AlphaLetters Paper Reviews – #6 and #7

From AlphaLetters, two reviews of quant working papers:

Article 6.
Category: Strategy, mutual fund holdings
Title: The Investment Value of Mutual Fund Portfolio Disclosure
Author: Russ Wermers, Tong Yao, and Jane Zhao
Source: University of Maryland working paper
Link: http://www.rhsmith.umd.edu/faculty/rwermers/holding_200610.pdf

Summary:
The strategy:
· long stocks that are overweighted (underweighted) by successful (unsuccessful) managers,
· short stocks that are underweighted (overweighted) by unsuccessful (successful) managers.

The risk adjusted (size, b/p, momentum) annual return is 7%+. This strategy has a low correlation with other known factors.

AlphaLetters comments:
1. Why important
This strategy is built on a convincing story that since skillful managers may continue to pick better stocks, people can benefit from their skills by studying their portfolio holdings.

The persistence of their good performance is substantiated by the findings of “Morningstar Mutual Fund Ratings Redux” (http://webpage.pace.edu/mmorey/publicationspdf/redux.pdf), where it is shown that the highly rated Morningstar mutual funds tend to outperform over the next few years.

2. Data
1980 – 2002 US equity mutual fund data (only include funds with the investment objectives of aggressive growth, growth, or growth and income) are from CDA/Spectrum and CRSP mutual fund database. Stock data are from CRSP/Compustat, and analyst earnings forecasts are from IBES.

3. Discussions
We reviewed a related paper, “Portfolio Manager Ownership and Mutual Fund Performance” (http://dc.lib.unc.edu/cgi-bin/showfile.exe?CISOROOT=/etd&CISOPTR=190), which shows that fund manager ownership can predict fund performance. This paper uses past performance as a predictor. We are curious to see whether the combination of these two can help us pick better funds and better stocks, especially when applied on more recent time period (this paper only covers till 2002).

We are also wondering whether there will be seasonality: will this strategy work better in certain quarters – say September is the fiscal year end for most mutual funds, will managers do differently in the third quarter each year? Contrasting this paper with “Trader Composition and the Cross-Section of Stock Returns” (http://ssrn.com/abstract=890656), which claims that anomalies (momentum, value, earning surprise) are stronger in the stocks with low FIT (fraction of institutional trading volume) in total volume, (since institution investors have more information), it would be interesting if we can refine FIT by applying it to trades of better funds.

————————————————————————-

Article 7.
Category: Value/Growth
Title: Do Investors Capture the Value Premium?
Author: Todd Houge, Tim Loughran
Source: SSRN working paper
Link: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=879291

Summary:
This paper juxtaposes 4 facts:
· In backtesting, value stocks significantly outperform growth stocks, as shown in various researches
· yet value mutual funds don’t outperform growth funds: from 1965-2001, value and growth funds return are 11.4% vs 11.3%(for large cap funds); 14.1% vs 14.5%(for small cap funds)
· yet value index doesn’t outperform growth index: from 1975-2002, large cap value index (S&P 500/Barra ) outperform growth index ~1%, same for all cap index (Russell 3K)
· In regression, value premium only shows up in small-cap universe, not in large cap.

Conclusion: over long horizon, value premium is hard to capture universe (though value does perform better in the 2000s so far)

AlphaLetters comments:
1. Why important
To us, these findings serve as a reminder that our backtesting results may be distorted – Past backtesting result does not guarantee (even) past realized returns. The question is: what is a robust back-testing methodology? What is a “robust” anomaly?

By providing three interesting (yet somewhat surprising) results, this paper makes us think about the nature of the value premium. It is true that in the 2000’s value performs far better, but in a longer time horizon its returns are comparable to growth for practitioners.

2. Data
1962-2001 mutual funds data are form CRSP Survivor Bias-Free U.S. Mutual Fund Database, stock data are from CRSP/Compustat.

3. Discussions
Table IV is a great example of how a lump -sum regression can be misleading. The author shows that although (in all stock universe) value factor is significant after controlling for size, this impact is gone when we limit the regression within large cap stocks. This is true for size effect as well. In our view, the key is that the stock market data we have are far from being normal/random. A robust backtesting system should go beyond and look into patterns within different segments.