Content uploaded by Edward Bishop Smith
Author content
All content in this area was uploaded by Edward Bishop Smith on May 19, 2015
Content may be subject to copyright.
Electronic copy available at: http://ssrn.com/abstract=1646331
1
Identities as Lenses: How Organizational Identity
Affects Audiences' Evaluation of Organizational Performance
(forthcoming, Administrative Science Quarterly)
Edward Bishop Smith
University of Michigan
Electronic copy available at: http://ssrn.com/abstract=1646331
2
This study calls into question the completeness of the argument that economic actors who fail to
conform to certain identity-based logics—such as the categorical structure of markets—garner
less attention and perform poorly, beginning with the observation that some nonconforming
actors seem to elicit considerable attention and thrive. By reconceptualizing organizational
identity as not just a signal of organizational legitimacy but also a lens used by evaluating
audiences to make sense of emerging information, I explore the micro, decision-making
foundations on which both conformist and nonconformist organizations may come to be favored.
Analyzing the association between organizational conformity and return on investment and
capital flows in the global hedge fund industry, 1994–2008, I find that investors allocate capital
more readily to nonconforming hedge funds following periods of short-term positive
performance. Contrary to prediction, nonconforming funds are also less severely penalized for
recent poor performance. Both "amplification" and "buffering" effects persist for funds with
nonconformist identities despite steady-state normative pressure toward conformity. I explore the
asymmetry of this outcome, and what it means for theories related to organizational identity and
legitimacy, in the discussion section.
Electronic copy available at: http://ssrn.com/abstract=1646331
3
The economic cost of organizational nonconformity is apparent from the disregard,
misunderstanding, and devaluation of those market participants that fail to align with prevailing
market logics (Meyer and Rowan, 1977; Hannan and Carroll, 1992; Zuckerman, 1999). One
logic that has been the focus of much recent empirical work centers on the notion of
organizational identity (Albert and Whetten, 1985; Hsu and Hannan 2005). Organizations that do
not present clear identities to evaluating audiences face the possibility of being miscategorized,
misunderstood, and, ultimately, ignored. The empirical association between identity and
audiences' evaluations has gained much attention because it underlies one of the primary insights
of organizational and economic sociology: organizational isomorphism in markets is, in part, the
result of the normative and cognitive constraints generated and applied by market audiences
about what constitutes an acceptable or legitimate organizational identity (DiMaggio and Powell,
1983; Thomas, Walker, and Zelditch, 1986).
Nevertheless, atypical organizations not only persist in many markets, they sometimes elicit
significant attention and thrive (Chen and MacMillan, 1992; Miller and Chen, 1996). The
emergence of microbreweries as a counter to large-scale commercial brewing (Carroll and
Swaminathan, 2000) and the divergence from tradition of French nouvelle cuisine (Rao, Monin,
and Durand, 2003) offer just two prominent examples of organizational nonconformity from the
recent literature. Examples from industry include the historical case of Apple Computer and its
revolutionary graphical user interface and, more recently, the unquestionably atypical shoe
company, Vibram, whose "fivefingers" shoes look more like gloves than traditional shoes. Such
instances of organizational atypicality present a problem for sociological theories of the
marketplace that existing research has only partially solved. Existing attempts to accommodate
4
the occurrence of organizational nonconformity look to variations in the level of constraint posed
by audiences' expectations and market, or categorical boundaries (e.g., Pontikes, 2008; Kovacs
and Hannan, 2010; Negro, Hannan, and Rao, 2010). For instance, Ruef and Patterson (2009)
argued that membership in multiple market categories—a variant of organizational
nonconformity—and the resulting ambiguity in one’s identity, is less damaging for organizations
when the categories themselves are emerging or in flux. Lounsbury and Rao (2004) established
that tolerance for heterogeneity among new market entrants in the mutual fund industry was
afforded only after market categories began to break down. Explanations like these, however, are
confined to either the most nascent organizational environments or to environments that are
already changing. In more stable settings in which "all audiences hold the same expectations,"
theory predicts that "violations of standards are met with particularly sharp devaluations" (Hsu
and Hannan, 2005: 476).
And yet even stable markets are populated by organizations that vary in the extent to which they
might be considered conformist or nonconformist. Apple is still regarded as something of an
atypical computer company, at the very least when compared to the likes of the Wintel standard.
Glove-like shoes do not appear to be turning the shoe industry upside down—traditional shoes
are still the norm—and yet Vibram does not seem to be experiencing the sort of devaluation that
existing theories would predict. To understand the success of nonconformity, then, we must
consider more than just how audiences use identity to sort organizations, deeming some
legitimate and others illegitimate. The longevity of some nonconformists may result from
differences in the way audiences evaluate emerging information when it is associated with
atypical organizations.
5
According to the traditional, sociological view, organizational identity is primarily a filtering,
screening, or sorting device. Once "sorted," conformist organizations are rewarded and
nonconformists ones are ignored. But identity may also be conceptualized as a lens through
which various kinds of information pass and take on meaning. Just as two lenses that vary in
shape can receive identical beams of light and yet refract that light in markedly dissimilar ways,
equivalent information may be differentially interpreted and reacted to when applied to
organizations with dissimilar identities. This observation—that the identity of something affects
the way new information about that something is interpreted—is not new to this paper. Homans
(1961) long ago observed that the reward an individual receives for performing a given role is a
function not only of the quality of the role performance, but also of the performer’s social
identity. Since then, variations on the theme can be readily found in disciplines as varied as
sociology (Burt, 1998; Zuckerman, 2004), management and organizational theory (e.g., Dutton
and Dukerich, 1991; Ashforth and Humphrey, 1997), behavioral economics (e.g., Kahneman,
Knetsch, and Thaler, 1990), and finance (e.g., Chevalier and Ellison, 1997; Berk and Green,
2004).
Given its ubiquity, then, it is surprising that this basic insight has not been more fully
incorporated into research linking organizational and categorical identity with audiences'
evaluations. There seems to be two reasons for this omission. First, existing research focuses
primarily on identifying a main effect of organizational identity. The result of this work is a
strong and unified statement that economic actors failing to conform to some kind of prevailing
order face certain penalties (e.g., Chen and Hambrick, 1995; Zuckerman, 1999). Second,
researchers in this area typically study organizational identity in implied states of equilibrium.
6
This is not to say that organizational identities are fixed in time (cf. Elsbach and Kramer, 1996;
Corley and Gioia, 2004). Rather, the relationship between organizational identity and audiences'
evaluations is typically analyzed in such a way that the regular, informational feedback (e.g.,
performance) of functional markets is ignored. As a result, whereas prior studies have focused
largely on the effects of organizational identity at the mean of other information-related
covariates—that is, holding them constant—viewing identity as a lens leads to a treatment of
identity that is about the active sensemaking and interpretation of such information.
In undertaking such a treatment of identity here, my aims are twofold. I use a comparative static
analysis to explore the relationship between organizational identity and performance and
audience evaluation and, in doing so, the micro foundation of the proliferation of organizational
nonconformity. In this respect, Zuckerman’s (1999) study of securities analysts offers a useful
base on which to build. He showed that organizations operating in multiple market segments
were more difficult to interpret, garnered less attention from the right kinds of analysts, and were
devalued in turn. His results are indicative of an “imperative” to present a clear, unambiguous
categorical identity. Failure to do so results in illegitimacy and makes it difficult to gather the
resources necessary for survival (see also Thomas, Walker, and Zelditch, 1986; Stryker, 1994).
In what is perhaps a relatively overlooked section of the same paper, however, Zuckerman
(1999: 1402-1403) also conjectured that “while conforming to audience expectations is generally
wise, the greatest returns likely flow to those who innovate by creating new categories and
corresponding interfaces.”
7
Conditions may even exist under which the gains to organizational nonconformity will outstrip
those of conformity (cf. Phillips and Zuckerman, 2001). This statement may seem obvious to
scholars in strategic management for whom organizational nonconformity presents far less of a
problem (e.g., Porter, 1985; Lieberman and Montgomery, 1988), but my aim is different. Rather
than demonstrate the competitive advantage of differentiation, I explore possible behavioral
mechanisms by which otherwise punishable nonconformity may be tolerated and even rewarded
by organizational evaluators. To identify and explore such conditions, this paper makes use of
an unusual analogy—the similarity of the movement of fishing boats to financial markets—and
suggests ways that organizational identity may moderate the relationship between organizational
performance and audience evaluation. I develop hypotheses about the amplification effect of
organizational nonconformity on the already positive (or negative) relationship between positive
(or negative) organizational performance and audiences' evaluation. I test the hypotheses in a
study of the global hedge fund industry in the period of 1994-2008.
REWARDS TO PERFORMANCE FOR CONFORMISTS AND NONCONFORMISTS
In the early days of sonar technology, finding fish in the ocean wasn’t easy. In addition to using
rudimentary sonar equipment, fishing boat captains—whose primary responsibility was to steer
the boat toward the fish—based their decisions on information gathered from ecological cues
(e.g., hovering seagulls, circling whales), technological cues (e.g., radio reports from research
vessels), and, importantly, social cues (e.g., following the movements of other fishing boats).
Highlighting the peculiarity of the social factor in particular, anthropologist Frederik Barth
(1966: 10) observed about the movement of fishing boats off the coast of his native Norway:
"The pattern of movement of vessels on the fishing banks is so extreme that it cannot fail to
8
strike an observer immediately: the several hundred vessels of the fleet constantly tend to
congregate in small areas of the immense, and potentially bountiful, expanse of sea; most
attention is concentrated on discovering the movements of other vessels, and most time is spent
chasing other vessels to such unplanned and fruitless rendezvous."
The imagery Barth’s observation brings to mind—tight clusters of boats intent on tracking not
fish but one another—has significant parallels to theories of the marketplace. One parallel is
organizational “herding,” or the oversaturation by firms of a particular market segment
(Brunnermeier, 2001). But whereas the notion of herding is typically used to explain the rise
(and eventual burst) of market bubbles, there is nothing bubble-like in Barth’s account.
Organizational clustering in markets need not be a temporary anomaly; it may reflect the steady
state, as it does in the fishing industry.
A second parallel is White’s theory of production markets (White, 1981, 2008; Leifer and White,
1987) wherein producers base their everyday production decisions not on an estimation of
consumer demand but rather on observations of peer producers. More generally, Barth’s
observations are an example of the importance of “social proof” (Rao, Greve, and Davis, 2001;
Cialdini, 2001) whereby increasing instances of a particular and, importantly, observable
outcome signal the appropriateness of that outcome. In both accounts, markets and fishing,
imperfect or unavailable information about demand—where consumers (or fish) are located and
how much they are willing to buy (or bite)—help to drive the decision-making process of
producers. Because peer behavior is more observable than consumer (and certainly fish)
9
behavior, mimicking peers’ decisions serves as a basis for organizational action (DiMaggio and
Powell, 1983).
For Barth, however, imperfect information about the location of fish is only the root of the
complexity in the series of decisions and evaluations that follow. Compounding the information
problem, the skipper is forced to make his decision—join the cluster, deviate from the cluster, or
else stay on the dock and forego any possibility of making a catch—within the context of
“important transactionally determined constraints” (Barth, 1966: 10). Transactional constraints
stem from two sources. The first is that of the skipper’s relationship with his crew: "Without
special information to justify the move, [if the skipper] decides to go elsewhere than where other
vessels go, he demands more trust in his transaction with the crew. They are asked to respect his
judgment, as opposed to that of the other skippers; they are thus asked to make greater
presentations of submission than they would otherwise have to do" (Barth, 1966: 10). In effect,
crew members aboard a vessel charting its own course are asked to ignore evident “social proof.”
