Naked Short Selling and the 2008 Financial Crisis

Regulatory and media concern has focused heavily on the potentially manipulative distortion of market prices associated with naked short selling. However, naked shorting can also have beneficial effects for liquidity and pricing efficiency. We empirically investigate the impact of naked short-selling on market quality, and find that naked shorting leads to significant reduction in positive pricing errors, the volatility of stock price returns, bid-ask spreads, and pricing error volatility. We study naked shorting surrounding the demise of financial institutions hardest hit by the financial crisis in 2008 and find no evidence that stock price declines were caused by naked shorting. We also find that naked short-selling intensifies after rather than before credit downgrade announcements during the 2008 financial crisis. In general, we find that naked short sellers respond to public news and intensify their activity after price declines rather than triggering these price declines. We study the impact of the SEC ban on naked short selling of financial securities during July and August 2008, and find that the ban did not slow the price decline of those securities and had a negative impact on liquidity and pricing efficiency. Finally, after examining the speeds of mean reversion of pricing errors and order imbalances, we infer that Regulation SHO was successful in curbing the impact of manipulative naked short selling, and this reduction in the impact of manipulative naked shorting has continued through the 2008 financial crisis. Overall, our empirical results are in sharp contrast with the extremely negative preconceptions that appear to exist among media commentators and market regulators in relation to naked shortselling.

1) Brief Background and Motivation

Short selling is the sale of a stock not owned by the seller. Generally, the stock is borrowed, or adequate borrowing arrangements are made, to ensure availability for delivery at settlement. Such short selling is “covered shorting”. On the other hand, “naked short selling” or “naked shorting” is a short sale in which the seller does not arrange to borrow, or even intend to borrow the securities to deliver to the buyer within the standard three-day settlement period. As a result, the seller fails to deliver securities to the buyer when delivery is due (known as a “failure to deliver” or “FTD”).

Regulators have understandably sought to curb naked short selling given that regular, intentional and widespread breach of settlement-related contractual delivery obligations can potentially disrupt the smooth functioning of financial markets, even though the Depository Trust and Clearing Corporation (DTCC) (electronic) system of a voluntary pool of lenders mitigates such disruptions, and makes naked shorting costly3. The Securities and Exchange Commission (SEC), through Regulation SHO, accordingly imposed major restrictions on naked short selling (like the locate requirement) effective from January 2005. More recently, in the wake of the heavy and rapid falls in the prices of financial sector stocks during the current financial crisis, there has been considerable concern about manipulative “bear raids” by naked short-sellers, and US regulators banned naked short selling for select financial institutions between July 21st and August 12th, 2008, since “false rumors can lead to a loss of confidence [and] panic selling, which may be further exacerbated by ‘naked’ short selling”, and as a result, “the prices of securities may artificially and unnecessarily decline well below the price level that would have resulted from the normal price discovery process”. In a similar spirit, Britain’s Financial Services Authority (FSA) imposed new regulations on short positions, and restrictive regulations on shorting were enacted in many other countries5. On the other hand, there have also been concerns about such restrictions, since short-sellers play “an important role in exposing the poor condition of some companies” and that “neither of the regulators has produced evidence [linking naked short selling to market manipulation] so far”.

Contemporaneously, there has been a surge of discussion in the media about the impact of naked shorting. Over 4,600 printed articles have appeared in English-language magazines and newspapers discussing naked shorting in the past 2 years. Of these, an extremely tiny minority depicts naked shorting as beneficial to market quality8. The vast majority focuses on the potential for stock price manipulation and on the creation of phantom shares. Other evidence of the current views against naked shorting include at least three investor associations lobbying for restrictions on naked shorting, at least three lawsuits by investor groups alleging stock price manipulation linked to naked shorting and at least twelve other lawsuits against the DTCC for allegedly facilitating naked short selling. Several senior managers of major companies targeted by naked short sellers have also been very vocal in their opposition to naked shorting, claiming that naked shorting led to their stock prices being artificially depressed. The SEC has received over 5,000 complaints alleging stock price manipulation through naked short selling between January 2007 and June 2008.

