High-Frequency Trading and Price Discovery Terrence Hendershott ( Jonathan Brogaard Ryan Riordan)

1. High-Frequency Trading andPrice DiscoveryTerrence Hendershott (Jonathan Brogaard Ryan Riordan)

2. Overview • Technology has been good for markets • Is every use of technology good? • How do we think about (evaluate) HFT? • What are costs/benefits of those closest to the market? • Analysis of Nasdaq HFT data in terms of price discovery • Statistical model of HFT and “information” and “noise” in prices • Challenges in interpreting these results • Cross-correlations between HFT and price changes • Explore what types of information HFT has (not in paper yet): • Macro news announcements • Market-level (as opposed to individual stocks) results • Limit order book imbalances High-Frequency Trading and Price Discovery

3. The old trading floor, NYSE 1873

4. More recent trading floor High-Frequency Trading and Price Discovery

5. Current trading floor Securities Market Algorithmic& High-Frequency Traders/Systems High-Frequency Trading and Price Discovery Algorithmic and High-Frequency Trading • Algo Trade Definition: “The use of computer algorithms to manage the trading process.‘‘ • (Hendershott et al. 2011); HFT similar, but for short holding periods

6. Is the evolution good?Some News Articles • “...rule changes are need to control risks...” • “…development have had negative effects…” • “HFT firms are accused of flooding markets with orders that are cancelled…, leading to volatility” • Ultra fast trading needs curbs – global regulators • July 7, 2011 - London • “...the stock market is more prone than ever to large intraday moves with little or no fundamental catalyst.” • “locusts … feeding off the equity market.” • It's hard to imagine a better illustration (of social uselessness) than high-frequency trading. The stock market is supposed to allocate capital to its most productive uses, for example by helping companies with good ideas raise money. But it's hard to see how traders who place their orders one-thirtieth of a second faster than anyone else do anything to improve that social function. • Paul KrugmanAugust 2, 2009 - New York Times • “High-frequency traders generated about $21 billion in profits last year.” • “...use rapid-fire computers to essentially force slower investors to give up profits, then disappear before anyone knows what happened. ”

7. Longer perspective on technological changes affecting liquidity Technological advances have made the markets better: Evidence from trends and specific events/introduction High-Frequency Trading and Price Discovery

8. Liquidity (Trading Costs/Spreads) on NYSE Decline is from lower averse selection (less information in trading) -- Hendershott et al. 2011 How well can we measure liquidity as trade size -> 0? Transitory impact of large orders? High-Frequency Trading and Price Discovery

9. Algorithmic Trading on NYSE Hendershott et al. 2011 High-Frequency Trading and Price Discovery

10. Correlation is not causality: we need a “natural experiment”: Autoquote High-Frequency Trading and Price Discovery

11. Algo Trading is a superset of HFTWhat is HFT? HFT is always AT – but AT is not always HFT Typical properties of HFT: • High-speedtrading • Sophisticatedcomputerprograms • Useofco-locationservicesanddatafeeds • Short time framesforestablishingandliquidatingpositions, high tradingvolumeandintensity (SEC, 34-61358, Concept Release on Equity Market Structure) • HFT is a mixture of the use of technology and trading strategies (do they differ?) • What is new is direct access, increased electronic information sources, and more computing High-Frequency Trading and Price Discovery

12. HFT Strategies (SEC) • Anything new here? • Short-lived strategies with tight risk management • Based on public information, technology, and sophisticated tools • Is the focus the technology, strategies, or interaction? • Cross asset and cross market arbitrage • Statistical arbitrage Passive Market Making Arbitrage • Make money at the bid-ask spread and liquidity rebates Strategies Structural Directional • News Trading • Liquidity Detection (order anticipation) • Momentum trading and ignition • Latency Arbitrage • Flash orders High-Frequency Trading and Price Discovery

13. HFT and Price Discovery (beyond liquidity) United Airlines files for Ch. 11 to cut costs - Bloomberg United Airlines Share Price Sept. 19th, 2002 Sept. 6th, 2008 10:58 Sept. 8th, 2008 Sept. 8th, 2008 High-Frequency Trading and Price Discovery

14. Price Discovery: Efficient Price & Noise • Explicitly model observed prices each minute as efficient price (martingale) and noise (pricing error) • Observed price is martingale (efficient price) plus noise: • Trading makes cov(s,w) <> 0 • Standard market microstructure approach: Innovations in martingale are related to innovations in trading: • Transitory component also relates to trading: • Identification from Cov(m,u)=0 (Hasbrouck (1993) and others) High-Frequency Trading and Price Discovery