The second constraint relates to the skipper’s calculation of his anticipated payoffs. Certain risks
are weighed against only potential rewards:
The skipper…risks more by not joining the cluster: if a few vessels among many make a catch,
the crew…can claim that it might have been them, had the skipper only given them the chance. If
the vessel on the other hand follows the rest, they are no worse off than most, and the onus of
failure does not fall on the skipper. Secondly, the absence of a catch matters less, so long as other
10
vessels also fail—the measure of a skipper’s competence and success is not absolute, but relative
to the catch of other vessels. (Barth, 1966: 10)
In Barth’s estimation, the risk inherent in nonconformity—as well as the associated trust and
obedience demanded from crew members—amplifies a crew's evaluation of its captain. Barth
contends that conditional on making a catch, the additional risk associated with deviating from
the cluster begets additional reward. In the absence of a catch, however, much like the release of
a pressure value, the supplemental trust and obedience required on board boats located outside
the cluster rapidly transition to anger and rejection of the captain. Even though many of the boats
in the cluster will also inevitably fail to make a catch, the crew on the vessel charting its own
course is left to wonder if they may have been among the successful. Refuting social proof and
failing to make a catch may have dire consequences for the captain.
Thinking in terms of an organizational analog to the scene Barth describes is useful in
identifying the mechanisms to which he alludes and constructing testable hypotheses about the
behavior of market actors. In the context of financial markets, specifically, we can think of fund
managers as comparable to boat captains, investors as similar to a boat’s crew (i.e., evaluating
audience), and fish akin to returns (i.e., performance). Boat captains have two goals: to attract the
best crew and catch the most fish. In constructing and managing their portfolios, fund managers
similarly strive to attract investors and to generate returns. Moreover, like boat captains, fund
managers use ecological cues (e.g., density of the market), technological cues (e.g., algorithms,
historical analysis, Bloomberg terminals, wire reports, etc.), and social cues (e.g., index
benchmarking, comparisons to fund products already in the market) to make their decisions.
11
At the point of critical decision making—pushing out from the dock each day or launching a new
fund—the captain and fund manager have two options. They can opt to join the cluster, should
one exist, or, they can strike out on their own. As Barth notes, doing the latter demands more
trust (and submission) from an audience. But this isn’t all. As a consequence of the captain’s or
fund manager’s initial decision, he or she (and the boat or fund, by extension) acquires an
identity that is either conformist or nonconformist.1 The person who joins the cluster is
conformist and typical. The one who deviates from it is nonconformist and atypical. Two
performance outcomes are possible for either type. Conformist and nonconformists alike may
catch fish (or produce positive returns) or fail to catch fish (or produce no returns or negative
returns). Evaluations (by crew members or investors) ensue in real time. Captains deemed highly
capable will be able to recruit the best crew. Fund managers seen as possessing the skills
necessary to generate above-normal returns will benefit from greater inflows of investor capital
and a reduction in the rate and size of capital redemptions. The conceptualization of identity as
conformist or nonconformist can have different effects on audiences' evaluations depending on
whether identity is used as a signal for evaluating an entity or whether it is used as a lens through
which to interpret information about an entity. In other words, evaluations are likely to be
contingent on the focal entity’s identity relative to other entities in the competitive environment.
Identity as a Signal
If nonconformity engenders misunderstanding, illegitimacy, and ultimately disregard, then there
should be a positive association between the conformity signaled by an organization’s identity
and its evaluation by an audience (e.g., Suchman, 1995; Miller and Chen, 1996; Henderson,
12
1999; Zuckerman, 1999; Rao, Monin, and Durand, 2003; Hsu, 2006). In the context of financial
markets and the hedge fund industry, there are several reasons to expect a positive effect of
organizational conformity on audiences' evaluations. First, investors' search costs should be
lower for conformist, or typical funds. Even though managers of atypical funds might offset this
additional cost by lowering their fees, typical funds should still be more visible to investors than
atypical ones. Second, atypical funds are further disadvantaged due to the difficulties associated
with educating potential investors about things with which they are not already familiar (cf.
Miller and Chen, 1996: 1217). One theme that emerged in a series of interviews I conducted with
hedge fund managers in the three major alternative investing markets—the U.S., U.K., and Asia
Pacific—was the inconvenience of having to teach potential investors about hedge fund
investment strategies, which are more complex than traditional investment vehicles such as
mutual funds. Hedge funds can employ complex trading strategies such as short selling and
operate in non-traditional asset markets such as real estate, derivatives, or even art. One London-
based manager noted, “It’s kind of a sad part of our industry…the gap between what managers
do and the knowledge from the management side and either a consultant or an investor. It’s
pretty big. I’m usually fairly disappointed when I walk out of the room in terms of that person’s
ability to truly grasp what we’re up to.” To overcome the resulting knowledge asymmetries, and
in the absence of social proof, managers devote significant time and resources to educating
potential investors. By contrast, a manager of a more typical fund can say, “We do the same
basic trades as the majority of long/short equity funds, only we do them better,” but atypical fund
managers must spend more time educating investors to generate comparable levels of
investment.
13
The same sorts of category-based constraints identified in other markets (e.g., Zuckerman, 1999;
Hsu, 2006) should function in the hedge fund industry. Search costs aside, many investors
simply lack the cognitive flexibility necessary to comprehend and appreciate funds that deviate
significantly from the majority of other funds in a given reference group. Industry consultants
and “cap-intro” (intermediaries responsible for introducing end capital investors to fund
managers) specialists are likely to account for some of this effect. Another fund manager I
interviewed offered an telling example: “We trade this GTAA product. It’s a top-down, country
level, global macro strategy. We decided to start trading single stocks within Euroland, so large-
cap equities within continental Europe, and the consultants did not want that bolted into the
GTAA strategy. They wanted it as a separate product, because we wouldn’t fit as nicely into
their GTAA box.” In spite of the fact that the manager believed the combination would offer
distinct benefits to clients, the additional focus was ultimately launched as a separate fund
product. A third fund manager echoed these sentiments even more directly, noting, “We feel like
the…consultants are the biggest bottleneck in having to keep strategies within a well-defined
box.” Identity that is well defined, however, acts as a signal that allows audiences to evaluate it
easily and should lead to a positive evaluation:
Hypothesis 1 (H1): The higher the typicality of an organization's identity, the more positively it
should be evaluated by an audience.
Identity as a Lens
The analogy to Barth’s fishing boats is even more useful when identity is conceptualized as a
moderator, or a lens through which certain kinds of information pass, take on meaning, and are
14
ultimately evaluated. In this respect, identity may be likened to a “sensemaking device” as the
phrase is used in organizational psychology (Fiske and Taylor, 1991; Weick 1995; Gioia and
Thomas, 1996; see also Dutton and Dukerich, 1991). Organizational identities offer a tool by
which to assign differential meaning and draw differential interpretations of otherwise
comparable information.
According to Barth’s observations, we should expect successful nonconformists to be the ones
evaluators most revere and unsuccessful nonconformists to be the ones they most despise, such
that organizational nonconformity should amplify both positive and negative reactions to
positive and negative emerging information, respectively. In the former case, at least, Barth’s
expectation is consistent with both Zuckerman’s (1999) aforementioned caveat and
Schumpeter’s proposition well before him: the greatest gains may flow to those that successfully
innovate. From the standpoint of the audience or evaluator, this statement implies that one is
drawn to the new and different if and when the new and different demonstrates competence. For
the producer, the same statement might be interpreted to mean that the additional risk assumed
by the act of violating social proof amounts to additional reward when things go right.
Prior research in both finance and economic sociology complements the basic risk-reward
account just offered. In studies of mutual and hedge fund industries, for instance, research has
shown that fund age significantly attenuates the association between fund returns and investor
capital flows in both positive and negative directions (Chevalier and Ellison, 1997; Berk and
Green, 2004). The argument relies on a kind of Bayesian learning thesis and suggests that
because investors have more information about older funds than younger funds, they should be
15
less affected by new information about the former. As a result, younger funds benefit more,
receiving additional capital inflows from investors, following recent positive performance and
are penalized more, facing additional capital outflows, following recent negative performance. In
both cases, positive and negative, new information simply means more to those doing the
evaluating.
In economic sociology, research has demonstrated a similar effect with respect to organizational
status (e.g., Podolny and Stuart, 1995; Stuart, Hoang, and Hybels, 1999). Much like the young
funds above, new information generally tends to affect the perception and evaluation of low-
status organizations much more than high-status organizations, unless the high-status
organization is being scrutinized for a particular transgression. Status also tends to be a more
salient indicator of firm survival—and, presumably as a precursor, audiences' evaluations—in
uncrowded market niches (Podolny, Stuart, and Hannan, 1996). This result is due in part to the
inherent uncertainty involved in evaluating firms in new or marginal market segments. In a
related application, Zuckerman (2004) linked uncertainty at the level of individual stocks to
trading volatility, suggesting that greater diversity in the set of mental models used by investors
to assess “incoherent stocks” should lead to different interpretations of the same information. By
extension, organizational nonconformity—itself an important determinant of uncertainty for an
evaluating audience—is likely to amplify both the positive and negative associations between
positive and negative performance and audiences' evaluations.
Hypothesis 2 (H2): Organizational nonconformity should moderate the effect of performance on
subsequent evaluation such that successful nonconformists are excessively rewarded.
16
Hypothesis 3 (H3): Organizational nonconformity should moderate the effect of performance on
subsequent evaluation such that unsuccessful nonconformists are excessively penalized.
METHODS
Hedge Funds
The choice of hedge funds as a research setting was not an arbitrary one. Following Sirri and
Tufano (1998: 1589), I view the fund industry as a “laboratory in which to study the actions of
individual investors who buy fund shares.” Based on a small number of variables, we can learn a
lot about investors' preferences for certain kinds of organizational offerings. Moreover, we can
examine how these preferences change, react to, and are shaped by various conditions.
Despite a lack of attention in the sociological literature (for an exception, see Hardie and
MacKenzie 2007), the hedge fund industry is a useful setting in which to test both the direct and
indirect effects of categorical identity on audiences' evaluations for a number of reasons. First,
identity, fund performance, and investor evaluation are readily identifiable and, importantly,
analytically distinguishable. I define identity as a fund’s correspondence to the central tendency
of a specified reference set of other hedge funds (cf, Miller and Chen, 1996). Reference sets
comprise funds that self-identify as being in the same primary investment style. Typical funds
have a high level of correspondence to the resulting central tendency. Atypical funds do not.
Performance is a fund’s returns. Audience evaluation is measured by capital flows into and out
of a fund on behalf of investors (i.e., investment). Second, unlike many industries, the hedge
fund industry is relatively free from legal and other formal institutional pressures that might
17
otherwise complicate analysis. This specific feature should contribute to greater variation along
the identity dimension I call typicality. Finally, although no comprehensive database exists that
covers the entire hedge fund industry, existing data, perhaps because it is targeted foremost at
practitioners, tends to be thorough and systematic, not to mention obtainable by other researchers
for further exploration or replication.
Data
Because most hedge funds are not obligated to report to a regulatory body, data on the industry
must be obtained from one or more private data collection services. I used data from the Tremont
Advisors Statistical Services (TASS) hedge fund database. Liang (2000) suggested using the
TASS database for academic research, in particular because it is the most complete and most
accurate with respect to information on returns, assets, fees, and fund characteristics. The sample
I used consists of monthly return data and total estimated assets for funds from January 1994 to
December 2008. I used January 1994 as a starting point because this was the first year for which
TASS maintained a “graveyard” database. Prior to 1994, liquidated funds and funds that stopped
reporting to TASS were dropped from the database. Including of data prior to 1994 in an analysis
such as the one here would result in an obvious survivor bias (see Ackerman, McEnally, and
Ravenscraft, 1999; Capocci and Hubner, 2004, for a more complete discussion of this particular
bias with respect to performance).
