Information Content when Mutual Funds Deviate from Benchmarks Hao Jiang, Marno Verbeek and Yu Wang Rotterdam School of Management Erasmus University www.rsm.nl/mverbeek [email protected]Inquire Europe-UK, 26 March 2012 Marno Verbeek, Rotterdam School of Management 1/33
33
Embed
Information Content when Mutual Funds Deviate from Benchmarks Hao Jiang, Marno Verbeek and Yu Wang Rotterdam School of Management Erasmus University .
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Marno Verbeek, Rotterdam School of Management 1/33
Information Content when Mutual Funds Deviate from
Marno Verbeek, Rotterdam School of Management 2/33
Are mutual funds informed investors?Do they help impounding information into prices?
We evaluate the role that active managed (equity) mutual funds play in determining security prices.
Do deviations from benchmarks by active mutual funds contain information about future stock returns?
If so, what types of funds and stocks are most likely to exhibit such informed mutual fund portfolio tilting?
What are the implications for mutual fund performance and asset pricing?
• 3
Measure of Deviations from Benchmarks
: the portfolio weight of stock i in fund j during quarter t;
: the portfolio weight of stock i in fund j’s benchmark index
in quarter t;
: the number of funds whose investment universe includes stock i. (The stock is either held by the fund or in the benchmark index of the fund).
Thus, DFB is a stock-level measure that reflects the average amount of over/underweighting of the stock relative to the benchmark(s).
1
(1)
, , , ,iN
j bi t i t i t i
j
DFB w w N
,j
i tw
,bi tw
iN
• 4
Selecting Mutual Funds’ Performance Benchmarks
For each fund, select one index from the following 19 indexes:
S&P 500, S&P 400, S&P 600, S&P 500/Barra Value and Growth, Russell 1000, Russell 2000, Russell 3000, Russell Midcap, Value & Growth Wilshire 5000, and Wilshire 4500.
(Cremers and Petajisto, 2009)
We also experiment with alternative ways of constructing benchmarks, with very similar results.
Marno Verbeek, Rotterdam School of Management 5/33
Matching funds and benchmarks
• We start from a universe of 19 benchmarks.
• For each fund in each quarter, we select the benchmark that minimizes the distance in portfolio weights between the fund and the index.
• Distance is measured by “Active Share” (Cremers and Petajisto, RFS, 2009):
• Thus: benchmarks are not self-reported and a fund can have different benchmarks over time.
Marno Verbeek, Rotterdam School of Management 6/33
Data and Sample Selection
• Merging five major sources of financial data:– CRSP Survivor-Bias-Free US MFDB (fund returns, fees, …)
• Final sample contains 2,750 US domestic equity mutual funds spanning 1980-2008.
MFLINKS NCUSIPHoldingsMFDB Stocks
Analyst forecasts
Accounting info
Marno Verbeek, Rotterdam School of Management 7/33
Data and Sample Selection• Using the investment objective codes provided by
CRSP and TFN, we eliminate bond funds, balanced funds, asset allocation funds, international funds, precious metal funds, and sector funds.
• We also require that funds hold on average 80%-105% in common equity.
• We (manually) exclude index funds by dropping fund names with “INDEX”, “IND”, etc.
• Multiple share classes in CRSP MFDB confer ownership in the same underlying pool of assets. We select the share class with the longest history of data.
Marno Verbeek, Rotterdam School of Management 8/33
Does DFB predict returns?I: Equally-weighted and value-weighted portfolio sorts.
Portfolio performance measured by:
• Raw returns
• CAPM alphas
• 3, 4 and 5-factor alphas, controlling for size, value, momentum and the Pastor and Stambaugh (2003) liquidity factor.
• DGTW-adjusted: performance relative to a characteristics-based benchmark (consisting of stocks matched on size, book-to-market and momentum). [Daniel et al., 1993]
Marno Verbeek, Rotterdam School of Management 9/33
Return forecasting power of DFB: Decile Portfolios
1 2 3 4 5 6 7 8 9 10
-0.40
-0.20
0.00
0.20
0.40
0.60
0.80
1.00
1.20
1.40
Average Return Carhart Alpha DGTW-Adj Return
Return forecasting power of DFB: EW Decile Portfolios
Marno Verbeek, Rotterdam School of Management 11/33
What are the characteristics of stocks with extreme DFB? • What are the stocks with heavy mutual fund bets? • At end of each quarter, we sort stocks by their DFB,
calculate the cross-sectional averages of the characteristics, and then average over time.
Per DFB decile we report:• Average portfolio weights in benchmark• % stocks outside the benchmarks• Scores (from 1-10) on size, book-to-market, previous
Marno Verbeek, Rotterdam School of Management 16/33
So…• DFB strongly predict returns
• A portfolio that buys stocks in Decile 10 (high DFB) and sells stocks in Decile 1 generates an average return of 0.74% (EW) or 0.56% (VW) per month.
