Information Sharing within the Networks of Delegated Portfolio Managers: Evidence from Plan Sponsors and Their Subadvisers

ABSTRACT

The authors study information sharing among delegated portfolio managers through networks connected by investment mandates between plan sponsors and their subadvisers. Specifically, they identify similarity in returns, holdings, and trading between mutual funds operated by subadvisers, and test whether such similarity is stronger when two funds share a mandate network. The authors find evidence consistent with information sharing among these delegated portfolio managers. A mutual fund on average shares more similar returns, holdings, and trading with funds in subadvisory mandate networks than with funds outside the networks. Preliminary evidence suggests that information about both general investment styles and individual firms is transferred within mandate networks.

KEYWORDS:

• Investment mandate
• Network
• Information sharing
• Correlated trading
• Return co-movement

Acknowledgments

The authors are grateful for very helpful comments from participants at the 2013 FMA Asian Conference, 2013 FMA International Meeting, and brown-bag seminar at UNSW. They also thank David McLellan for providing the access to and helpful information about the iiSEARCHES database.

Notes

1. Once a plan sponsor has decided the amount and asset class of an investment mandate, it puts out an RFP and the search for investment companies begins. The RFP usually introduces the plan sponsor and the specific mandate, and also specifies the services to be provided by the hired investment company and the minimum qualifications of the hired investment company.

2. For example, Chicago Teachers' Pension Fund requested their hired investment companies “to report to the Board monthly, in writing on the composition and relative performance of the investments in the designated portfolio; the economic and investment outlook for the near and long term; significant changes in the portfolio during the month; and the reasons for any significant differences between the performance of the portfolio and the appropriate market indices or other performance benchmarks established by CTPF and the investment manager.” A similar case is presented in the RFPs by Teachers' Retirement System of Louisiana. Please refer to Part 3 of the RFP by Chicago Teachers' Pension Fund and Appendix B (I) of the RFP by Teachers' Retirement System of Louisiana. The two RFPs could be found via the following links: https://www.callan.com/about/rfp/ctpf/em_market/Emerging%20Markets%20CTPF%20rfp%20draft%203-13-12.pdf and http://trsl.org/uploads/File/SFPs/SFP_Mid%20Cap%20Growth%202013.pdf.

3. For example, CalPERS requested their hired investment companies to “provide advice on market conditions, including positive and/or negative trends and various security-related issues.” New Mexico State Investment Council Specified in their RFP that their hired investment companies should “advise the SIC and appropriate staff on equity-related issues” and “advise the SIC when specific segments of the equity markets are particularly attractive and be willing to manage an opportunistic portfolio in those segments for the SIC as part of this RFP.” Please refer to section III of the RFP by CalPERS and section V of the RFP by New Mexico State Investment Council. The two RFPs could be found via the following links: http://www.calpers.ca.gov/eip-docs/business/opportunities/2005-3865/rfp-2005-3865.pdf andhttp://www.sic.state.nm.us/PDF%20files/20111107%20RFP%2012-0020%20Large-Cap%20Domestic%20Equity.pdf.

4. For example, Firefighters' Retirement System of Louisiana specified in their RFP that the hired investment company should “provide on-going education to trustees and staff if requested.” University of Kentucky endowment also requests that “the successful Contractor(s) will also be required to provide educational and ongoing advisory services to Investment Committee members and investment staff.” Please refer to section B of the RFP by Firefighters' Retirement System of Louisiana and section 7.1 of the RFP by University of Kentucky endowment. The first RFP can be found via the following link: http://www.lafirefightersret.com/pdf/RFPRiskParity080613.docx. The second RFP is not available online anymore. The document is available on request.

5. Most plan sponsors, such as public and corporate pension plans, endowments, and foundations, normally do not manage mutual funds. In our final sample, there are four plan sponsors that manage equity mutual funds. Thus, we focus only on mutual funds managed by the hired investment companies for our analyses and only use plan sponsor information to identify mandate networks.

6. See “Investment outsourcing nears $1 trillion, P&I survey finds”, Pensions & Investments, 5 July 2013. The article can be accessed at http://www.pionline.com/article/20130705/REG/130709944/investment-outsourcing-nears-1-trillion-pi-survey-finds.

7. For a sample of 56 public defined-benefit plans covered by both our mandate database and the annual performance data made available by the Center for Retirement Research at Boston College, we find that higher commonality in asset holdings among investment companies hired by a plan sponsor is associated with increased volatility in the plan's investment returns. However, the reliability of the finding and the causality inference is an issue due to the very small sample size and weak power of the test.

