Imported from previous forum
[ original email was from John Harris - john.harris@bondmart.com ]
During our last call I promised (or threatened :-)) to post some thoughts on how considerations of market structure might influence development of the new protocol. Please see these thoughts below. Recommendations are denoted with three asterisks ("***"). Comments would be most appreciated.
When individuals or firms (each a “person” for the sake of this discussion) undertake to effect transactions in the market, they act either as agents or principals. Agents acting as such bring together principals and may be paid for doing so, but otherwise have no financial interest in the orders they represent or execute.
*** The new protocol must allow for the identification of agents and principals in conjunction with the orders they represent or own, respectively. This may be via explicit or implicit means (implicit almost certainly being more efficient).
Not every transaction involves an agent but every transaction has one or more principals on each side of it. That is to say, every order is “owned” by at least one person and every transaction is “proprietary” on each of its sides, i.e., for the account of one person or more. The notion of a “non-proprietary trade” is absurd. All arguments to the contrary are really about the intent of proprietary trades.
*** Implicitly or explicitly, the new protocol should link persons identified as agents or principals to their respective roles.
Any trade effected for the account of two or more persons is a block trade. Defining block trades in terms of a quantity of goods traded is arbitrary and incorrect. Every block trade commences with the submission of a block order on one or both of its sides. Every block order is submitted by at least one person acting as agent, though nothing prevents one of the principals in a block trade from serving as agent for other principals.
*** The new protocol should employ the concept of allocation to all orders. That is, every order must be allocated to one or more accounts. An order allocated to two or more accounts shall be defined as a block order. It may be desirable, especially for orders allocated to a large number of accounts, to separate the allocation from the order. In such case, a device such as an allocation message could be employed. This allocation message would have a unique identifier. The order would have a pointer to that identifier. The new protocol could also support the use of standing allocation instructions. Doing so would reduce consumption of network resources. By separating allocation from order messages, it becomes possible to correct or amend allocations without correcting or amending the order that gives rise to the allocation. This may provide for a cleaner audit trail.
The allocation of block trades to accounts is non-trivial. Market participants employ various methods, including explicit allocation, percentage allocation, and weighted allocation.
*** It may be desirable for the protocol to define and specify the operation of such allocation methods.
A broker acting as such is an agent. A dealer acting as such is a principal. Exchanges act only as agents.
When trading on behalf of their clients, investment advisors act as agents. Some market participants accept and execute block orders from investment advisors without knowing for which principals the advisors act in advance of such acceptance and execution. Doing so is an irresponsible practice, predicated on an unwarranted assumption.
*** The new protocol should discourage the submission of unallocated orders and the execution of unallocated trades.
What separates exchanges from other agents is that exchanges exercise no discretion in bringing together or executing orders from buyers and sellers. Other agents always exercise discretion.
An order is an express willingness to enter into an agreement on definite terms. The consummation of an agreement to trade requires at least two parties, with at least one on each side of the trade. Principals may grant to agents the authority to effect transactions on their behalf, in accordance with their instructions. Principals may wish to limit the dissemination of order information to one or an otherwise finite number of agents or other principals.
*** The new protocol should provide principals a means of limiting exposure of their orders to other market participants and for a verifiable chain of execution control for each order, so that principals may ensure that their instructions are followed.
Many divide the market into “buyside” and “sellside,” but these terms are specious. Colloquially they connote “customers” and “dealers,” respectively. But at various times customers and dealers alike may be buyers or sellers. Most commonly, when customers effect trades with dealers as such, they do so directly and without the services of an agent.
*** The syntax of the new protocol should convey unambiguously the technical and, where necessary, economic and legal roles that agents and principals play with respect to orders and executions. It may reasonably speak of “senders” and “receivers” of orders, “buyers” and “sellers” of goods, or “producers” and “consumers” of information. It should avoid unnecessary reliance on transient regulatory or colloquial terminology.
The market thus consists of agents and principals. Some agents act as exchanges, a distinction meaningful for protocol design. With respect to their settlement obligations with respect to one another, clearinghouses sometimes replace or “novate” the principals in a transaction, each with respect to the other. So, post-trade, clearinghouses may also serve as principals – this, too, is meaningful for protocol design.
Now that we have identified the types of actors in the market, let us address the ways in which these actors discover the prices at which they may and in fact do trade. These ways are (1) auctions and (2) negotiations. Auctions may be continuous or periodic. The classic exchange operates on the principle of the continuous, two-sided auction. The classic interaction between a “dealer” and a “customer” is negotiation. When non-exchange agents bring together principals, they typically do so through negotiation.
