The Breakdown of Standard Microstructure Techniques: And What to Do About it

Abstract

U.S. equity markets have explosively increased their trade and quote frequency and the decline of the dominance of the NYSE has increased the importance of National Best Bid and Offer (NBBO) quotes. We address three methodology issues that arise in the computation of the NBBO: (1) millisecond versus second timestamps, (2) withdrawn quotes, and (3) cancelled quotes. We find that each of these three issues is a significant and independent source of distortion in measures of market quality. We find that the potential cost per year of poor routing decisions based on these distortions is $8.4 Billion. These distortions and costs are so massive that standard microstructure techniques essentially fail. We test fourteen different methods for calculating the NBBO based on different combinations of three data clean-up techniques, two alternative quote sources, and three quote timing techniques. We conclude that the first best solution is to use the Complete DTAQ NBBO, found by combining the NBBO and Quotes files in the Daily Trade And Quote (DTAQ) database, because this is the only way to avoid major distortions on most performance criteria and the only way to use the full sample. If a researcher is financially constrained to using only the Monthly Trade And Quote (MTAQ) database, then the second best solution is to use two clean-up techniques (Withdrawn Quotes and exclude the remaining NBBO Crossed and Locked observations) and use Interpolated Time as the quote timing technique. Looking to the future, we anticipate the ultimate demise of the NBBO and propose to replace it with a Relative Best Bid and Offer (RBBO) that is different for each market center.