However, it can also be forcefully argued that both covered and naked shorting should be beneficial for pricing efficiency and for liquidity. First, both covered and naked short-selling should potentially contribute to the price discovery process by enabling value-traders to more quickly and easily bring the prices of overpriced securities in line with their “true value”, as argued by Miller (1977) and Diamond and Verrecchia (1987), among others. And second, financial intermediaries and other liquidity suppliers should be able to provide liquidity more effectively and expeditiously in the presence of both covered and naked short-selling. While one would expect naked shorting to be at least as beneficial as covered shorting in this context, and arguably more so, the SEC has, because of the fear of manipulative naked shorting, relaxed covered-shorting restrictions but increased naked-shorting restrictions, through, for example, the “locate” and “close-out” requirements under Regulation SHO, removal of the uptick rule, and the temporary naked shorting “bans” during the 2008 financial crisis.

There is a significant body of literature focusing on the impact on market quality of short selling. This literature largely suggests and does actually find that short-selling is potentially beneficial both for pricing efficiency and liquidity, but also offers evidence linking short selling to price manipulation. However, this literature does not distinguish between covered and naked short-selling. In particular, existing research does not provide any evidence specifically on the effect of naked shorting on liquidity, price distortions and pricing efficiency, and the extent to which the effects of naked shorting exist after controlling for covered-shorting.  Our study fills this void in existing literature.

Using data on fails to deliver for the first half of 2007, we estimate that naked shorting has affected about 91% of NYSE securities, about 71% of NASDAQ securities and about 60% of AMEX and ARCA securities. Using a random sample of 300 NYSE securities, and a vector autoregressive model to control for causality and endogenous interrelationships between market quality metrics, we find that an increase in naked short selling leads to lower positive pricing errors, lower pricing error volatility, reduced stock price volatility and lower order imbalances.  We also observe that naked short selling intensifies after positive order imbalances. These results are consistent with market makers employing naked short selling to provide liquidity when it is otherwise scarce and value arbitrageurs enhancing pricing efficiency through naked short selling. As we are more likely to find negative effects associated with naked short selling when it is most intense, we also focus on a sample of the most naked shorted securities. Even in this sample, we find a positive impact on market quality: a one percentage point increase in the incidence of naked short selling leads to approximately a 2.8% reduction in returns volatility, a 1% reduction in bid-ask spreads, a 4.3% decline in pricing error volatility and a 4.6 % decline in positive pricing errors.

Since naked short sellers have been widely accused in the media for having contributed to the financial crisis by precipitating manipulative price declines of financial firms in 2008, we analyze a few high-profile individual cases of financial firms that experienced dramatic stock price declines. In particular, we analyze naked short selling in Bear Sterns Companies Inc. (Bear Stearns), Lehman Brothers Holdings Inc. (Lehman), Merrill Lynch & Co. Inc. (Merrill), and American Insurance Group (AIG). We find that, except for one instance in June 2008 of possible stock price manipulation through naked shorting in relation to Lehman Brothers Holdings Inc., most of the time, naked short selling was too low to reasonably “cause” significant stock price distortions, and when naked shorting did become abnormally heavy, it was after dramatic price declines, not before, indicating that naked short sellers were responding to public domain information about the firms, rather than being responsible for triggering the observed precipitous price decline. We further analyze how naked short selling changes around public news of credit rating downgrades, and again find that naked short selling increases after rather than before the credit downgrade announcement, again consistent with naked short sellers responding to public information, rather than being responsible for triggering price declines.

We also analyze the market impact of the SEC naked short selling ban for 19 financial securities between July 15 and August 12, 2008. For the period during which the ban was enacted, we find significantly higher absolute pricing errors and significantly lower trading volumes, indicating that the naked short selling ban hampered price discovery and reduced liquidity. Returns, albeit negative, were not significantly affected, indicating that the ban failed to slow the price decline of the related securities.

