From: josh [josh@josh.com] Sent: Monday, December 03, 2001 11:33 AM To: rule-comments@sec.gov Cc: cameron@isld.com; matt@isld.com; ed@datek.com Subject: Comments on File No. S7-14-01 Request for Comment on the Effects of Decimal Trading in Subpennies Release No. 34-44568 File No. S7-14-01 -INTRODUCTION The transition to decimal prices has noticeably improved market efficiency and quality. Further reductions in the minimum increment size are likely to bring further, albeit exponentially smaller, improvements. Moving to smaller increments, like any significant change to an existing market structure, does pose interesting challenges. Technology exists today to mitigate many of these challenges. As the designer of Island ECN's market structure and system software, I think my experiences with sub-penny trading may provide insight into these issues. I hope this comment letter proves useful. The opinions expressed here are my own and not necessarily those of the Island ECN. -HOW FINE IS FINE ENOUGH? Finer increments always improve market efficiency. In an ideal market, participants would be able to place their orders at their true prices and all possible transactions would occur. It would take infinite resources to support infinitely precise prices, so real markets must balance the efficiency of finer increments with the cost of processing those finer increments. On NASDAQ, currently about on in ten trades are reported at prices with 3 or more decimal places. Only about 1 in 50 NASDAQ trades are reported with 4 decimal places of precision (the maximum precision currently supported). So, while there is significant activity at 4 places, the large falloff between 3 and 4 places indicates that the activity at 5 places would be insignificant. Four decimal places is also the smallest increment that results in a change in the net settlement amount of a single round lot (100 shares). For example, an order to buy 100 shares at 25.00001 and an order to buy 100 shares at 25.00002 both result in the buyer transferring the same net amount of $2,500.00 to the seller. Some tricky market structure problems arise when same-sized orders with different per share prices can settle to the same net amount of money. While these problems can be mitigated by having the minimum increment change based on the size of the order, there are costs associated with supporting multiple minimum increments. These costs favor a single global minimum increment when that increment is fine enough to accommodate the vast majority of all trading, as is currently the case. It is also possible to create synthetic finer increments for larger orders by entering them as several smaller orders priced across a spectrum of supported values, reducing the need for super-fine increments that would only be useful on very large orders. Therefore, I believe that four decimal digits is the correct choice for a global minimum increment. -ARE SUB-PENNY INCREMENTS CONFUSING? Decimal prices are easier to read because we use a 10-based number system, not because they have fewer digits than their fractional counterparts do. The old fraction-based prices were confusing because comparing the relative sizes of fractions with different denominators often requires several multiplication steps. To figure out that 127/256ths is less than <, you must normalize the denominators before comparing the numerators. This is confusing. Adding additional digits to decimal-based prices does not make them more confusing, just longer. Fear of increased confusion is not a compelling argument against smaller decimal increments. -ARE FLICKERING QUOTES BAD? More increments can mean more frequent price changes simply because there are more possible prices. Quickly changing, accurate, timely prices are desirable features of an efficient market. People do have trouble reading numbers that update more often than once per second, but there are many ways to present quickly changing information in a way that people can comprehend and absorb. Graphical displays can replace flickering digits with fluid motion. Human brains are well equipped for recognizing patterns in changing shapes. Just as today's common high/low indicators can condense a full day's worth of trading activity into just two numbers, shorter term high/low-type indicators can give a concise and stable numerical indication of short-term price variations while hiding minute intra-second fluctuations. Very short-term moving averages can also dampen distracting flicker for those more interested in a general sense of the current market position. There is tremendous opportunity for innovation in the design of the user interfaces that present market data to traders and investors. Competitive forces will insure that all market participants have access to market data in a format that specifically suits their diverse needs. Dealing with flickering quotes is a user-interface design challenge, not a market structure design constraint. -DO FINER INCREMENTS OBSCURE MARKET DEPTH? Today, a standard National Best Bid and Offer quote consists of four numbers; the Best Bid Price, the Best Bid Size, the Best Ask Price, and the Best Ask Size. This format has not changed in more than a century. As technology and market structure have improved over the years, the four-number system has grown less useful as a gauge of buying and selling interest. With the introduction of decimal-based prices, this format is useless. We must replace the antiquated four-number system and acknowledge that there is not (and has never been) a single Best Bid Price and Best Ask Price. Quotes only have meaning for a given size transaction. The current four-number system only gives information about the arbitrary size that happens to exist at the top price level at any given moment. A simple solution is to begin disseminating a "tiered" quote showing the Best Bid and Best Ask at several representative transaction sizes. This eliminates all of the current problems where better-priced small orders obscure worse-priced larger orders. One natural series of transaction sizes is the powers of ten ? 100 shares, 1,000 shares, and 10,000 shares. A powers-of-ten "tiered quote" might look like this? IBM 112.98 ? 113.12 ? 113.13 x 113.25 ? 113.31 ? 