This post was originally published on my GoodeValue.com blog on 5/3/2009. Due to blog moves it was not correctly moved to this blog so I have reposted it.
In the wake of the stock market rout of 2008, the SEC is considering reinstating the uptick rule [pdf] or some variation thereof and the SEC has asked for comments from market participants (see the submitted comments). Having researched the question thoroughly (as a short seller it behooves me to be prepared for any such changes), I found that while there has been some good research on the effects of the uptick rule, there is little information in the press and most of the bloggers and journalists who have written about the uptick rule have given biased or incomplete accounts of what the uptick rule is and how it has affected markets. This post is an attempt to correct the misinformation. In addition to reading this article, I encourage you to download and read the original research that I cite in forming your own opinion on the uptick rule.
A History of Stock Market Manipulation
While bear raids are legend on Wall Street, there is much less evidence of bearish stock price manipulation than there is of bullish manipulation, even back in the wild days before there was much stock market regulation. So while tens millions of dollars of illicit profits are made each year in pump and dump schemes there have only been a couple of ‘short and distort’ schemes prosecuted in the last couple decades (in fact, of the two cases of which I am aware, Anthony Elgindy’s case involved no distortion; his crime was conspiring with FBI agents to gain information on companies actually being investigated by the FBI).
For some information on stock market manipulation back in the early 1900s I suggest reading Edwin Lefevre’s Reminiscences of a Stock Operator. Allen’s article Stock price manipulation [pdf] also has a few interesting tales, including the story of how members of the City Council of New York conspired to manipulate the stock price of the Harlem Rail company but were outsmarted by Cornelius Vanderbilt. While there is little evidence of manipulation on the part of short sellers, there is plenty of evidence of manipulation by longs to squeeze short sellers. A recent example is how Porsche cornered the stock of Volkswagen in October 2008 without anyone knowing, causing a huge short squeeze (which could not have happened in the United States due to the requirement that large stockholders disclose trades in 13d or 13f filings; for background on past stock corners see this NY Times article from 1989 and the follow-up article).
A Brief History of the Uptick Rule
In response to the stock market crash in late October 1929, the US Senate formed the Pecora Commission, which revealed many problems with banks and led directly to the Glass-Steagall act and the Securities acts of 1933 (which required that securities be registered) and 1934 (which formed the SEC to regulate stock exchanges). Short sellers were often blamed in the media and by politicians for the stock market crash. In response to this, the NYSE collected data on short interest, which was only at 0.15% of outstanding shares on November 12, 1929, compared to 2.3% on September 13, 2002 (Jones, Shorting restrictions, liquidity, and returns [pdf]).
Opponents of short selling continued to call for measures to restrict the practice or ban it outright. To prevent politicians from completely banning short selling, the NYSE ruled on October 6, 1931 that short selling on a downtick was ‘demoralizing’, thus banning the practice as demoralizing trades were already prohibited. Interest in restricting short selling continued among politicians and on April 1, 1932, a new rule went into effect that required written authorization for brokers to borrow customers’ stock to lend to short sellers. While this caused a temporary short squeeze, Wall Street has adapted, and now most brokerages’ margin account documents grant the brokerage the right to borrow all securities held in the account without payment of interest (a broker cannot borrow stock from a cash account).
Starting on February 8, 1938 short selling was restricted even further when the SEC ruled that short sells could take place only on a strict uptick (at a higher price than the previous transaction); this applied to all exchange-listed stocks, including the regional stock exchanges. This greatly restricted short selling; because it was so restrictive the SEC amended the rule to allow short selling on upticks or zero-plus ticks (this rule went into effect on March 20, 1939); the uptick rule was not changed after this point until Reg SHO came along in the early 2000s. Before addressing recent research on the effects of the uptick rule, we must take a detour to the Nasdaq (formerly an acronym, NASDAQ), which was not a stock exchange prior to August 1, 2006. It was simply the National Association of Securities Dealers Automated Quotation system for OTC securities (SEC OEA, Economic Analysis of Short Sale Price Restrictions [pdf]). Nasdaq stocks were thus not subject to the uptick rule.
In 1994, the NASD introduced a bid test for Nasdaq National Market stocks (not the smaller companies listed on the Nasdaq SmallCap Market), which required short sales of Nasdaq stocks to be above the best bid if the bid is lower than the previous best bid. This is less restrictive than the uptick rule. Furthermore, the bid test was never instituted by certain popular ECNs, including Inet and Arca; this made it easy for short sellers to avoid the bid test (Diether, Lee, & Werner, It’s SHO Time! [pdf]). So if short sellers who have primarily traded Nasdaq stocks think the uptick rule will not affect them because they had no problems in the past, they are wrong; the SEC has indicated that whatever rule it promulgates will apply to Nasdaq as well as NYSE and Amex stocks and to all the ECNs.
