The Supply and Demand of Liquidity: Understanding and Measuring Institutional Trade Costs Donald B. Keim Wharton School University of Pennsylvania WRDS Advanced Research Scholar Program August 21, 2018
Understanding, Measuring and Controlling Trade Costs Measurement and control of trade (or implementation) costs is a fundamental issue facing both portfolio managers and those who evaluate their performance. It is important for portfolio managers to understand the magnitude and determinants of their trading costs. Better able to measure and control their trade costs and, hence, improve investment performance. Investment performance can be defined by the identity: Investment Performance = Alpha from Investment Strategy Implementation Costs The magnitude of Implementation Costs can easily be an order of magnitude larger than the magnitude of Alpha from Investment Strategy. Adding value by reducing implementation costs (which are indeed manageable) is far easier than adding value by increasing alpha (which is extremely difficult). Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 1
The Components of Institutional Trading Costs (1) Commissions (Explicit Costs) Payment to broker for execution can range from less than 1 to 15 per share In 2014, commissions were 0.146% of trade value for U.S. institutional equity trades (Greenwich Associates) (2) Market Maker (Bid-Ask) Spread Compensation to market maker for providing immediacy of execution. The less liquid the market in which the stock trades, the larger the spread. Spreads can range from less than 0.05% for very liquid stocks to larger than 2% for very illiquid stocks. (3) Market Impact/Price Concession The amount by which your trade moves the stock price away from its current level Market impact is likely to increase with trade size and market illiquidity: (4) Opportunity Cost The cost of not executing, and missing potential opportunity, once the decision to invest is made. the information motivating the trade may become stale. Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 2
Measurement and Magnitude of Implementation/Trade Costs We use a measure of trade costs (Perold 1988, Keim and Madhavan 1997) that incorporates temporary, permanent, and opportunity costs, plus commissions: Implementation Cost = [(P trade / P d ) 1] + (Commissions / P d ) where P trade is the trade price and P d is the price at the time the decision to trade is made. Implementation costs increase in trade illiquidity (e.g., market cap) and trade size. For example: Average Total Trading Costs (in %) by Market Cap Quintile for Common Stock Trades by 21 Institutions NYSE & AMEX Stocks Nasdaq Stocks Average Std Error # obs Average Std Error # obs Buyer Initiated Trades Largest 0.31 0.02 10,960 0.24 0.11 1,155 2 0.43 0.03 7,989 0.54 0.09 1,934 3 0.64 0.06 4,137 0.92 0.08 2,929 4 1.00 0.07 2,115 1.52 0.09 2,720 Smallest 1.78 0.12 834 2.85 0.13 1,801 Seller Initiated Trades Largest 0.26 0.02 10,901 0.16 0.12 960 2 0.63 0.04 4,738 0.85 0.18 853 3 1.02 0.09 2,296 1.18 0.12 1,517 4 1.33 0.16 1,112 1.73 0.15 1,613 Smallest 2.03 0.23 568 2.91 0.23 1,106 Source: Keim and Madhavan (1998) Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 3
Updated Estimates of Institutional Trade Costs (Frazzini, Israel, Moskowitz (2018)) Trade Costs (in basis points) are computed using the method described above Pro: Sample is much larger and more recent: 11.0 million orders worth $1.7 trillion Con: All trades made by one firm, and the firm has very good traders (how representative?) Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 4
Demanding vs. Supplying Liquidity A first look Typically, think of institutional trades as being liquidity demanding, and hence costly But depending on market conditions, investment style, trade ideas might supply liquidity lower, possibly negative, trade costs (pseudo market making) Investment strategy/style tells a lot about whether liquidity is demanded or supplied (e.g., Keim and Madhavan (1997)). Consider three investment styles: Momentum trend chasers who buy stocks in rising markets and sell stocks in falling markets. Buys (sells) more expensive in rising (falling) markets (liquidity demanding) Value investment decisions are dependent on long-term fundamentals; buy stocks when undervalued (e.g., low P/E), sell when overvalued. Allows for patient trading (and possibly liquidity provision) Diversified/Index objective to maintain a well-diversified balanced portfolio, or to minimize tracking error relative to an index; Trades might range from liquidity demanding to liquidity supplying Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 5
