1. Letter, 2. Government Response, 3. Data: Canada

U.S. Treasury Market

July 14, 2000              

To Whom It May Concern:

The United States Treasury market is generally considered the world’s second most efficient financial market (after foreign exchange). As the chairman of the SEC put it, “Treasury markets today exhibit an extraordinary combination of high liquidity and low transaction costs.”[1] While the current system does handle many simple Treasury trades very efficiently, many large ones arising from transactions in other markets are subjected to significant costs because of dealer front-running, market manipulation, and gouging.[2] These practices are well known to most experienced market participants, but as I have not seen them publicly discussed or acted upon by regulators I would like to try to describe both their general characteristics and some market data I recorded to illustrate their application in specific transactions.[3] I will conclude with a few thoughts on how these practices could be reduced in the Treasury market and an extrapolation to other financial markets.

By all appearances, the Treasury market approaches the requirements of economists’ perfect competition in perfect markets: price-takers (the market has hundreds of participants contributing to liquidity) executing costless transactions (the bid-offer spread averages about one-quarter of a basis point, customers pay no commissions to trade with dealers, and dealers pay less than $10 per million to trade with interdealer brokers [IDBs]) in perfect communication (live bid, offer, and traded prices are available through Telerate, Reuters, and Bloomberg), thus achieving instantaneous equilibrium (screens go blank for only a few seconds after major news releases).[4] An experienced money manager seeking to buy or sell a small to moderate amount of a benchmark Treasury (the most recently auctioned bill, note, or bond for a given maturity) can pick up one of his direct wires to primary dealers and, within a few seconds, usually execute a trade off of the same offers or bids he observes on the live pages of Cantor Fitzgerald (the largest Treasury broker) or GovPX (a display-only composite of the best prices at the other IDBs). If he does not like the prices he gets by calling one dealer, he can put several in competition on each trade.

Large trades carry the potential for front-running, though it should more accurately be called “simul-running,” because of the speed of benchmark trades (some traders increase their lead-time by hitting bids [selling] or lifting offers [buying] in the brokers’ screens as soon as a customer’s direct wire flashes based on an educated guess—frequently obvious—as to which way he is going). Competition on large trades may bring more than razor-fine markets. Traders usually do not like to miss on these inquiries because large market shares require hitting on them. So every time a customer buys a block by putting four dealers in competition, three traders have an interest in the market quickly moving against the winning dealer in order to caution other dealers against aggressive pricing on future trades and show their own firms’ trading managers and salespeople that it was a good miss (knowing when and how to miss, preferably gracefully, is an essential skill for a market maker). If the purchase is large enough and the market is well bid, one or more of the losing traders may attempt to push it up, pressuring the winning dealer to cover his sale at a loss. Some customers try to police such “back-running” since it leads to inferior pricing, but it is even harder to detect and deter than front-running.

If he decides that policing his dealers is more trouble than the information and liquidity (and entertainment) they provide is worth,[5] he can open an account with Cantor (the other IDBs restrict who may trade with them) and bypass dealers whenever he chooses. There is a problem with this tactic: at any given moment other IDBs may have a better or collectively deeper (good for more size) bid or offer than Cantor.

The market for non-benchmark (off-the-run or OTR) Treasuries is only slightly more difficult for customers to navigate. OTRs trade on yield spreads to adjacent benchmark Treasuries. Most IDBs display active, and frequently very tight (approximately two-tenths of a basis point) dealer markets in swap boxes. However, these spread markets are displayed in price terms versus a fixed price for the appropriate benchmark (updated when the outright market moves significantly) and therefore do not appear in GovPX, preventing customers from seeing live markets in OTRs. Cantor, whose screens are directly visible by non-dealers, has periodically attempted posting outright OTR prices derived from swap spreads but has had limited success because dealers (still Cantor’s dominant customers) are reluctant to give Cantor the necessary spread markets and quick to protest outright OTR markets based on others’ spread markets.

