Download The Role of Organizational Structure

The Role of Organizational Structure:
Between Hierarchy and Specialization
Massimo Massa *
Lei Zhang **
* Department of Finance, INSEAD, Boulevard de Constance, 77300 Fontainebleau France, Tel: +33-160724481, Email: [email protected]
** Division of Banking and Finance, Nanyang Business School, 50 Nanyang Avenue, 639798 Singapore.
Tel: +65-81824202, Email: [email protected]
The Role of Organizational Structure:
Between Hierarchy and Specialization
Abstract:
We study how organizational complexity affects managerial behavior, breaking it down into vertical hierarchy and
degree of specialization. We exploit a novel dataset on the organizational structure of the US asset management
industry. We show that more hierarchical structures reduce the incentive to collect soft information by investing less in
closely-located firms. This reduces portfolio concentration and performance and increases herding. In contrast, a higher
degree of specializations and variety of competences makes asset managers provide more unique trading strategies and
herd less with each other. Changes in fund structure quickly find their way into the behavior of fund managers.
JEL classification: G23, G30, G32
Keywords: organizational structure, hierarchy, specialization, mutual funds, performance, herding,
proximity investment.
1
The Role of Organizational Structure:
Between Hierarchy and Specialization
Abstract:
We study how organizational complexity affects managerial behavior, breaking it down into vertical hierarchy and
degree of specialization. We exploit a novel dataset on the organizational structure of the US asset management
industry. We show that more hierarchical structures reduce the incentive to collect soft information by investing less in
closely-located firms. This reduces portfolio concentration and performance and increases herding. In contrast, a higher
degree of specializations and variety of competences makes asset managers provide more unique trading strategies and
herd less with each other. Changes in fund structure quickly find their way into the behavior of fund managers.
JEL classification: G23, G30, G32
Keywords: organizational structure, hierarchy, specialization, mutual funds, performance, herding,
proximity investment.
Introduction
The literature on the theory of organizations has extensively studied the way the organizational
structure affects the behavior of the members of the organization. However, very little empirical
evidence exists on what part of the organizational structure matters and what is its main channel of
influence. There are two alternative intuitions: on the one hand, the vertical chain of control
provides the main differentiating effect. On the other hand, it is the degree of different specialties
among the members that mostly matters.
The vertical chain is related to and defines the level of “bureaucracy” within the organization. A
vertical structure creates a chain of command that relates the HQ to the lowest level business units,
through a series of intermediate layers that take care of executing HQ directives at a more
disaggregated level. The allocation of authority and tasks to the managers of the intermediate layers
in the organization affects both the incentive of the managers as well as the ability of the HQ to
monitor them.
At the same time, the degree of specialization matters as well. The fact that the members have
different and specialistic competences affects the incentives. For example, the fact that managers
share the same qualifications improves the “congruence” (Dessein, 2002). This should be reflected
in higher delegation and ameliorated coordination. The lower informational asymmetry may help
the decision-making process and help to generate more information and provide more expertise. On
the other hand, different qualifications may make it more difficult to reach coordination.
The fact that the degree of specialization affects the ability of HQ to monitor and interferes with
the effectiveness of the hierarchy and the fact that vertical structure and specialties are directly
related suggest that any test of their role of the organization should account for them jointly. Not
being able to control for one of the two dimensions may provide spurious results. The goal of this
paper is to use a novel dataset on vertical structure and specialties to jointly estimate their role and
impact on managerial behavior.
We focus on the asset management industry. This provides an ideal test case. Indeed, asset
management companies are organized as complex organizations in which a strong trade-off exists
between incentives and control. Funds are in general run by teams of managers1. These managers
tend to have different competencies and specialties. At the same time, managing teams are parts of a
more complex multi-layer hierarchical (“vertical”) structure with a CEO at the top, the Head of
Fixed Income below and then the Portfolio Manager(s) at the more operative level.
1
While all the funds are managed by teams of managers, they may choose to report to the market that they are solemanaged for marketing purposes.