Non-Experimental Research on the Uptick Rule
The uptick rule prevents short sellers from hitting the bid most of the time and requires them to use limit orders above the best bid. This causes short sellers to achieve better prices than they would absent the uptick rule (because their limit orders are on average executed above the midpoint of bid and offer), but conversely many short sale orders are never filled and those that do fill take more time to fill than they would absent the uptick rule (Alexander & Peterson, Short Selling on the NYSE and the Effects of the Uptick Rule [pdf]). In fact, the uptick rule was found to delay or prevent the execution of over 90% of short sale orders on the NYSE. The lack of a bid test for NasdaqSC stocks while NasdaqNM stocks had a bid test allowed Ferri, Christophe, and Angel (A short look at bear raids Testing the bid test [pdf]) to examine the effects of the bid test on the tech bust of 2000-2001.
They found that there was less short selling in the NasdaqSC stocks than in the NasdaqNM stocks and that highly-shorted NasdaqSC stocks did not suffer a greater frequency of stock price crashes than similar NasdaqNM stocks. These conclusions are not very useful in the uptick/upbid rule debate, considering the weakness of the bid test and the ease with which it could be avoided. One intriguing finding from the study is that in every instance (out of thirteen) where a Nasdaq stock fell by over 10% in one day and over 20% of the sales that day were short sales, bad news such as a poor earnings report (and not rumors or a bear raid) explained the stock price decline. In the debate over the uptick rule, it is important to remember that tick size affects the restrictiveness of any uptick or upbid rule. After tick sizes were reduced from 1/8 dollar to 1/16 dollar in 1997, short sales became easier and executions faster (Alexander & Peterson, Implications of a reduction in tick size on short-sell order execution [pdf]). With decimalization in early 2001 the uptick and upbid rules became even less effective.
Reg SHO & The Great Experiment
Reg SHO (short for Regulation on Short Selling or Regulation Short; those SEC folks are creative!), besides changing the rules regarding naked short selling (which is beyond the purview of this article) established an experiment wherein from May 2, 2005 to August 5, 2007 one-third of all the stocks in the Russell 3000 index were exempted from the uptick and upbid rules. Detailed tick-level data (with short sales identified by the exchanges) was collected on all the stocks, with the 2000 stocks for which the bid or tick test was not removed serving as controls for the pilot group of stocks that had the rules lifted. The SEC Office of Economic Analysis analyzed the data and made it available to other economists as well. The outside economists and the SEC used varying methodologies to analyze the data; due to the number of different analyses conducted I will not describe them in detail, referring the interested reader instead to the SEC OEA report [pdf] for a good discussion of all the analyses. I provide only summaries of the results below.
The researchers were interested in the effect of the uptick/upbid tests on liquidity, volatility, the prevalence of ‘bear raids’, stock returns, and short interest. Generally speaking, different researchers reached similar conclusions. All researchers found that differences between uptick-test and non-uptick-test stocks were greater than between bid-test and non-bid-test stocks; for that reason when I will discuss only the uptick test stocks below. Multiple studies found that short-term (5-minute level) volatility increased significantly in pilot stocks, as would be expected, because short sellers could demand liquidity rather than supplying it. This increased the likelihood of small downward moves as short sellers hit the bid, exhausting all the orders at the best bid, then filling at the new lower best bid; conversely, upward volatility increased because large buyers were more likely to exhaust the best offer and drive the price higher without short sellers’ limit orders there. In fact, control stocks, because short sellers were forced to offer liquidity, had 10% greater liquidity on the offer than on the bid; this imbalance was corrected once the uptick rule was removed and test stocks thus had more balanced bid/offer liquidity (Diether, Lee, & Werner, It’s SHO Time! [pdf]).
Longer-term volatility (at the 30-minute level or longer) did not increase in pilot stocks. As to stock returns, the SEC study found no significant difference in returns over the first 6 months of the Reg SHO study. A more recent study of the same data by Harmon and Bar-Yam of the New England Complex Systems Institute or NECSI (Technical Report on SEC Uptick Repeal Pilot [pdf]) came to a different conclusion using different statistical methods. They found that pilot stocks–both Nasdaq and NYSE stocks– fell significantly relative to control stocks. I disagree with many of the methods they used, inlcuding fitting to a normal distribution and trimming outliers (with long-term stock returns having very fat tails–Mandlebrot has argued they are best described by the Cauchy distribution rather than a normal distribution–both those methods are inappropriate).*
The more important problem with their finding is that pilot Nasdaq stocks fell by 2.09% relative to control stocks, while pilot NYSE stocks fell by 2.38% relative to control stocks. As I described above, the Nasdaq bid test was generally ineffective and it could easily be avoided by sending orders to ECNs. Therefore, similarly negative adjusted returns for both NYSE and Nasdaq stocks seems likely to be a spurious finding. Furthermore, even if this finding were robust, it could indicate either that stocks in general were overvalued due to the uptick/upbid rules or that they became undervalued when the uptick/upbid rules were removed. I must wonder why the NECSI researchers did not compute the stock return data for the rest of the Reg SHO pilot period, instead sticking to the data from the first 6 months that was used by the SEC. As to bear raids, different researchers searched for them in different ways. Some used variance and semi-variance. These measures capture normal volatility more so than extreme events like bear raids.