While GovPX does display continuously updated estimates of OTR bids and offers corresponding to current market levels of the appropriate benchmarks, this is a poor substitute for the live price-action seen by dealers on the individual IDB screens. (The same complaint applies, to a lesser extent, for benchmark Treasuries: GovPX shows that a 100-19 bid for the ten-year was hit; it does not show whether all the 100-19 bids were hit. This can be a meaningful distinction.) Consider the customer trying to determine if his request for a bid on a block of an OTR is being front-run. Since GovPX does not display swap-box trades, the customer has no way of seeing if the trader at the first dealer he calls immediately hits the best swap-box bid for that security, yet the bids he receives from each dealer he subsequently calls (whose traders will know exactly what happened) will suffer as a result.[6] Whether the trader at the first dealer bids the customer off of the swap-box spread he sold or off of a wider spread depends on several factors: how many of the OTR he was able to sell through the IDB at the original spread, whether he likes owning the OTR at that spread, how good a shopper he judges the customer, and whether he is under pressure from his boss to execute that customer’s inquiries or achieve broad market-share targets. If he does choose to bid off the original spread, hitting the swap-box bid may have at least reduced the probability that another dealer will match his bid. Note that the first dealer’s front-running may help non-front-running dealers by reducing the price required to win the trade. For most customers, awareness of OTR front-running depends on another dealer reporting swap-box activity to them. However, since customers have no way to prove which dealer (or dealers, for more than one may trade ahead of a customer on a given inquiry) is guilty, attempts at disciplining dealers are necessarily imprecise.

Despite these problems, the experienced money manager who follows OTR spreads continually, not just when he needs to trade them, does not telegraph his intentions, and uses dealers intelligently can buy or sell OTR Treasuries on very tight markets. Indeed, many traders grumble that they are required to provide favored customers with more liquidity than they themselves receive from other dealers through the IDBs. (They also point out that dealers are not the only ones to employ questionable tactics: some customers simultaneously execute identical large trades with several dealers, virtually guaranteeing that the market quickly moves against them.)

If an informed customer need not be at a severe disadvantage to his dealers in executing stand-alone Treasury trades, the same cannot be said for the customer who enters the Treasury market as part of another transaction, such as an interest-rate swap, corporate-debt hedge, or municipal-bond defeasance, for these structured transactions (“deals”) frequently have characteristics conducive to dealer front-running, market manipulation, and gouging. They tend to be:

  1. Large: A front-runner seeks to move the market against the customer; it takes a lot of notes or bonds (few deals require bills) to move the Treasury market. Structured transactions frequently require the purchase or sale of $250 million ten-year notes, for instance, and trades requiring a billion dollars or more of Treasuries are not extraordinary.
  2. Complicated: Many structured transactions require the dealer to execute at least three different trades, each of which may allow for front-running and manipulating a different market and extracting extraordinary bid-offer spreads, as the implied prices of each of the subtrades may be difficult for the customer to discern. Even when the component prices are disclosed, in some markets there may be no live quotes available to the customer to monitor for front-running or markups at pricing (interest-rate swaps, for example, are traded between dealers by voice brokers).
  3. Scheduled: Many large stand-alone Treasury trades occur spontaneously and therefore cannot be front-run for longer than the duration of the inquiry (e.g., employment data print stronger than expected and a hedge fund asks for a bid on $400 million five-years). Structured transactions, by their very nature, require advance notice to coordinate their multiple legs, review the customer’s credit, and address regulatory issues in other markets. If a trader knows he will be offering $500 million thirty-year bonds at 1:20 Tuesday afternoon, he can time purchases from the IDBs to try to produce the maximum market price precisely then. If the deal takes an OTR Treasury, that security’s price on spread can be pushed up as well (this is one situation where a dealer may want to post an OTR market at Cantor). Timed front-running is not without risk: if a bearish comment from a Federal Reserve Governor appears on Telerate (or a hedge fund starts to liquidate a large position) at 1:10, after the trader is already long $300 million thirty-years, he will most likely make the sale for the deal at well below his average cost. Salespeople or syndicate desks, in communication with the trader, can sometimes influence the timing of the deal, providing an escape hatch if the market starts turning against his position. In other situations, the customer independently determines the timing, but still provides far more advance warning than for a typical stand-alone Treasury trade. The scheduled deal also provides more opportunities for market manipulation by showing offered-side bids or lifting above-market offers during the minute or so of actual pricing.
  4. Price-insensitive: Many Treasury-only customers will walk away from a trade rather than buy too high or sell too low, particularly if they think front-running is to blame. Many derivative users of Treasuries, however, execute the Treasury leg of the deal regardless of market levels. This is particularly true when someone else foots the bill for any front-running, market manipulation, or gouging (yield burning provides the classic example of this principle). For instance, a corporate treasurer who sold Treasuries short against future debt issuance usually covers the short the day his company sells the new bonds, regardless of where the market is. If his rate-lock is unwound simultaneously with the pricing of the new issue, the three-basis-point (say) front-running markup in the cost of the repurchased Treasuries is exactly offset by a three-basis-point reduction in the interest rate set on the corporate, absorbed by investors who have agreed to buy it outright, rather than on spread to Treasuries. (Of course any front-running costs incurred in setting the rate lock were borne solely by the corporation.) Other customers do pay for front-running, but for reasons discussed later, simply care less about a basis point than the typical money manager.
  5. Executed by the inexperienced or indifferent: Most Treasury market investors or speculators buy or sell bonds many times a day and are rewarded at least partially in relation to their results (there are exceptions: some state treasurers and Japanese bank traders come to mind). They therefore frequently know and care when dealers take advantage of them. The treasurer of a manufacturing corporation that issues debt every few years may not have a Telerate on his desk and may not personally gain or lose from a five-basis point change in his firm’s cost of financing. This is an ideal candidate for gouging as well as front-running (the third technique, screen manipulation, would probably be wasted on a customer without the experience or information necessary to follow the market).

These features of deals give the dealer a large advantage over the customer in an associated Treasury transaction. This advantage may not be reduced by putting dealers in competition, for though three dealers will compete aggressively in offering a billion dollars of ten-years at 11:00 for a rate-lock unwind, they will cooperate nicely (usually, though not always, without any direct communication) in pushing the whole market up (and the ten-year up most of all) until that moment. And because telling three dealers a piece of market information is not too different from posting it on Telerate (which, incidentally, has a service that monitors deal flows, sometimes quite accurately), other dealers and even customers may join in the front-running. Thus, pricing a deal in competition may reduce the bid-offer spread (or eliminate it: a dealer long $400 million ten-years from lower levels may offer on the bid side, knowing that the market is artificially high and that the two other dealers will have securities for sale if they miss the trade) but increase the market level at the time of pricing by far more than the bid-offer savings. This may be a rational tradeoff for a customer whose front-running losses are absorbed by a third party (buyers of a new issue, for instance) but who must pay out of his own pocket for any difference between the level his deal is priced off of and what he pays to buy back his hedge.[7] Other customers who negotiate in advance small bid-offer spreads for their Treasury transactions do not understand that the costs of front-running usually far exceed bid-offer. A guaranteed small bid-offer for a large deal virtually requires the dealer to front-run, for selling a billion ten-years on the offered side of a fair market is rarely prudent.

A single informed dealer might not explicitly tell anyone about an impending trade (because he judges the information more valuable if held exclusively, because any disclosure might get back to the customer, or because of ethics), but this may not prevent group front-running. First, each dealer has many traders, all paid roughly in proportion to their individual profits, and therefore likely to want to participate in any front-running (traders must be alert to intradealer front-running). Second, some traders choose to signal their deal flows to other traders (dealers and speculators) by placing on-the-market (or even offered-side) bids for unusually large amounts with an IDB (usually Cantor, whose prices are the most widely distributed), presumably after quietly accumulating a significant front-running position. While this behavior is perhaps partially explained by egotism, it can also assist in achieving the maximum run up into the pricing by cautioning other dealers that there is a large, price-insensitive buyer in the market whom it would be foolhardy to oppose (and perhaps profitable to join). The actual pricing is signaled by a flurry of strong bids and high offers being lifted, sometimes followed by the opposite signal (in this example, a large on-the-market offer), for many traders like to go from long to short on the pricing of a deal (e.g., they are long $700 million when they make the sale of a billion, leaving them short $300 million), expecting the market to give back at least some of its deal-pricing gains.