Average Price Impacts (implementation shortfall) vs. Investment Style: Prior excess stock return (PriorXRet) is the return on the traded stock, measured over the three weeks prior to the initiation of trade, less the market return over the same period. Price impact is the ratio of the average trade price to the closing stock price on the day before the order was initiated, minus 1.0, in excess of the market return over the interval of the trade. Price impacts for sells are multiplied by (-1) so they can be interpreted as costs and compared to the impacts for buys. Aggregate investment flow is the sum of the dollar value of all trades in the respective category ($bill). Number of orders is in parentheses. Results are for the 1996-97 and 2000 periods. Buys Sells PriorXRet <0 PriorXRet > 0 PriorXRet <0 PriorXRet > 0 Diversified -0.197 0.908 1.413-0.494 $9.44 $10.91 $8.81 $9.18 (11,842) (12,803) (9,316) (9,921) Value -0.155 0.851 0.889-0.366 $32.31 $24.77 $23.03 $34.82 (13,231) (12,301) (12,454) (13,598) Momentum 0.494 1.892 2.238 0.319 $14.19 $30.08 $23.93 $17.10 (14,159) (17,959) (13,557) (12,218) Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 6
Active Mutual Funds & the Generation of Investment Ideas (From Passive Aggressive Trading by Christoffersen, Keim, Musto and Rzeźnik) We now consider how an inventory of investment ideas, and its depletion, can result in a tradeoff between demanding and supplying liquidity and how this tradeoff affects trade costs and fund performance. We examine this through the lens of active stock mutual funds, with a special database my coauthors and I painstakingly constructed. Active mutual funds are generally viewed as liquidity demanders Develop a (presumably) profitable trade idea Pay the going price for immediacy to put it on Later, pay the going price for immediacy to take it off The generation of profitable trade ideas positive alpha The cost of implementing the idea is determined by the price of immediacy Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 7
Dynamic inventory of trade ideas Active institutional investors build an inventory of trade ideas Research produces trade ideas, the value of which presumably expires quickly Putting these ideas to work requires some amount of immediacy costly demand for liquidity Ex ante, willingness to incurring such costs the research ideas have value high expected near-term return Alpha-generating ideas need to be replenished after prior ideas are used up by trades But inventory of ideas depletes when fund inflows exceed idea production inflows need to be invested, but the fund has no positive-alpha ideas in which to invest Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 8
Demanding vs. Supplying Liquidity, again What does a fund do when the cost of a trade exceeds its value? Demand immediacy/liquidity and make the trade anyway? Not recommended Instead, supply immediacy/liquidity to someone else (e.g., if buying, take the other side of someone s sell) But have to consider: Is the stock within your investable universe? Adverse selection? (Does the seller know something that you don t?) Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 9
Testable Hypotheses Do inflows deplete buy ideas? Do outflows deplete sell ideas? Increased inflows shift from demanding to supplying immediacy when buying? Similarly, do increased outflows lead to liquidity provision when selling? Does such shifting from demanding to supplying liquidity show up in trade costs? Does the demand for liquidity reflect profitable trade ideas that ultimately show up in near-term performance on individual stock trades? What is the net effect of profitable trade ideas, costly demand for immediacy, and the benefits of liquidity provision on fund performance? Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 10