• These differences are statistically significant.
• Risk adjustment has only limited impact. All five versions of alpha are highly significant.
• Cross-sectional regressions confirm this. Controlling for other predictive factors has little impact.
We argue that this is (partly) an information-based story.
Marno Verbeek, Rotterdam School of Management 17/33
Some evidence that is consistent with our information-based story.
Stock characteristics
• DFB is more profitable for midcap stocks, stocks with high idiosyncratic volatility, and for stocks that have relatively low numbers of mutual funds investing in it.
Fund characteristics
• DFB is more profitable for funds with high past performing funds (skilled funds), growth funds
Marno Verbeek, Rotterdam School of Management 18/33
Further evidence supporting the information story(1) Stock characteristics
DFB strategy per size-quartile
Small Q2 Q3 Large0.00
0.10
0.20
0.30
0.40
0.50
0.60
Average 4-factor % alpha on DFB strategy
Small Q2 Q3 Large0.00
0.50
1.00
1.50
2.00
2.50
3.00
3.50
T-statistics
Marno Verbeek, Rotterdam School of Management 19/33
Further evidence supporting the information story(1) Stock characteristics
DFB strategy per volatility-quartile
Idiosyncratic volatility: firm-specific information
Low Q2 Q3 High High-Low0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
Average 4-factor % alpha on DFB strategy
Equal-Weight Value-Weight
Marno Verbeek, Rotterdam School of Management 20/33
Further evidence supporting the information story(1) Stock characteristics
Number of funds: competition for private information
Low Q2 Q3 High High-Low
-0.80
-0.60
-0.40
-0.20
0.00
0.20
0.40
0.60
0.80
Average 4-factor % alpha on DFB strategy
Equal-Weight Value-Weight
Marno Verbeek, Rotterdam School of Management 21/33
(2) Fund characteristics
Past two-year alphas: high vs low
Low 2 High High_low0.00
0.10
0.20
0.30
0.40
0.50
0.60
Average 4-factor % alpha on DFB strategy
Equal-Weight Value-Weight
Marno Verbeek, Rotterdam School of Management 22/33
(2) Fund characteristics
Investment objectives: Growth vs Income
Aggressive Growth Growth Income
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
Average 4-factor % alpha on DFB strategy
Equal-Weight Value-Weight
Marno Verbeek, Rotterdam School of Management 23/33
Information
• If mutual funds have superior information, what kind of information do we expect them to have?
• In stock markets, one of the most important pieces of information is about corporate earnings.
• Can DFB predict earnings surprises?
(3) DFB and earnings news
Quarters t+1 t+2 t+3 t+4
Earnings Surprise Scaled by Actual Earnings (%)
Q1 0.159 0.393 0.462 0.453
(0.32) (1.05) (1.26) (1.19)
Q5 2.47 1.84 1.262 0.858
(5.62) (4.35) (2.81) (1.85)
Q5-Q1 2.353*** 1.447*** 0.800*** 0.405**
(6.03) (9.05) (5.69) (2.22)
Q5-Q1 (Momentum-Adj) 1.384*** 0.474 0.468 0.467
(5.31) (0.93) (1.38) (1.12)
Earnings Surprise Scaled by Price (%)
Q1 -0.004 0.002 0.003 0.003
(-0.39) (0.30) (0.44) (0.42)
Q5 0.033 0.025 0.015 0.01
(5.60) (4.27) (2.44) (1.46)
Q5-Q1 0.038*** 0.023*** 0.012*** 0.007**
(4.06) (6.93) (6.36) (2.10)
Q5-Q1 (Momentum-Adj) 0.024*** -0.01 0.004* 0.062
(4.23) (-0.74) (1.70) (1.02)Marno Verbeek, Rotterdam School of Management
24/33
Marno Verbeek, Rotterdam School of Management 25/33
Marno Verbeek, Rotterdam School of Management 32/33
Relation to previous literature
Micro-foundations of mutual funds’ information advantages
– Coval and Moskowitz (2001): Geography and mutual funds
– Cohen, Frazzini, and Malloy (2008): Shared educational background
– It will be interesting to explore the link!
Best ideas of fund managers: Cohen, Polk, and Silli (2010)
Extracting beliefs from fund holdings: Shumway, Szefler, and Yuan (2009)
Marno Verbeek, Rotterdam School of Management 33/33
Conclusions
A stock-level measure that seeks to aggregate various pieces of information scattered among fund managers, as revealed through their over- and under-weighting decisions, strongly forecasts cross-sectional variation in future returns.
Strong evidence in favor of an information-based interpretation of this return forecasting power.
Mutual funds tend to be “informed investors” whose investment activities help information transfer into security prices.