But even if the sole informed dealer keeps the transaction to himself, the nature and timing of the underlying deal (a corporate-bond issuance, for example) is frequently public information. Anyone with a new-issue calendar will know the date of the deal; anyone in contact with a member of the syndicate can track the approximate time. Determining the implications for the Treasury market is frequently trivial (e.g., Eurobond issuers may customarily swap their fixed-rate debt for floating and therefore have to buy Treasuries to match their fixed payments).

How prevalent are these practices? Dealers front-run, manipulate, and gouge only as the situation—market, customer, and transaction—allows. Most stand-alone Treasury trades simply do not afford much opportunity for these practices. In addition, detection carries the risk of temporarily or permanently losing the customer, which traders have strong incentives to avoid. A trader’s ethics may also restrain such practices. So while a novice customer will usually pay for his inexperience, a stand-alone Treasury trade between a dealer and a skilled customer is close to a fair—if needlessly wearing—game.

For large structured trades, however, the expected profits from these practices are much larger and the risk of being caught is much lower. Furthermore, their size frequently dictates the head trader’s management, which usually removes any ethical considerations, for head traders are selected, at least partially, based on their willingness and ability to maximize profits from such transactions. These practices should therefore be more common than with stand-alone trades. Both my direct observation and reading of the IDB screens suggest this is the case.

While gouging is invisible to all but the gouger, front-running and market manipulation are frequently evident to any close observer of the market through two telltales:

  1. An unusually transparent and aggressive large buyer or seller in the IDBs: A trader who has executed a large customer trade usually tries to operate discretely in the IDBs to prevent other traders from pushing the market against him before he has covered his position. A trader who knows he is going to sell $400 million ten-years (and is already long $100 million of them), on the other hand, wants to encourage the market to trade up and may therefore choose to advertise the impending large purchase in the IDB screens.
  2. Deal pricing at market extremes: The intraday timing of most deals should be independent of market levels. If it is not—if deals that require the purchase of Treasuries are disproportionately priced at localized highs and deals that require their sale are disproportionately priced at localized lows—(successful) front-running may be the cause. (As I mentioned earlier, front-running is not always successful and therefore mid-market pricing does not prove its absence.)

Although most experienced Treasury traders recognize these phenomena as signs of possible front-running and manipulation, I thought it might be useful to present them to interested parties outside the bond market. With this goal in mind, in 1998 I attempted to record Treasury market data for periods I thought might contain large structured trades and information from other markets that might identify the underlying transaction and its characteristics. Combining these two sources of information, I then looked for one or both signs of possible front-running and manipulation. There are two ways to try to connect Treasury market action with a particular deal: with foreknowledge of the underlying deal and its timing (obtained from internal sources, other dealers, or electronic information services), watch the screens for aggressive activity; or having observed unusual screen activity, watch Telerate for announcements of priced deals. With either method (the first is easier), the goal is to find Treasury action that corresponds to the terms and timing of a deal. For example, the screens might show increasingly aggressive buying of the ten-year Treasury for the half hour leading up to the pricing of a billion dollar ten-year corporate bond. If a single dealer noisily purchases $400 million ten-years at Cantor over that period (brokers may reveal the presence of a dominant buyer or seller and his total purchases or sales; they are not supposed to reveal his identity), he may be front-running a rate-lock unwind by the issuer. If he made his last purchase at precisely the price the deal is spread off of and at precisely the time of pricing, that probability grows.

