Liquidity and Return Reversals

Transcription

1 Liquidity and Return Reversals Kent Daniel Columbia University Graduate School of Business No Free Lunch Seminar November 19, 2013

2 The Financial Crisis Market Making Past-Winner & Loser Portfolios Feb-08 - Mar Normalized Portfolio Value Mar 2008 past-winners past-losers May 2008 Jul 2008 Sep 2008 Nov 2008 Jan 2009 date Mar 2009 May 2009 Jul 2009 Sep 2009

3 The Financial Crisis Market Making Past-Winner & Loser Portfolios Feb-08 - Oct Normalized Portfolio Value Mar 2008 past-winners past-losers May 2008 Jul 2008 Sep 2008 Nov 2008 Jan 2009 date Mar 2009 May 2009 Jul 2009 Sep 2009

4 The Financial Crisis Market Making Past-Winner & Loser Portfolios Mar-09 - Oct-09 past-winners past-losers 4 Normalized Portfolio Value Mar 2008 May 2008 Jul 2008 Sep 2008 Nov 2008 Jan 2009 date Mar 2009 May 2009 Jul 2009 Sep 2009

5 The Financial Crisis Market Making Liquidity Provision in the Financial Crisis In the 6 months after March 9, 2009, beaten down stocks returned 460%. in contrast, The portfolio of the stocks that had best weathered the crisis was up only 23% over this 6 month period Some have argued that a contributing factor to the low prices/high subsequent returns of the beaten down past-loser stocks was inadequte liquidity provision: The capital of those who would ordinarily buy these stocks was already depleted. They had no more dry powder. This is essentially the argument used by Bernanke for the Fed s liquidity provision efforts: See speech/bernanke htm

6 Liquitidy Provision The Financial Crisis Market Making Liquidity Provision or Market Making is a business. Individuals enter this business to earn profits We are interested in how trying how various economic factors might cause market makers to step back from markets: Risk Technology Access to capital Competition Here, we re going to do dig into this questions by studying the evolution of the profitability of a short-term reversal strategy. You can view this as a simple form of a statistical arbitrage strategy.

7 The Financial Crisis Market Making From

8 The Financial Crisis Market Making The bid-ask spread: Ask Price Bid Price Ask Price

9 The Financial Crisis Market Making Ask Price Spread (=0.25) midpoint Bid Price If you knew that the value of AAPL would stay at $ , you could make lots of money as a market-maker buying that the bid, and selling at the ask. But, there are problems with this reasoning: often, someone will know before you when the value of

10 The Financial Crisis Market Making Ask Price Spread (=0.25) midpoint Bid Price The question we are interested in is the extent to which market makers quotes always reflect the value of AAPL. Market makers don t want to build up big inventories long or short of a stock. e.g., if they end up buying a lot of AAPL as the price falls, they will lower their quotes enough to induce traders to buy their inventory.

11 We re going to examine a strategy where we jump in and provide liquidity to market makers when they have lost money. If market stress is affecting their quotes, this should be a profitable strategy. Effectively, the market makers are paying us to take the inventory (long or short) off their hands. If these profits are big, it implies that price distortions are large. We ll measure profitability using the annualized Sharpe-Ratio. Note that, since we can t directly observe inventories, we ll have to estimate them using returns.

12 Forecasting the Returns tu r i,t tu r i,t-4 w r i,t-3 th r i,t-2 f r i,t-1 m r i,t We ll assume that, if the price has fallen, the market makers have bought Our strategy will be to sell to them when the price has fallen. We ll examine this strategy by first regressing today s returns on past returns. r i,t = γ 0 + γ 1 r i,t 1 + γ 2 r i,t γ 29 r i,t 29 + ɛ i,t If we re right about MM behavior, we ll see γ i < 0.

13 Strategy Sample Liquidity Provision We look at the 100 largest firms, by market capitalization, traded in the US. Looking at just these largest firms ensures that there is sufficient of trading volume every day. Our primary sample is we also examine the longer period.

14 Forecasting the Returns r i,t = γ 0 + γ 1 r i,t 1 + γ 2 r i,t γ 29 r i,t 29 r i,t = γ 0 + γ 1 u i,t 1 + γ 2 u i,t γ 29 u i,t 29 To estimate the return pattern, we run a Fama-MacBeth regression: For each of the 10,000 days in our sample, regress the returns of the 100 firms on the past 29 days of returns. Then, we average the regression coefficients from these 10,000 observations, and calculate t-statistics. However: in the regression we use residual returns: u i,t = r i,t ˆβ i,t 1 r m,t We also zero the coefficient around earnings announcement dates.

15 Fama MacBeth Regression Results 0.01 Estimated short-term reversal effect ( ) Fama-MacBeth coefficient Fama-MacBeth t-statistic l 1 l 6 l11 l16 l21 l26 lag (trading days) mean-beta expn. fit 30 35

16 Estimating Firm Risk 140 Cross-Sectional Residual Volatility and h ɛ, Jan Mar-2013 h ɛ,t 120 σ xs,t Annualized Volatility (%/ year) date

17 Estimating the Individual Firm Risk 180 Cross-Sectional Residual Volatility and h ɛ, Jun Dec h ɛ,t σ xs,t 140 Annualized Volatility (%/ year) Aug 2007 Nov 2007 Feb 2008 May 2008 Aug 2008 Nov 2008 date Feb 2009 May 2009 Aug 2009 Nov 2009

18 Implementing the Strategy r i,t r i,t-4 r i,t-3 r i,t-2 r i,t-1 r i,t To implement the strategy, we assume that, historically, we would have weighted each stock by its expected return, scaled by the forecast residual risk (variance). We re going to assume that we hedge out any market risk. We ll also skip 1 full day between time we make our forecasts, and the time we trade.

19 10 STR Strategy -- Rolling Sharpe-Ratio 8 SR over preceding 1 year end-of-interval date

20 , STR Strategy -- Rolling Sharpe-Ratio 6 SR over preceding 1 year

21 , NYSE and NASDAQ Firms STR Strategy -- Rolling Sharpe-Ratio NYSE Firms NASDAQ Firms 8 SR over preceding 1 year end-of-interval date

22 Value Weighted Daily Turnover Average daily turnover for top 100 US stocks, 1927: : Average Daily Turnover (%) -- 2 month average date