Tick Size Constraints, High Frequency Trading and Liquidity

Transcription

1 Tick Size Constraints, High Frequency Trading and Liquidity Chen Yao University of Warwick Mao Ye University of Illinois at Urbana-Champaign December 8, 2014

2 What Are Tick Size Constraints Standard Walrasian equilibrium Continuous price Reality Discrete prices SEC rule 612: 1 penny tick size Prohibits stock exchanges from displaying orders in an increment smaller than $0.01 if the quotation, order, or indication of interest is priced equal to or greater than $1.00 per share. Price control Price competition on liquidity provision is constrained

3 Price Control and Non-Price Competition Consequences of price control (Rockoff, 2008) Queuing, black markets, evading, and rationing Queuing First come, first served High frequency liquidity provision Queuing Compete for the position at the front of the queue At the constrained quotes

4 Price vs. Time Priority NASDAQ market Liquidity is provided by limit orders Price priority limit orders offering better prices execute first limit sells at lower prices limit buys at higher prices Time priority Limit orders at the same price are executed in the order in which they have been submitted

5 Cost to Establish Price Priority: Relative Tick Size Citigroup HSBC Price $3.3 $59 Relative Tick Size 30 basis points 1.69 basis points Citigroup HSBC

6 Price vs. Time Priority HFT willing to quote proportional spread of 30 basis points Non-HFT willing to quote proportional spread of 15 basis points

7 Relative Tick Size: 30 BPS A A 30 BPS A Willing to quote at 15 BPS 30 BPS Willing to quote at 15 BPS B B has time priority over (when they quote at the same price)

8 Relative Tick Size: 15 BPS A 30 BPS A Willing able to quote to quote at at BPS BPS B B has price priority over

9 Contribution: Tick Size Constraints Channel Speed allows HFTs to establish time priority when price competition is constrained Large relative tick size Increases the cost to establish price priority HFTs and non-hfts quote the same price Time priority determines execution precedence Large relative size leads to a large proportion of HFT liquidity provision relative to non-hft

10 Two Existing Channels on HFTs Price competition channel Speed allows HFTs to provide better liquidity Avoid pick-off risk (Hendershott, Jones and Menkveld, 2011) Better management of inventory (Brogaard et al, 2013) Low cost of operation Information channel Speed allows HFTs to adversely select non-hfts Fast access to information or Fast reaction to public information (Biais, Foucault and Moinas (2013) and Budish, Cramton and Shim (2013))

11 Compared to Price Competition Channel We find that HFTs do not quote better prices than non-hfts Suggest the existence of other economic forces that encourage non-hfts to establish price priority A large relative tick size increase the chances that HFTs and non-hfts quote the same price Facilitate HFTs to establish time priority HFT liquidity provision is most active in low priced stocks

12 Compared to Information Channel We identify A non-informational driver of speed competition ETF Splits increase and reverse splits decrease HFT activity Control group: ETFs track the same index but do not splits/reverse splits Same fundamental information Non-informational source of profit for speed competition Large relative tick size leads to higher profits of liquidity provision

13 Roadmap Relative tick size and HFT liquidity provision Double sorting Regression analysis Diff-in-diff test Robustness checks Relative tick size and profit of liquidity provision

14 Main Hypothesis on HFT Liquidity Provision Larger relative tick size causes more HFT liquidity provision relative to non-hft Challenge: endogeneity (Roberts and Whited, 2012) Omitted variables Fail to control variables correlated with price as well as HFT liquidity provision Reverse causality HFT liquidity provision reduces nominal price

15 Double sorting Identification Strategy Nominal share price is exogenous after controlling for market cap (Benartzi, Michaely, Thaler and Weld, 2009) Regressions analysis Variables that are correlated with nominal price Variables that affect HFT liquidity provision Diff-in-diff regression of ETFs splits Pilot: ETFs that split/reverse splits Control: ETFs tracking the same index but are not treated

16 Roadmap Relative tick size and HFT liquidity provision Double sorting Regression analysis Diff-in-diff test Robustness checks Relative tick size and profit of liquidity provision