114.02 This quote indicates that you could sell 100 shares of IBM for 113.25 per share, or 1,000 shares for 113.12 per share, or 10,000 shares for 112.98 per share. Similarly, you could purchase 100 shares of IBM for 113.25 per share and so on. While this quote looks similar to a standard four-number quote, it is very different. Each number represents the average price per share to transact at each size level. Even a tiered quote that only supplied two tiers (and thus four prices) provides much more information about the market than the current standard four-number price-size-price-size quote. A retail investor wanting to trade 100 shares need only look at the innermost bid and ask to see the best prices available to him. An institutional trader would look to the outermost 10,000 share prices to get an idea of the slippage he could expect for a large order. Tiered quotes are a huge improvement over current top-of-book quotes and are appropriate for limited-capability media like newspapers. For visual displays, however, there are even better ways of presenting market information that convey full market depth. Fundamentally, quote information is two-dimensional in nature and any attempt to map it to fewer dimensions will necessarily discard information. I believe that the optimal quote display is a graph with the number of shares to execute on the y-axis and average price per share on the x-axis. In this representation, you can instantly determine what the exact execution price would be for any size marketable order. The shape of the buy and sell curves also conveys information about the depth and liquidity in the stock. You can see real-time prototypes of my 2-dimensional quote depth representation at http://josh.com/quote As markets become faster and more efficient, the marketplace will adopt new tools and technologies to deliver market data in a format that accurately represents buying and selling interest. The inadequacy of the current obsolete National Best Bid and Offer style quote is not a justification for limiting the size of the minimum quoting increment. -DO FINE INCREMENTS DISCOURAGE NON-MARKETABLE LIMIT ORDERS? There is a "price of disclosure" associated with entering a displayed non-marketable order into any market. This cost is fundamental and results directly from revealing intent to transact without any guarantee that a transaction will occur. Market participants can capitalize on the information disclosed to the market though displayed non-marketable orders. A parasite is able to price his orders using information gained from the original displayed order without incurring the cost of acquiring that information. Market parasites are nothing new. Anytime a market participant who is not currently driving the inside market offers to execute an order based on the inside market in the security, they are in effect purchasing priority over the order or orders that are driving the inside market. Finer increments make priority jumping explicit and more efficient. When constrained by artificially large increments, market participants tend to enter into private priority jumping arrangements where the incentive payments are typically not included in the price of the executed orders and thus are hidden from the marketplace. In efficient markets, competitive forces quickly find an equilibrium that thwarts parasitic pricing. Parasites find themselves competing with one another and ultimately must add information to the marketplace to survive, at which point they are no longer parasites. The potential for parasitic pricing is a feature of all efficient markets. Attempts to deter parasitic pricing with artificially large increments only result in making parasites less visible to the market. One measure of the fitness of a market structure is its ability to converge to an equilibrium point where parasites cannot enjoy a net information advantage. -REGULATORY EFFECTS OF FINER INCREMENTS Unfortunately, some rules are "hard-coded" with expectations about the size of minimum pricing increment in the marketplace. Applying these rules to markets with reduced minimum increments often has unintended consequences. In computer programming, "hard-coding" is a sure sign of poor quality software. It is usually in indication that the programmer did not fully understand the general case of the problem. This heuristic also seems to hold true with rulemaking. The NASDAQ and NYSE short-sale rules are bad rules. Tinkering with the various increments relating to these rules is folly; the rules are ill conceived and ineffective under any price increment. See my previous comment letter on the SEC Short Sale Concept Release file number S7-24-99 for more information. The so-called "stepping ahead" rules in both the Listed and NASDAQ markets also must be reconsidered. These rules serve only to provide regulatory cover to conflicted market participants who act as both principal and agent simultaneously. There is no minimum increment that can overcome this inherent conflict. It is impossible to protect your customer's best interests while simultaneously executing trades for yourself in the same stock. Smaller increments are beneficial because they highlight the absurdity of the current arrangement. -PRACTICAL IMPACTS OF FINER INCREMENTS Any market participant who uses the NASDAQ SelectNet system must already support execution prices with up to four decimal digits. NASDAQ's trade reporting and dissemination services also currently support up to four decimal places. NSCC currently clears trades with up to six decimal places. For trade execution, reporting, and clearance, sub-penny prices are commonplace. The national consolidated quotes, however, are currently restricted to pennies. Expanding the national quote feeds to four decimal places of resolution would not have a significant impact on capacity. As I argued prior to decimalization, moving to smaller increments actually tends to slightly reduce the total number of updates to a four-number best bid and offer quote because each individual quote includes fewer aggregated orders. Both SIAC and NASDAQ already include support for sub-penny quotes in their data feed protocols. The transition to four digit quotes under the existing quotation systems would likely only require a few months of testing. -CONCLUSIONS All U.S. marketplaces should move swiftly to support four decimal places of precision for both quoting and trade reporting. Market data vendors must be freed to replace the current four-number style consolidated quote with a quote that actually reveals buying and selling interest in the security. The NASDAQ and NYSE short-sale rules should be abrogated. As customers get better access to more precise market data, they will demand that their agents obey they fiduciary obligations and avoid conflicts of interest, eliminating the need for ineffective customer limit order protection regulation. A greater number of possible transactions will occur, accuracy of market data will improve, and market efficiency and quality will go up, keeping U.S. markets strongly competitive in the global marketplace. Josh Levine P.O. Box 362 New York, NY 10268