act of 1934,
15 U.S.C. § (2009).[6] Regulation SHO and Rule 10a-1,
17 CFR 240, 242 (2007) .[7] R. C. Pozen, Y. Bar-Yam, There’s a better way to prevent ‘bear raids’,
Wall Street Journal (November 18, 2008) http://online.wsj.com/article/SB122697410070336091.html .[8] Y. Bar-Yam, D. Harmon, V. Misra, J. Ornstein, Regulation of short selling: The uptickrule and market stability, report presented at the SEC Division of Trading and Mar-kets February 24, 2010,
NECSI report (2010) .[9] Y. Bar-Yam, D. Harmon, Technical report on SEC uptick repeal pilot,
NECSI report (2008).[10] Data Explorers ( ).[11] R. N. Mantegna, H. E. Stanley,
An Introduction to Econophysics: Correlations and complexityin finance (Cambridge University Press, Cambridge, 1999).[12] N. N. Taleb,
The Black Swan: The impact of the highly improbable (Random House, NewYork, 2007).[13] E. F. Fama, Efficient capital markets: A review of theory and empirical work,
The Journal ofFinance , 383 (1970).[14] E. F. Fama, K. R. French, The capital asset pricing model: Theory and evidence, The Journal f Economic Perspectives , 25 (2004).[15] Hostile reactions to CIBC’s Citi report, The New York Times: Dealbook (November 5, 2007).[16] S. Rosenbush, The analyst who rocked Citi,
Bloomberg: Business Week (November 26, 2007).[17] M. Whitney, Is Citigroup’s dividend safe? Downgrading stock due to capital concerns,
CIBCequity markets: Change in recommendation (October 31, 2007).[18] W. F. Sharpe, Capital asset prices: A theory of market equilibrium under conditions of risk,
The Journal of Finance , 425 (1964).[19] J. Engelberg, A. V. Reed, M. C. Ringgenberg, How are shorts informed? Short sellers, news,and information processing, University of North Carolina working paper (2010).[20] P. Asquith, L. Meulbroek, An empirical investigation of short interest,
MIT Working Paper (1995).[21] A. J. Senchack, L. T. Starks, Short-sale restrictions and market reaction to short-interestannouncements,
Journal of Financial and Quantitative Analysis , 2 (1993).[22] E. Boehmer, C. M. Jones, X. Zhang, Which shorts are informed? Journal of Finance , 491(2008).[23] H. Desai, K. Ramesh, S. R. Thiagarajan, B. V. Balachandran, An investigation of the in-formational role of short interest in the NASDAQ market, The Journal of Finance , 2263(2002).[24] F. A. Longstaff, The subprime credit crisis and contagion in financial markets, Journal ofFinancial Economics , 436 (2010).[25] R. J. Caballero, E. Farhi, P.-O. Gourinchas, Financial crash, commodity prices, and globalimbalances, NBER Working Paper No. 14521 (2008).[26] Press release: August 10, 2007,
Federal Reserve Board of Governors (August 16, 2007) .[27] I. Goldstein, A. Guembel, Manipulation and the allocational role of prices,
Review of EconomicStudies , 1 (2008).[28] N. Khanna, R. D. Mathews, Bear raids and short sale bans: Is government interventionjustifiable? Michigan State University Working Paper (2009).[29] D. Wilchins, J. Stempel, Citigroup CEO Prince to resign: Reports,
Reuters (November 2,2007).[30] Financial Crisis Inquiry Commission,
The Financial Crisis Inquiry Report: Final report of the ational commission on the causes of the financial and economic crisis in the United States (US Government Printing Office, 2011).[31] A. H. Pessin, Fundamentals of the Securities Industry (New York Institute of Finance, NewYork, 1978).[32] M. L. Mah-Hui, Old wine in a new bottle: Subprime mortgage crisis — causes and conse-quences,
The Levy Economics Institute of Bard College Working Paper No. 532 (2008).[33] O. J. Blanchard, The crisis: Basic mechanisms and appropriate policies,
IMF Working PaperNo. 09/80 (2009).[34] V. V. Acharya, M. P. Richardson, Causes of the financial crisis,
Critical Review , 2-3 (2009).[35] M. F. Hellwig, Systemic risk in the financial sector: An analysis of the subprime-mortgagefinancial crisis, De Economist , 2 (2009).[36] Amendments to Regulation SHO,
17 CFR 242 (2010) .[37] A. G. Haldane, R. M. May, Systemic risk in banking ecosystems,
Nature , 351 (2011).[38] N. F. Johnson, Proposing policy by analogy is risky,
Nature , 302 (2011).[39] T. Lux, Network theory is sorely required,
Nature , 303 (2011).[40] H. M. Markowitz,
Portfolio Selection: Efficient diversification of investment (Wiley, NewYork, 1959).[41] S. A. Ross, The arbitrage theory of capital asset pricing,
Journal of Economic Theory , 341(1973).[42] J. C. Cox, J. E. Ingersoll, S. A. Ross, A theory of the term structure of interest rates, Econo-metrica , 385 (1985).[43] A. E. Khandani, A. W. Lo, What happened to the quants in August 2007? Evidence fromfactors and transactions data, Journal of Financial Markets , 1 (2011).[44] W. B. Arthur, J. H. Holland, B. LeBaron, R. G. Palmer, P. Tayler, Asset pricing underendogenous expectation in an artificial stock market, in The Economy as an Evolving ComplexSystem II , W. B. Arthur, D. Lane, S. Durlauf, eds. (Addison-Wesley, Redwood City, 1997)p. 1544.[45] T. Lux, M. Marchesi, Scaling and criticality in a stochastic multi-agent model of a financialmarket,