17 Data: NASDAQ HFT Dataset Snapshots of limit order book The depth at best price from HFT and non-hft in each minute Trading volume with liquidity providers identified as HFTs or non-hfts 120 stratified sample of stocks in October, 2010

18 Price Competition vs. Tick Size Constraints Small relative tick size Reduces the constraints to establish price priority Should increase the proportion of liquidity provision from traders who can quote better price Implication from price competition channel Small relative tick size should increases proportion of liquidity provision from HFTs Findings under tick size constraint channel Small relative tick size facilitates non-hft to establish price priority Inconsistent with the price competition channel Suggests the existence of economic forces other than price competition channel

19 Tick Size Constraints and Price Priority Large Cap Middle Cap Small Cap (1) (2) (3) (4) Relative Tick Size HFT Non-HFT HFT & Only Only Non-HFT Ratio Large (Low Price) 1.60% 2.50% 95.90% 1.55 Medium (Medium Price) 11.90% 18.60% 69.60% 1.57 Small (High Price) 16.80% 37.70% 45.50% 2.25 Large (Low Price) 18.00% 15.20% 66.80% 0.84 Medium (Medium Price) 20.00% 56.60% 23.40% 2.83 Small (High Price) 20.70% 63.70% 15.70% 3.08 Large (Low Price) 11.30% 54.70% 34.10% 4.86 Medium (Medium Price) 20.20% 55.80% 24.00% 2.77 Small (High Price) 18.60% 70.70% 10.70% 3.8 Total 15.40% 41.70% 42.90% 2.62

20 Tick Size Constraints and Time Priority Low-priced stocks High tick size constraints Higher probability that HFTs and non-hfts quote same price HFTs can establish time priority more easily Prediction Percentage of volume with HFTs as liquidity providers increases in relative tick size

21 Percentage of Volume with HFTs as the Liquidity Providers 49,29% HFTs are most active for the group with the least price differentiation 50,00% 45,00% 40,00% 35,00% 30,00% 25,00% 20,00% 15,00% 10,00% 5,00% 0,00% 23,40% Large tick size 39,15% 19,93% 18,74% Medium tick size 38,48% 23,56% 22,34% Small tick size 35,53% Medium Cap Small Cap Large Cap Small Cap Medium Cap Large Cap

22 Roadmap Relative tick size and HFT liquidity provision Double sorting Regression analysis Diff-in-diff test More robustness checks Relative tick size and profit of liquidity provision

23 Omitted Variable Bias Causal relationship we aim to establish Large relative tick size increases proportion of liquidity provided by HFT Biases occur if we fail to control variables correlated with both Nominal prices (relative tick size) HFT liquidity provision We search for control variables affecting at least one of them Benartzi, Michaely, Thaler and Weld (2009)

24 Factor Affecting Nominal Prices Marketability hypothesis lower price appeals to individual investors Optimal tick size hypothesis firms choose optimal relative tick size through split Signaling hypothesis Firms use stock splits to signal good news Catering hypothesis Low price predicts distress risk

25 Factors Affecting HFT Liquidity Provision Probability of informed trading (PIN) Control for information asymmetry Volatility and turnover Hendershott, Jones, and Menkveld (2011) Past Returns

26 Tick Size Constraints and Time Priority Execution due to price vs. time priority HFT price priority Non-HFT limit orders offer worse price at the time of HFT execution HFT time priority Non-HFT limit orders offer identical price at the time of HFT execution Non-HFT price and time priority are similarly defined TimePriority i,t,n : proportion of trades due to time priority i: firm t: time n: non-hft or HFT (two observations each firm)

27 Specification and Results TimePriority i,t,n = β 1 tick it + β 2 HFTdummy i,t,n + β 3 tick it HFTdummy i,t,n +u j,t + Γ X i,t +ε i,t,n Dependent Variable: Proportional of Volume due to Time Priority tick 3.374*** (17.94) HFTdummy 0.077*** (8.53) tick * HFTdummy 0.439** (2.09) R N 1074 Other Controls Yes Industry*time FE Yes More trades due to time priority with large relative tick size