Alexander and Peterson ((How) do price tests affect short selling [pdf]) also compare the daily price change (known to stock traders as the average daily range or ADR; this is the day’s high minus the day’s low) of pilot and control stocks, finding no differences. They also looked at price runs following a short sale, a much more powerful method of identifying bear raids. There was no tendency for there to be runs of more than two short sales executing at lower prices in pilot stocks relative to control stocks, on either the NYSE or the Nasdaq. The SEC study found no increase in short interest or put option open interest in pilot stocks. While there were more short sales in pilot stocks, the increase in short sales was explained by a decreasing size in short sales. Because short sellers did not have to worry about the uptick rule preventing them from selling quickly, they were more likely to split up larger orders to disguise their true size (just as large buyers do).
Conclusions & Caveats
The conclusion of all the economists whom I have cited above or listed in my bibliography below, as well as the SEC’s Office of Economic Analysis, is that the uptick rule does not fulfill its goal of preventing bear raids while allowing short sellers to sell into rising markets. Rather, there remains little evidence of bear raids, while the uptick rule restricts short selling even when a stock is rising. I recommend reading the transcript of the SEC’s roundtable discussion on the results [pdf] to see the opinions of the relevant people. There are a couple caveats to the Reg SHO pilot study. It took place as stock markets were generally rising and volatility was low. If bear raids do exist, it is likely that they would be more prevalent during bear markets when confidence is low and investors are more easily scared. None of the above studies can address whether bear raids have taken place over the last year.
I am of course against reinstating the uptick rule. It is a poor rule that restricts all short selling (even though speculative short selling is a less common use of short selling than are arbitrage, hedging, and market neutral strategies) and if the SEC really wishes to prevent bear raids, it would make more sense to use some sort of individual stock circuit breaker, such as the one proposed by the major US stock exchanges that would institute a bid test only after a stock has fallen by a certain percentage, such as 10%. Such a rule would allow most short selling while restricting it when a stock is falling precipitously. Furthermore, a circuit breaker that automatically halts trading in a stock for 30 minutes following a huge drop or spike (say 20%) would be a good addition to the rule proposed by the stock exchanges.
SEC spotlight on short sales SEC OEA: Economic Analysis of the Short Sale Price Restrictions Under the Regulation SHO Pilot [pdf]
Transcript of the SEC Roundtable on the Regulation SHO Pilot [pdf]
A short look at bear raids Testing the bid test [pdf]
(How) do price tests affect short selling [pdf]
Implications of a reduction in tick size on short-sell order execution [pdf]
Its SHO time! Short-Sale Price Tests and Market Quality [pdf]
Limited arbitrage and short sales restrictions: evidence from the options markets [pdf] Market bubbles and wasteful avoidance: tax and regulatory constraints on short sales [pdf] Nibbling at the Edges – Regulation of Short Selling Short Selling on the New York Stock Exchange and the Effects of the Uptick Rule [pdf]
Shorting restrictions, liquidity, and returns [pdf]
Short-Sale Constraints Good or Bad News for the Stock Market [pdf]
Stock price manipulation [pdf]
The uptick rule of short sale regulation (This paper examines the rate of decline of Reg SHO pilot and control stocks following negative earnings reports and finds no difference) Short selling and the informational efficiency of prices Technical Report on SEC Uptick Repeal Pilot [pdf]
*I should also mention that combining the NYSE and Nasdaq stock data (one of the other methods used by the NECSI researchers) is also innappropriate, given the facts I outlined above after the asterisk (I apologize for making the reader jump around). The one method the researchers used that was legitimate was using non-market-adjusted data. In general, I have to say that I was unimpressed with the NECSI report. It is obvious that the authors are not social scientists and are thus unfamiliar with certain difficulties inherent in social science research. Their other main analysis compared the number of large stock drops in 2000 and 2001 to those in 2007 and 2008. It is not appropriate to compare stock returns during a mild recession and the bursting of a bubble in a small subset of stocks and a period during which most large financial companies require government support to stay afloat and GDP in many countries decreases at rates not seen since the Great Depression as every single stock sector got pummeled.