Throttling hyperactive robots- order to trade ratios at the Oslo Stock Exchange. Discussion

Transcription

1 Throttling hyperactive robots- order to trade ratios at the Oslo Stock Exchange Kjell Jorgensen, Johannes Skjeltorp, and Bernt Arne Odegaard Discussion Clara Vega Board of Governors 1

2 Summary of the Paper Oslo Stock Exchange data Every quote and transaction in the OSE Flag Automated vs Non-Automated Evaluate impact of penalizing traders with high order-totrade ratio. Event: May 25, 2012 OSE announces fee for traders violating an order-to-trade ratio exceeding 70. Fee implemented on September 1, Not all order changes count. Penalize less HFT liquidity providers. Finding: No change in market quality Policy implication:? 2

3 Comment 1 Friederich and Payne (2013) find that a penalty for high order-to-trade ratio in the Italian Stock Exchange decreases bid-ask spreads and depth at the top of the book. Why is there a difference in the results? Authors explain that the policy in Italy is for all types of orders. In contrast, the Oslo rule penalizes less HFT liquidity providers. Is this the true explanation? 3

4 Comment 1 (continued) Differences in the findings could also be due to: differences in the empirical specification and differences in markets. Need to convince the reader these other explanations do not play a role. Otherwise, the interpretation of the results are very different. The remainder of my talk suggests ways of increasing the power of your test. 4

5 Differences between Italian SE and Oslo SE Italian Stock Exchange is the main exchange, is OSE the main exchange? According to Thomson Reuters: Italian Stock Exchange (60-70 percent of trading volume) What about Oslo Stock Exchange? Authors show, OSE has 54% of trading volume of StatOil (not taking into account Asian and US exchanges). If we take into account those exchanges OSE has less than 50% of the trading volume. Suggestion: Show results for sample of stocks that are mainly traded in the OSE. 5

6 Increase the power of the test further Convince the reader penalization matters Restrict your sample to stocks with high order-to-trade ratios. Can the exchange tell you which stocks had traders whose order-to-trade ratio exceeded 70 prior to the regulation? 6

7 Increase the power of the test further 7

8 Differences in empirical specification What is the right control group? Friederich and Payne (2013) use Stoxx 50 Europe index stocks that are NOT Italian. Reason: Liquidity (bid-ask spreads) of same stock across exchanges is highly correlated (Biais, Bisiere, and Spatt (2010) find that market liquidity in the main exchange is what matters, other exchanges match bid-ask spread). Suggestion: Consider another control group. Non-Italian, non-canadian firms that are in the same industry/similar market size, traded in an exchange with no penalties for high order-to-trade ratios. 8

9 Liquidity of same stock across exchanges is correlated 9

10 Differences in empirical specification (2) Authors report diff-in-diff mean. Instead estimate a more robust regression specification: i=1 treatment, 0 control Treatment = 1 group affected by order-to-trade penalization After= 1 after event, zero otherwise X = exogenous variables = Implied volatility I would discard data from May to September 10

11 Differences in empirical specification (3) Potential controls are: implied volatility, tick size divided by stock price, market cap, etc. 11

12 Can you estimate HFT? Comment 2 12

13 Comment 2 (continued) Estimate HFT, and perhaps you ll find a bigger impact on firms with high HFT participation. 13

14 Discuss theory more. Comment 3 Bid-ask spreads are a function of: order processing, inventory control, adverse selection costs. Fees that penalize high order-to-trade ratios of fast traders that demand liquidity are likely to: decrease adverse selection costs we expect such policy would decrease bid-ask spreads. Authors do not find this to be the case. 14

15 Comment 4 What is the policy implication? Penalizing high order-to-trade ratios has no impact on the market quality We should adopt this policy because high number of orders imposes a negative externality to data processors/exchange. 15

16 Minor Comment Discussion highlights that policy penalizes HFT index arbitrage traders the most. Why is this a good thing? I would suspect policy makers want to penalize quotestuffing and spoofing, not index arbitrage trades. 16

17 Summary of Comments The authors have a great data set and ask an important question. Need to convince the reader that the reason why we do not find an impact is because the policy is different from that implemented in Italy, not because of differences in the sample/specification. Increase the power of your tests. Identify situations where the high order-to-trade fee matters the most. 17