28 Specification and Results TimePriority i,t,n = β 1 tick it + β 2 HFTdummy i,t,n + β 3 tick it HFTdummy i,t,n +u j,t + Γ X i,t +ε i,t,n Dependent Variable: Proportional of Volume due to Time Priority tick 3.374*** (17.94) HFTdummy 0.077*** (8.53) tick * HFTdummy 0.439** (2.09) R N 1074 Other Controls Yes Industry*time FE Yes More trades due to time priority for HFT

29 Specification and Results TimePriority i,t,n = β 1 tick it + β 2 HFTdummy i,t,n + β 3 tick it HFTdummy i,t,n +u j,t + Γ X i,t +ε i,t,n Dependent Variable: Proportional of Volume due to Time Priority tick 3.374*** (17.94) HFTdummy 0.077*** (8.53) tick * HFTdummy 0.439** (2.09) R N 1074 Other Controls Yes Industry*time FE Yes Large relative tick size increases the HFT proportion of trades due to time priority relative to that of non-hft

30 Roadmap Relative tick size and HFT liquidity provision Double sorting Regression analysis Diff-in-diff test More robustness checks Relative tick size and profit of liquidity provision

31 Diff-in-Diff Regression Leveraged ETFs ETFs amplifying the return of the underlying index Appear in pairs: Bear and Bull Dow Jones 30 UDOW +300% SDOW-300% Similar issuance prices Issuers conduct splits/reverse splits after large price divergence

32 Empirical Design Treatment group: ETFs split/reverse split Control group: ETFs do not split/reverse split Share the same underlying fundamentals with the treatment group Dependent variables Proxy for HFT liquidity provision: RunInProcess Hasbrouck and Saar (2013) Correlation: 0.77 Liquidity measure: quoted spread, effective spread and depth

33 Regression Specification y i,t,j = u i,t + γ ij + ρ D i,t,j + θ return i,t,j + ε i,t,j u i,t Index by time fixed effect Control for common fundamentals γ ij : ETF fixed effect Control for time invariant difference between two ETFs Eg. : clientele before splits D i,t,j : Treatment dummy Treatment group: 1 after splits and 0 before splits Control group: always 0

34 Splits Without Tick Size Constraints Price Normal spread Reverse splits Price Normal spread Proportional spread Should not change Cost to trade the same dollar amount should not be affected HFT liquidity provision Should not change Because of the same fundamentals

35 Splits (1) (2) (3) (7) Qtspd pqtspd Depth1 RunsInProc (in cent) (in bps) (in mn) (in.1sec) Dummy treatment *** 1.007* *** (-16.02) (1.94) (1.39) (3.42) return ** ** (-2.40) (-2.11) (-0.13) (-0.63) Constant *** *** 0.129*** 1.856*** (8.39) (14.06) (6.23) (9.15) R N Index*time FE Y Y Y Y ETF FE Y Y Y Y

36 Economic Mechanism after Splits $ HFT: Non-HFT: $ $ $50.02 $ $ $50.01 $ $50.00

37 Reverse Splits (1) (2) (3) (7) Qtspd pqtspd Depth1 RunsInProc (in cent) (in bps) (in mn) (in.1sec) Dummy treatment 1.175*** *** *** *** (8.41) (-13.48) (-6.02) (-17.08) return ** 0.878** (-1.56) (-2.48) (2.19) (-1.28) Constant 3.190*** 9.260*** 0.547*** *** (8.79) (18.42) (3.95) (12.71) R N Index*time FE Y Y Y Y ETF FE Y Y Y Y

38 Economic Mechanism after Reverse Splits HFT: Non-HFT: $ $ $50.02 $ $50.01 $ $50.00 $100.00

39 Roadmap Relative tick size and HFT liquidity provision Double sorting Regression analysis Diff-in-diff test More robustness checks Relative tick size and profit of liquidity provision

40 Alternative Hypotheses HFTs prefer low-priced stocks for other reasons Possibility 1: small capital requirement to trade same amount of shares Possibility 2: clientele effects (low priced stocks have more retail traders) Falsification test Under alternative hypotheses, liquidity taking activities are affected the same way We find HFT liquidity taking activity does not increase with relative tick size

