Finding Fortune: How Do Institutional Investors Pick Asset Managers?
Gregory W. Brown
University of North Carolina (UNC) at Chapel Hill – Finance Area
Oleg Gredil
Tulane University — Finance Area
Preetesh Kantak
University of North Carolina (UNC) at Chapel Hill – Finance Area
March 15, 2016
Abstract:
This paper studies how professional asset allocators such as endowments, fund-of-funds, or pension funds select fund managers for investments. We develop a simple model of their due-diligence process to motivate predictions about the timing of investment decisions. We then test these predictions using a unique dataset with detailed information on the interactions between a large institutional investor and 1,093 hedge funds over the course of 8 years. Soft information conveyed during the meetings with fund managers strongly influences the decisions. A one standard deviation increase in our proxy for positive soft information doubles the probability of fund selection and reduces the due-diligence time by 20%. Contrary to prior research, we find no evidence that relying on these subjective judgements is wasteful. Instead, in a matched sample, conditioned on the fund characteristics and past performance, the 12-month average peer-adjusted returns are 1.5% higher for the selected funds.
Finding Fortune: How Do Institutional Investors Pick Asset Managers? – Introduction
In the U.S. alone, over $15 trillion in assets are allocated by institutional asset managers such as pensions funds, insurance companies, endowments, funds-of-funds, and foundations. Investment of their assets is delegated to managers of traditional long-only investment strategies in equity and fixed income markets. But, increasingly allocations include ‘alternatives’ such as hedge funds and private equity funds which are allowed substantial discretion in determining specific investments. Overall, the vast majority of delegated asset management involves active investment strategies.1 The past 20 years has seen substantial growth in all types of delegated asset management (including traditional “40-Act” mutual funds in the U.S.), but the growth in delegated asset management by institutional investors has been more than twice as large.2 Despite the enormous size of the market for delegated asset management among institutional investors, often referred to as “allocators”, relatively little is known about the process by which these institutions make decisions. This paper seeks to provide a framework for better understanding the allocator’s problem and investigates the specific process for choosing funds of a long-short equity portfolio for a large institutional investor.
While there is a mature literature examining the optimal contract between investment principal and asset manager, few papers examine how the principals allocate capital among investment managers (henceforth “fund managers“) in less abstract settings. Institutional allocators appear to hire and fire managers based on past excess returns (Goyal andWahal, 2008) and also appear to monitor managers better than retail investors (Evans and Fahlenbrach, 2012). However, little is documented about the due diligence process these asset allocators undertake and how additional information, acquired through substantial expenditures on in-house research and consultants, guides their choices. The investors problem can be framed as a tension between acting quickly using widely available “hard information” that is fairly cheap to obtain but of relatively low quality versus expending resources (time, labor, and fees) to obtain additional “soft information” that will better allow the allocator to identify the quality of the fund manager. That is, the allocator may earn greater excess return (i.e., alpha) by informing their choice with soft information, but the costs of collection will also lower net returns.
The allocator’s problem stems from the same type of asymmetric information condition present in many principal-agent relationships. With the limited information available to them, allocators are unable to determine “good” from “bad” fund managers, which all have an incentive to represent themselves as good managers to attract investments in their funds. For example, Korteweg and Sorensen (2014) show that limited partners in private equity firms need a sequence of about 25 funds to identify top quartile firms (only 3% of managers have such extensive track records) while Pastor et al. (2015) find that performance deteriorates over a typical mutual fund’s lifetime.3 Another consideration for allocators is that waiting to learn a fund’s type through collection of only hard information increases the chance that other asset allocators may identify the skilled managers first and the opportunity to invest will no longer be available (e.g., the hedge fund closes to new investors after an influx of capital or a successful venture capital firm only provides allocations to its previous fund investors). Berk and Green (2004) suggests that this is the equilibrium when there exists a declining return to scale to a fund’s investment strategy.
We develop a model for their investment decision; our objective is to model the allocator’s “hazard” functions, i.e., P(investment j non-investment in previous periods). The model is similar to ones used to characterize the industry diffusion of new technologies, e.g., characterization of adoption curves or optimal contracts to incentivize technology development.5. It can also be viewed as an infinite-period extension of Hermalin et al.’s (1998) work on governance and CEO monitoring. The intuition developed through the model results in four testable predictions. First, better “hard” (or auditable) information leads to faster investment decisions. Second, more precise auditable information also leads to faster investment decisions. Third, better “soft” (or private) information leads to faster investment decisions. And fourth, more precise private information leads to faster decisions.
Our empirical tests involve a proprietary dataset obtained from a large institutional investor that allocates to a portfolio of long-short equity hedge funds. We document in detail the process by which the allocator identifies prospective mangers, interviews representatives of the funds (typically multiple times), and analyze the resulting information. We investigate how the hard and soft information produced by the allocator on 1,093 funds impacts manager selection decisions over a 7-year period from 2005-2012. Our unique dataset allows us to observe how the research staff of the allocator conducts internal analysis of hedge fund managers’ pitchbooks and manager meetings via notes made in an internal research system. We use these notes to construct proxies for soft information by conducting textual analysis of all the entries related to each fund in our sample. Additionally, we test whether the use of soft information has a positive effect on realized excess returns.
Our empirical results suggest that both the level and uncertainty of soft information are strong predictors of the probability of selecting a fund (also called “admitting” or “accepting” a fund, a concept we formalize later). A one standard deviation higher level of our soft information proxy leads to almost a doubling of the chance of admitting a fund at any point in time. Given our empirical specification, this leads directly to a shorter time for a recommendation (i.e., a lower due-diligence time). A one standard deviation decrease in our proxy for soft information uncertainty leads to an almost 20% (8 month) drop in due-diligence time. As predicted, the uncertainty around hard information is a significant predictor of the time to admission as well. A one standard deviation increase in hard information uncertainty leads to roughly an 11 month increase in the due-diligence period. Interestingly, the level of hard information, while statistically significant, is not as strong a predictor of fund selection as soft information and realized return uncertainty. We also find evidence that these recommendations add pecuniary value to the allocator. In the 12 months following admission, gross alpha is approximately 1.5% higher for admitted than non-admitted funds. Interestingly, in keeping with Berk and Green (2004), this alpha decays to an insignificant level in the second year after admission.
See full PDF below.