Two remaining biases that must be acknowledged are self-selection with respect to funds’
likelihoods of reporting to TASS and “instant history” (Fung and Hsieh, 2000; Lawson and
Peterson, 2008). Studies primarily concerned with measuring the average or aggregate
18
performance of the hedge fund industry are attuned to a self-selection bias, as over reporting by
well-performing funds will result in an overstatement of performance. Underreporting by well-
performing funds will have the opposite effect. The analyses in this paper, though not concerned
with industry-level performance, may be similarly biased if hedge funds differ in their
likelihoods of reporting along a dimension correlated with my measure of fund typicality. There
is no systematic way to account for this possibility, but there are several reasons to believe that
any bias resulting from it is likely to be small. Funds with categorical identities that are exact
replicas of those already in the market (i.e., funds having maximum typicality at the time of
market entry) may have an incentive to veil their complete lack of differentiation. Similarly,
highly atypical funds may choose not to report in order to hide from investors the fact that they
are radically different from competing funds. The lower likelihood of reporting among funds at
both extremes of the distribution may offset any resulting bias (see Fung and Hsieh, 2000, for a
comparable argument with respect to returns).
A third important bias, “instant history,” results from the back-filling of returns for funds whose
managers choose to report to a database such as TASS sometime after the fund's inception.
Back-filling returns will upwardly bias return estimates if funds misrepresent or smooth past
returns data or, more commonly, only report past returns when they reflect well on the fund. The
TASS data show some evidence of an instant history bias. Among the funds used in the analyses
here, average monthly returns for funds in their first year are 0.73 percent higher than in all other
years. This figure is slightly smaller than the one Fung and Hsieh (2000) calculated, indicating
that hedge funds opting to report to TASS in more recent years were more likely to do so from
inception. Following convention, I dropped from the analysis the first 12 months of data from
19
each of the funds. Including these months has no discernible effect on the substantive claims of
the paper.
After excluding funds based on the guidelines above, and dropping from the dataset any funds
with significant omissions or irregularities in their reported assets under management, 6,562
funds remained with which to perform most of the analyses in this paper. When aggregated at the
quarter level, as are all of the analyses, this number equates to 92,226 fund-quarter observations.
I note variations to this N when applicable.
Dependent Variable: Audience Evaluation
I used capital flows to measure investors’ evaluations of a given fund or fund manager. Capital
flows refer to increases, by way of additional investment, and decreases, by way of capital
redemption or divestment, in the amount of invested capital in a fund. Flows were calculated as
the intertemporal change in total net assets (TNA) above and beyond what can be attributed to
returns (R). Algebraically,
,
where is fund i’s total net assets at time t, and is fund i's return over the prior period.
has a theoretical lower bound of -100 percent that corresponds to the scenario in which
all investors redeem all capital allocated to fund i in the period between t-1 and t. Following Sirri
and Tufano (1998) and others (e.g., Fung et al., 2008), I assume by construction of the measure
that flows are realized at the end of each period. Periods are quarters.
20
Although nearly all prior studies reveal a strong association between fund performance and
capital flows (for studies using data on mutual funds, see Chevalier and Ellison, 1997; Sirri and
Tufano, 1998; for hedge funds, see Agarwal, Daniel, and Naik, 2007; Fung et al., 2008), this
association is not necessarily symmetric with respect to positive and negative returns. Capital
inflows respond more to positive returns than capital outflows do to negative returns. There are
several reasons for this asymmetry. One is purely mechanical: whereas capital outflows, usually
associated with negative returns, are attributable only to current investors in a fund, capital
inflows result from both current and new investors. Another relates to the practice, common
among hedge funds, of restricting the flow of investor capital. Such “liquidity terms” are
typically in the form of lockup periods—invested capital must stay in a fund for a predetermined
number of months—and redemption notice periods—investors must give a manager sufficient
warning before withdrawing invested capital. In addition, several other factors also affect capital
flows. Lower-fee funds grow faster than higher-fee funds (e.g., Sirri and Tufano, 1998). Funds
with more strict liquidity terms reduce capital outflows. Beating a specified (even an incorrectly
specified) benchmark increases capital flows (e.g., Sensoy, 2009, Fung et al., 2008). Finally, size
predicts capital flows, although evidence on the effect of fund size is mixed (see Getmansky, Lo,
and Makarov, 2004)).
Independent Variables
Hedge fund identity. Research in organizational theory and economic sociology has employed
three basic and overlapping identity concepts: identity as a set of core organizational features
(Albert and Whetten, 1985; Whetten, 2006), identity as membership claims to particular social
21
categories (Tajfel and Turner, 1986; White, 2008), and identity as dynamic and imputed via
constituent audiences over the course of an organization’s life (Hsu, 2006; Briscoe and Safford,
2008). The measurement of hedge fund identity I used combines elements from each of these
three conceptions. I refer to the measure as fund typicality.
[Insert table 1 about here]
The process of identification begins with group membership. In the case of hedge funds, groups
are primary investment styles. Table 1 describes each style. There are several reasons to believe
that primary investment styles constitute the most important and visible dimension on which
categorization occurs. First, for the purpose of marketing to investors and inclusion in various
databases, all funds self-select into a single style grouping, so a fund’s primary investment style
is typically the first thing a potential investor will know about the fund from its offering
documents. This initial exposure is important on a cognitive level as it has been shown to have
significant and long-lasting effects on the propensity of consumers (investors, in this case) to use
stated categories to derive a product’s or organization’s reference set (Leclerc, Hsee, and Nunes,
2005). Second, research in finance consistently and definitively illustrates that style categories
are important predictors of risk, return, rates of fund survival and attrition, and levels of returns'
correlation with market indices (see Lo 2008, chap. 1-2, for a comprehensive review of literature
and findings). As a result, all empirical studies of hedge funds treat style-based categories as
distinct and important groups and adjust for them (i.e., use fixed effects) accordingly. Finally,
funds typically remain in a single categorical group for their entire lifespan. In the event that an
22
individual fund fits into more than one category, a manager can choose to designate the fund
Multi-Strategy.
[Insert figures 1 and 2 about here]
In addition to this broad classificatory schema, funds also provide more fine-grained information
on their particular investment approaches (i.e., sub-strategies and/or specific trading approaches)
and strategic focuses. TASS collects a total of 33 investment approach and strategic focus
attributes. In addition to offering a dimension on which to differentiate funds within a given style
category, these attributes are also useful for further demonstrating the importance of primary
styles in segmenting or categorizing the industry. Figures 1 and 2 are intended to illustrate this
point. Figure 1 uses the Fruchterman-Reingold (1991) algorithm to plot the cluster distribution of
a sample cross section of data on funds from 1995, in which individual funds (denoted as
individual nodes and shaded by primary investment style) are positioned according to their
composition of attributes. Funds are shown as clustered when there is a significant (95 percent)
overlap in their investment approach and strategic focus attributes, which generates a graph with
an approximately 5 percent density. The resulting plot indicates that many, though not all,
primary styles are composed of funds that are relatively similar to one another and relatively
dissimilar from funds in other styles. The exceptions are Multi-Strategy, Funds of Funds, and
Emerging Markets, which show no discernible clustering. Figure 2 is a multidimensional scaling
plot that uses a cross section of the data from 2005. Unlike figure 1, the nodes in figure 2 are
primary styles. The number printed inside the nodes is the number of funds in a style during the
given cross section. Node size is proportional to the amount of heterogeneity among funds of a
23
given style with respect to their investment approach and strategic focus attributes. Large nodes
thus denote styles with significant heterogeneity among funds. Smaller nodes indicate
homogeneity among funds. As one might expect, Multi-Strategy, Funds of Funds, and Emerging
Markets prove to be the most heterogeneous styles. Style nodes are positioned relative to one
another according to their similarity, as determined by the composition of attributes of the
underlying individual funds. In addition to visually illustrating the salience of the primary style,
figure 2 indicates the possibility that heterogeneous styles (i.e., large nodes) may not constitute
meaningful reference groups, a possibility I explore after testing the three primary hypotheses.
By combining coarse- (style) and fine-grained (attributes) information, it is possible to grade
hedge funds on the extent to which they match up against other funds in the same primary style
grouping. The idea that organizational identity can be graded in terms of likenesses or typicality
follows from Rosch’s (1973) notion of webs of sameness and is conceptually similar to Miller
and Chen’s (1996: 1210) measure of organizational nonconformity as deviation from “industry
central tendencies or de facto norms,” as well as Hannan’s (2010) more recent formalizations of
organizational “grades of membership.”
Finally, I treated identities as dynamic over time and dependent on changes among other funds in
a focal fund’s proximity. Complete control over one’s identity is afforded only at the moment of
market entry. Subsequent valuations of a fund’s level of typicality are the result of comparisons
with other entities in the environment and thus depend on the movements of those entities. The
measurement technique I used is readily generalizable to other settings, though for the sake of
simplicity and interpretability, I discuss it formally below in the specific context of hedge funds.
24
Typical funds have combinations of attributes that are similar to an average, hypothetical fund of
the same style. Atypical funds do not. If we represent each hedge fund in the database as a vector
of binary indicators that correspond to its strategic focus and investment approach attributes, then
we can compare it against all other funds in its primary style category at a given point in time. I
did this by first constructing a two-mode, rectangular matrix for every style-quarter combination,
in which columns represent unique funds and rows are strategic and investment focus attributes.
The number of rows in the resulting matrix is fixed and equals 33, or the number of possible
attributes from TASS. The number of columns is equal to the number of funds in a particular
style category in operation during a given quarter.
For each style-quarter combination, I generated a hypothetical fund by averaging across the rows
in the rectangular matrix. If 100 funds of style X exist in quarter T, and 90 of them indicate the
presence of attribute Y, then in the resulting hypothetical fund, the vector element corresponding
to attribute Y would equal .9. If only 10 of the 100 indicate the presence of attribute Z, by
comparison, then the corresponding hypothetical fund vector element Z would be .1. For
illustration, the columns in table 2 are the resulting 33 attribute vectors from TASS for the
Convertible Arbitrage style during the first quarter of each of the years shown.
[Insert table 2 about here]
At this level of analysis—using the 33 attributes as opposed to, for instance, the actual securities
held by a fund—a fund’s vector of attributes can be treated as fixed over time.2 One reason given
25
in the finance literature as to why this assumption is tenable is that fund managers will more
readily close down and reopen a fund than make changes to a focal fund’s underlying sub-
strategies and focuses (Agarwal, Daniel, and Naik, 2007; Klebanov, 2008: 7). This feature has an
interesting effect on the measure of fund identity I propose. After the point of market entry, fund
identities respond solely to changes in the market around a focal fund. As a fund’s reference
group moves away from it (in terms of sub-strategies and focuses) through the entry of new
funds and exit of old funds, its degree of typicality will decline. By contrast, a fund will become
more typical when new funds in its reference group cluster around it. Typical funds are like the
boats in Barth's cluster. Atypical funds sail alone.
Formally, fund typicality was measured as the amount of overlap between the fund vector and
the hypothetical average, or central tendency vector for the fund’s primary category. There are
several techniques for computing similarities between vectors. The one I used is a Dice
coefficient (Dice, 1945):
where A and B represent the individual fund and the average fund, respectively, p is equal to the
number of attributes that are present for both A and B, and a and b are equal to the number of
attributes present in A but not B, and B but not A, respectively. Vector elements in A are binary.
An actual focal fund either has the attribute or it does not. Vector elements in B have upper and
lower bounds of 0 and 1, but may take any fractional value in between as well, as demonstrated
in table 2, above. The intersection of vector elements across A and B will naturally take the
26
fractional value in B, should one exist. The similarity between a focal fund A and the
hypothetical fund B will thus be greater when two corresponding elements across the vectors are
1 and .9 versus 1 and .1. Because of this implicit weighting, investors (who I expect to both
perceive and respond to the fund typicality) need not conceptualize the complexity of typicality
in 33 dimensions but may zero in on a few of the more important attributes to distill a fund’s
approximate level of conformity. This measurement feature enhances the robustness of the
typicality construct by allowing investors to use any number of (unobservable) cognitive
schemas to evaluate a given fund. Having said that, the substantive empirical results presented in
the next section can be replicated using measures of typicality computed with the Jaccard
coefficient and Euclidean distance.
