How to Price Quote Stuffing Out of the Market

by: Craig Pirrong

My first substantive post on SWP, in January of 2006, was about the pricing of message traffic on electronic trading systems. The post was motivated by a system failure on the Tokyo Stock Exchange, when the exchange’s trading system shut down under the weight of an avalanche of orders. I discussed why exchanges typically do not price system capacity (bandwidth), and how this can lead to problems, including system failures.

The controversy swirling around high frequency trading (HFT) has me thinking about the subject again. In particular, the furor over “quote stuffing.” Quote stuffing is like a dedicated denial of service (DDOS) attack on a website. Somebody submits a hugh number of orders, away from the market, and then cancels them quickly.

Quote stuffing can be done with the intent of slowing down the system on the target exchange. This can create a discrepancy between the quotes on the targeted exchange, and the quotes on other exchanges trading the same instrument, but not targeted. This triggers other HFTs to submit orders to the exchanges, to exploit the apparent–but not real–price difference. The stuffer knows this is likely to happen, and can submit orders to the other exchange(s) and trade at the expense of those duped by the mirage of a price difference.

A couple of comments. First, this is a consequence of a fragmented market structure. The SEC chose an “information and linkages” approach over a central limit order book alternative. The resulting fragmentation creates the incentive to exploit–or create–price discrepancies or apparent price discrepancies.

Second, free bandwidth encourages these strategies. There have been proposals to impose transactions taxes to address this problem, but that would throw out the baby with the bath. Not all orders cause the problem.

“Do it yourself” latency is the problem. So charge for it. Levy fees on message traffic that slows down system operation. Zero prices for submitting quotes, and prices that don’t vary with their impact on system performance, encourage overuse of system resources, and also encourage opportunistic consumption of this free resource in order to extract rents from others.

The hard question is what price to charge for latency. The cost of latency depends, in part, on the cost of the actions that others take because of the false signals emitted by systems that slow down unexpectedly and unbeknownst to them. These firms/individuals submit orders that they would not otherwise submit, and may end up with transactions that they regret once they realize that they were triggered by false signals. It is difficult to put a number on this cost. But it isn’t zero, so the cost of latency-inducing orders shouldn’t be zero either.

It may also be worthwhile to discriminate among orders. For instance, a blizzard of orders far away from the market doesn’t really improve liquidity, but can impair system performance as badly as a blizzard of orders near the market. The benefit of the away-from-the-market orders is smaller than the more competitive ones. Thus, it makes sense to charge more for away from the market orders that induce latency than orders that are close to the market. The further away from the market, the bigger the charge. This would force quote stuffers to submit orders that have a greater chance of being executed–which raises the risks and potential costs of the strategy–or to eschew the strategy altogether.

The basic lesson here isn’t really news. Bad things happen when scarce resources aren’t priced. So price them.

The important thing to keep in mind is to price what is really scarce. Taxing all orders and taxing them by the same amount is non-sensical, as different orders impose different costs and provide different benefits, and these costs and benefits can vary over time. To deter quote stuffing it is sufficient to impose fees on orders that actually impair system performance (induce latency). To reduce the possibility that “good” orders are punished, tax orders differentially: for two sets of orders that impose the same latency, charge less competitive orders higher fees than more competitive orders.

To address the concerns with HFT, regulators and exchanges should play Bob Barker, and get the prices right. One mystery–something that puzzled me back in 2006 when I first thought about this–is why don’t the exchanges do it themselves. Why do they treat their scarce system resources like free goods? Is it too costly to design and implement a pricing system? Is it because the costs of overconsumption of system resources are borne by others? If so, what is the best way to address the collective action problem?

I would hope, though, that in addressing this problem that exchanges and regulators focus on the central economic issue, which is pricing of a resource. They should attempt to design more discriminating pricing systems, rather than impose indiscriminate rules and constraints on conduct that are likely to be both over- and under-inclusive.