A tale of two types: Generalists vs. specialists in asset management

Abstract

Management companies assign some portfolio managers to run funds within a single investment objective (i.e., specialists), and others to run funds across several investment objectives (i.e., generalists). Our results show that funds achieve higher performance when they appoint superior pickers as specialists and market timers as generalists. We argue that these decisions are the result of a better match of manager mandates with the way information is collected and processed in each investment strategy. Overall our results are consistent with decision-making in fund families that add value to their investors by aiming to optimally assign or reassign portfolio managers.

Introduction

The vast studies on mutual funds have traditionally focused on performance, along with incentives of fund managers and their alignment with the incentives of investors. Authors more recently examine the impact of management company organization on mutual funds, including investment strategies, risk-taking, and incentives of asset management companies. Papers in this literature include Khorana and Servaes (1999), Nanda et al. (2004), Gervais et al. (2005), and Gaspar et al. (2006). We consider a different aspect of mutual funds that only recently has started to attract substantial interest in the literature: the allocation of human capital within fund families.

We start with the observation that some money managers run individual funds with a well-defined, relatively narrow investment mandate (e.g., growth domestic equity) while other money managers simultaneously run several funds with very different mandates (e.g., growth-oriented and value-oriented) funds. Does it take the same investment approach to follow a narrow mandate and to manage a mix of different investment objectives? Or, are particular investment strategies better suited for one type of mandate?

We show there is indeed an optimal assignment of managers as specialists or generalists, depending on their investment strategy. We also show that many funds allocate managers accordingly, and this has a positive effect on returns.

The two different investment strategies that we consider are security picking and market timing. Many studies examine these strategies, independently or in comparison, although authors mostly try to measure performance within each investment strategy. However, in this paper, we are mainly interested in identifying whether portfolio managers engage in stock picking or market timing. With that goal, we use some of the tools suggested in that literature.

In principle, managers could try to combine strategies or switch from one to another on a regular basis. Yet, empirical evidence suggests this does not happen frequently, at least for short- to medium-term horizons. Analysts’ personal attributes might also be a significant factor in the decision between picker or timer because there is a natural fit suggested by characteristics like temperament, abilities, or education. For example, traditionally, managers with more quantitative education tended to be timers.

In a nutshell, stock picking calls for identifying what individual securities are potential outperformers within their category, and investing a meaningful share of the assets of the fund in them, thereby embracing exposure to their idiosyncratic risk. On the other hand, market timing calls for forecasting market trends and investing in any security correlated with expected positive trends, thereby seeking systematic risk exposure.

To further fix the terminology, we describe as specialists money managers who run securities within a narrow mandate, and generalists as the money managers who run securities that simultaneously meet different investment objectives. Whether a manager is a specialist or a generalist is easily observable. We also identify managers as pickers or timers. This labeling can be very challenging, depending on the information available, and we admittedly use different approaches to identify managers as either or none, but this is a noisier process. The specialist/generalist and picker/timer classifications are independent of one another. Whether a money manager is a specialist or a generalist is the result of the company assignment. The money manager can have a preference and negotiate an assignment with the fund family, but it is up to the fund family to deploy human capital. After that point, the money manager chooses the investment strategy and decides to be a picker or a timer.

Given this framework, the primary question we ask is whether there is a benefit for a specialist to follow a picking or a timing strategy, and similarly for a generalist. Naturally, a picker with outstanding skills will perform better than a picker less skilled. However, our analysis focuses on the fit between the investment strategy —picking or timing— and the mandate —specialist or generalist. We find that a specialist mandate is better for a picking investment strategy, while a generalist mandate is better for a timing investment strategy.

Our explanation of this finding is that there is an overlap or match between picker and specialist regarding the type of information required and the methodology to process it, and similarly between timer and generalist. This information channel is the reason why pickers are a better fit for specialist assignments and timers for generalist assignments.

Pickers rely on detailed information at the micro-level of the firms whose securities they are considering for investment: financial statements, management composition, organization and experience, consumer base, and brand value. Their goal is to gain exposure to a specific idiosyncratic risk expected to be rewarded with excess risk-adjusted return (i.e., alpha). Timers, on the other hand, forecast trends in markets or subsets of the market-driven by different economic or statistical variables (i.e., factors). For that purpose, timers rely on information of a macroeconomic nature, along with statistical and forecasting models to predict factor trends: GDP growth, interest rates and term structure, monetary variables, and industry share of the economy.

In other words, pickers require depth, while timers are interested in breadth. Note that not only the nature of the information they need is different, but the methodology required to process it is also different. That is, pickers depend more on financial statement analysis and valuation techniques, while timers depend more on forecasting models and statistical analysis. Thus, there are different types of human capital suitable for pickers and timers. The fixed cost associated with acquiring and processing different types of information partially explains the difficulty in switching investment strategies.

Specialists, by definition, manage securities within a specific investment objective. Therefore, their job requires analysis of relatively focused data. Their universe of securities under consideration is also well defined. Generalists, however, consider a broader range of information, and their possible universe of securities for investment is, in general, also broader.

Two considerations are worth noting. First, there is no reason to think that specialists and generalists have any different capacity to process information. Differences lie in the type of information they process. Information is more focused, more homogeneous, in the case of specialists, and broader, more heterogeneous, in the case of generalists. Second, we do not argue that specialists could not access the information generalists receive or the other way around. Given their job profile, specialists (or generalists) and their teams (financial analysts, junior portfolio managers) are better as a result of practice acquired through their work, but also probably as a result of background, at processing the type of focused information that is most useful for their job function.