Conformity is captured in SA,B, which is a measurement of the similarity between the focal fund
vector and the hypothetical fund vector, though because all funds in a given categorical group
(and at the same time) are compared with the same average vector, the resulting value may be
interpreted as fund A’s degree of typicality. TA is bound by zero and one. For any given cross
section, funds in the same reference style will be arrayed with respect to their degree of
typicality. Although it may appear counterintuitive to the reader, zero level typicality is possible
(though very rare in the data) because all funds are obliged to identify with one primary style
category, even if the combination of attributes they report are similar to none of them.
Descriptive statistics for a single cross section in the data are presented in table 3. Funds are
grouped in this table by primary style.
[Insert table 3 about here]
27
Performance. I computed several measures of fund performance following conventional
approaches in the finance literature. Variations across measures amount to the degree of risk
adjustment included in the measures themselves. For instance, in some cases it may be useful to
know a fund’s returns in excess of one or more specified benchmarks. If a fund reports a net-of-
fees return of 6 percent for a period when the S&P500 returned 10 percent, we might think of the
fund's actual returns as something more like -4 percent. By contrast, if a fund posts 6 percent
returns over a period of time when the S&P500 fell 5 percent, then it might be more accurate to
think of the fund as having returns around 11 percent. In addition to the S&P500, researchers
have used several other benchmarking factors against which to study returns, including various
foreign exchanges, as well as treasury and corporate bonds, and alternative investment rates.
Whether to adjust returns is a potentially complex decision when modeling investor behaviors
such as capital allocation. If we believe that investors respond primarily to raw, net-of-fees return
figures, then returns should not be adjusted using any benchmarking factors. If there is reason to
believe that investors consider returns only in comparison with the returns of one or more
investment benchmarks, then returns should be adjusted accordingly. For simplicity and
interpretation, I used raw, net-of-fees returns in the analyses presented in this paper. However,
because I also included time-period fixed effects in all of the models, the choice to use raw
returns data does not indicate a strong assumption about investors’ propensities to benchmark.
Period fixed effects have the advantage of soaking up all variations in the dependent variable that
are the result of macro conditions as well as contemporaneous movements of equity markets and
thus greatly diminish the importance of the benchmarking question. Results of the primary
28
analyses are consistent when reanalyzed using standard, single-factor—S&P 500—risk-adjusted
returns.
Hypothesis Testing
Having defined the three primary variables for the analysis—fund identity as typicality,
performance information as returns, and evaluation as capital flows—it is useful to restate the
hypotheses derived above using language specific to the hedge fund industry. First, there is the
main effect: (H1) The higher a hedge fund’s level of typicality, the greater its capital inflows.
Second, there are the interactions: (H2) The effect of recent positive returns on capital inflows is
greater among atypical funds, and, (H3) the effect of recent negative returns on capital outflows
is greater among atypical funds. Hypothesis 2 reflects hedge fund managers who strike out on
their own and are successful. Investors are not only drawn to positive returns, they are drawn to
positive returns in unexpected places. Hypothesis 3 calls to mind the disappointment, disdain,
and eventual flight by investors in funds whose managers dismissed the social information
available to them and failed. In either case, whether joining the cluster or going alone, typicality
should amplify the association between performance and evaluation.
Model
I analyzed the link between typicality, and returns and flows using a linear time-series regression
framework applied to fourteen years of fund-level data. In each of the analyses, I fit some
variation of the following model:
,
29
where represents the capital flows into or out of fund i in period t, and are
the typicality and percentage returns of fund i in the preceding period, and represents
the interaction between the two. Following convention, is split into two variables, one
accounting for positive returns and one for negative. In the former case, observations having
negative returns are set to zero. In the latter, observations having positive returns are set to zero.
This approach is standard in the financial literature (Chevalier and Ellison, 1997; Abdellaoui,
Bleichrodt, and Paraschiv, 2007; Lo, 2008) and is used to adjust for the potential nonlinear
association between returns and capital flows that occurs around the point of zero returns. Note
that one should expect positive coefficient estimates for both returns variables. On the right-hand
side of the zero point, a positive coefficient indicates that increasing (more positive) returns
should result in greater (more positive) capital flows. Equivalently, a positive coefficient on the
left-hand side indicates that increasing (less negative) returns should result in greater (less
negative) capital flows. The result of this adjustment from a regression standpoint is that all
models include two returns parameters and two interactions between returns (positive and
negative) and fund typicality, or:
All models contain several control variables. These variables include fund age measured in years
since inception, fund size measured as the log of total fund assets, percentage management and
incentive fees, liquidity terms measured in days of lockup and redemption notice periods, and
fund family size.3 Fund family size is simply a count of the number of additional funds currently
30
operated by the same management company. In addition to these, two additional variables are
included as measures of funds’ long-term performance. One is a rolling 12-month returns
average that is calculated over the year-long period ending two quarters before the focal quarter
so that there is no overlap between and this more long-term metric. The other is the
standard deviation of the monthly returns (i.e., volatility). These adjustments are useful for two
reasons. First, they amount to an additional control for managers' skill. To the greatest extent
possible, I wanted to ensure against the possibility that an effect of fund typicality on capital
flows is due to managers' skill and not fund typicality. In a robustness test, I assessed this even
more strongly by including a fixed effect for every unique fund in the data. Second,
incorporating longer-term performance parameters emphasizes the short-term nature of the
hypothesized effects.
Importantly, all models include fixed effects for primary style categories. Fixed effects eliminate
all between-style variation. This adjustment is necessary given evidence, noted above, that styles
yield systematically different risk-return profiles. Additionally, I adjusted for all types of
temporal heterogeneity, such as macroeconomic conditions, by including quarterly fixed effects.
Because the models were estimated on the entire dataset, standard errors are clustered at the level
of individual funds and are heteroskedasticity-consistent. This is necessary, as individual fund-
quarter observations may not be independent of one another.
RESULTS
[Insert table 4 about here]
31
Table 4 includes descriptive statistics for all variables used in the analyses and the correlations
among those variables. Following convention, I eliminated observations having capital flows
greater or less than three standard deviations from the mean. Results do not differ when these
observations are included in the analysis, but their exclusion guards against undue influence of
outlying data points. Table 5 presents results from the primary analyses.
[Insert table 5 about here]
Model 1 includes only the control variables. Model 2 tests hypothesis 1 and assesses the effect of
recent returns on capital flows. Model 3 tests hypotheses 2 and 3. Model 4 tests the robustness of
the results in model 3 by reestimating the regression as a change model. This approach involves
including a lagged dependent variable on the right-hand side of the regression equation. To
further control against unobservable and unmeasured differences in quality at the level of the
fund manager, I included a fixed effect at the level of the fund in model 5. Models 6 and 7 are
additional robustness checks designed to account for possible endogeneity in the prior
specifications.
The results of model 1 are consistent with prior findings. Fund age and size have a negative
effect on capital flows. The two effects of fees are opposite each other. Whereas higher
management fees—typically between 1 percent and 2 percent of total assets—are associated with
increased capital flows, incentive fees—typically around 20 percent of total profits—have a
negative effect on capital flows. Both liquidity restrictions have a positive effect on capital
32
flows—only the redemption notice period is robustly significant at a 5 percent level—which are
likely driven by the negative association between strict liquidity and capital outflows. Being part
of a large fund family decreases capital inflows, but only marginally. Finally, the rolling 12-
month return and volatility measures have large positive and large negative effects on capital
flows, respectively. Controlling for volatility, funds posting strong long-term returns are
rewarded. Controlling for returns, increased volatility reduces capital inflows.
Model 2 introduces prior quarter returns as well as fund typicality. The results support both
implicit and stated hypotheses and are consistent with prior research on both the mutual fund and
hedge fund industries. Capital flows respond more to positive recent returns (.447, t=19.43) than
to negative recent returns (.122, t = 4.52). A 1-percent increase in positive returns is associated
with about a half-percent increase in subsequent inflows. A 1-percent decrease in negative
returns generates just under an eighth-percent decrease in outflows. As for fund identity, there is
a positive effect of fund typicality on capital flows (2.485, t = 3.06). Other things held constant,
atypical funds elicit fewer capital inflows. This result supports the argument that investors—
either directly or via consultants—do consider something like my measure of typicality when
making capital allocation decisions. In the absence of this main effect, justifying this paper’s
main claim—that investors interpret market information through the lens that identity offers—
would be more difficult.
Model 3 presents a direct test of the moderating capacity of typicality by including the two
interaction terms, typicality by prior quarter positive returns and typicality by prior quarter
negative returns. According to H2, we should find a negative coefficient estimate on the
33
interaction term with positive performance. Atypical funds should be more rewarded for positive
recent performance than typical funds. Results strongly support this proposition (-.731, t = -
4.46). H3 suggests this effect should be symmetric about the zero returns point: whereas atypical
funds are in fact more rewarded for positive returns, they should also be more penalized for
negative returns. The coefficient estimate on the second interaction term does not support this
expectation (.668, t = 3.77). Rather, atypical funds appear to be less penalized, or buffered, with
respect to capital outflows for equivalently poor recent performance. Figure 3 illustrates visually
that atypical funds are both more rewarded and less penalized for positive and negative returns,
respectively.
[Insert figure 3 about here]
Both results, the expected amplification for positive returns and the unexpected buffering for
negative ones, are significant even when adding a lagged dependent variable to the right-hand
side of the regression equation (model 4) and when including a fixed effect at the level of the
fund (model 5). With a lagged dependent variable as a regressor, the other variables in the model
now predict quarter-by-quarter change in capital flows into and out of a fund, rather than the
level of capital flows in a focal quarter. Results indicate that flows are correlated from one
quarter to the next, even after accounting for the effect of the control variables.4 Importantly, the
changes in both models 4 and 5 do not affect the primary findings from model 3.
What the interaction terms alone conceal, however, is the main effect of fund typicality. This
omission is potentially important and should be accounted for to make sense of the complex
34
relationship between fund typicality, and fund performance and capital flows. In fact, if the main
effect of typicality on capital flows is large enough, then despite the implications of model 3,
atypical funds may never be in a position to benefit from their atypicality. The greatest inflows
should accrue to atypical funds that produce excessively positive returns in the prior quarter. The
greatest outflows, by contrast, are found among the most typical funds that post excessively
negative returns in the prior quarter. A different pattern takes shape, however, around the point
of zero returns. If we consider only a narrow distribution of quarterly returns, then the main
effect of typicality outweighs the potential for either excess reward or buffering among atypical
funds.
Due to the negative correlation between redemption notice period and fund typicality, there is a
potential for liquidity restrictions, rather than investors' decision making, to drive the buffering
effect observed among atypical funds: capital may flow from atypical funds more slowly
following periods of negative returns simply because investors in these funds are more restricted,
on average. But certain features of the data and analysis render this alternative unlikely. The
distribution of redemption notice periods across all funds in the database indicates that more than
98 percent of them have redemption notice periods that are less than 90 days, or a single quarter.
The median is 30 days. As a result, even though redemption notice periods are slightly longer
among atypical funds, we might still reasonably assume that capital outflows that are the result
of losses in a given quarter should be realized by the end of the following quarter, or 90 days
later. Nevertheless, model 6 addresses this alternative explanation by modeling capital flows at
time t+1 on fund typicality, returns, and the interaction between them at time t-1, controlling for
returns at the intermediate period, time t. Model 6 thus accounts for an additional 90 days of
35
“drag” in the association between the interaction terms and outcome. Results are statistically
weaker but generally consistent with model 3.
An additional identification concern arises largely as a consequence of the mathematical
dependency in the assignment of fund typicality, that is, change in a fund’s level of typicality is
determined by the entry and exit of other funds in its primary style category. An endogeneity
concern presents itself and is rooted in the fact that fund foundings are probably driven in part by
observations of which funds were relatively more successful in the past. This statement is
consistent, of course, with the premise that entry into a market will continue until the marginal
actor in that market fails to realize an economic profit. As a result, endogeneity will be a concern
if fund typicality in any way causes fund performance. In supplemental analyses available from
the author it is clear that this association is not present (beta = 0.165, p = 0.439), thus eliminating
much of the basis for concern. Model 7 addresses this endogeneity concern even more directly,
however. Due to the sheer size of the sample and moderate rate of turnover, most variation in
typicality is between funds. A simple control for the concern just raised then is to fix fund
typicality at the point of fund entry, thus alleviating any possible confounding arising from the
complex interdependencies between performance, changes in fund typicality as a result of
changes in the competitive landscape, and capital flows.5 The results of model 7 are again
consistent with model 3.