1
A more vertical structure lends itself to better risk management by helping reduce managerial
moral hazard and lowering the incentives to take (un)necessary risk. However, a more vertical
structure, by reducing the discretion of the portfolio manager, also lowers his incentives to collect
difficult-to-transfer information (“soft information”) – i.e., the one based on direct personal
interaction with the managers of the firm (Stein, 2002). Indeed, it makes it more difficult to transfer
information and tends to bureaucratically produce the wrong kind of information. Flat structures are
better in the case of soft information. “A decentralized approach – with small, single-manager firms
– is most likely to be attractive when information about projects is “soft” and cannot be credibly
transmitted. In contrast, large hierarchies perform better when information can be costlessly
“hardened” and passed along inside the firm” (Stein, 2002). This would also reduce the tendency to
engage in proximity investment and collection of information. Given that the availability of more
information also induces the manager to tilt his portfolio towards the set of assets over which he has
more information (e.g., Kacperczyk et al., 2005), less hierarchical structures tend to hold more
concentrated portfolios.
In the asset management industry performance is positively related to the collection of soft
information on the firms located close-by (Coval and Moskowitz, 1999 and 2001, Chen et al., 2004)
and to the risk taking behavior of the manager. A more hierarchical structure, by lowering the
incentives to collect soft information and to invest in geographically close assets – the ones more
likely to provide superior performance (Chen et al., 2004) – effectively reduces performance.
Therefore, the net effect is that more hierarchical structures should deliver worse performance.
At the same time, the area of competence of the manager is critical. Indeed, a manager
specialized in currencies would have a different view of the world with respect to managers whose
core competency is just fixed income. More functional competencies improve information and help
to develop “original thinking” and “unique solutions”. However, they also make coordination and
consensus reaching more difficult (Crawford and Sobel 1982, Cremer 1993, Aghion and Tirole
1997, Dessein 2002). For example, Van den Steen (2004) argues that shared beliefs within an
organization “decrease agency problems, lower the need to monitor and facilitate coordination”.
Blau (1977), Westphal and Zajac (1995), Bourgeois, et al., (1997), Carter, et al., 2003) and Adams
and Ferreira (2005) analyze the impact of board diversity/similarity on firm value. In general, the
idea is that the main benefit of a diverse team is that “team members are able to provide different
perspectives on important issues, which may reduce the probability of complacency in decisionmaking. ... Diversity may add value by bringing different perspectives, experiences and opinions to
the table” (Adams and Ferreira, 2005).
Therefore, specialties are more likely to provide more “unique” view of the world as outcome
of the decision process among many different managers. This should make funds less likely to herd
2
with other funds. Specialties should also improve cautiousness and help reduce managerial moral
hazard and incentives to take (un)necessary risk. Uniqueness may arise from the fact that the
interaction of different (differently informed) brains may generate something very different from
the common solution as well as from the fact that the lower risk taking may simply reduce the
incentive to “herd with the pack” if this entails sizable risk. More specialties, by making the
decision process more complicated, may reduce managerial moral hazard and risk taking. This
effect is compounded by the lower incentives due to the lack of attribution of performance to the
individual managers. Overall, both effects may negatively affect performance.
Overall, these considerations suggest the existence of a relation between organizational structure
and investment strategies and performance. If the goal is just performance maximization, a flatter
structure with lower variety of specialties should be the optimal one. However, the goal is not
limited to performance maximization but also geared to risk/managerial moral hazard control. A
more hierarchical structure characterized by a variety of specialties would instead allow a better
control of managerial behavior.
The existence of an optimal degree of hierarchy/specialty within each asset management
organization is dictated by the trade-off between performance and risk management. While we
cannot observe what drives the specific optimum for each mutual fund or insurance family, we can
use family specific characteristics to explain – instrument – the level of hierarchy of the funds
belonging to the family and then relate it to observable fund policies. In particular, we will focus on
three policies: proximity investment, portfolio concentration and herding and their relation to
performance.