15. Price Discovery: Interpretation • ktrading’s (order flow) relation to efficient price • + informed about change in efficient price • - adversely selected on efficient price • ytrading’s (order flow) relation to noise • + direction of trading correlates with more noise • Transitory price impact, risk mgt., manipulation, order anticipation • - trading against noise, making prices more efficient • Statistical arbitrage via identifying transitory effects • Decompose overall HFT order flow into Liquidity Demand and Supply • Examine highest volatility days (stability?) High-Frequency Trading and Price Discovery

16. Nasdaq HFT Data • 120 stocks: small, medium, large cap • Identifies 26 independent HFT firms • Via unique “port” which firms use • Does not identify large integrated firms, e.g., Goldman • All data is aggregate across all HFT firms • Is this a representative sample? If not, why not? • HFT firms worried about regulatory response • Data is available via NDA with Nasdaq • Nasdaq had 20-40% market share • HFT is 42% of volume in large stock, 12% in small stocks • Market/limit order volume similar in large, less limit in small High-Frequency Trading and Price Discovery

17. Nasdaq HFT (SSM) Price Discovery Results • Overall HFT trade to make prices more efficient • Results remain (and are stronger) on high-volatility days • Market order trading is responsible for this • + correlation with efficient price, - with pricing error/noise • Consistent with forecasting both parts of returns • Limit order trading coefficients have opposite signs • - correlation with efficient price, + with pricing error/noise • Efficient price: standard adverse selection of liquidity providers • Noise: risk mgt., adverse selection, manipulation, order anticipation? • Do these passive trades make money? • Yes, earn the spread and liquidity rebates • Overall HFT profitability per $ is low: ~$0.03/$10,000 traded High-Frequency Trading and Price Discovery

18. Interpretation of Statistical Model • Is HFT being informed good or bad? • HFT liquidity demand imposes adverse selection • But, also trade against noise in prices • It is possible to do one without the other? • How important is this information? How long-lived? • Would it get into price anyway? • HFT liquidity supply looks a lot like market making (good) • How could we measure excess intermediation (bad)? • Next, cross-correlations for individual stocks and market-wide and macro news announcements High-Frequency Trading and Price Discovery

19. Individual Stocks:HFT and Future Stock Returns (1sec periods) • HFT Demand positively predict stock returns for a few seconds • HFT Supply negatively predict stock returns for a few seconds • Demand effect dominates overall High-Frequency Trading and Price Discovery

20. Event Study: Macro Announcements • Aligning return and HFT times is challenge (Nasdaq BBO for subset) • HFT Supply is adversely selected; HFT Demand is transmitting info • Overall Supply dominates; how to think about this HFT Demand? • Regressions show HFT Demand is not associated with noise High-Frequency Trading and Price Discovery

21. Event Study: Macro Announcements • Similar to Negative announcements • Predominantly HFT is providing liquidity at most stressful time • 209 Announcements in: Construction Spending, Consumer Confidence, Existing Home Sales, Factory Orders, ISM Manufacturing, ISM Services, Leading Indicators, Wholesale Inventories High-Frequency Trading and Price Discovery

22. Market-wide:HFT and Future Stock Returns (1sec periods) • HFT Demand positively predict market-wide returns for a few seconds • HFT Supply negatively predict market-wide returns for a few seconds • Supply effect dominates overall (also true in a market-level SSM) • Past market returns predict HFT similarly (E-mini predicts HFT Demand) • HFT has a longer lasting role in market-wide price discovery High-Frequency Trading and Price Discovery

23. Another source of HFT Information • Limit order book imbalance LOBI (Size at offer minus Size at bid) • LOBI predicts future returns (-), HFT Demand (-), HFT Supply (+) • Not much in the small stocks High-Frequency Trading and Price Discovery

24. Sources of HFT Information • Other sources of “Public” information in past prices, orders? • In other assets (market-wide, announcement results) • Large stock leading small stocks • HFT Demand predicts returns • How to think about short-lived nature of (public) information? • Foucault, Hombert, and Rosu (2012) • Reducing the transitory part is good • Trading on (soon to be) public information is not • How do we know what information will get into price without HFT? High-Frequency Trading and Price Discovery

25. Conclusions/Thoughts • Technology has improved markets • Algorithms: prices more efficient and markets more liquid • HFT is technology applied to certain strategies • Make prices more efficient; true on high-volatility days also • What are the benefits of getting info into prices in seconds? • Does this improve financing and investment decisions? • Should traders closest to the market mechanism be informed? • Focus attention on HFT liquidity demanding strategies? • Are long-term investors losing out (zero sum trading)? • If so, will HFT become competitive and offer technology services to long-term investors? • Will market structures evolve to support LFT (without HFT)? • What is the optimal configuration of intermediation sector? • Free entry or regulated monopoly/oligopoly? • Specific regulations on liquidity supply and demand? High-Frequency Trading and Price Discovery