[Insert tables 6 and 7 about here]
36
Models 8 and 9 in table 6, as well as models 10 through 13 in table 7, add additional nuance to
the principal empirical results in model 3. In table 6, I investigate the duration of the buffering
effect by sampling from funds that have two and three consecutive quarters with negative
returns. The model on the left-hand side of table 6 replicates model 3 from table 5 and serves as a
point of comparison. Positive returns and the interaction between positive returns and typicality
drop out of models 8 and 9, as inclusion in regressions is contingent on having negative returns.
Together, models 8 and 9 indicate that the buffering effect afforded to atypical but poorly
performing funds extends to funds with two consecutive quarters of negative returns (both the
magnitude and significance level of the effect is attenuated, however), but not to those with three
such bad quarters. All prior controls are included in models 8 and 9. Coefficient estimates of
these additional controls are comparable to the models in table 5 and are thus omitted.
Finally, models 10 through 13 incorporate Ruef and Patterson’s (2009) argument that categorical
identity effects should be far smaller or even non-existent when the category itself is not well
defined (see also Kovacs and Hannan, 2010; Negro, Hannan, and Rao, 2010). In the case of
hedge funds, this means that we might expect the effects established above (in models using
fixed effects for style categories) to be attenuated or altogether absent among funds in styles that
are devoid of anything that might be considered meaningfully typical or average in the first
place. Of course, by nature of the typicality measurement, all style groups are ascribed a nominal
typical representation at the very least. We can assess how relevant or real this representation is
in two basic ways. First, eigenvalue decomposition of each style-quarter matrix allows for a
quick investigation of the principle components of each matrix. Principle components are useful
for assessing the amount of variation (or overlap) among the unique fund vectors that make up a
37
style-quarter matrix. If significant variance is detected, then the derived average vector against
which all actual fund vectors are compared may be less relevant.
A second, simpler approach that results in a nearly equivalent ranking of fund styles entails
looking at the average level of typicality in each style-quarter combination. Low-average
typicality is characteristic of greater variance among the individual funds in a given style. For
example, of the 11 styles in the study, the multi-strategy style, which should have the most
variation by definition, has the lowest average level of typicality. The results in table 7 are those
derived from regressions run on four subsamples of the data. Models 10 and 11 used funds from
the three style categories with the highest average measure of typicality, namely, Long/Short
Equity and Managed Futures. Models 12 and 13 were run using funds from the two style
categories with the lowest average measure of fund typicality, namely, Funds of Funds, and
Multi-Strategy. A brief visual scan of the resulting coefficients offers support for Ruef and
Patterson’s (2009) hypothesis. H2 and H3 are indeed stronger and more systematic in models 10
and 11.
Whereas the preceding check brings to light an important scope condition—that the direct and
indirect effects of categorical identity are greater when categories have internal consistency or
cohesion—the diminished effect in the latter groups may simply be due to measurement error
rather than audiences not using organizational identities to make sense of emerging information.
The measure of typicality I employed uses the central tendency or average fund as the point of
reference in each style-quarter grouping. As a result, the lack of a typicality effect in models 12
and 13 does not necessarily imply that identity is irrelevant in these style categories, only that the
38
average fund is an irrelevant reference point. In heterogeneous populations, an alternative
measure of typicality—for instance, one using an exemplar fund as a reference point rather than
a hypothetical average fund—may yield results consistent with those in more homogeneous
populations. The difficulty here, of course, and one that is beyond the scope of this paper, is
determining what constitutes a meaningful reference point for audiences evaluating a more
heterogeneous set of entities.
DISCUSSION AND CONCLUSION
The results of the study support the main effect, or identity-as-signal hypothesis. Investors
generally reward conformist hedge funds with conformity measured as the degree of fund's
typicality vis-à-vis its primary style grouping. The identity-as-lens hypotheses are also
supported, though the presumed amplification effect is supported in one direction only. The
combination of atypicality and positive performance indeed begets additional reward. Contrary
to prediction, however, atypical funds with poor recent performance are penalized less than
comparably performing but typical funds. The absence of symmetry suggests that mechanisms
other than those proposed above may in fact drive audiences' behavior, at least in the context of
the hedge fund industry.
Had the outcomes of the analysis been symmetric—that is, had atypical funds been rewarded
more and penalized more for positive and negative performance—the mechanisms previously
suggested to link identity, performance, and evaluation would suffice. Atypicality would be
beneficial, as was predicted at the outset, but only following positive performance. That
atypicality also buffers funds from the consequences of poor performance, however, suggests the
39
need for an alternative theory. Two possible anchoring points for such a theory are the concepts
of commitment and comparison, known in one form or another across sociological,
psychological, and economic literatures.
Towards a Theory of Beneficial Atypicality
Two plausible mechanisms can be distilled from the hedge fund case to underpin a theory of
beneficial atypicality. The first is escalating commitment. Staw (1976, 1997) first proposed the
concept to describe people’s tendency to use information about their own prior investment in
something to justify their continued investment in it. In other words, given equivalently negative
information about two comparable things, such as hedge funds, people are more apt to abandon
the one in which they initially invested less. Behavioral economists and psychologists have
referred to this effect as the sunk-cost fallacy.
The analogy to fishing is again useful here when we shift attention from risk and reward to
investment, payment, and cost. If investing a given amount of capital in an atypical fund is more
costly than investing the same amount in a typical one, a commitment bias should be more
readily observable among investors in atypical funds. Investing in atypical funds, or
equivalently, endorsing non-conformist organizations, involves two types of additional costs.
First, finding atypical funds in which to invest should involve additional search costs. Second,
much like the crew members aboard fishing boats located outside the visible cluster, investing in
atypical funds requires additional trust from investors. Investors thus incur greater sunk costs—
tangible search and intangible trust, neither of which can be recovered—when allocating capital
to an atypical fund. The greater aggregate costs incurred by these investors may result in an
40
escalation of commitment. Consistent with Staw’s original thesis, the escalation of commitment
among atypical funds is apparent in investors' relative hesitancy to redeem capital following
periods of negative performance. The results of the study suggest that atypical fund investors
displayed commitment through one and two consecutive quarters of negative returns, though the
effect disappeared after three such quarters.
The second mechanism relates to social comparison and substitutability. Just as atypical funds
are more costly to find and invest in, typical funds are more readily substitutable. As one fund of
funds manager I interviewed noted, “We’re definitely more quick to fire a manager in a more
traditional fund because it’s really easy to switch.” A second fund manager, commenting on the
advantages of offering a “niche product,” elaborated on this basic premise: “If I’m an investor
I’m going to stick it out more with non-traditional investments. If I decided to invest in, umm, I
don’t know, Brazilian hardwoods, I’m not going to run for the hills at the first sign of trouble.
But if I’m also invested in a pretty traditional long-short fund and things aren’t looking real
good, I can invest in about 10 others guys I know who are probably doing the exact same thing.”
The availability of substitutes implies that typical funds may in fact occupy more fragile
positions in the larger structure of the hedge fund industry (cf. Bothner, Smith, and White, 2010).
Should less-than-positive information emerge, investors can easily and relatively cheaply scan
the environment for alternative funds and, lockup periods notwithstanding, shift their capital
accordingly. Atypical funds, in contrast, fall outside the realm of ready social comparison.
41
The commitment and comparison mechanisms have contrasting implications for the typicality
construct and for the concept of organizational identity more generally. Under the commitment
mechanism, typicality is little more than a proxy for the unobservable cost of investment or
endorsement. Identity does not affect audiences' behavior, per se; rather, it offers clues to the
researcher about the nature and price of exchange between an organization and a given audience.
The salience of identity reasserts itself, however, when we consider social comparison. Despite
claiming a common label—for example, long-short equity, global macro, event driven, and so
forth—and thus inviting comparisons against more global benchmarks, in the absence of local,
identity-based comparisons, atypical organizations may be relatively free from the scrutiny of
social comparison. As a result, atypical organizations may be less affected by the devaluation—
capital redemption and reallocation in the case of hedge funds—that typical organizations face
following periods of poor performance.
Additional research is necessary to test hypotheses based on both the commitment and
comparison mechanisms. Testing the former, in particular, would require data at the level of the
investor that is not typically available in the hedge fund industry. Future research may therefore
benefit by adopting more qualitative methods to test the propositions that the additional costs
involved in endorsing an atypical entity (1) translate to greater commitment and (2) are manifest
as greater short-term tolerance for poor performance.
A comparison hypothesis, by contrast, may be testable with the sort of data used in this study,
though additional research will be necessary to explore the possible link between this paper’s
unexpected empirical result and the kind of cognitive processes suggested to underlie the
42
comparison hypothesis. Performance theory offers a useful starting point. Performance theory
implies that investors may evaluate returns relative to other indexed benchmarks. Standard
benchmarks in financial markets include the S&P 500 as well as various style-specific indexes
constructed and maintained by data providers such as TASS. According to the comparison
hypothesis, when an atypical organization performs poorly it benefits from audiences' reluctance
to evaluate that performance on a relative basis. Accordingly, while any kind of poor
performance, relative or absolute, should result in devaluation, it may be that for relative poor
performance of the same magnitude, a resulting devaluation will be noticeably smaller for
atypical organizations than typical ones.
To rephrase, atypical organizations may benefit from having audiences that use a very wide
"aperture" lens of comparison when things go well but a much narrower aperture lens when
things go poorly. By contrast, a mid-sized and non-variable aperture lens may be employed by
audiences evaluating all information, positive and negative, about more typical organizations.
Positive information in a wide-aperture lens world elicits excessive positive evaluations. As the
aperture of comparison narrows, however, both positive and negative information have
diminishing effects on audiences' evaluations. This explanation not only captures the significant
"kink" in the performance-capital flow relationship among atypical hedge funds, but it also does
well to explain the near linear (and heretofore unobserved) performance-capital flow relationship
among typical hedge funds revealed in this study.
Using financial data to draw definitive conclusions about the comparison or “variable-aperture”
hypothesis is not without its challenges. For starters, several researchers have downplayed the
43
extent to which multiple calculations of relative returns may be used along with data on capital
flows to impute and measure the presence of behavioral biases in investors' decision making
(e.g., Sirri and Tufano, 1998). Fully and satisfactorily addressing this skepticism is beyond the
scope of the current paper. Nevertheless, in future research it may be feasible to test the validity
of the variable-aperture concept by analyzing audience evaluation with respect to performance
measures that have been adjusted for one or more identity- (typicality-) based benchmarks.
Legitimacy and Isomorphism
In spite of their ubiquity, markets structured and reproduced by the social process of mimicry are
a peculiar phenomenon. Like fishing, the probability of realizing gains in zero-sum or near-zero-
sum markets ultimately must be inversely related to the number of actors competing for the same
gains. This principal observation underlies the concept of density dependence in organizational
ecology (Hannan and Freeman, 1977). Only so many fish are in the sea, after all. And yet both
fishing boat captains and organizational leaders watch and mimic one another all the time. How
then should we make sense of such behavior when it leads to excessive clustering in the market?
In addition to being a response to uncertainty (DiMaggio and Powell, 1983), the most prominent
answer to emerge in the sociological literature centers on the notion of legitimacy. The need for
legitimacy is so strong a force that organizations must strive to be legitimate before ever
expecting to be profitable. If typicality indeed constitutes a kind of legitimacy then the results of
this study offer an avenue by which to situate in a probabilistic frame concerns about
organizational legitimacy and the strategic choice to conform or not conform. Should
management anticipate performance around the average point for their industry, for instance,
44
they may be better off building an organization that conforms to the tendencies of other
organizations in the market. Should they aim for greater volatility—generating performance or
returns outside that band—they may be better off managing an atypical organization.