In this paper, we study these factors by focusing on two main players of the asset management
industry: the mutual funds and the funds managed by insurance companies (both life insurance and
property and damage insurance). We use information on portfolio composition, performance as well
as organizational structure of all the US mutual funds and insurance-managed funds investing in US
corporate bonds (the “funds”). For each fund we have information on the organizational structure of
the “entity” that manages it. We know the identity, the functions and the roles of all the members of
the management team. Each member is characterized in terms of his functional attributions (e.g.,
Chairman, CEO, CFO, Fixed Income Head, Portfolio Manager, Trader, Analyst) as well as his areas
of competence (e.g., market sector, credit sector, geographical focus). This allows us to construct a
measure of organizational structure: “hierarchy”. Hierarchy is defined in terms of the number of
layers, from the top down, in which the organization is structured. For example, a vertical structure
presided by the CEO, with a fixed income head and two portfolio managers has a hierarchy equal to
3, while a flat structure with only two portfolio managers has a hierarchy equal to 1. We also
construct a measure of employee “specialty” of the fund managers which allows us to distinguish
3
from the potential confounding effects due to heterogeneity of tasks, qualification, and competence
of functional attribution existing within the structure.
We start by focusing on the determinants of the organizational structure. We argue that the
overall objective of the financial family managing the fund (mutual fund family or insurance
family) is not limited to performance maximization, but is also geared to risk/managerial moral
hazard control. Therefore, the choice of the degree of hierarchy and specialty can be explained in
terms of the characteristics of the financial family the fund belongs to. This lets us identify the
exogenous determinants of hierarchy and specialty.
We then relate hierarchy and specialty to proximity investment. We effectively run a horse-race
between specialty and hierarchy. Hierarchy seems to be the most relevant dimension. More
hierarchical structures tend to invest less in firms located close to them. An additional layer in the
structure increases the average holdings-weighted manager-bond distance by 8%. At the same time,
hierarchy increases the tendency to herd more and to hold less concentrated portfolios. An
additional layer in hierarchy increases herding by 14% and reduces portfolio concentration by 50%.
This has a direct impact on performance: more vertical structures display worse performance. An
additional layer in hierarchy reduces the average performance by 21 basis points per month.
Specialty directly affects herding. A higher degree of specialty reduces the tendency to herd. A one
standard deviation increase in specialty reduces herding by 10%.
Overall, the organizational structure affects performance slightly more for mutual funds than
insurance firms, while it impacts proximity investment, herding and portfolio concentration more
for insurance companies than mutual funds. These findings indicate that the organizational structure
is important in determining asset management strategies as well as performance. More importantly,
they show that the most relevant dimension is related to the degree vertical hierarchy. In contrast,
specialty does not seem to be very relevant except along the herding dimension. Here, it not only
offsets the tendency to herd induced by a more vertical structure, but more than outweighs it. This
suggests that the overall herding behavior is the outcome of a trade-off between vertical structure
and degree of specialty within the organization.
It is worth noting that we will focus on bond funds. Recent findings have discovered evidence
of proximity investment also for this class of asset managers. Massa et al., (2007) document the
existence of local bias and proximity investment in bond investing by both mutual funds and
insurance companies. This happens on a scale similar if not greater than in the case of equity. In the
case of bonds, soft information is mostly about financial conditions and distress of the firm, as
opposed to movements in the government yield curve. This is confirmed as most of the performance
benefits of a flatter organizational structure – i.e., the one more related to the collection of soft
information and proximity investment – are concentrated in the investment in low quality bonds.
4
Our findings contribute to several strands of literature. First, we relate to the recent literature on
coordination (e.g., Marshack and Radner (1972). Cremer (1980), Genakoplos and Milgrom (1991)
and Vayanos (2002) study the optimal grouping of subunits into units in the presence of
interdependencies. Dessein and Santos (2006) study the trade-off between ex ante coordination,
through rules, and ex post coordination, through communication. Cremer et al., (2007) study the
limits to firm scope due to the loss of specificity in organizational languages as firm scope grows.