My data have far too many limitations to constitute a conclusive scientific study of front-running. My information is:

  1. Circumstantial: Even in my example—never mind in far less clear cases—I usually have no way of proving any connection between the Treasury screen activity and the corporate deal. Indeed, I usually have no way of knowing if there even is a rate-lock to unwind.
  2. Anonymous: The anonymity of the IDBs means that I cannot show that a large screen buyer is one of the lead underwriters of the deal (corporations usually execute rate-locks with one of them) or that the large buyers at two different IDBs are actually the same dealer.
  3. Incomplete: I was generally able to monitor only publicly-disclosed deals, primarily corporate bond offerings, and even within this subset missed a great number of them. For those deals I did follow (I tried to focus on the larger ones), I frequently missed important trades.
  4. Sometimes inaccurate: While I did print out hard copies of real-time IDB screens, I also recorded much information by hand, frequently relying on brokers for the price, size, timing, and number of counterparties of significant trades. There are surely many errors in both the brokers’ reporting and my recording of these data.
  5. Difficult to summarize: I have one or more pages of raw market data, with notes, for each deal I followed. I have not attempted to condense or summarize it.

Even with these weaknesses, I think the data strongly suggest that the Treasury component of large structured transactions is frequently subject to front-running and market manipulation. Although there were many deals that were neither priced at market extremes nor correlated with unusual screen activity (some of which simply had no Treasury trade to front-run—I had no way of eliminating deals without a Treasury component),[8] the number of large deals that were was too great to attribute to coincidence.

While my raw data do not provide an estimate of the cost of these practices, they do allow me to comfortably extrapolate across the whole market from deals I observed directly: at least a nine-figure sum is annually transferred from issuers and investors to dealers through front-running, market manipulation, and gouging on the Treasury component of structured transactions. (I am not including the cost of these practices in stand-alone Treasury trades or in the other markets involved in structured transactions.) My very rough estimate is that issuers pay two-thirds and investors one-third of this amount.

I do not know whether any of these practices are illegal. Securities Regulation in a Nutshell, the extent of my reading on the subject, does not even mention front-running and states that markups up to 5% are generally permissible. (Any laws against front-running presumably must apply only to dealers who acted on information obtained directly from the injured party, not to those who learn of a deal by other means.) However, to the untrained eye, SEA Rule 10b-5 appears to cover manipulation of the sort I am describing. I do know the practices are inefficient—many people spend their days engrossed in this zero-sum chicanery—and unfair.

Nor do I know what evidence would be required to prosecute those acts that are illegal. While the circumstantial case for front-running could be made much stronger (ask the issuing corporation which dealer executed their rate-lock; ask the IDBs to identify the dominant screen buyer), would it ever be enough to prove that the trader had not simply gotten long because he liked the market? Short of frequent prosecutions, it is hard to see what could offset the powerful incentives to front-run. Traders are rewarded for the frequently difficult task of staying ahead of customers—getting long before customers decide to buy and short before they decide to sell.[9] It is unrealistic to expect them to refrain from doing this when it is easy. Equally unrealistic is the idea (the “Chinese Wall”) that dealers will keep valuable information away from their traders: dealers are structured to funnel information as quickly as possible to those who can profit from it. Supervision would only be effective if it were so oppressive as to be impracticable: a full-time, market-savvy official on each trading desk with unrestricted access to all phone lines and trading blotters. Small improvements could be achieved by:

  1. Requiring all IDBs to make their pages (including swap boxes) available to non-dealers to help them detect front-running. It is hard to see why the holders of a government-granted franchise (the primary dealership) trading government-issued debt were allowed to determine the form of interdealer trade-data dissemination (GovPX). (As my earlier discussion of third-party front-running should make clear, I do not think dealer-customer trade data should be publicly disseminated.)
  2. Removing accounting incentives for corporations to hedge with derivatives (e.g., interest-rate swaps) instead of with the underlying securities (Treasuries and Eurodollar futures), for simple trades are less vulnerable to front-running. Better yet, streamline the registration of new issues to the point where established companies can come to market immediately, reducing the need for hedging.