Finally, we examine changes in the impact of manipulative naked shorting, in contrast to naked shorting that improves pricing efficiency or liquidity. We find that negative pricing errors and negative order imbalances mean-revert at a significantly faster rate after Regulation SHO, indicating a lower impact of manipulative naked short selling after Regulation SHO; and importantly, this reduced impact of manipulative naked shorting has continued through the 2008 financial crisis.

The remainder of the paper is structured as follows. Section 2 briefly reviews extant research on naked short selling. Section 3 develops our hypotheses. Section 4 defines the measures and variables we use for naked and covered shorting, pricing efficiency, and liquidity.

Section 5 documents our empirical methods and results: provides salient descriptive statistics; investigates the impact of naked shorting on market quality; analyzes the role of naked short sellers around major economic events over the financial crisis period, e.g., the demise of Bear Stearns and Lehman Brothers, the near-demise of Merrill and AIG, and around credit rating downgrade announcements; examines the impact of the ban on naked short selling of select financial securities in July and August 2008, and analyzes changes in the impact of manipulative naked short-selling in the context of Regulation SHO and the 2008 financial crisis. Section 6 presents concluding remarks.

2) Extant Research on Naked Short-Selling

A large body of existing research examines the relation between short selling and market quality. Diamond and Verrecchia (1987) conclude, on the basis of their theoretical model, that short-selling constraints do not bias prices upwards, but reduce the speed of adjustment of prices to private information, and hence reduce pricing efficiency. Abreu and Brunnermeier (2002, 2003) and Scheinkmann and Xiong (2003) show theoretically that constraints on short selling are linked to bubbles and excess market volatility. Early studies supporting the view that short selling reduces overpricing and increases market efficiency include Miller (1977) and Harrison and Kreps (1978). Pope and Yadav (1994) find that spot-market short-selling restrictions make index futures more negatively mispriced, leading to a pricing bias. Asquith and Meulbroek (1996), Aitken et al. (1998), Danielsen and Sorescu (2001), Jones and Lamont (2002), Gezy et. al. (2002), Ofek and Richardson (2003) and Reed (2007) provide evidence that stock prices do not fully incorporate information in the presence of short sale constraints. Daouk and Charoenrook (2005) study the effects of changing restrictions on short selling in 111 countries and conclude that allowing short selling improves market quality. Bris et. al. (2007) similarly analyze equity markets around the world and find that prices incorporate negative information faster in markets where short sales are allowed. Diether et al. (2007) document that the temporary suspension of the uptick rule that restricted short-selling did not negatively affect market quality, and conclude that the suspension can be made permanent18. Boehmer et al. (2008) use proprietary NYSE order data to find that short sellers are, on average, better informed, and contribute to efficient pricing. Shkilko et al. (2008) focus on short selling during intraday liquidity crises and observe that short-sellers have price-destabilizing effects: they conclude that short-selling is often used as a tool for price manipulation and short selling restrictions would improve market quality.

The literature focusing specifically on naked short selling is much less developed. There are two academic “thought-pieces” without any theoretical models or empirical evidence: Christian et al. (2006) provide a descriptive review of naked shorting from a legal perspective; and, Uchimoto et. al. (2005) discuss the impact of naked short selling on ETF trading,
conjecturing adverse impacts on capital formation and pricing efficiency, particularly for small and emerging companies. There are also two theoretical papers on naked shorting without any empirical evidence. Finnerty (2005) develops a theoretical model for market equilibrium in the presence of short selling, both covered and naked shorting, and concludes that naked short selling is likely to be used as an instrument for market manipulation. On the other hand, Culp and Heaton (2007) offer a theoretical model of the effects of naked shorting on markets, and, in the context of an extensive analysis of the DTCC settlement system, conclude that “naked shorting is not fundamentally different from traditional short selling and is unlikely to have detrimental effects on capital markets”.