To study the link between organizational structure and strategies, we loosely rely on Stein’s (2002)
theory on organizations. It posits that different structures perform differently in terms of generating
information about investment projects and allocating capital to these projects. We contribute to this
literature by providing a first direct test of the role played by vertical hierarchy and degree of
specialty within the organization and showing their distinct role.
Second, we contribute to the literature that studies the choice between team and sole
management as well as the one between internal performance generation and outsourcing and the
impact of size and structure on asset management. Chen et al. (2004) study how size affects the
behavior of the funds managers and provide the first existing model of the impact of the
organizational structure on asset management. Baer et al., (2005) argue that teams have a
production technology that differs from that of single managers, while Massa et al. (2005) show
that the marketing implications of the choice between team and managers. Chen et al. (2005) and
Del Guercio et al. (2007) investigate the effects of outsourcing on the incentives and performance
of mutual funds, showing that funds managed externally significantly under-perform those run
internally. We complement these findings by focusing on the trade-off between hierarchy and
specialty. We directly relate to the literature on the organizational structure and contractual
relationship with the mutual fund industry (Kuhnen, 2004, 2009).
Our analysis is not limited to the mutual fund industry but also focuses on the insurance
industry. This is, to our knowledge, the first paper that also analyzes the strategies and performance
of the insurance companies and directly compares them to the mutual funds. Also, we focus on a
hitherto relatively unexplored area: the funds specialized in corporate bonds.
Third, we relate to the vast literature on mutual fund performance. Its goal has been to
determine whether it is possible to identify consistently overperforming funds (e.g., Brown and
Goetzmann, 1995, Elton, et al., 1996, Carhart, 1997). Our paper contributes to this literature by
presenting evidence for one of the drivers of fund performance – the fund organizational structure.
Again, our findings have bearings not only for the mutual fund industry but for the overall asset
management industry, including the insurance one.
Fourth, we relate to the literature on proximity investment (Coval and Moskowitz, 1999, 2001,
Chen et al., 2004). Our findings help to explain the positive relationship between proximity
5
investment and performance already documented in the mutual fund literature by showing that one
of the important factors that induce some funds to invest in close-by bonds is the organizational
structure of the fund.
Fifth, we relate to the literature on herding. Lakonishok et al. (1992), Grinblatt et al. (1995) and
Wermers (1999) document herding among pension fund and mutual fund managers. We contribute
to this literature in two ways. First, we provide evidence of herding for the insurance companies and
in the bond market. Second and more importantly, we show how organizational structure affects
herding for the mutual funds and asset management in general.
Finally, we relate to the literature on the economics of mutual fund families. Nanda et al.
(2004) document the positive spillover that having a ‘star’ fund provides to all the funds belonging
to the same family and the strategies played by the families to generate star funds. Khorana and
Servaes (1999) study the determinants of mutual fund stars, while Mamaysky and Spiegel (2001)
provide a first equilibrium model of the mutual fund industry, arguing that families generate funds
to allow investors to overcome their hedging needs. More recently, Guedj and Papastaikoudi (2004)
show that performance persists at the family level, especially large fund families, suggesting that
families purposefully allocate resources across funds in an unequal way, while Gaspar et al. (2004)
provide evidence of cross-fund subsidization at the family level. We contribute by explaining the
organizational structure of the fund with the characteristics of the family the fund belongs to.
Moreover, we extend the definition and analysis of “family coordination” to the families of
insurance-managed funds. This is, to our knowledge, the first paper to tackle this topic.
The rest of the paper is organized as follows. Section II discusses the data and provides
summary statistics. Section III presents the empirical results. A brief conclusion follows.