Measuring Organizational Typicality
The measurement strategies developed for this study may be usefully employed elsewhere.
Using as a starting point several insights from research on the social construction and
consequences of categories in markets, the measure of typicality introduced here offers a general
and readily replicable framework. Though the style categories in the hedge fund industry are
relatively discrete, it is likely that not all action in markets is organized in terms of nested, well-
known categorical sets. Often, the structural properties economic agents employ are better
understood as webs of sameness and difference (Rosch, 1973). These webs, as well as the central
tendencies within them (Miller and Chen, 1996), can change over time. As a result, whereas
basing a measurement of identity on specific boundary or category spanning may make sense in
the short run or in the most static environments, employing a more fluid measurement like the
one used here may be more appropriate when studying the effects of categorical identities in
industries over longer periods of time.
Having said that, the measure of typicality used here has several limitations of its own. Besides
the implicit weighting in the measurement—deviation from elements more common contribute
more to the overall level of atypicality—I do not account for or explore alternative weighting
options. For instance, the most relevant reference point in hedge fund style categories may in fact
be a function of the average, whereby larger or more successful funds are preferentially
45
weighted. Alternatively, investors may converge on a single, exemplar fund against which to
compare a given, focal fund. The results do offer empirical support for the measurement used,
but further studies might investigate how variations in the assessment of typicality, and
organizational identity more generally, affect the primary outcome.
Decision Making, Performance, and Incentives in the Context of Identity
By focusing on audiences' evaluations, this study also adds a distinctive sociological voice to
recent research on behavior and decision making in behavioral economics. First, the results
extend prior findings by adding an important nuance to the nonlinearity between performance
and evaluation. The conventional interpretation of this nonlinearity is that audiences respond
more to positive information than they do to negative information. My results indicate that it is
particularly true of audiences for atypical organizations. In the case of typical organizations, the
relationship between performance and evaluation more closely approximates a linear one. The
fact that the nonlinearity is concentrated among atypical organizations suggests a distinct
competitive advantage for organizational nonconformity.
Finally, I believe a focus on organizational incentives may shed additional light on the empirical
results. One of the hedge fund managers with whom I spoke shared an office colloquialism
"Negative Upside"—a term he and his partners used to describe pension fund managers’
dilemmas in choosing to allocate capital. Investing in either typical or atypical funds can produce
positive returns—results showed no effect of typicality on returns—but investing in atypical
funds is more likely to cost the pension manager his or her job should things go wrong.
Moreover, excessive positive returns often have only a marginal impact on a pension manager’s
46
total compensation. Given these considerations, pension managers should be more likely to
invest in typical funds. Understanding the conditions under which a certain type of audience
member—one that, like pension fund managers, has an incentive to endorse the standard
offering—chooses to endorse an atypical thing, then, should provide additional insight into how
people use organizational and categorical identity in their decision-making processes.
47
REFERENCES
Abdellaoui, M., H. Bleichrodt, and C. Paraschiv
2007 "Loss aversion under prospect theory: A parameter-free measurement." Management
Science, 53: 1659-1674.
Ackermann, C., R. McEnally, and D. Ravenscraft
1999 "The performance of hedge funds: Risk, return, and incentives." Journal of Finance, 54:
844-874.
Agarwal, V., N. Daniel, and N. Naik
2007 “Role of managerial incentives and discretion in hedge fund performance.” Working
paper, Georgia State University, Purdue University, and London Business School.
Albert, S., and D. A. Whetten
1985 "Organizational identity." In L.L. Cummings and B.M. Staw (eds.) Research in
Organizational Behavior, 7: 263-295. Greenwich, CT: JAI Press.
Ashforth, B. E., and R. H. Humphrey
1997 "The ubiquity and potency of labeling organizations." Organization Science, 8: 43-58.
Barth, F.
1966 “Models of social organization.” Occasional Paper No. 23, Royal Anthropological
Institute of Great Britain and Ireland, London.
Berk, J. B., and R. C. Green
2004 "Mutual fund flows and performance in rational markets." Journal of Political Economy,
112: 1269-1295.
Bothner, M., E. B. Smith, and H. White
48
2010 “A model of robust positions in social networks.” American Journal of Sociology, 116:
943-992
Briscoe, F., and S. Safford
2008 "The Nixon-in-China effect: Activism, imitation, and the institutionalization of
contentious practices." Administrative Science Quarterly, 53: 460-491.
Brunnermeier, M. K.
2001 Asset Pricing under Asymmetric Information: Bubbles, Crashes, Technical Analysis and
Herding. Oxford: Oxford University Press.
Burt, R. S.
1998 "The gender of social capital." Rationality and Society, 10: 5-46.
Capocci, D., and G. Hubner
2004 "Analysis of hedge fund performance." Journal of Empirical Finance, 11: 55-89.
Carroll, G. R., and A. Swaminathan
2000 "Why the microbrewery movement? Organizational dynamics of resource partitioning in
the U.S. brewing industry." American Journal of Sociology, 106: 715-762.
Chen, M., and D. C. Hambrick
1995 "Speed, stealth and selective attack: How small firms differ from large firms in
competitive behavior." Academy of Management Journal, 38: 453-483.
Chen, M., and I. MacMillian
1992 "Nonresponse and delayed response to competitive moves: The roles of competitor
dependence and irreversibility." Academy of Management Journal, 35: 539-570.
Chevalier, J. A., and G. Ellison
49
1997 "Risk taking by mutual funds as a response to incentives." Journal of Political Economy,
105: 1167-1200.
Cialdini, R. B.
2001 Influence: Science and Practice. Boston: Allyn & Bacon.
Corley, K. G., and D. A. Gioia
2004 "Identity ambiguity and change in the wake of a corporate spin-off." Administrative
Science Quarterly, 49: 173-208.
Dice, L. R.
1945 "Measures of the amount of ecologic association between species." Ecology, 26: 297-
302.
DiMaggio, P. J., and W. W. Powell
1983 "The iron cage revisited: Institutional isomorphism and collective rationality in
organizational fields." American Sociological Review, 48: 147-160.
Dutton, J. E., and J. M., Dukerich
1991 "Keeping an eye on the mirror: Image and identity in organizational adaptation."
Academy of Management Journal, 34: 517-554.
Elsbach, K. D., and R. M. Kramer
1996 "Members' responses to organizational identity threats: Encountering and countering the
Business Week rankings." Administrative Science Quarterly, 41: 422-476.
Fiske, S. T., and S. E. Taylor,
1991 Social Cognition (2nd ed.) New York: McGraw Hill.
Fruchterman, T. M. J., and E. M. Reingold,
50
1991 "Graph drawing by force-directed placement." Software - Practice and Experience, 21:
1129-1164.
Fung, W., and D.A. Hsieh
2000 "Performance characteristics of hedge funds and commodity funds: Natural vs. spurious
biases." Journal of Quantitative Financial Analysis, 35: 291-307.
Fung, W., D. A. Hsieh, N. Naik, and T. Ramadorai
2008 "Hedge funds: Performance, risk, and capital formation." Journal of Finance, 63: 1777-
1803.
Getmansky, M., A. W. Lo, and I. Makarov
2004 "An econometric model of serial correlation and illiquidity in hedge fund returns."
Journal of Financial Economics, 74: 529-610.
Gioia, D. A., and J. B. Thomas
1996 "Identity, image, and issue interpretation: Sensemaking during strategic change in
academia." Administrative Science Quarterly, 41: 370-403.
Glynn, M. A., P. Barr, and M. T. Dacin
2000 "Pluralism and the problem of variety." Academy of Management Review, 25: 726-734.
Hannan, M. T.
2010 “Partiality of memberships in categories and audiences.” Annual Review of Sociology,
36: 159-181.
Hannan, M. T., and G. R. Carroll
1992 Dynamics of Organizational Populations: Density, Legitimation, and Competition. New
York: Oxford University Press.
Hannan, M. T., and J. Freeman
51
1977 "The population ecology of organizations." American Journal of Sociology, 82: 929-964.
Hardie, I., and D. MacKenzie
2007 "Assembling an economic actor: The agencement of a hedge fund." Sociological Review,
55: 57-80.
Henderson, A. D.
1999 "Firm strategy and age dependence: A contingent view of the liabilities of newness,
adolescence, and obsolescence." Administrative Science Quarterly, 44: 281-314.
Homans, G.
1961 Social Behavior: Its Elementary Forms. London: Routledge and Kegan Paul.
Hsu, G.
2006 "Jacks of all trades and masters of none: Audiences' reactions to spanning genres in
feature film production." Administrative Science Quarterly, 51: 420-450.
Hsu, G., and M. T. Hannan
2005 "Identities, genres, and organizational forms." Organization Science, 16: 474-490.
Kahneman, D. J. Knetsch, and R. Thaler
1990 “Experimental tests of the endowment effect and the Coase theorem." Journal of Political
Economy, 98: 1325-1348.
Klebanov, M. M.
2008 “Betas, characteristics and the cross-section of hedge fund returns.” Working paper,
Booth School of Business, University of Chicago.
Kovacs, B., and M. Hannan
2010 "The consequences of category spanning depend on contrast." In G. Hsu, G. Negro, and
O. Kocak (eds.) Research in the Sociology of Organizations, 31: 175-201. Bingley, UK:
Emerald Books.
52
Lawson, D., and D. Peterson
2008 "Do hedge funds arbitrage market anomalies?" Working paper, Department of Finance,
Florida State University.
Leclerc, F., C. Hsee, and J. Nunes
2005 "Narrow focusing: Why the relative position of a good in its category matters more than
it should." Marketing Science, 24: 194-205.
Leifer, E., and H. White
1987 "A Structural Approach to Markets." In M Mizruchi and M Schwartz (eds.).
Intercorporate Relations: The Structural Analysis of Business: 85-108. Cambridge:
Cambridge University Press.
Liang, B.
2000 "Hedge funds: The living and the dead." Journal of Financial and Quantitative Analysis,
35: 309-325.
Lieberman, M., and D. Montgomery
1988 "First mover advantage." Strategic Management Journal, 9: 441-458.
Lo, A. W.
2008 Hedge Funds: An Analytic Perspective. Princeton, N.J.: Princeton University Press.
Lounsbury, M., and H. Rao
2004 "Sources of durability and change in market classifications: A study of the reconstitution
of product categories in the American mutual fund industry, 1977-1985." Social Forces,
82: 969-99.
Meyer, J. W., and B. Rowan
53
1977 "Institutionalized organizations: Formal structure as myth and ceremony." American
Journal of Sociology, 83: 340-363.
Miles, R. E., and C. C. Snow
1978 Organizational Strategy, Structure, and Process. New York: McGraw-Hill.
Miller, D., and M. Chen
1996 "Nonconformity in competitive repertoires: A sociological view of markets." Social
Forces, 74: 1209-1234.
Negro, G., M. Hannan, and H. Rao
2010 "Categorical contrast and audience appeal: Niche width and critical success in
winemaking." Industrial and Corporate Change, 19: 1397-1425.
Phillips, D., and E. Zuckerman
2001 "Middle-status conformity: Theoretical restatement and empirical demonstration in two
markets." American Journal of Sociology, 107: 379-429.
Podolny, J. M.
1993 "A status-based model of market competition." American Journal of Sociology, 98: 829-
872.
2001 "Networks as pipes and prisms of the market.” American Journal of Sociology, 107: 33-
60.
Podolny, J. M., and T. Stuart
1995 "A role-based ecology of technological change." American Journal of Sociology, 100:
1224-1260.
Podolny, J. M., T. Stuart, and M. Hannan
1996 "Networks, knowledge, and niches: Competition in the worldwide semiconductor
industry, 1984-1991." American Journal of Sociology, 102: 659-689.
54
Pontikes, E. G.
2008 “Fitting in or starting new? An analysis of invention, constraint, and the emergence of
new categories in the software industry.” Working paper, Booth School of Business,
University of Chicago.
Porter, M. E.
