Trends in Institutional Stock Ownership and dS Some IImplications li ti Marshall Blume and Donald Keim The Wharton School University of Pennsylvania S i Q Spring Q-Group G C Conference f April 2, 2008 April 2, 2008 Spring Q-Group Conference 1 The Impact of Changing Stock Ownership ● Since World War II, institutional investors have increased their percentage holdings of US equities. ● Importantly, over time the composition of institutional portfolios has diverged from the composition of the market portfolio. ● The h increased i d presence off institutions i i i in i the h stockk market has possible implications for − the h li liquidity idi off the h market k − the behavior of stock prices − the returns on institutional portfolios April 2, 2008 Spring Q-Group Conference 2 What we do ● Update stylized facts on institutional stock ownership, extending the evidence byy 10 yyears to 2006 → a period of considerable change in equity markets ● Analyze (somewhat preliminary at this point) the implications of this changing ownership for − equity i market k liquidity li idi − daily return momentum − institutional returns April 2, 2008 Spring Q-Group Conference 3 Preview of Conclusions ● Over time, institutions & mutual funds increased their holdings of smaller stocks and decreased their holdings of larger stocks. → Now N underweight d i ht the th largest l t stocks t k andd overweight the smaller stocks relative to market weights. ● The increasing liquidity in the stock market generally, generally and in smaller cap stocks in particular, over the past two decades mirrors the increase in institutional ownership of common stocks. → Today virtually no stale prices, except for very smallest stocks ● Daily return predictability is time varying ● Any excess returns of institutional investors in the past 25 years are primarily due to security selection within market cap deciles, not to allocation across deciles (and not significant). April 2, 2008 Spring Q-Group Conference 4 Institutional Stock Ownership – Literature • Overall trends in equity holdings • Goldsmith (1971) – 1850-1952 • Friedman (1996) • Trends in composition of equity holdings • Del Guercia (1996) – banks hold safer stocks • Falkenstein (1996) – generally tilted towards larger stocks, except small cap funds • Gompers and Metrick (2001) – over the 1980 to 1996 period institutions increased their demand for large stocks and decreased their demand for small stocks April 2, 2008 Spring Q-Group Conference 5 Institutional Ownership of Common Stock 1900-2006 100 Households Perrcent of Total Stock Outstanding 90 Institutions 80 70 60 50 40 30 20 10 0 1900 1912 1922 1929 1939 1945 1950 1960 1970 1980 1990 1994 2006 Sources: Goldsmith (1971), FRB Flow of Funds April 2, 2008 Spring Q-Group Conference 6 Data • Institutional Holdings (Thomson/CDA via WRDS) – Quarterlyy institutional holdings g from 13f filings g – Year end 1980 through year end 2006 – All institutions with over $100 million in discretionary assets must report at quarter end common stock positions of the greater of 10,000 shares or $200,000 – Some forms are submitted to the SEC late; some are submitted b itt d on time ti andd are, att the th requestt off the th institution, made public a year later • Mutual Fund Holdings (Thomson/CDA via WRDS) – Quarterly mutual fund holdings from S12 filings – Year end 1980 through g yyear end 2006 April 2, 2008 Spring Q-Group Conference 7 Data (cont.) ( ) • Trade and Quote (TAQ) data, Jan 1993 to Dec 2006 − Last transactions prices, volumes, and trade times on primary markets for all NYSE, AMEX and Nasdaq stocks • CRSP Monthly M thl andd Daily D il Returns R t