8. The amount of outsourced assets grew secularly over years. According to the two surveys conducted by Pensions & Investments in 2011 and 2013, the sector grew by 59% over the two-year period, from $586 billion to $955 billion. We find a similar trend of increasing commonality in mutual funds' returns over years – the average R² of a time-series regression of fund returns on Fama-French three-factor (market excess return, small minus big size return, and high minus low book-to-market return) and Carhart's momentum factor increased from about 0.83 in 1995 to more than 0.95 in 2010 for all actively managed U.S. equity funds in our mutual fund sample.

9. The correlated trading and its effect on return comovement attract attentions from both academics and practitioners. For example, in a recent Financial Analyst Journal article, Greenwood and Sosner [2007] examined the excessive correlated trading and return comovement around the redefinition of Nikkei 255 index. It is also speculated that the significant losses of some high-profile and highly successful quantitative long/short equity hedge funds during the week of August 6, 2007, were caused by similarity in portfolio holdings of the sector and the liquidation series triggered by the fire-sale of only one or few funds (Khandani and Lo[(2011]).

10. The mandate records fall into four categories: potential, new, completed, and discontinued. A potential mandate is an expression of intent that a plan sponsor might award a mandate in the future. A new mandate is created when there is an outstanding RFP. A completed mandate is a confirmed hiring decision of specific investment companies. The hiring decision can take one of the following three forms: (a) a new hiring, in which a new company is hired without replacing any existing company; (b) a rehiring, in which a new asset allocation is awarded to an existing company and no existing company is replaced; or (c) a replacement, in which a newly hired company or another current company replaces an existing company and takes over its asset allocation. A discontinued mandate can also involve one of the following three situations. First, the selection process for a potential new company is completed but no mandate is awarded. Second, an existing investment company is fired but no new company is hired. Third, part of the asset allocated to an existing company is withdrawn but no new company is hired to take over the withdrawn asset allocation.

11. It is possible that multiple investment managers are hired under one investment mandate, and thus the total number of hiring decisions is larger than that of investment mandates. Some plan sponsors have missing values for assets under management. Similarly, some investment mandates have missing values for the size of mandated assets. As such, the number of plan sponsors and investment mandates reported here differs from that in Table 1, which counts only the plan sponsors and investment mandates with nonmissing assets. However, we do not exclude those plan sponsors and investment mandates with missing assets from our sample because they are used only for the identification of mandate networks.

12. Our assumption of a three-year effective period for an investment mandate may sound arbitrary. However, based on the 176 hiring decisions for which we could obtain terminating dates, the mean (median) effective period is 4.7 (4.5) years. Thus, this assumption seems to be conservative.

13. We match the investment companies in the panel data with the fund management companies in the mutual fund data by company names. For each investment company, we first apply an algorithm to find a small list of fund management companies with close names, and then manually check each matched pair to pick up correct matches.

14. Again, the total numbers of plan sponsors and investment mandates included in the final sample are different from the numbers presented in panel B of Table 1 because of missing values for plan sponsor size and mandate size.

15. We conduct our tests about return correlations using net fund returns. Theoretically, gross fund returns should be the performance variable to examine. However, we still decide to report the results based on net fund returns due to two considerations: first, occasionally expense ratios are missing in the data, which causes unnecessary loss of observations; second, the results are similar whether using gross or net fund returns. The results using gross fund returns are available on request.

16. We require at least 12 monthly observations to calculate pairwise correlations. To mitigate the impact of outliers, we winsorize both the pairwise correlations and the fund-level average correlations at the 5th and the 95th percentile values. The winsorization of variables only improves the statistical significance of estimated coefficients, but does not change the sign or the magnitude of the coefficients qualitatively. Results without winsorization are available upon request.

17. We also try an alternative measure of total turnover—the sum of net turnovers on style portfolios—which makes the common trading measures larger (due to smaller total turnover measures) and more importantly, results in more discernible differences in common trading measures between inside-network fund pairs and outside-network fund pairs.

18. Similarly, to mitigate the impact of outliers, we winsorize both the quarterly commonality measures and the fund-level average commonalities at the 5th and the 95th percentile values. Again, the winsorization of variables only improves the statistical significance of estimated coefficients, but does not change the sign or the magnitude of the coefficients qualitatively. Results without winsorization are available upon request.