We are aware of three empirical studies specifically on naked shorting. First, Evans et al. (2008) use data from one market maker to link FTDs to hard-to-borrow situations and examine the possibility of arbitrage based on misalignments between the option and stock markets. Second, Boni (2006) analyzes delivery failures in U.S. equity markets. She finds that, prior to Regulation SHO, most U.S. equity issues, listed and unlisted, experienced at least a small percentage of failures-to-deliver each day. A substantial fraction of issues (42% of listed stocks and 47% of unlisted stocks) had persistent fails of 5 days or more. She conjectures that Regulation SHO would lead to less liquidity, increased price volatility and temporary short squeezes, but does not offer any empirical evidence on the issue19. And third, Edwards and Hanley (2008) examine the effect of short selling constraints and of naked short selling on the short-term performance of IPOs, and find that IPOs with greater naked shorting are more accurately priced.

However, we are not aware of any empirical research that investigates the link between naked shorting and market quality: we fill this gap in the literature with the present paper. Our aim is to empirically investigate naked shorting in the context of manipulative price distortions, pricing efficiency and liquidity; and thereby offer insights into the tradeoff market regulators face. Our expectation is that naked short selling potentially increases the risk of manipulative episodes, but also leads to an improvement in market quality by reducing pricing efficiency and increasing liquidity.

3) Development of Hypotheses

As highlighted in the introduction and the footnotes therein, the media and the CEOs of affected firms have vociferously and persistently accused naked short-sellers of undertaking
“bear raids” and thereby causing associated stock prices to decline. Regulators have also sometimes accused naked short sellers of depressing stock prices20. Accordingly, our first hypothesis is the following:

H1: Naked short selling depresses stock prices.

While the large body of literature that has examined the relationship between short sales and pricing efficiency has sometimes documented contrasting results, a growing consensus is emerging that short sellers enhance price efficiency. Two recent studies are especially relevant here. First, Diether et al. (2007) use the recently available data on daily short sales to find that short sellers correct overreaction in stock prices. Second, Boehmer et al. (2008) use proprietary NYSE order data to find that short sellers, especially institutional short sellers, act as value arbitragers and correct overpriced securities to bring about permanent price effects and hence contribute to efficient pricing. While there is no empirical evidence in regard to naked short selling, Culp and Heaton (2007) argue that naked and covered short sales should have similar impacts on security markets. Similar to a covered short sale, a naked short sale (which eventually leads to a FTD) starts off with a short seller agreeing to sell a security to a buyer at the prevailing market price; but, unlike a covered short sale, the naked short seller fails to deliver the security to the buyer (on day t+3) because she has not borrowed the security from a lender21. Ordinarily, the DTCC system triggers immediate delivery to the buyer through borrowing from a voluntary pool of lenders, and even where the pool is empty, the buyer, by holding the selling price of the stock as collateral and also maintaining an exposure to the security, effectively becomes the lender of the security to the naked shorter. Accordingly, Culp and Heaton (2007) argue that, except for this nuance of a change in roles, covered and naked short sales are functionally indistinguishable. In view of this functional similarity, and the emerging consensus of an improvement in price discovery associated with short selling in general, our hypothesis is:

H2: Naked short selling reduces positive pricing errors.

Given that we know from Hasbrouck (1993) that informationally efficient markets display lower dispersion of pricing errors, our hypothesis is:

H3: Naked short selling reduces the volatility of pricing errors.

Naked short selling is often employed by market makers and other liquidity providers to quickly and efficiently fulfill orders. The SEC accordingly asserts on its website that “in certain circumstances, naked short selling contributes to market liquidity. For example, broker-dealers that make a market in a security generally stand ready to buy and sell the security on a regular and continuous basis at a publicly quoted price, even when there are no other buyers or sellers. Thus, market makers must sell a security to a buyer even when there are temporary shortages of that security available in the market. This may occur, for example, if there is a sudden surge in buying interest in that security, or if few investors are selling the security at that time. Because it may take a market maker considerable time to purchase or arrange to borrow the security, a market maker engaged in bona fide market making, particularly in a fast-moving market, may need to sell the security short without having arranged to borrow shares. This is especially true for market makers in thinly traded, illiquid stocks”22. In fact, naked short selling is a mechanism to improve the efficiency of security-lending markets since the option to fail becomes particularly valuable when borrowing is too expensive for covered short sellers, which is exactly when liquidity is most needed (Evans, Geczy, Musto and Reed, 2008). In this context, we expect naked short sales to increase liquidity (by decreasing bid-ask spreads) and to reduce order imbalances. This leads us to our next set of hypotheses:

H4: Naked short selling reduces bid-ask spreads.