Files, Fil 1980 – 2006 (also shares outstanding, volume) • We exclude ETFs, closed end funds, and ADRs • TAQ and CRSP data are matched by a WRDS file • Stock and institutional/mutual fund data are combined by matching cusips of individual stocks April 2, 2008 Spring Q-Group Conference 8 A Caveat when using WRDS-distributed Thomson/CDA Data Axis Title e Cisco Stock Price‐‐1999 140 120 100 80 60 40 20 0 0 06/16 Thomson Adjustment (fdate) 30 J 30-Jun 12300 64.4375 0 05/25 0 05/04 0 04/13 0 03/22 0 03/01 Manager 220 Original Report (rdate) D t Date 31 M 31-Mar Shares 6150 Price 109.563 WRDS Adjustment (rdate) 31 M 31-Mar 12300 109.5625 → WRDS doubled the market value of Cisco April 2, 2008 Spring Q-Group Conference 9 Distribution of stock holdings across market market-cap cap deciles ● Partition stocks into ten equal-capitalization deciles at the end of each year from 1980 to 2006 → each decile contains (approximately) 10% of the total market value ● Compute institutional (13f) and mutual fund (S12) holdings within eachh off the h ddeciles il at eachh year endd andd produce d annuall distributions di ib i of holdings. ● Compare value of institutional holdings to total market value of each decile and compute under- or over-weighting of institutional holdings. April 2, 2008 Spring Q-Group Conference 10 Equal-Mkt-Cap Deciles vs. Traditional (equal # names) Mkt-Cap Deciles → Largest decile represents disproportionately large percent (60%) of market and Smallest decile represents disproportionately small percent (1.6%) of market April 2, 2008 Spring Q-Group Conference 11 Table 1 Percentage Distribution of Stock Ownership by Market Cap Deciles: All Stocks, Institutions, and Mutual Funds Year End 1980 and 2006 A. 1980 All Stocks All Institutions All Mutual Funds Market Cap Decile 4th 5th 6th Largest 2nd 3rd 9.74 9.99 5.26 10.25 11.02 6.72 9.92 13.51 9.40 9.91 12.07 9.34 10.10 10.99 11.06 4 4 9 13 20 33 35 68 9.77 9 77 8.38 7.02 9.77 9 77 8.63 8.71 10.13 10 13 8.72 8.28 6 6 11 17 19 36 Total for 9th Smallest All Stocks 7th 8th 10.07 11.67 12.50 9.96 10.47 12.49 10.04 9.69 13.47 10.01 7.12 12.67 10.01 3.47 7.09 $1,375 $473 $45 59 127 87 214 133 347 230 577 514 1,091 3,753 4,844 4844 10.10 10 10 10.35 10.76 10.16 10 16 9.54 9.83 9.99 9 99 10.53 10.83 10.04 10 04 10.95 11.14 10.01 10 01 11.43 12.08 10.02 10 02 11.26 11.83 10.00 10 00 $18,841 $18 841 10.20 $12,594 9.52 $3,680 31 67 52 119 81 200 124 324 240 564 584 1,148 4,186 , 5,334 Memo: All Stocks Number Cumulative B. 2006 All Stocks All Institutions All Mutual Funds Memo: All Stocks Number Cumulative 5334 → Market value of stocks is highly concentrated → Institutional / Fund portfolios deviate from market weights, and change over time April 2, 2008 Spring Q-Group Conference 12 Table 2 Percentage Distributions of Stock Ownership by Deciles: All Stocks, Institutions, and Mutual Funds Year end 1980 and 2006 Largest 2nd 3rd Market Cap Decile 4th 5th 6th 7th 8th 9th Smallest Total A. Percent Over and Underweight 1980 All Institutions All Mutual Funds 2.7 -45.9 7.4 -34.5 36.2 -5.2 21.9 -5.7 8.8 9.5 15.9 24.1 5.1 25.4 -3.5 34.1 -28.9 26.6 -65.3 -29.2 2006 All Institutions All Mutual Funds -14.3 -28.2 -11.7 -10.8 -13.9 -18.2 2.5 6.5 -6.1 -3.2 5.4 8.4 9.1 10.9 14.1 20.7 12.4 18.0 2.0 -4.9 B. Percentage Ownership 1980 All Institutions All Mutual Funds 35.3 1.8 37.0 2.2 46.9 3.1 41.9 3.1 37.4 3.6 39.9 4.1 36.2 4.1 33.2 4.4 24.5 4.2 11.9 2.3 34.4 3.3 2006 All Institutions All Mutual Funds 57.3 57 3 14.0 59.0 59 0 17.4 57.5 57 5 16.0 68.5 68 5 20.8 62.8 62 8 18.9 70.5 70 5 21.2 72.9 72 9 21.7 76.3 76 3 23.6 75.1 75 1 23.0 68.2 68 2 18.6 66.8 66 8 19.5 → Byy 2006, largest g 30% of