Two scholars associated with the University of California, Berkeley, have argued in a recent paper that there is less to latency arbitrage, or at least to a certain paradigmatic sort of latency arb, than meets the eye.
Robert P. Bartlett III and Justin McCrary used data from the Securities Information Processors (SIPs) to look at reporting lags and the question whether fast traders can and do profitably exploit stale quotes. The proposition that they do pick off stale quotes is one of the theses of Michael Lewis’ 2014 book, Flash Boys and one of the concerns behind the creation of the IEX, newly minted as a full exchange.
Bartlett and McCrary, in “How Rigged are Stock Markets?” a UC Berkeley Public Law Research Paper, conclude that “there is little scope in equity markets currently for latency arbitrage arising from stale SIP quotes.” That statement is carefully limited in at least two ways. First, the adverb “currently” suggests that there could have been such scope in the past, even in the recent past. The microsecond timestamps on the SIP data are themselves a new innovation, so the study could not examine the question of stale-price based arb prior to August 2015.
The Bartlett/McCrary analysis might, then, be applied retroactively by someone trying to show that Lewis was addressing a non-problem. But it also might also be cited to congratulate the heroes of his book on having applied a market solution to the problem it depicted. The authors suggest that “the emergence of venues such as IEX… may have simply made these SIP-oriented arbitrage strategies increasingly infeasible.”
The second qualification inherent in the above sentence is this: not all latency arb is stale-price arb. The authors of this paper acknowledge other ways in which latency can be employed, ways that the methods of their study do not reach. So their results, as they caution, “should not be over-interpreted.”
Still, it is important news if it is in fact the case that stale prices do not, or if they once but no longer do, provide the basis of profitable arb trades.
Liquidity taking trades at the SIP NBBO rather than at the Direct NBBO actually benefit in bottom line terms: they get a benefit of $0.0002 per share from the staleness. This suggests, Bartlett and McCrary say, that “the widespread concerns about the risk to liquidity takers posed by [the hypothetical] latency advantage of SIP prices are exaggerated or perhaps even misplaced.”
That small benefit is consistent across the various exchanges for which these scholars measured, except for the Chicago Stock Exchange, where it disappears. But no benefit to the liquidity providers arises there, either. Rather, ”traders were, on average, effectively indifferent to having their trades proceed at the SIP NBBO or the Direct NBBO” at the CSE.
Another IEX Related Point
While we’re on this subject … the IEX has applied for a patent for their discretionary peg (DPEG) calculation.
The DPEG addresses the issue, not of stale prices, but of “crumbling” prices, that is, an NBBO at a moment when “quotes for a particular security are experiencing rapid changes or transitions.” In those contexts, high frequency traders may have an advantage that would not be measured by the SIP-versus-direct NBBO distinction.
The patent application proposes a new type of trading order that “may be allowed to trade at more aggressive price levels if the market is relatively stable, and… less aggressive price levels when the market is unstable.”
The process works by stipulating a Crumbling Quote Indicator value “based on a predetermined algorithm and the received price data.” Orders would be processed differently once the CQI passed a certain threshold than they would while it was still in the stable terrain south of that threshold.
A good idea? Well … one that has drawn an important imitator already.