Existing Studies Rough proxies for trades from quarterly SEC disclosures: Grinblatt & Titman (1989); Wermers (1999,2000); Chalmers, Edelen & Kadlec (1999); Edelen, Evans & Kadlec (2013); Da, Gao & Jagganathan (2011) Able to assess fund-specific influences on inferred costs, and can compare inferred costs to fund performance. But changes in holdings are poor proxy for actual trades, and do not allow precise measurement of price impacts and other costs associated with actual trades. Anonymous transaction-level proprietary data: Chan & Lakonishok (1995); Keim & Madhavan (1997); Anand et al (2012); Frazzini, Israel & Moskowitz (2018); DiMascio, Lines & Naik (2016); Abel-Noser studies Can assess trade-specific and invest-style influences on costs and performance, but not able to examine fund-specific influences (e.g., flows, fund size). Fund anonymity precludes comparisons of fund performance with trade costs. Forensic matching of anonymous (Abel-Noser) transaction data with actual funds: Busse, Chordia, Jiang & Tang (2016); Xing (2017) Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 11
The Trade Data Until June 2005, Ontario Securities Commission required a Statement of Portfolio Transactions from all Canadian funds disclosed all trades plus periodic holdings. All funds reported all individual trades But the required form was ambiguous and, as a result, there was no consistency in the way institutions filled out the form. These public filings are available from SEDAR (www.sedar.com) in pdf form. Had to assemble the database ourselves from pdfs very painful. Limitations of the data: No (within-day) time-stamp No indication of whether a trade was part of a larger desired order quantity But we can observe whether there were nearby trades by the same fund in the same stock in the same direction, thereby permitting us to classify a trade as part of a possibly larger package Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 12
Building the Database Detailed trade data for January 2001 through December 2003 identity of traded stock, number of shares traded, trade price (less commissions) Merge the individual stock trade data with CRSP and TAQ data from U.S., and Datastream and TAQ data from Canada, using Cusips to match. Sample includes only Canadian and U.S. equities All values are converted to Canadian dollars (daily rate from Bank of Canada) Merge the above data at the fund level with fund-specific data from Canadian Morningstar using Morningstar id numbers. Share classes aggregated within funds Of 293 initial sample funds, we are able to match 199 with Canadian Morningstar 103 Canadian Equity Funds, 43 US Equity, 18 International, 35 Specialty represents 30% of TNA of entire Canadian mutual fund industry in 2004 After filtering and matching: 107,058 buys ($34.7 bill) and 80,202 sells ($34.5 bill) Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 13
Summary Statistics for Mutual Funds and Implementation Costs (Passive Aggressive Trading by Christoffersen, Keim, Musto and Rzeźnik) Implementation Costs (%) Total Value Mean Std Dev # obs ($bill) Buy Trades 0.11 0.23 107,058 34.7 Sell Trades 0.39 0.25 80,202 34.5 Mutual Fund Characteristics - Our Sample Total Net Assets ($mill) 406.28 862.99 Total Net Assets - Family ($bill) 20.83 20.03 Net Returns (%) 0.29 4.81 # of Funds 199 Mutual Fund Characteristics - Morningstar Universe Total Net Assets ($mill) 349.83 862.96 Total Net Assets - Family ($bill) 17.09 14.61 Net Returns (%) 0.41 4.01 # of Funds 677 Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 14
Variables Used in the Analysis: Stock-Specific Variables LogMktCap log of the market capitalization of the traded stock CanUS an indicator variable that equals one if a Canadian stock, zero if a U.S. stock Amihud impact of dollar trading volume on stock price illiquidity (Amihud (2002)) Spread average % spread over prior 20 trading days XSRet Pre 1 Mo excess return on stock over month ending the day before the trade XSRet Post 1 Day excess return on stock for the day after the trade XSRet Post 1 Wk excess return on stock over week beginning the day after the trade Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 15
Trade-Specific Variables Trade cost (TC) implementation shortfall, using same-day open price as benchmark decision to trade occurred between prior-day close and before same-day open (No intra-day time stamp for trade submission or transaction) Tr Size/Vol size of the trade relative to average shares traded over 20 trading days prior to the trade. ClosePastTrade one if the fund traded the stock at any time during the week prior to the trade, zero otherwise proxies for trade breakup Fund-Specific variables logtna the log of the total net assets of the fund measured at the end of the month of the trade logtnasponsor the log of the total net assets of the fund family measured at the end of the month of the trade Net Flow (TNAt TNAt-1 (1+Rt)) / TNAt-1 where TNAt is defined as above and Rt is the net-of-fee rate of return for the fund during month t Performance net-of-fee return on the fund during the month of the trade Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 16