the market is underweighted g byy institutions / funds and smaller-cap stocks (approximately the bottom half) are over-weighted April 2, 2008 Spring Q-Group Conference 13 Time Trend of 4 Cap-Decile groups Under & Overweighting - All Institutions (%) 30 20 10 0 -10 -20 -30 30 -40 -50 -60 60 Largest 4 April 2, 2008 5th to 8th 9th Decile Spring Q-Group Conference 2006 2004 2002 2000 1998 1996 1994 1992 1990 1988 1986 1984 1982 1980 -70 10th Decile 14 Table 3 Percentage Distributions of Stock Ownership in the Smallest Market Cap Decile All Stocks, Institutions, and Mutual Funds (Year End 1980 and 2006) Ninth Decile (Smallest 10% of Market Value) Next 1% Next 0.5% Next 0.25% Smallest 0.25% Entire Decile Largest 8% A. Percent Over and Underweight 1980 All Institutions All Mutual Funds 2006 All Institutions All Mutual Funds 17.1 18.8 -62.5 -66.7 -68.1 -74.7 -79.9 -92.5 -82.2 -93.3 8.5 8 5 12.2 -18 18.6 6 -34.0 -37 37.7 7 -52.6 -54 54.4 4 -68.3 -69 69.2 2 -81.5 B. Percentage Ownership 1980 All Institutions All Mutual Funds 14.0 2.8 4.5 0.8 3.8 0.6 2.4 0.2 2.1 0.2 11.9 2.3 2006 All Institutions All Mutual Funds 74.0 20 9 20.9 55.5 12 3 12.3 42.5 88 8.8 31.1 59 5.9 21.0 34 3.4 68.2 18 6 18.6 C. Stocks in Sub-Decile as a Percent of Total Stocks in Market 1980 2006 30.2 29.2 11.7 11.1 9.8 9.7 7.6 7.9 18.1 20.6 77.5 78.5 → Only the smallest 2% of stocks are under-weighted by institutions / funds April 2, 2008 Spring Q-Group Conference 15 Comparison with Previous Results • Gompers p and Metrick report p an average g correlation from 1980 through 1996 between institutional ownership and log(market value) of 0 625 0.625 • We recalculated this correlation each year from 1980 through g 2006. • Values ranged from 0.61 to 0.67 → same results • But the results from Equal-Cap decile distributions shown h above b are inconsistent i i with i h these h results… l • What is happening? April 2, 2008 Spring Q-Group Conference 16 Figure 3A. Predicted and Actual Institutional Ownership of Securities by Market Value Deciles 1980 06 0.6 Sizes of disks reflect number of 0.5 Labels next to disks are residuals from the predicted 0.062 0.037 0.073 Perccent Ins stitutional Ow wnership 0.4 -0.085 -0.147 0.029 0.056 0.3 0 056 0.056 0.012 0.2 -0.010 0.1 0 7 9 11 13 15 17 19 -0.1 Log of Market Value → The estimated regression fits the data reasonably well April 2, 2008 Spring Q-Group Conference 17 Figure 3B. Predicted and Actual Institutional Ownership of Securities by Market Value Deciles 2006 14 1.4 Sizes of disks reflect number of securities in that market value decile 1.2 Percent Ins stitutional Ow wnership Labels next to disks are residuals from the predicted values -0.603 -0.542 1 -0.496 -0.335 -0.335 -0.216 -0.151 0.028 0.8 -0.049 0 049 0.6 0.019 0.4 0.2 0 10 11 12 13 14 15 16 17 18 19 20 Log g of Market Value → Estimated regression line driven by stocks in the smallest two deciles April 2, 2008 Spring Q-Group Conference 18 Table 4 Percentage Distribution of Stock Ownership by Cap Deciles All Stocks, Institutions, and Diversified Equity Mutual Funds; Grouped by Morningstar Equity Style Categories Year End 2006 Largest 2nd 3rd Market Cap Decile 4th 5th 6th 7th 8th 17 15 -2 2 18 18 9 -2 -86 -98 -84 -8 -16 -7 -9 25 154 146 160 180 -81 152 9th Smallest Percent Over and Underweight S&P 500 Index Funds Large Value Large Blend Large Growth Mid Value Mid Blend Mid Growth Small Value Small Blend Small Growth 45 49 19 11 -38 -99 -97 -99 -100 -100 -100 35 35 34 29 -26 -95 -96 -96 -100 -100 -100 29 30 25 18 23 -88 -89 -96 -100 -100 -100 26 33 22 30 44 -95 -86 -90 -100 -100 -100 10 7 43 8 13 -88 -75 -88 -100 -100 -99 23 21 8 27 50 -67 -60 -64 -98 -99 -97 -76 -74 -52 -40 -34 268 270 255 602 159 626 -99 -99 -90 -76 -77 90 76 120 617 → Relative to the S&P, Large-Cap funds underweight the largest cap stocks → Mid-Cap funds tilt toward the smallest 4 deciles, relative to market weights → Small Value and Small Growth funds do not hold any stocks from smallest decile → Small Blend funds concentrate their portfolios in the 9thh and 10thh deciles April 2, 2008 Spring Q-Group Conference 19 Institutional Stock Ownership & Equity Market Liquidity ● Some stylized facts: − Institutions trade more than individuals − Institutional I tit ti l stock t k ownership hi increased i d over last l t 25 years − Big institutional shift to smaller stocks over last 25 years ● Two conjectures: (1) as institutions increase their overall ownership of stocks (figure 1), overall equity market liquidity should increase, holding constant other influences on liquidity (2) as institutions increase their relative holdings of smaller stocks ((figure g 2), ), the relative liquidity q y of smaller stocks should increase ● We examine three liquidity measures: bid-ask spreads, yearly turnover rates, and stale prices April 2, 2008 Spring Q-Group Conference 20 Figure 4 Median Bid-Ask Spreads (%) for Market-Cap Quintiles with equal # of Stocks (Quintile cutoffs based on all NYSE NYSE, AMEX and Nasdaq Stocks) 12 Largest Quintile 2 3 4 Smallest Quintile 10 8 6 4 2 0 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 ● Even during regimes when minimum tick size was constant at eighths, sixteenths and pennies, percent spreads were falling ● Largest declines for the 3 smallest quintiles (~ the smallest decile in analysis above) → smallest ll stocks k most sensitive i i to changes h in i factors f affecting ff i liquidity li idi → Decline due to increased institutional holdings of these stocks over the period? April 2, 2008 Spring Q-Group Conference 21 Figure 5 Turnover – Annual Ratio of (Shares Traded/Total Outstanding) Grouped Market-Cap Market Cap Deciles 400% 350% Largest 4 Deciles 300% Deciles 5 to 8 Decile 9 250% Smallest Decile 200% 150% 100% 50% 0% 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 ● General increase in turnover through 2001 is consistent with first conjecture, but subsequent decline afterward despite increased institutional ownership is inconsistent ● Ratio of smallest-cap turnover to largest-cap turnover increased from 1.83 in 1980 to only 1.94 in 2006, despite the large institutional shift toward smaller-cap stocks → Inconsistent with second conjecture April 2, 2008 Spring Q-Group Conference 22 Percentage of Stocks that Traded Within the Last Five minutes of the Trade Day Grouped Deciles 1.0 1 0 0.9 0.8 0.7 0.6 05 0.5 0.4 0.3 0.2 0.1 0.0 Largest 4 Deciles Deciles 5 to 8 Decile 9 Smallest Decile 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2004 2005 2006 Sub-Groups of the Smallest Decile 1.0 Largest 8% Next 1% Next0.5% Next 0.25% Smallest 0.25% 0.8 0.6 0.4 0.2 0.0 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 → By end of 2006 there were virtually no stale prices, except for very smallest stocks April 2, 2008 Spring Q-Group Conference 23 Percentage of Stocks that Traded on Day t Grouped p Deciles 1.00 0.95 0.90 0.85 Largest 4 Deciles Deciles 5 to 8 Decile 9 Smallest Decile 0.80 0.75 0.70 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Sub-Groups Sub Groups of the Smallest Decile 1.00 0.95 0.90 0.85 Largest 8% Next 1% Next 0.5% Next 0.25% Smallest 0.25% 0 25% 0.80 0.75 0.70 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 April 2, 2008 Spring Q-Group Conference 24 Institutional Stock Ownership & Stock Return Predictability ● Research