Recursive Ordering in a Multi-Venue VECM
Tests whether ordering choices in multi-venue IBM VECMs identify venue leadership or mainly expose specification sensitivity.
This project studies price discovery in fragmented equity markets using synchronized IBM prices from five U.S. trading venues. The empirical question is whether researcher-imposed ordering choices in a multi-venue vector error-correction model identify venue leadership or mainly expose specification sensitivity.
The benchmark is a cointegrated vector error-correction model estimated at one- to 60-second intervals. The analysis separates three objects that are easy to conflate in high-frequency market data: Hasbrouck information-share ordering, equation-order diagnostics, and event-time market chronology.
The current IBM evidence is deliberately negative. Every venue can rank first under some Cholesky order, equation-order wedges remain economically small, and microsecond-stamped event-time validation does not support a stable observed-leadership interpretation. The paper is therefore best read as a specification-sensitivity diagnostic for fragmented price discovery.
Current manuscript: non-peer-reviewed working paper.