Determinants of Trade Costs: a (univariate) First Look Expect: Larger trades cost more (but larger trades are broken up to camouflage intent) Large inflows depletion of trade ideas shift to liquidity provision Trade costs are decreasing in flows What we find (Fig 1): Buy costs increasing in Tr Size/Vol up to ~7 th decile, then decline Sell costs decreasing in Tr Size/Vol larger trades gravitate toward liquidity provision (or broken up) Buy costs decline w/ lagged flows (Fig 2 in paper) larger flows result in shift to liquidity provision for buys Sell costs unrelated to lagged flows Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 17
Determinants of Trade Costs: Regression Model (Table 3) Trade costs Tr Size/Vol, LaggedFlows, + stock-, fund-, and trade-specific variables Estimated separately for Buys and Sells Include specification with Tr Size/Vol kink at 75% (based on Figure 1 above) Fixed effects for year/month Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 18
Key Regression Findings: Trade Costs... decrease in lagged flows for buys... unrelated to lagged flows for sells Inflows lead to substitution of liquidity supply for liquidity demand for buys... increase with trade size for smaller buys (< 75% kink) decrease with trade size for larger buys (> 75% kink)... unrelated to trade size for sells Liquidity demand increases in buy size; largest buys include liquidity provision... increase in Amihud illiquid stocks costlier to trade, especially for buys... decrease with ClosePastTrade breakup of large trades reduces costs... increase w/ higher past returns for buys buys more expensive in rising markets... increase w/ lower past returns for sells sells more expensive in falling markets... decrease in fund size both buys & sells greater trade efficiency Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 19
Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 20
Are Post-Trade Stock Returns Related to Trade Costs? (Tables 5 & 6) Valuable ideas liquidity demand higher trade cost higher post-trade return Depleted ideas liquidity supply low or negative cost flat post-trade returns How long a post-trade period? We measure post-trade returns for the traded stock in excess of the market return (TSX300 for Canadian shares, S&P500 for U.S. shares) for one day and one week We test this relation using similar regression design as above regress post-trade stock returns on trade costs and our other explanatory variables. Main Result: When a fund pays more to buy a stock, the post-trade return is higher When a fund pays more to sell a stock, the post-trade return is lower In predicted direction; stronger at a day than at a week Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 21
Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 22
Relation between Trade Costs and Fund Performance (Table 7) So far, we find: Funds incur trading costs for buys when they have trading ideas, otherwise substitute into liquidity provision. Providing liquidity not so feasible for sells, and we see little evidence of it. For both buys and sells, higher trade costs predict better post-trade performance; What is the net effect of all this on fund performance? Regress fund s month-t return on VW trade costs of fund s trades during month t Results: When estimated over all trades, no relation between performance and trade costs When estimated separately for buys and sells: Performance is increasing in the costs of buys Performance is decreasing or flat in the costs of sells Trade costs of buys can be viewed as an investment in performance Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 23
Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 24
Conclusion Fund builds up inventory of profitable trade ideas, flows deplete it Funds substitute from demanding to providing liquidity as inventory of ideas depletes The effect of this tradeoff is strong for buy costs, not for sell costs Higher Trade Costs More informed trades Better post-trade returns for both buys and sells Fund performance is increasing in trade costs of buys Fund performance is decreasing or flat in trade costs of sells Supply and Demand of Liquidity: Understanding & Measuring Institutional Trade Costs WARSP Aug 21, 2018 25