2. Data, Construction of Main Variables and Summary Statistics
The main dataset is the Lipper’s eMAXX fixed income database. eMAXX contains details of fixed
income holdings for nearly 20,000 U.S. and European insurance companies-managed funds, U.S.,
Canadian and European mutual funds, and U.S. public pension funds. It provides information on
quarterly ownership of more than 40,000 fixed-income issuers with $5.4 trillion in total fixed
income par amount from the first quarter of 1998 to the second quarter of 2005. Moreover, it has
detailed information on the structure of the fund managing entity, including the functions, roles,
competences and areas of specialty of its members. In our analysis, we will mostly focus on the
behavior of funds managed by either mutual fund families or insurance companies including life
and property insurance companies. We also include a residual category that includes public pension
funds and variable annuities. In the case of mutual funds, we focus on the fund, aggregating the
holdings of its different classes (i.e., A, B, C).
6
To ensure the consistency of the Lipper data on employee identities, we cross-check the
information with the other datasets: CRSP mutual funds database and Morningstar. CRSP and
Morningstar do report whether the fund is managed by a team or by a single manager. However, in
some cases, instead of reporting the names of the managers, team managed funds are recorded as
“Team Managed” or “Multiple Managers”. So, as a first check, we exploit the information
contained in the team-managed funds in either CRSP or Morningstar. We match Lipper with CRSP
and Morningstar using fund names. 84% of the funds with more than 1 employee in Lipper and
matched with CRSP are recorded as team-managed in CRSP, while 86% of the funds with more
than 1 employee in Lipper and matched with Morningstar are recorded as team-managed in
Morningstar.
Then, we consider the cases in which CRSP and Morningstar do actually report the names. Also
in this case, the names of the managers are consistent. Almost all the funds reported in CRSP
(Morningstar) as sole-managed have one employee in Lipper. More specifically, out of the funds
recorded as sole-managed in CRSP, 88% report only 1 employee in Lipper, while in the case of
funds recorded as single-managed in Morningstar, 89% report only 1 employee in Lipper. However,
only 51% (50%) of the team-managed funds in CRSP (Morningstar) have more than one employee
in Lipper. Considering that funds tend to under-report rather than over-report employee identities –
it is difficult to create fake names – to ensure data accuracy we proceed as follows to construct our
final sample. We will consider all the funds in Lipper with more than one employee and, for the
funds with only one employee, we will only select those matched with CRSP and recorded as
single-managed. We will refer it as our main sample. We will also perform robustness checks on the
subsample restricted to the funds having more than one employee. We will call it the “teammanaged subsample”.
We now define our proxy for the degree of hierarchy. It is constructed as the distinct number of
layers of the structure of the fund. We consider 6 layers. The first layer is made of the Chairman,
President and CEO; the second layer is made of the CFO and CIO; the third layer is made of the
Bond Department Head and the Fixed Income Head; the fourth layer is made of the portfolio
managers (Portfolio Manager General, Portfolio Manager Balanced, Portfolio Manager
Convertibles, Portfolio Manager Equity/Preferred Stocks, Portfolio Manager High Yield, Portfolio
Manager Investment Grade Corporate Bond, Portfolio Manager Private Placements, Portfolio
Manager Short-Term/ Money Market). The fifth layer is made of Traders. The sixth layer is made
of Research Analysts. A person may have multiple job functions. Therefore, if the distinct number
of hierarchies exceeds the number of employees, we use the number of employees as vertical layer.
7
Some descriptive statistics are reported in Table I, Panel A1. We report the number of
observations by the level of fund hierarchy. Hierarchy ranges between 1 and 42. In the main sample,
60.1% are mutual funds, 34.3% are insurance-owned funds and 5.6% are funds owned by other
institutions (annuities and pension funds). For mutual funds, the average number of fund hierarchy
is 1.308, which corresponds to 72.8% one-layer funds, 23.9% two-layer funds, 2.9% three-layer
funds and 0.4% four-layer funds. For insurance funds, the average number of fund hierarchy is
1.764, which corresponds to 39.8% one-layer fund, 45.9% two-layer funds, 12.3% three-layer funds
and 1.9% four-layer funds. It appears that in the main sample mutual funds have much lower
hierarchies than insurance-owned funds. The reason is that the sole-managed funds in Lipper that
can be matched with CRSP are mostly mutual funds instead of insurance-owned funds.