has documented short-term predictability in returns, and some attribute part of this predictability to stale prices. − Mutual fund predictability: Chalmers, Edelen, and Kadlec (2002); Goetzmann, Ivkovic, and Rouenhorst (2001); and Zitzewitz (2003) − Direct Tests of Stock Return Predictability: e.g., Lo & Mackinlay (1988,1989,1990), Kadlec & Patterson (1999) ● The recent increase in trade volume and institutional ownership – and the concurrent decrease in stale prices, particularly for smaller stocks – warrants a reexamination of the evidence on predictability. p y ● Using equal-cap decile groups, we examine daily predictability with: − Regressions using lagged S&P 500 futures price changes − Conditional returns (based on extreme prior-day S&P price changes) April 2, 2008 Spring Q-Group Conference 25 Table 5 Time Variation in Stock Return Predictability (w\ Stale Prices) Estimated slope coefficients from the following model estimated separately in each of the 14 annual periods: R i,t = a 0 + a 2 R(SP fut) t-1 + e i,t where R i,t is the return on cap-decile group i for day t and R(SP fut) t-1 is the S&P 500 futures price change over the interval from the close on day t –2 to 3:55 pm on day t -1. Values in bold indicate significance at the .05 level. year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Market Capitalization Decile Groups Largest 4 5 to 8 9 Smallest -0.054 -0.058 -0.018 0.081 -0.053 -0 056 -0.056 -0.030 0.037 0.009 -0.019 -0.186 -0.109 -0.138 -0.056 0.152 0.146 0.133 0.168 0.077 0 098 0.098 0.009 0.163 0.030 0.027 -0.090 -0.058 -0.035 0.086 0.303 0.293 0.241 0.210 0.162 0 226 0.226 0.103 0.265 0.116 0.078 0.026 0.027 0.054 0.183 0.304 0.267 0.242 0.229 0.164 0 243 0.243 0.163 0.255 0.144 0.031 0.070 -0.056 -0.004 0.162 → Time series of estimated slopes is inconsistent with patterns in stale prices April 2, 2008 Spring Q-Group Conference 26 Time Variation in Stock Return Predictability (No Stale Prices) Estimated slope coefficients from the following model estimated separately in each off the h 14 annuall periods: i d R i,t = a 0 + a 2 R(SP fut) t-1 + e i,t where R i,t is the return on cap-decile group i for day t and R(SP fut) t-1 is the S&P 500 futures price change over the interval from the close on day t –2 to 3 55 pm on dday t -1. 3:55 1 V Values l iin bold indicate i di significance i ifi at the h .05 05 level. l l year 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Market Capitalization Decile Groups Largest 4 5 to 8 9 Smallest -0.055 -0.060 -0.017 0.081 -0.063 -0.057 -0.028 0.027 0.000 -0.027 -0.188 0.188 -0.112 -0.139 -0.067 0.141 0.134 0.122 0.163 0.078 0.091 -0.005 0.162 0.029 0.026 -0.107 0.107 -0.062 -0.066 0.044 0.290 0.305 0.251 0.222 0.159 0.222 0.106 0.261 0.114 0.069 0.017 0.019 0.033 0.146 0.395 0.294 0.330 0.287 0.195 0.283 0.219 0.278 0.144 0.003 0.069 -0.081 -0.033 0.130 → Compared C d tto results lt with ith stale t l prices, i coefficients ffi i t here h are often ft larger l in i the th early l partt of sample, and little different thereafter - stale prices?? April 2, 2008 Spring Q-Group Conference 27 Daily Returns (basis points) conditional on prior-day return on S&P 500 S&P up 1% 80 60 40 20 0 -20 -40 Largest 4 Deciles Deciles 5 to 8 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Decile 9 Smallest Decile S&P down 1% (decile return mutliplied by -1) 80 60 40 20 0 -20 -40 Largest 4 Deciles Deciles 5 to 8 1990 19991 19992 19993 19994 19955 19966 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 Decile 9 April 2, 2008 Spring Q-Group Conference Smallest Decile 28 Institutional Stock Holdings & Institutional Returns ● Long academic literature linking