H5: Naked short selling reduces order imbalances.

We expect that improvements in both liquidity and pricing errors should translate into more orderly and less volatile markets. Accordingly, we expect to observe lower stock price volatility in the presence of naked shorting. Hence, we hypothesize that:

H6: Naked short selling reduces stock price volatility.

Naked short-sellers are typically thought of as undertaking “bear raids” to trigger downward price spirals, and specifically with the aim of achieving credit downgrades so as to also profit from simultaneous positions in the CDS market. In this context, we test the following hypotheses:

H7A: Naked short-sellers triggered the price crashes associated with Bear Stearns, Lehman, AIG and Merrill.

H7B: Naked short-sellers triggered credit downgrades.

We investigate the impact of SEC restrictions on naked short selling introduced in the wake of the 2008 financial crisis. Given that we expect naked short selling to improve liquidity, reduce the volatility of pricing errors, and accelerate price declines, we expect restrictions on naked short selling to have the opposite effect. Hence, we hypothesize that:

H8A: Restrictions on naked short selling reduce trading volumes.

H8B: Restrictions on naked short selling increase bid-ask spreads.

H8C: Restrictions on naked short selling increase pricing error volatility.

H8D: Restrictions on naked short selling increase returns.

We test the effect of Regulation SHO and the 2008 financial crisis on the impact of manipulative naked short-selling. In this context, we note that naked short-sellers can have three trading motivations. They could be value arbitrageurs aiming to profit from positive pricing errors; or dealers or other traders aiming to profit from supplying liquidity (by selling when order imbalances are positive and buying when they are negative); or the manipulative traders that Regulation SHO was intended to curb and regulate. If Regulation SHO has been effective, we would expect that the impact of manipulative naked shorting would have declined after Regulation SHO, leaving only the beneficial effects on liquidity and pricing efficiency.  Furthermore, we would not expect any significant differences in the wake of the 2008 financial crisis.

It is difficult to precisely define the criteria that can be used to identify and classify naked short-sellers as potentially manipulative. We define potentially manipulative naked shortsellers from the mind-set used in the financial crisis media coverage of naked short-selling, i.e., the mindset of bear-raids, based on whether naked shorting is undertaken at a time when pricing errors are positive or negative. If naked shorting is done when pricing errors are positive, it contributes positively to pricing efficiency. If it is done when pricing errors are negative, then it will arguably further amplify negative pricing errors in the next period, and thereby contribute to generating a bear-raid scenario. Hence, we would expect that the mean reversion of pricing errors, conditional on the pricing error being negative, will decrease in the presence of greater manipulative naked shorting.

We can apply similar criteria based on order imbalances. If naked shorting is done when order imbalances are positive, naked shorting will arguably have a positive liquidity-related
impact. However, following the same mind-set of bear raids, naked short-selling will be potentially manipulative if it is initiated at a time when order-imbalances are negative, since it will cause them to become even more negative. Hence, once again, we would expect that the mean reversion of order-imbalances, conditional on the order-imbalance being negative, will decrease in the presence of greater manipulative naked shorting.

Accordingly we test whether:

H9A:Regulation SHO increases the rate of mean reversion of negative pricing errors.

H9B:Regulation SHO increases the rate of mean reversion of negative order imbalances.

H9C:The financial crisis does not change mean reversion of negative pricing errors.

H9D:The financial crisis does not change mean reversion of negative order imbalances.

4) Definitions of Measures and Variables

4.1) Naked Short Selling

Our proxy for naked short selling is based on the outstanding number of fails to deliver (FTDs), daily data on which has been made available by the SEC under the Freedom of
Information Act (FOIA) since March 22, 200423. In this context, we note that, while every naked short sale does, by definition, result in an FTD, every FTD does not necessarily originate from a naked short sale. In particular, we can think of three specific factors that can make an FTDbased proxy imperfect.