1985 Competitive Advantage. Free Press.
Rao, H., H. R. Greve, and G. F. Davis
2001 "Fool's gold: Social proof in the initiation and abandonment of coverage by Wall Street
analysts." Administrative Science Quarterly, 46: 502-526.
Rao, H., P. Monin, and R. Durand
2003 "Institutional change in Toqueville: Nouvelle cuisine as an identity movement in French
gastronomy." American Journal of Sociology, 108: 795-843.
Rosch, E.
1973 "Natural categories." Cognitive Psychology, 4: 328-350.
Ruef, M., and K. Patterson
2009 "Credit and classification: The impact of industry boundaries in nineteenth century
America." Administrative Science Quarterly, 54: 486-520.
Sensoy, B. A.
2009 "Performance evaluation and self-designated benchmark indexes in the mutual fund
industry." Journal of Financial Economics, 92: 25-39.
Sirri, E., and P. Tufano
1998 "Costly search and mutual fund flows." Journal of Finance, 53: 1589-1622.
55
Staw, B. M.
1976 "Knee deep in the Big Muddy." Organizational Behavior and Human Decision Process,
35: 124-140.
1997 "The escalation of commitment: An update and appraisal." In Z. Shapira (ed.),
Organizational Decision Making: 191-215. New York: Cambridge University Press.
Stryker, R.
1994 "Rules, resources, and legitimacy processes: Some implications for social conflict, order,
and change." American Journal of Sociology, 99: 847-910.
Stuart, T., H. Hoang, and R. C. Hybels
1999 "Interorganizational endorsements and the performance of entrepreneurial ventures."
Administrative Science Quarterly, 44: 315-349.
Suchman, M. C.
1995 "Managing legitimacy: Strategic and institutional approaches." Academy of Management
Review, 20: 571-610.
Tajfel, H., and J. C. Turner
1986 "The social identity theory of inter-group behavior. " In S. Worchel and W. Austin (eds.)
Psychological of Intergroup Relations: 7-24. Chicago: Nelson-Hall.
Thomas, G. M., H. Walker, and M. Zelditch
1986 "Legitimacy and collective action." Social Forces, 65: 378-404.
Weick, K. E.
1995 Sensemaking in organizations. Thousand Oaks, CA: Sage.
Whetten, D. A.
2006 "Albert and Whetten revisited: Strengthening the concept of organizational identity."
Journal of Management Inquiry, 15: 219-234.
56
White, H.
1981 "Where do markets come from?" American Journal of Sociology, 87: .
2008 Identity and Control: A Structural Theory of Social Action. Princeton, NJ: Princeton
University Press.
Zuckerman, E. W.
1999 "The categorical imperative: Securities analysts and the illegitimacy discount." American
Journal of Sociology, 104: 1398-1438.
2004 “Structural incoherence and stock market volatility.” American Sociological Review, 69:
405-432.
57
Figure 1. Cluster plot of the TASS hedge fund database, 1995, with funds positioned
relative to one another according to similarity in strategic focus and investor approach
attributes.*
* Distinct clustering is visible for several styles: Long/Short Equity (dark shade in middle right), Fixed Income
Arbitrage (light shade in lower middle), Global Macro (mid-shade in lower left), Managed Futures (mid-light shade
in upper left), and Convertible Arbitrage (small cluster of black shaded nodes above LSE cluster).
58
Figure 2. MDS plot of the 2005 TASS hedge fund database, by primary investment style.*
* Nodes are primary investment styles. The number of individual funds is indicated within the node. Node size
represents the amount of heterogeneity among funds in the style category with respect to their underlying strategic
focus and investment approach attributes.
59
Figure 3. Relationship between prior quarterly returns and capital flows, using results
from model 3.*
* Trajectories are not adjusted for the main-effect parameter estimate of typicality, only the
parameter estimate of returns that is based on both the main returns effect and the interaction
between returns and typicality.
60
Table 1
Definitions from TASS of Primary Investment Styles (Definition of Fund of Funds from
Barclay's Hedge)
Style and definition
Convertible Arbitrage funds aim to profit from the purchase of convertible securities and
the subsequent shorting of the corresponding stock when a pricing error is made in the
conversion factor of the security. Managers typically build long positions of convertible and
other equity hybrid securities and then hedge the equity component of the long securities
positions by shorting the underlying stock or options. The number of shares sold short usually
reflects a delta neutral or market neutral ratio. As a result, under normal market conditions,
the arbitrageur generally expects the combined position to be insensitive to fluctuations in the
price of the underlying stock.
Dedicated Short Bias funds take more short positions than long positions and earn returns
by maintaining net short exposure in long and short equities. Detailed individual company
research typically forms the core alpha generation driver of dedicated short bias managers,
and a focus on companies with weak cash flow generation is common. To [facilitate] the
short sale, the manager borrows the stock from a counterparty and sells it in the market. Short
positions are sometimes implemented by selling forward. Risk management consists of
offsetting long positions and stop-loss strategies.
Emerging Markets funds invest in currencies, debt instruments, equities, and other
instruments of countries with “emerging” or developing markets (typically measured by GDP
per capita). Such countries are considered to be in a transitional phase between developing
and developed status. Examples of emerging markets include China, India, Latin America,
much of Southeast Asia, parts of Eastern Europe, and parts of Africa. There are a number of
sub-sectors, including arbitrage, credit and event driven, fixed income bias, and equity bias.
Equity Market Neutral funds take both long and short positions in stocks while minimizing
exposure to the systematic risk of the market (i.e., a beta of zero is desired). Funds seek to
exploit investment opportunities unique to a specific group of stocks while maintaining a
neutral exposure to broad groups of stocks defined, for example, by sector, industry, market
capitalization, country, or region. There are a number of sub-sectors, including statistical
arbitrage, quantitative long/short, fundamental long/short, and index arbitrage. Managers
often apply leverage to enhance returns.
Event Driven funds invest in various asset classes and seek to profit from potential
mispricing of securities related to a specific corporate or market event. Such events can
include mergers, bankruptcies, financial or operational stress, restructurings, asset sales,
recapitalizations, spin-offs, litigation, regulatory and legislative changes, as well as other
types of corporate events. Event Driven funds can invest in equities, fixed income
instruments (investment grade, high yield, bank debt, convertible debt and distressed),
options, and various other derivatives. Many managers use a combination of strategies and
adjust exposures based on the opportunity sets in each sub-sector.
61
A Fund of Funds hedge fund is an investment vehicle whose portfolio consists of shares in a
number of hedge funds. How the underlying hedge funds are chosen can vary. A fund of
hedge funds may invest only in hedge funds using a particular management strategy. Or a
fund of hedge funds may invest in hedge funds using many different strategies in an attempt
to gain exposure to all of them. The benefit of owning any fund of fund is experienced
management and diversification. A portfolio manager uses his or her experience and skill to
select the best underlying funds based on past performance and other factors.
Fixed Income Arbitrage funds attempt to generate profits by exploiting inefficiencies and
price anomalies between related fixed income securities. Funds limit volatility by hedging out
exposure to the market and interest rate risk. Strategies include leveraging long and short
positions in similar fixed income securities that are related either mathematically or
economically. The sector includes credit-yield-curve relative-value-trading involving interest
rate swaps, government securities and futures; volatility trading involving options; and
mortgage-backed securities arbitrage (the mortgage-backed market is primarily U.S.-based
and over the counter).
Global Macro funds focus on identifying extreme price valuations, and leverage is often
applied on the anticipated price movements in equity, currency, interest rate, and commodity
markets. Managers typically employ a top-down global approach to concentrate on
forecasting how political trends and global macroeconomic events affect the valuation of
financial instruments. Profits are made by correctly anticipating price movements in global
markets and having the flexibility to use a broad investment mandate, with the ability to hold
positions in practically any market with any instrument. These approaches may be a
systematic trend following models, or discretionary.
Long/Short Equity funds invest on both long and short sides of equity markets, generally
focusing on diversifying or hedging across particular sectors, regions, or market
capitalizations. Managers have the flexibility to shift from value to growth; small to medium
to large capitalization stocks; and net long to net short. Managers can also trade equity
futures and options as well as equity-related securities and debt, or build portfolios that are
more concentrated than traditional long-only equity funds.
Managed Futures funds (often referred to as Commodity Trading Advisors or CTAs) focus
on investing in listed bond, equity, commodity futures, and currency markets, globally.
Managers tend to employ systematic trading programs that largely rely upon historical price
data and market trends. A significant amount of leverage is employed since the strategy
involves the use of futures contracts. CTAs do not have a particular bias toward being net
long or net short any particular market.
Multi-Strategy funds are characterized by their ability to allocate capital based on perceived
opportunities among several hedge fund strategies. Through the diversification of capital,
managers seek to deliver consistently positive returns regardless of the directional movement
in equity, interest rate, or currency markets. The added diversification benefits reduce the risk
profile and help smooth returns, reduce volatility, and decrease asset-class and single-strategy
risks. Strategies adopted in a multi-strategy fund may include, but are not limited to,
convertible bond arbitrage, equity long/short, statistical arbitrage, and merger arbitrage.
62
Table 2
Average of "Central Tendency" Fund Vector as Computed for All Convertible Arbitrage Funds in TASS during Q1, 1994–2008, Using
the Strategic and Investment Focus Attributes
Convertible Arbitrage
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
Arbitrage
.60
.67
.72
.74
.83
.82
.73
.78
.75
.77
.77
.73
.70
.72
.69
Bottom Up
.67
.54
.47
.46
.40
.38
.36
.35
.35
.33
.33
.39
.37
.35
.41
Contrarian
.00
.00
.00
.00
.00
.00
.06
.08
.07
.06
.06
.04
.05
.04
.06
Directional
.13
.08
.06
.06
.04
.03
.01
.02
.02
.01
.02
.03
.03
.04
.04
Discretionary
.20
.17
.16
.14
.06
.05
.09
.10
.12
.13
.14
.17
.19
.15
.10
Diversified
.20
.21
.19
.20
.12
.11
.12
.09
.07
.08
.10
.12
.09
.12
.14
Fundamental
.40
.33
.25
.26
.19
.18
.22
.26
.27
.27
.30
.36
.36
.32
.43
Long Bias
.20
.12
.09
.11
.12
.11
.10
.10
.12
.13
.14
.15
.16
.16
.24
Market Neutral
.53
.54
.62
.63
.63
.57
.55
.53
.50
.46
.44
.37
.34
.34
.31
Non-Directional
.47
.58
.62
.66
.63
.59
.56
.50
.43
.41
.40
.36
.36
.34
.27
Opportunistic
.13
.21
.16
.14
.10
.10
.16
.13
.12
.11
.13
.13
.15
.13
.16
Other
.13
.12
.09
.09
.10
.08
.05
.04
.03
.03
.02
.03
.03
.04
.00
Relative Value
.33
.38
.28
.29
.37
.36
.42
.38
.38
.36
.38
.35
.34
.35
.37
Short Bias
.13
.12
.09
.14
.17
.15
.13
.10
.12
.16
.15
.15
.16
.15
.24
Systematic Quant
.20
.17
.09
.11
.12
.11
.21
.25
.22
.19
.19
.13
.11
.12
.12
Technical
.13
.12
.12
.14
.08
.07
.09
.09
.10
.10
.08
.05
.05
.06
.06
Top Down Macro
.20
.25
.22
.20
.17
.18
.18
.15
.15
.16
.15
.18
.19
.21
.35
Trend Follower
.00
.04
.03
.03
.00
.02
.01
.01
.00
.00
.00
.00
.00
.00
.00
Bankruptcy
.00
.00
.00
.00
.04
.03
.06
.05
.04
.04
.04
.03
.04
.02
.04
Capital Structure Arbitrage
.00
.00
.03
.06
.12
.13
.22
.26
.25
.26
.28
.23
.17
.13
.18
Distressed Bonds
.00
.00
.03
.03
.06
.05
.08
.08
.07
.06
.07
.06
.07
.05
.04
Distressed Markets
.13
.17
.12
.11
.08
.07
.09
.08
.08
.07
.08
.09
.07
.05
.04
Equity Derivative Arbitrage
.00
.04
.06
.06
.08
.07
.06
.11
.13
.14
.14
.15
.14
.14
.12
High Yield Bonds
.27
.25
.19
.17
.12
.11
.16
.16
.19
.17
.17
.21
.21
.15
.10
Merger Arbitrage Risk Arbitrage
.60
.58
.50
.49
.50
.41
.38
.29
.26
.23
.20
.19
.16
.15
.16
Mortgage Backed Securities
.07
.08
.06
.06
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
Multi-Strategy
.00
.00
.00
.00
.00
.00
.04
.03
.07
.07
.08
.09
.08
.05
.00
Pairs Trading
.00
.00
.00
.00
.00
.03
.05
.05
.03
.03
.03
.02
.02
.02
.02
Regulation D
.00
.00
.00
.00
.00
.02
.01
.01
.02
.02
.02
.03
.03
.01
.00
Share Holder Activist
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
Socially Responsible
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
.00
63
Special Situations
.20
.21
.19
.20
.15
.13
.14
.11
.12
.14
.14
.13
.13
.13
.12
Statistical Arbitrage
.00
.04
.03
.03
.00
.02
.01
.01
.01
.01
.01
.01
.02
.02
.04
64
Table 3
Cross Section of Fund Typicality by Style (Year = 2000)
Style
Funds
Mean
typicality
S.D.