trading to returns g speculation p and pprices, S&P listingg effect − block trading, ● Investment industry has for many years used investment flows to explain and predict returns − e.g., Biryini Associates ● Gompers G and d Metrick M t i k (2001) conclude l d th thatt institutional i tit ti l flows explain “part of the disappearance of the historical small-company stock premium.” ● Do institutional stock holdings forecast equity returns? April 2, 2008 Spring Q-Group Conference 29 Attribution Analysis y Decompose the institutional excess returns into excess returns earned from security selection (deviation from market weights on securities within market-cap deciles) and from asset allocation (deviation from g for market-capp deciles): ) market weights Security Selection Asset Allocation Lower case letters represent attributes of the aggregate of institutional portfolios and upper case letters attributes of the market portfolio. Subscripts refer to deciles of market cap. April 2, 2008 Spring Q-Group Conference 30 Total Excess Return by Quarter (basis points) 200 150 100 50 0 -50 -100 -150 150 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 April 2, 2008 Spring Q-Group Conference 31 Security and Decile Excess Returns (basis points) 200 decile security 150 100 50 0 -50 -100 -150 150 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 April 2, 2008 Spring Q-Group Conference 32 Attribution Analysis Comparison of Institutional Returns to Benchmark Returns The table reports means and t-values (in italics ) of the quarterly total excess returns (in basis points) for institutional investors in our sample, and the decomposition into security selection and asset allocation, for several time periods. Excess Returns (basis points) Security Selection Asset Allocation 1981 1998 1981-1998 10.09 10 09 2.37 68.99 4.56 -11.85 -3.30 0.33 0 33 0.12 4.07 0.45 5.14 1.95 10.41 10 41 2.29 73.06 4.71 -6.71 -1.56 1981-2006 7.50 1.64 3.07 1.26 10.57 2.17 1981-2006 (exc 1999-2000) 4.60 1.14 1.53 0.59 6.13 1.44 1999-2000 2001-2006 April 2, 2008 Total Spring Q-Group Conference 33 Attribution Analysis - Discussion ● 1999 and 2000 aside, institutional performance during 19812006 is marked by substantial variation ● Positive value added from security selection in the 1981-1998 period is offset by negative contribution from security selection in the 2001 2001-2006 2006 period. period ● Positive contribution from asset allocation in the 2001-2006 pperiod likelyy reflects the overweighting g g of smaller-capp stocks and underweighting of larger-cap stocks over this time period relative to market weights (see figure 2) → a period i d when h small-cap ll stocks t k significantly i ifi tl outperformed t f d large-cap stocks. ● Excess returns of institutions are not significant in this period April 2, 2008 Spring Q-Group Conference 34 Conclusions ● Over time, institutions & mutual funds increased their holdings of smaller stocks and decreased their holdings of larger stocks. → Now N underweight d i ht the th largest l t stocks t k andd overweight the smaller stocks relative to market weights. ● The increasing liquidity in the stock market generally, generally and in smaller cap stocks in particular, over the past two decades mirrors the increase in institutional ownership of common stocks. → Today virtually no stale prices, except for very smallest stocks ● Daily return predictability is time varying ● Any excess returns of institutional investors in the past 25 years are primarily due to security selection within market cap deciles, not to allocation across deciles (and not significant). April 2, 2008 Spring Q-Group Conference 35