If we look at the team-managed subsample, from Panel A2 of Table I, for mutual funds, the
average number of fund hierarchy is 1.657, which corresponds to 42.1% one-layer funds, 50.8%
two-layer funds, 6.2% three-layer funds and 0.9% four-layer funds. For insurance-owned funds, the
average number of fund hierarchy is 1.818, which corresponds to 35.5% one-layer fund, 49.2%
two-layer funds, 13.2% three-layer funds and 2.1% four-layer funds. We can conclude that there is
considerable cross-sectional variation in the sample which allows us to explore the impact of
organizational hierarchy on fund investment behavior and fund performance.
“Employee specialty” proxies for the number of “competences” represented in the organization.
We want to separate the effect of hierarchy from the fact that funds are managed by experts in many
fields. Indeed, the degree of specialty of the asset managers may directly affect performance. A
higher degree of employee specialty affects managerial behavior. It increases the ability to beat the
peers and potentially discourages herding. Also, the presence of different areas of specialty would
make it less optimal to concentrate on few bonds and to invest just mostly in closely located assets.
At the same time, the presence of managers specialized in different areas may also make it more
difficult to reach consensus – e.g., a high-yield expert is likely to view the markets from a
perspective very different from that of a credit-derivative expert.
We define our variable of specialty in the following way. First, we consider the areas of
competence. They are: Market Sector, Credit Sector, Geographical Focus and Job Title. Market
Sector comprises the following areas of specialty: Asset Backed Securities, Corporate Bonds,
Government Bonds, Mortgage Backed Bonds, Local/regional Bonds, US firms investing nondomestically, Combination of the above. Credit Sector comprises the following areas of specialty:
Any Corporate Sector, Any Municipal Sector, Any Specialty, Canadian Dollar, Euro, U.S. Dollar,
and Combination of Above. Geographical Focus comprises the following areas of specialty: Any
Country, Canada, Emerging Markets, Euroland, Any State/Territories, United States, Non-domestic,
2
We drop outliers such as funds with more than 4 vertical layers which represent less than 0.05% of the sample.
8
and Combination of Above. Finally, Job Title is defined as above. That is, it comprises: Chairman,
President, Chief Executive Officer, Chief Financial Officer, Chief Investment Officer, Bond
Department Head, Fixed-Income Head, Portfolio Manager-General, Portfolio Manager-Balanced,
Portfolio Manager-Convertibles, Portfolio Manager-Equity/Preferred Stock, Portfolio ManagerHigh Yield, Portfolio Manager-Investment Grade Corp Bonds, Portfolio Manager-Private
Placements, Portfolio Manager-Short-Term Bonds and Money Market. Then, we calculate the
number of total combinations of the above four sectors. Finally, we define the number of specialties
as the number of total combinations of the above four sectors. We define employee specialty as the
number of specialties divided by the number of employees. Therefore, our proxy of specialty
controls for heterogeneity in the degree of specialty, qualification and competence of functional
attribution existing within the structure.
Also for this variable, there is considerable cross-sectional variation in the sample.. Specialty
ranges between 0.5 and 6. For mutual funds, the average degree of specialty is 1.577, while for the
insurance funds, the average degree of specialty is 1.245. In the main sample, mutual funds have
higher degree of specialty than insurance-owned fundsIf we look at the team-managed subsample,
for mutual funds, the average degree of specialty is 1.382, while for the insurance funds, the
average degree of specialty is 1.241.
Next we describe our main control variables. First, we control for the effect due to the existence
of a team as opposed to a sole portfolio manager (Baer et al., 2005, Massa et al., 2005), so we
define a team dummy. This is a dummy variable equal to 1 if the fund has more than 1 employee
and 0 otherwise.