Nature , 498 (1999).[46] E. Samanidou, E. Zschischang, D. Stauffer, T. Lux, Agent-based models of financial markets, ep. Prog. Phys. , 409 (2007).[47] M. Levy, H. Levy, S. Solomon, A microscopic model of the stock market, Economics Letters , 103 (1994).[48] R. Cont, J. P. Bouchaud, Herd behavior and aggregate fluctuations in financial markets, Macroeconomic Dynamics , 170 (2000).[49] R. Donangelo, K. Sneppen, Self-organization of value and demand, Physica A , 572 (2000).[50] X. Gabaix, P. Gopikrishnan, V. Pierou, H. E. Stanley, A theory of power law distributions infinancial market fluctuations,
Nature , 267 (2003).[51] Y. Liu, P. Gopikrishnan, P. Cizeau, M. Meyer et. al., Statistical properties of the volatility ofprice fluctuations,
Physical Review E , 1390 (1999).[52] D. Sornette, Why Stock Markets Crash: Critical events in complex financial systems (Prince-ton University Press, Princeton, 2002).[53] S. Solomon, M. Levy, Spontaneous scaling emergence in generic stochastic systems,
Interna-tional Journal of Modern Physics C , 745 (1996).[54] F. Allen, D. Gale, Stock-price manipulation, Review of Financial Studies , 503 (1992).[55] F. Allen, G. Gorton, Stock price manipulation, market microstructure and asymmetric infor-mation., European Economic Review , 624 (1992).[56] R. A. Jarrow, Market manipulation, bubbles, corners, and short squeezes, Journal of Financialand Quantitative Analysis , 311 (1992).[57] A. S. Kyle, Continuous auctions and insider trading, Econometrica , 1315 (1985).[58] R. Benabou, G. Laroque, Using privileged information to manipulate markets: Insiders, gurus,and credibility, The Quarterly Journal of Economics , 921 (1992).[59] P. Kumar, D. J. Seppi, Futures manipulation with cash settlement,
Journal of Finance ,1485 (1992).[60] R. Aggarwal, G. Wu, Stock market manipulations, Journal of Business , 1915 (2006).[61] M. T. Bradshaw, S. A. Richardson, R. G. Sloan, Pump and dump: An empirical analysis ofthe relation between corporate financing activities and sell-side analyst research, Universityof Pennsylvania Working Paper (2003).[62] M. Minenna, The detection of market abuse on financial markets: A quantitative approach,
Quaderni di Finanza (2003).[63] R. Cholewiski, Real-time market abuse detection with a stochastic parameter model, Central uropean Journal of Economic Modelling and Econometrics , 261 (2009).[64] O. De Bandt, P. Hartmann, Systemic risk: A survey European Central Bank Working PaperNo. 35 (2000).[65] S. V. Buldyrev, R. Parshani, G. Paul, H. E. Stanley, S. Havlin, Catastrophic cascade of failuresin interdependent networks,
Nature , 1025 (2010).[66] P. Schulte, L. Alegret, I. Arenillas, J. A. Arz et. al, The Chicxulub asteroid impact and massextinction at the Cretaceous-Paleogene boundary,
Science , 1214 (2010).[67] N. C. Arens, I. D. West, Press-pulse: a general theory of mass extinction?
Paleobiology ,456 (2008).[68] D. Harmon, B. C. Stacey, Yavni Bar-Yam, Yaneer Bar-Yam, Networks of economic marketinterdependence and systemic risk, arXiv (2010).[69] Schedule 13D,
17 CFR 240.13d-101 (2007) .[70] Short interest,
NASDAQ .[71] H. Mittal, Are you playing in a toxic dark pool? A guide to preventing information leakage,
The Journal of Trading , 20 (2008).[72] L. Harris, Trading and exchanges: Market microstructure for practitioners (Oxford UniversityPress USA, 2002).[73] M. C. Faulkner, An introduction to securities lending,
Handbook of Finance (2008) http://onlinelibrary.wiley.com/doi/10.1002/9780470404324.hof001073/full .[74] US equity short positions and securities lending data, Data Explorers (2011). [75] TAQ NYSE Short Sales, NYSE Technologies .[76] Daily TAQ (Trade and Quote), NYSE Technologies .[77] Exhibit 8, The Firing of an SEC Attorney and the Investigation of Pequot Capital Man-agement. Joint report by the United States Senate Committee on Finance and the SenateJudiciary Committee (August 3, 2007).[78] N. Mehta, Sponsored Access Comes of Age, Traders Magazine (March, 2009)
79] K. D. Freeman, Economic warfare: Risks and responses, Cross Consulting and Services, LLC(June, 2009).[80] U.S. Securities and Exchange Commission, Exchange Act Release No. 63241 (November 3,2010) [81] P. Chapman, Brokers see challenges ahead to meet sponsored access mandates
Traders Magazine Online News (July 19, 2011) [82] U.S. Securities and Exchange Commission, Exchange Act Release No. 61379 (January 19,2010)