But these will not address the fundamental problem. Dealers, fraught with conflicts of interest (market maker vs. speculator and underwriter vs. distributor, among others), will treat customers unfairly as long as customers have to deal with them. Fortunately, that need not be for long. Electronic exchanges providing equal, anonymous access for all market participants should be permitted to provide an alternative to dealers (and human exchanges) in both primary and secondary markets as quickly as possible.[10] Most customers will migrate to such exchanges. The competitive threat and market information provided by such exchanges should improve the terms on which the remaining customers transact their business with dealers.

Finally, though my data are limited to the Treasury market, I would be surprised if similar unethical practices were not common in any financial market with the same conflicts of interest. For instance, both market structure and second-hand information suggest that front-running is equally prevalent in the interest-rate swap market. Please contact me if you would like to examine my data (I am afraid it requires a tour guide) or discuss my conclusions.


1. Arthur Levitt’s testimony before the House Subcommittee on Finance and Hazardous Materials, March 18, 1999.

2. This letter is about markets for institutions, not individuals.

3. I traded Treasuries for a primary dealer for more than a decade. Changes in the market’s operation since then may have rendered some details in this account inaccurate.

4. I rely here on my perhaps outdated undergraduate microeconomics text, Price Theory and Applications, by Jack Hirshleifer.

5. At least one customer goes so far as to ask his floor broker at the Chicago Board of Trade which accounts sold or bought bond futures—widely used by dealers for both front-running and legitimate hedging and speculation—before his cash Treasury trade was completed. Since a trader need not use his firm’s futures commission merchant, this is an imperfect check.

6. Of course the outright component of a large OTR sale can be front-run as well: the dealer sells the OTR on spread and sells a benchmark Treasury (or bond futures) outright. Since OTR inquiries take longer to price than benchmark inquiries (another advantage for dealers from the use of swap boxes), they are frequently better candidates for front-running than benchmark inquiries, where a bid or offer is expected virtually instantaneously.

7. Why do investors buy corporate issues at their pricing if the market is frequently artificially inflated? Perhaps because the deals are usually underpriced enough on spread to make up for the outright market’s expensiveness.

8. There were also many instances of unusual screen activity for which I could not find an underlying deal. Some of these were simply aggressive Treasury trading (executing orders for a hedge fund, for example); others probably did represent front-running of deals that did not appear on news services.

9. Many dealers try to strip bid-offer from their market makers’ trading revenues to measure their value-added. Profits from front-running can account for a significant portion of this amount.

10. The same conflicts of interest that make dealer misconduct inevitable should be considered when deciding whether dealers should be permitted ownership or management of these exchanges.

[Primary dealers are banks and securities firms selected by the Federal Reserve Bank of New York as counterparties for its purchase and sale of U.S. Government securities in implementing monetary policy. There were 31 primary dealers in 1998: ABN AMRO, Aubrey Lanston, Bear Stearns, BT Alex. Brown, Barclays, Chase, CIBC Oppenheimer, Credit Suisse First Boston, Daiwa, Deutsche Bank, DLJ, Dresdner Kleinwort Benson, First Chicago, Fuji, Goldman Sachs, Greenwich Capital, HSBC, J.P. Morgan, Lehman Brothers, Merrill Lynch, Morgan Stanley, NationsBanc Montgomery, Nesbitt Burns, Nikko, Nomura, Paine Webber, Paribas, Prudential, Salomon Smith Barney, Warburg Dillon Read, and Zions Bank. There are 18 primary dealers currently: BNP Paribas, Banc of America, Barclays, Cantor Fitzgerald, Citigroup, Credit Suisse, Daiwa, Deutsche Bank, Goldman Sachs, HSBC, Jefferies, J.P. Morgan, Mizuho, Morgan Stanley, Nomura, RBC, RBS, and UBS.] 1. Letter, 2. Government Response, 3. Data: Canada

Please e-mail if you have comments or questions or would like to be notified of new content.