Locks and Crosses in the Foreign-Exchange Electronic Communication - PowerPoint PPT Presentation

Motivation Dataset Three Hypotheses on the Origin of Locks & Crosses Conclusion Locks and Crosses in the Foreign-Exchange Electronic Communication Networks Ly Tran Last updated: Apr 30, 2015 Ly Tran Locks and Crosses 1/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion A Normal Limit-Order Book EUR, Mar 03, 2008, 9:26 am 1.5176 1.5174 Price 1.5172 1.517 1.5168 1 5 10 20 40 80 160 320 Depth amount Bid Ask Mid-quote Spread > 0, or Bid < Ask Ly Tran Locks and Crosses 2/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion Locks & Crosses in Limit-Order Book Dataset EUR, May 07, 2008, 11:12 am EUR, May 07, 2008, 4:28 pm 1.544 1.5386 1.5438 1.5384 1.5436 1.5382 Price Price 1.5434 1.538 1.5432 1.5378 1.543 1.5376 1 5 10 20 40 80 160 320 1 5 10 20 40 80 160 320 Depth amount Depth amount Bid Ask Mid-quote Bid Ask Mid-quote Spread = 0,or Bid = Ask Spread < 0,or Bid > Ask Ly Tran Locks and Crosses 3/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion Locks & Crosses not Data Errors L&C Locks Crosses Sources FX Dataset 32% 8% 24% Nasdaq, Pre-open 35% 11% 24% Cao et al. 2000 Nasdaq, Reg. hours 11% 9% 2% Shkilko et al. 2008 NYSE, Reg. hours 5% 3% 2% Shkilko et al. 2008 ”Abnormality” only observed at and inside the quotes Ly Tran Locks and Crosses 4/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion Locks & Crosses in Practice Security Exchange Commission Rule 4613 (2000): Nasdaq, trading hours Rule 610 (2007): prohibits locks & crosses Proponents: rules resolve issues of non-positive spreads Confusion and unfairness to market participants Slow trading: halt auto-execution, trigger manual execution Opponents: rules create unintended consequences Lack of transparency; breeding new order types Increase in high-frequency trading Ly Tran Locks and Crosses 5/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion Locks & Crosses in Academic Literature Nasdaq preopening: signals to price discovery Nasdaq & NYSE trading hours Friction among networks Maker-taker pricing Order-in-transit Ly Tran Locks and Crosses 6/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion Contributions Three hypotheses on locks and crosses Locks Crosses Information � Market conditions � � Make-taker fees � Implications Locks & crosses in unregulated market Limit-order books carry information Ly Tran Locks and Crosses 7/25

Motivation Observation of ”Abnormality” Dataset Locks & Crosses by Different Views Three Hypotheses on the Origin of Locks & Crosses Contributions Conclusion Outline Motivation 1 Observation of ”Abnormality” Locks & Crosses by Different Views Contributions Dataset 2 Three Hypotheses on the Origin of Locks & Crosses 3 Information hypothesis Market Fragmentation Maker-Taker Pricing Conclusion 4 Ly Tran Locks and Crosses 8/25

Motivation Dataset Three Hypotheses on the Origin of Locks & Crosses Conclusion Large High-Frequency Dataset Minute-by-minute February 28th - June 3rd, 2008 7 currencies vis-a-vis USD EUR, JPY, GBP, AUD, CHF, CAD, NZD Aggregated from 7 distinct ECNs Reuters, EBS, Hotspot, Lava, Currenex, FXCM, and eSpeed 16 series per currency Best bid, best offer Best marginal price at which one could buy or sell 5mn, 10mn, 20mn, 40mn, 80mn, 160mn, 320mn Ly Tran Locks and Crosses 9/25

Motivation Dataset Three Hypotheses on the Origin of Locks & Crosses Conclusion Snapshot of Limit-Order Book EUR, Mar 03, 2008, 9:26 am 1.5176 1.5174 Price 1.5172 1.517 1.5168 1 5 10 20 40 80 160 320 Depth amount Bid Ask Mid-quote Total almost 1 million observations Ly Tran Locks and Crosses 10/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Three Hypotheses Information hypothesis Locks and crosses are created to convey information held by traders Market fragmentation hypothesis In fragmented market, market conditions affect locks’ and crosses’ occurrences Maker-taker pricing hypothesis Pricing schemes that favor liquidity makers over liquidity takers encourage locks and crosses occurrences Ly Tran Locks and Crosses 11/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Information Hypothesis Hypothesis Locks and crosses are created to convey information held by traders Examine relationship Price movement ”during” & after normal, lock, cross Locks & crosses: size, net depth Future price dynamics: returns, volatility Ly Tran Locks and Crosses 12/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Intensity of Price Contribution Intuition Does price tend to move more during locks, crosses, or normal markets? Measure � � � ∆ P c , mk � � � m m ∈ t IPC c , mk � = t ��� � m ∈ t I c , mk �� m ∈ t | ∆ P c m | m c Currency ∆ P Change in price m , t Time subscripts Market type lock, cross, normal mk I m k Index, = 1 if market is mk Ly Tran Locks and Crosses 13/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Crosses Contribute More to Price Changes Lock/Norm Cross/Norm Findings Total 1.11 2.13*** Crosses contribute more than normal and locked EUR 1.07 2.54*** markets JPY 1.05 2.48*** Locks is not different from normal markets GBP 1.15 2.05*** Literature AUD 0.97 2.28*** Locks and crosses CHF 1.08 1.46** contribute more than normal markets CAD 1.15 1.45* Crosses contribute more NZD 1.34* 2.66*** than locks Ly Tran Locks and Crosses 14/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Net Depth Measures Order Imbalance bid - OverlapDepth ask Net depth = OverlapDepth Best bid Best ask ask bid OverlapDepth OverlapDepth Ly Tran Locks and Crosses 15/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Cross Net Depth Predicts Future Returns In-sample predictive power for future returns rets c , mk m + T = α + β c , mk netdep c , mk + ε c m m Cross b c,mk 95% CI 10 Coefficients , units of .01 5 0 n r r r r r r r r y u u u u u u u u i a m o o o o o o o o d h h h h h h h h 5 1 2 1 5 2 3 6 9 2 1 1 / 1 . 1 Ly Tran Locks and Crosses 16/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Lock Net Depth Not Associated with Returns Lock 8 b c,mk 95% CI 6 Coefficients , units of .01 4 2 0 -2 n r r r r r r r r y u u u u u u u u i a m o o o o o o o o d h h h h h h h h 5 1 2 1 5 2 3 6 9 2 1 / . 1 1 1 Non-significant coefficients Ly Tran Locks and Crosses 17/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Cross Size Anticipates Volatility CrossSize = | Spread | Spread < 0 vol c , T m +1 = φ + γ T CrossSize c m + u c m Cross Size Coefficients g .3 T 95% CI .25 .2 Coefficients .15 .1 .05 r r r r r r r r y u u u u u u u u a o o o o o o o o d h h h h h h h h 1 2 1 5 2 3 6 9 2 / . 1 1 1 Ly Tran Locks and Crosses 18/25

Motivation Information hypothesis Dataset Market Fragmentation Three Hypotheses on the Origin of Locks & Crosses Maker-Taker Pricing Conclusion Locks, Crosses, and Price Discovery Cross variables convey information ⇑ Cross size → ⇑ Future volatility ⇑ Cross net depth → ⇑ Future returns Consistent with information hypothesis Locks not associated with future price dynamics Does not support information hypothesis Ly Tran Locks and Crosses 19/25