We conclude that, given information availability, along with the specific methodology required to process it, a picker benefits from a specialist assignment, as this implies concentration on a particular type of securities, characterized by a single investment objective, and provides an environment more appropriate to source and process the kind of information that improves the productivity of the picker. This is because the limited number of firms, as well as their relatively homogeneous nature compared to firms across different investment styles, as is the case for generalists, permits detailed analysis and, possibly, spillover across companies.

Timers, on the other hand, are better suited to generalist assignments that, in general, involve a larger number of companies with a broader range of characteristics. They benefit from exposure to multiple investment objectives because they receive insights about a broad range of macroeconomic variables and their interactions (i.e., wide information span), plus they have access to more diverse factors to work with. This generates positive informational externalities for their type of ability.

The natural next question is whether one can observe this type of allocation among fund families. We find that, consistent with an optimal deployment of human resources, timer-specialists are more likely to be redeployed in generalist mandates, and picker-generalists are more likely to be redeployed in specialist mandates. When this happens, overall manager performance improves, consistent with the notion that this is an optimal allocation of human capital. At the same time, we do not observe a similar likelihood of reassignments of timer-generalists or picker-specialists.

In the process, we explore why sometimes mutual funds misallocate timers and pickers, and we find that there are factors that can, at least partially, explain it. For example, fund families that are expanding their offerings might charge a picker with running several funds with different investment objectives, which might make sense in terms of cost-saving, at least temporarily. Overall we find evidence of optimality in the way fund families deploy their talent.

One of the most controversial subjects in financial economics is whether asset managers can achieve returns higher than the market on a risk-adjusted basis. Berk and Green (2004) argue that there is such a thing as portfolio management ability, which helps explain mutual fund flows. However, we do not attempt to contribute directly to the performance evaluation literature, as our principal interest is in the organizational structure of mutual funds.¹ Our primary finding is that there is an optimal assignment of managers as specialists or generalists, depending on their investment strategy, whether stock picking or market timing.²

We also show that many funds allocate managers accordingly, which has a positive effect on returns. We complement the work of Fang et al. (2014), who show that fund families allocate their skilled managers to funds targeting inefficient markets; and Berk et al. (2017), who find that funds reallocate resources to increase added value. Our results are also related to work by Ibert et al. (2018), who show that fund managers’ compensation depends on the overall performance of the family. This result is consistent with optimization across the whole range of funds and in line with our findings on the allocation of human capital.

In this paper, we study the allocation of human capital within a mutual fund family or, equivalently, the distribution of the total amount of assets managed by a fund family across the talent within the family. A similar problem is the delegation of investment management by very large investors, such as pension funds and sovereign wealth funds, to different external asset managers. This practice, referred to by Sharpe (1981) as decentralized investment management, is part of a literature that has recently grown with papers such as Binsbergen van et al. (2008) and Blake et al. (2013). While this external allocation, as opposed to the internal allocation we study in this paper, has specific characteristics, some of them the subject of the studies just mentioned, it also shares some elements with the problem of mutual funds families, such as the allocation of asset managers among different mandates. In particular, the choice between a fund manager with a “generalist mandate” across multiple asset classes, or multiple managers, each with a “specialist mandate” within each asset class.

There is a large body of literature on the importance of the right deployment of personnel and optimal compensation; yet, there is little evidence on whether firms add value when they allocate employees to positions. The mutual fund industry offers several advantages over most of the organizational and labor economics literature to study human capital allocation. First, there is a direct and immediate link between the manager’s actions (investment decisions) and their outcome (return). Second, while it is often challenging to measure employees’ output, mutual fund performance is public information. Lastly, we also observe the inputs, that is, type of skill and assignment of mutual fund managers.³

In the first stage of our analysis, we establish whether managers follow a stock picking or a market timing strategy. For that purpose, we focus on domestic equity, which is the largest subset within the universe of funds in the US. This makes our analysis more directly comparable to previous studies, which focus predominantly on US domestic equity mutual funds. We use several approaches. The first is the return-based model proposed by Treynor and Mazuy (1966), augmented with multi-risk factors. We also use an alternative approach suggested by Henriksson and Merton (1981). We estimate the manager’s investment strategy following the conditional method of Ferson and Schadt (1996) to control for some of the predictability in the data. Finally, we use daily returns to account for intra-month information, as in Bollen and Busse (2001).

These tests were originally designed to identify managers with outstanding ability, whether as stock pickers or market timers. As we are interested in identifying managers’ investment strategies rather than their performance, these are just some of the criteria we use. Our focus on domestic equity funds allows us to take advantage of the fact that, for this category, there are data available on portfolio holdings. We thus can use other methodologies to sort out managers’ investment strategies.

In one of the approaches based on equity holdings, Daniel et al. (1997) split the return of a fund between a “characteristic selectivity” component (CS), and a “characteristic timing” part (CT). CS is associated with stock selection while CT corresponds to market timing, which arises when fund managers try to exploit time-varying expected returns of size, book-to-market, or momentum benchmark portfolios. Also for domestic equity funds, we estimate the “active share” and “tracking error” of Cremers and Petajisto (2009), which measure ex ante the manager’s intention to engage in stock selection or factor timing.⁴

The paper is organized as follows. First, we describe the data and define our main variables in Section 2 We present our main results in Section 3 We show that pickers display better performance as specialists and timers as generalists, and examine how these manager assignments affect portfolio investment decisions. We study the allocation of portfolio managers among different mandates in Section 4 We undertake several robustness tests addressing possible endogeneity problems in Section 5 We conclude in Section 6