*** For efficiency, the new protocol should rely on the specification of order destinations rather than price discovery methods. The method of price discovery is a feature of the destination, not of the order. To send an order to a classic exchange is implicitly to request execution under a continuous, two-sided auction. To send an order to a “dealer” is implicitly to request negotiated execution.
A cross is a trade effected by reference to a price obtained from a third party. Crosses may be consummated via auction or negotiation. A cross is effectively a price specification.
*** The new protocol should provide a means for users to provide cross instructions as a price specification.
Multi-leg orders – orders involving two or more goods – may be effected in only two ways: either (1) each leg is executed independently through separate transactions or (2) the legs are executed simultaneously in a single transaction at a known, aggregate price. In the case of the former, the aggregate price is unknown until each leg is executed, but the price of each leg is known immediately upon execution. In the case of the latter, the price of each leg must be assigned after the fact or else the price of at least all but one of the legs must be assigned arbitrarily, before the fact.
*** The new protocol should facilitate the reporting of constituent prices after the fact of execution for multi-leg trades in which the legs are executed simultaneously at a known, aggregate price. The protocol may require such reporting within a prescribed period of time.
When exchanges seek to provide execution of all legs in a single transaction at a known, aggregate price, they do so typically in a single order book. To do so across multiple order books or exchanges requires coordination between or among books or exchanges and the pausing of trading in all involved books or exchanges while the multi-leg trade is effected. Others seeking executions of their orders in those other books or exchanges would suffer through these pauses, so seldom or never do exchanges
*** The new protocol should assume no coordination of multi-leg executions either across exchanges or, within a single exchange, across order books.
More broadly, the objects or “goods” of the financial-trading realm – whether single- or multi-leg – fall into one of three broad categories: (1) securities, (2) commodities, and (3) contracts. All derivatives are either securities or contracts. All currencies are commodities. Repurchase agreements, loans of securities, and other financing arrangements are contracts.
Rigorous, unambiguous, unique identification of financial goods is an ongoing challenge. Some have estimated that market participants spend billions of dollars per year on identification schemes and errors in such identification.
*** The new protocol must provide a means of identifying the objects of trade in such a way that messages are as efficient as possible and not unduly burdened with identification data. One approach would be to promote the creation of central repository. Another would be to promote the use of a single standard. Yet another would be to allow destination identifiers to stand also for object identifiers. Thus, an exchange operating e.g., 200 order books would have 200 object identifiers, one for each of those books. Included in the book definition would be an identifier for the object. An order sent to that destination implicitly identifies the good that is the subject of the order.
Apologies for the length and formatting of this post. A wiki or similar facility would be helpful at this stage in the working group’s activities.
John,
your description gives a good overview of many of the fundamentals in trading. Looking at your recommendations, I did not see any gaps related to what the current FIX protocol has to offer. Is the idea to limit the FIX protocol to a well-defined subset of its current capabilities for the purpose of high performance trading environments?
Regards,
Hanno.
[ original email was from John Harris - john.harris@bondmart.com ]
Hanno,
To my knowledge, there is no consensus view (at least as yet) of what the new protocol should be, how it should differ from existing FIX, etc. I think it is fair to say, however, that many people - whether purely from their experiences with existing FIX or observation of inefficiencies or capability gaps when applying FIX within certain market segments - think a “clean sheet” protocol-development effort is timely and in order. I would guess that some may support at present only what they would posit as relatively modest, incremental improvements, e.g., binary instead of text. Others already prefer more radical departures or have grander visions.
Certainly we are at a stage of articulating competing visions, information sharing, learning from one another, groping through poorly-lit places.
I hope that multiple, well-made proposals will emerge, even if contradictory and engendering of passionate debate.
In the coming days, I will flesh out some of my ideas for comment (or insult) and hope others will do the same. As you saw, I oppose unallocated trades and would prefer a protocol in which such trades are defined as non-compliant. As I understand it, unallocated trades are compliant with FIX at present. More to come…
Best,
John
John,
your description gives a good overview of many of the fundamentals in trading. Looking at your recommendations, I did not see any gaps related to what the current FIX protocol has to offer. Is the idea to limit the FIX protocol to a well-defined subset of its current capabilities for the purpose of high performance trading environments?
Regards,
Hanno.