Min.
Max.
Convertible Arbitrage
56
0.298
0.088
0.022
0.435
Dedicated Short Bias
16
0.290
0.247
0
0.578
Emerging Markets
200
0.281
0.100
0
0.438
Equity Market Neutral
170
0.301
0.102
0
0.456
Event Driven
200
0.335
0.125
0
0.504
Fixed Income Arbitrage
137
0.286
0.106
0
0.447
Fund of Funds
922
0.277
0.122
0
0.453
Global Macro
151
0.309
0.127
0
0.497
Long/Short Equity
978
0.331
0.130
0
0.527
Managed Futures
230
0.348
0.132
0
0.510
Multi-Strategy
321
0.272
0.098
0
0.439
65
Table 4
Descriptive Statistics and
Correlations
Variable
Mean
S.D.
1
2
3
4
5
6
7
8
9
10
11
12
1. Capital flow
3.82
22.45
2. Age
3.68
2.90
-.19
3. Log assets
17.17
1.94
-.05
.30
4. Management fee
1.45
0.63
.00
-.02
-.03
5. Incentive fee
16.77
6.77
.01
-.03
-.01
.06
6. Lockup period
3.40
6.43
.04
-.02
.03
-.07
.11
7. Redemption period
36.77
26.87
.06
.00
.13
-.12
-.07
.29
8. Number in family
3.80
5.82
-.02
.00
.12
.05
.12
-.06
-.07
9. Rate of return, 12
0.81
1.57
.23
-.08
-.01
.01
.08
.06
.04
-.01
10. Volatility, 12
3.01
2.85
-.03
-.03
-.18
.08
.16
.02
-.16
-.02
.16
11. (+) Return qtr.
3.46
4.58
.17
-.04
-.05
.02
.10
.05
-.01
-.01
.47
.28
12. (-) Return qtr.
-1.43
3.57
.15
-.03
.06
-.03
-.03
.01
.07
.00
.34
-.37
.30
13. Typicality
0.34
0.13
.02
.01
.00
.00
.16
.02
-.07
.04
.05
.13
.06
-.03
66
Table 5
Robust OLS Regression Predicting Capital Flows at t+1*
Variable
1
2
3
4
5
6
7
Capital flow
0.266
0.185•••
(0.005)
(0.005)
Age
-1.247•••
-1.260•••
-1.261•••
-0.547•••
-1.061•••
-1.099•••
-1.272•••
(0.038)
(0.038)
(0.038)
(0.030)
(0.307)
(0.028)
(0.039)
Log assets
-0.129••
-0.131••
-0.132••
-0.415•••
-4.153•••
-0.040
-0.131••
(0.052)
(0.052)
(0.052)
(0.043)
(0.121)
(0.041)
(0.052)
Management fee
0.468••
0.463••
0.458••
0.239
0.386•••
0.457••
(0.190)
(0.188)
(0.188)
(0.149)
(0.122)
(0.188)
Incentive fee
-0.041••
-0.047••
-0.048••
-0.048•••
-0.055•••
-0.048••
(0.019)
(0.019)
(0.019)
(0.015)
(0.013)
(0.019)
Lockup period
0.024
0.019
0.019
0.015
0.007
0.020
(0.017)
(0.017)
(0.017)
(0.013)
(0.012)
(0.017)
Redemption period
0.035•••
0.034•••
0.034•••
0.026•••
0.032•••
0.033•••
(0.004)
(0.004)
(0.004)
(0.003)
(0.003)
(0.004)
Number in family
-0.046••
-0.045••
-0.045••
-0.030•
-0.046•••
-0.045••
(0.023)
(0.023)
(0.023)
(0.018)
(0.013)
(0.023)
Rate of return, 12
3.198•••
2.532•••
2.530•••
1.805•••
1.797•••
2.282•••
2.527•••
(0.094)
(0.092)
(0.092)
(0.082)
(0.073)
(0.082)
(0.092)
Volatility, 12
-0.473•••
-0.552•••
-0.552•••
-0.368•••
-0.454•••
-0.531•••
-0.550•••
(0.042)
(0.046)
(0.046)
(0.039)
(0.058)
(0.042)
(0.046)
(+) Return qtr.
0.447•••
0.712•••
0.654•••
0.557•••
0.528•••
0.721•••
(0.023)
(0.065)
(0.061)
(0.056)
(0.021)
(0.063)
(-) Return qtr.
0.122•••
-0.114
-0.072
-0.057
0.137•••
-0.054
(0.027)
(0.070)
(0.064)
(0.068)
(0.028)
(0.068)
Typicality†
2.485•••
6.064•••
5.223•••
16.527•••
5.302•••
(0.812)
(1.048)
(0.901)
(4.386)
(0.979)
(+) Return qtr. × Typicality†
-0.731•••
-0.605•••
-0.544•••
-0.724•••
(0.164)
(0.152)
(0.140)
(0.152)
(-) Return qtr. × Typicality†
0.668•••
0.628•••
0.592•••
0.480•••
(0.177)
(0.161)
(0.171)
(0.164)
(+) Return qtr., t-1
0.389•••
(0.050)
67
(-) Return qtr., t-1
0.123•
(0.070)
Typicality, t-1
3.829•••
(0.803)
(+) Return qtr., t-1 × Typicality, t-1
-0.280••
(0.124)
(-) Return qtr., t-1 × Typicality, t-1
0.343••
(0.174)
N, observations
92,226
92,226
92,226
83,245
83,245
84,446
92,226
N, funds
6562
6562
6562
6353
6353
6367
6562
Quarterly fixed effect
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Style fixed effect
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Fund fixed effect
No
No
No
No
Yes
No
No
R2
0.107
0.113
0.114
0.170
0.277
0.118
0.113
• p < .10; •• p < .05; ••• p < .01.
* Heteroskedasticity-consistent robust standard errors are in parentheses, clustered by fund.
† In model 7, typicality is fixed as the point of fund entry, which affects the typicality term and the two interactions involving typicality.
68
Table 6
Robust OLS Regression Testing the Duration of the Buffering Effect*
3
8
9
Variable
All funds†
Two qtr. (-)
Returns†
Three qtr. (-)
Returns†
(+) Return qtr.
0.712•••
(0.065)
(-) Return qtr.
-0.114
0.028
0.298
(0.070)
(0.119)
(0.189)
Typicality
6.064•••
4.574•
1.779
(1.048)
(2.532)
(4.632)
(+) Return qtr. × Typicality
-0.731•••
(0.164)
(-) Return qtr. × Typicality
0.668•••
0.683••
0.508
(0.177)
(0.302)
(0.496)
N, observations
92,226
8,783
3,116
N, funds
6562
3634
1729
Quarterly fixed effect
Yes
Yes
Yes
Style fixed effect
Yes
Yes
Yes
R2
0.114
0.078
0.094
• p < .10; •• p < .05; ••• p < .01.
* Heteroskedasticity-consistent robust standard errors are in parentheses, clustered by fund.
† Prior controls from models in table 5 are included in the models but omitted from the table.
69
Table 7
Robust OLS Regression Testing the Duration of the Buffering Effect*
10
11
12
13
Variable
LSE†
MF†
FOF†
MS†
(+) Return qtr.
0.718•••
0.718•••
0.474•••
0.524••
(0.088)
(0.199)
(0.130)
(0.219)
(-) Return qtr.
-0.013
-0.181
-0.254•
-0.219
(0.093)
(0.225)
(0.139)
(0.339)
Typicality
7.029•••
4.479
2.833•
-1.815
(1.425)
(3.191)
(1.638)
(4.073)
(+) Return qtr. × Typicality
-0.921•••
-0.873••
0.131
1.214•
(0.209)
(0.432)
(0.367)
(0.736)
(-) Return qtr. × Typicality
0.500••
0.825•
0.537
0.484
(0.225)
(0.499)
(0.396)
(1.011)
N, observations
29,140
7,417
18,793
5,809
N, funds
2072
508
1465
362
Quarterly fixed effect
Yes
Yes
Yes
Yes
R2
0.146
0.092
0.099
0.122
• p < .10; •• p < .05; ••• p < .01.
* Heteroskedasticity-consistent robust standard errors are in parentheses, clustered by fund.
† Prior controls from models in table 5 are included in the models but omitted from the table.
70
1 An anonymous reviewer pointed out the similarities between this conceptualization of identity
and corporate strategy more generally (e.g., Miles and Snow, 1978). Henderson (1999), for
instance, classified technological firms into "standards-based" (conformists) and "proprietary"
(nonconformists) groups. Several network studies of organizations have adopted similar
treatments (e.g., Podolny and Stuart, 1995; Podolny, Stuart, and Hannan, 1996).
2 This particular feature of the hedge fund industry—relatively fixed organizational attributes
over time—need not hold for the measure of typicality to be applied to other settings.
3 A reviewer noted the importance of controlling for total net assets to account for possible floor
or ceiling effects, in addition to capturing an important fund characteristic. If, for instance,
atypical funds are generally smaller, then greater inflows following positive performance might
indicate more room to grow among atypical funds, less room to grow among comparably
performing typical funds, or a combination of the two. On the opposite side of the performance
spectrum, atypical funds might actually benefit from a size-based floor effect: if a small, atypical
fund performs poorly, it may be difficult for investors to fault it more than before the failure was
observed. In addition to using the single control for fund size, I also ran models (not shown but
available from the author) including an interaction between fund size and fund typicality. The
coefficient on the interaction term was far from significant, thus indicating that the results
described in the next section are not the result of size-based floor or ceiling effects.
4 Positive first-order autocorrelation is present in the data, as expected. I tested for
autocorrelation using the Durbin-Watson statistic ( = 1.22) for models not including the lagged
dependent variable. For models with a lagged dependent variable, I also assessed autocorrelation
using a Wooldridge test for autocorrelation in panel data (Durbin-Watson is an invalid measure
of autocorrelation because including a lagged dependent variable constitutes a strong violation of
the exogeneity assumption). Because the presence of positive first-order autocorrelation when
left uncorrected may downwardly bias estimated standard errors, I ran a series of Prais-Winsten
AR(1) and Cochrane-Orcutt AR(1) models, in addition to model 4 in the text, to account for first-
order autocorrelation. Results of these additional models were entirely consistent with those in
the text. Models and details are available from the author.
5 This result is not intended to imply that any sort of “imprinting”—for example, that the
typicality of a fund at the point of entry is imprinted on the minds of investors and remains
constant even as a fund’s actual level of typicality changes—is occurring in the hedge fund
industry. Because there is little within-fund variation with respect to typicality, it is hard to know
to what extent imprinting occurs in this context, if at all.