First, as highlighted by the SEC, “human or mechanical errors or processing delays can result from transferring securities in physical certificate rather than book-entry form, thus causing a FTD. However, we believe such errors and delays should be random and not systematically related to any of our hypotheses and therefore may add noise but should not affect any of our conclusions. Further, former SEC commissioner Roel Campos stated in an interview that “The majority of these failures-to-deliver are not the result of honest mistakes or bad processing”

Second, Edwards and Hanley (2008) suggest that FTDs “in price supported IPOs may arise from the mechanism of the offering process”. Accordingly, to avoid the possibility of IPO related FTDs, we exclude securities that started trading during our sample interval.

Third, it can be conjectured that a reported FTD may be caused by the trader that fails to receive (and hence not caused by naked shorting). In this context, we note that Evans et al. (2008) find that the number of FTDs is strongly related to rebate rates, indicating that FTDs originate largely from (naked) short transactions; and Boni (2006) shows that the number of FTDs is related to the number of short sales, and offers evidence that market makers ‘strategically’ fail to deliver when borrowing costs are high, again pointing to FTDs being governed by (naked) short selling. Still, we undertake our own empirical analysis to more formally test for this conjecture, and, consistent with extant research, do not find any support for it. Specifically, as per the regression results we report in Table 2, we find that the number of new FTDs each day is strongly and significantly related (p-value<<0.01) to the daily contemporaneous trading volume arising from short sales (that includes and should arguably be correlated with the number of naked shorts), and is not significantly related to the ‘non-short’ daily trading volume arising from regular sales (which should arguably be the source of any FTDs that are driven by failures to receive)25.

We accordingly proxy the intensity of naked short selling by an FTD-based measure: the Outstanding Naked Short Ratio (ONSR) defined for each day T as the estimated cumulative naked short sales till day T scaled by the total number of shares outstanding (obtained from CRSP). To calculate the cumulative naked short sales till day T, we adjust the outstanding FTDs data from SEC by adding the naked short sales that have already taken place but have not yet been observed because they will show up only after settlements are duly completed over days T+1, T+2 and T+326. We also compute the New Naked Short Ratio on day T as the number of new FTDs on day T divided by the number of shares traded on day T; and the Naked to All Shorts Ratio as the ONSR divided by the short interest27.

Finally, to additionally reassure ourself that our FTD-based ONSR measure is a good proxy for naked shorting, we examine whether major time-series dips and surges in ONSR correspond to periods during which we have good economic reason to independently expect such dips and surges. Accordingly, Figure 1 presents a plot of market-wide average ONSR across all securities, and cumulative S&P 500 index returns for 2007 and 2008. Two features are immediately apparent. First, the low frequency crests in ONSR correspond roughly to the low frequency troughs in cumulative S&P 500 index returns, and vice-versa. Second, and more importantly, ONSR drops sharply in two periods where we would have independently expected it to drop, and in at least the second period, the sharp drop cannot be related to returns since the change in cumulative returns is in a direction opposite to what it is in other periods. Specifically, ONSR drops sharply between July 21 and August 12, 2008 when the SEC heavily increased restrictions on naked short-selling. And it drops sharply again between September 19 and October 9, 2008, when the SEC banned short sales on 799 financial stocks. Clearly, Figure 1 provides strong independent support for our FTD-based ONSR measure being a good proxy for naked short-selling.

4,672 NYSE, AMEX, NASDAQ or ARCA securities have at least one FTD observation exceeding 10,000 shares over our 2007 sample period, and constitute our sample of what we hereafter refer to as “securities with naked shorts”. They represent about 91% of NYSE-listed securities, 61% of AMEX securities, 71% of NASDAQ securities, and 60% of ARCA securities.

In order to control for the effects of covered short selling, we also compute the Covered Short Ratio (CSR), the daily number of covered-shorted shares scaled by the daily trading volume in shares. Covered-shorted shares are estimated as the difference between the total number of shorted shares (obtained from the Regulation SHO data in TAQ) and the number of new FTDs on the day.