41 Other Robustness Checks Active vs. Passive improvement of best quotes Non-HFT may be present on the best quotes because HFT withdraw Eg: stale quotes Our results hold even if we only consider the case that best quotes are actively improved The tick size constraints channel Not restricted to stocks with 1 cent spread

42 Roadmap Relative tick size and HFT liquidity provision Double sorting Regression analysis Diff-in-diff test More robustness checks Relative tick size and profit of liquidity provision

43 Microstructure on Real Economy Arms race in speed directly affect real resource allocation Physical capital: investment in facilities to reduce latency Human capital: competition for human talents Indirect channels for market structure to affect real economy Asset pricing channel (through affecting cost of capital) Liquidity Information risk Ambiguity Corporate finance channel Liquidity Price discovery

44 Source of Profits for Arms Race in Speed Literature: information advantage Debate: whether information advantage is fair This paper: tick size and time priority Two predictions in the literature Large relative tick size leads to higher rents for liquidity provision Harris (1994) and Foucault, Pagano, and Röell (2013) Time priority creates higher profit Sandås (2001) and Biais, Hillion and Spatt (1995) Speed allocates rents from large relative tick size We empirically test these two predictions

45 Profit Measure Total Profit (Broggard, Hendershott and Riordan (2013)) Cash flows obtains through liquidity provision Cumulative value changes for inventory Unit profit Total profit divided by dollar volume (in basis points) We use different intervals of inventory clearance

46 Profit and Relative Tick Size Profit i,t,n = β 1 tick it + β 2 HFTdummy i,t,n + β 3 tick it HFTdummy i,t,n +u j,t + Γ X i,t +ε i,t,n Dep. Var Unit Profit (bps) (5 minutes) (30 minutes) (1 hour) (Daily) tick *** *** *** ** (5.45) (3.97) (3.43) (2.31) HFTdummy 0.762*** 0.421** (6.43) (2.29) (1.06) (-0.18) tick * HFTdummy 7.054** (2.21) (0.47) (-0.04) (-1.13) R N Other Controls Y Y Y Y Industry*time FE Y Y Y Y

47 Do HFTs Have Higher Profits? Profit i,t,n = β 1 tick it + β 2 HFTdummy i,t,n + β 3 tick it HFTdummy i,t,n +u j,t + Γ X i,t +ε i,t,n Dep. Var Unit Profit (bps) (5 minutes) (30 minutes) (1 hour) (Daily) tick *** *** *** ** (5.45) (3.97) (3.43) (2.31) HFTdummy 0.762*** 0.421** (6.43) (2.29) (1.06) (-0.18) tick * HFTdummy 7.054** (2.21) (0.47) (-0.04) (-1.13) R N Other Controls Y Y Y Y Industry*time FE Y Y Y Y

48 Does Difference Increases in Tick Size? Profit i,t,n = β 1 tick it + β 2 HFTdummy i,t,n + β 3 tick it HFTdummy i,t,n +u j,t + Γ X i,t +ε i,t,n Dep. Var Unit Profit (bps) (5 minutes) (30 minutes) (1 hour) (Daily) tick *** *** *** ** (5.45) (3.97) (3.43) (2.31) HFTdummy 0.762*** 0.421** (6.43) (2.29) (1.06) (-0.18) tick * HFTdummy 7.054** (2.21) (0.47) (-0.04) (-1.13) R N Other Controls Y Y Y Y Industry*time FE Y Y Y Y

49 Conclusion HFTs do not quote better prices than non-hfts HFTs are more active in stocks with large relative tick size Price competition is more constrained Non-informational channel of speed competition Splits/reverse splits do not increase/decrease the amount of information of an ETF relative to its pair But HFT liquidity provision activity changes Profit of liquidity provision increases in relative tick size

50 Debates on HFT Policy Implications Whether to pursue additional regulation on HFT This paper: HFT can be consequence of existing regulation Deregulation instead of more regulation? Tick size SEC recently announced pilot program to increase tick size for less liquid stocks SEC argument: wider tick size increase liquidity and controls HFT and finally increase IPO We encourage SEC considering a pilot program to decrease tick size for liquid stocks