To control for portfolio turnover, we construct a measure that captures how frequently a fund
rotates its portfolio. Let us denote the set of bond issues held by fund i by Q. The turnover ratio of
fund i at quarter t is:
Q
Turnover
i ,t
=
∑
k =1
V i ,k ,t − V k ,i ,t − 1 ( 1 + R k ,t )
Q
∑
k =1
，
V k ,i ,t + V k ,i ,t − 1
2
where Rk ,t and Vi ,k ,t represent the return and the par amount of bond issue k held by investor i at
quarter t. This definition follows those commonly used to assess overall equity portfolio rotation
(Barber and Odean, 2000, Gaspar et al., 2005).
The return of the fund is constructed as the cumulative monthly fund return during each quarter.
Fund monthly return refers to the raw return from its bond portfolio. The data on bond returns are
obtained from Bloomberg. For the case of the mutual funds, we also use the CRSP Mutual Funds
return data. The results do not differ from the ones we report. Fund return volatility is defined as
9
follows. For each fund at month t, we calculate return volatility as the standard deviation of monthly
returns in the prior 20 months.
We also include variables meant to capture fund size and family size. They are the logarithm of
the par-amount of bond holdings of each fund (fund family). To alleviate the concern that part of
the investment purpose of the insurance companies is asset-liability matching, we include a variable
representing fund portfolio maturity. It is the logarithm of the value-weighted average maturity of
all the bonds held by each fund. To control for the effect of fund domicile, we include a Financial
Center Dummy. It takes a value of 1 if the fund is located at either of the following cities: New
York, Chicago, Los Angeles, Boston and San Francisco.
Finally, we also control for the fraction invested by the fund in investment-grade bonds. This is
defined as the fraction invested in bonds with S&P’s bond rating not below BBB. Moreover, we
always include fund type dummies that proxy for whether the fund belongs to a mutual fund family,
a life insurance company, a property insurance company, is a variable annuity or pension fund. The
detailed definitions of these variables can be found in the Appendix. We report the summary
statistics in Panel B of Table I. In general, funds run by mutual fund families have larger size,
higher turnover ratio, higher return volatility and lower fraction in investment-grade bonds than
insurance-owned funds.
With the measure of organizational structure and other control variables at hand, our next step is
to see the impact of fund hierarchy on portfolio strategies and fund performance. In particular, we
will look at fund proximity investing, portfolio concentration and herding. We briefly report the
summary statistics of those variables in Panel B of Table I. The detailed definitions of these
variables are given in the next section as well as in the Appendix.
3. Empirical Results
We start our analysis with the relation between organizational structure and proximity investing.
Then, we link it to portfolio concentration and herding. Finally, we focus on how the organizational
structure affects fund performance.
A. Organizational Structure and Proximity Investing
We now link structure to proximity investment. For each fund, we define its distance from the firms
whose bonds it holds ( Disi ,t ). It measures the distance between the fund and its bond portfolio. If
we denote the set of bond issues held by fund i by Q and wi , j ,t the fraction invested in bond issue j,
the fund-bond distance is:
10
Disi ,t = log[3963 * ∑ wi , j ,t arccos( sin(lati ) sin(lat j ) + cos(lati ) cos(lat j ) cos(loni − lon j ))] ,
j∈Q
where ( lati , loni ), ( lat j , lon j ) are the (latitude, longitude) for fund i and issuer j in radian degrees.3
Given the fact that in the main sample there is an imbalance toward single-layer mutual funds,
for the purpose of robustness checks we resort to a matching sample technique, where, for each
multi-hierarchy fund, we match it with some other single-hierarchy fund similar in terms of fund
type (if insurance or mutual fund) and size, but different in terms of fund structure. We construct
two matching samples, one within fund family and one across fund families. Both matching
samples are more balanced in terms of fund hierarchy. Each one of the two matching procedures
has some advantages. By matching within fund family, we have better control for the unobserved
common – presumably collected at the family level – information set of fund managers, and for the
potential coordinated behavior among funds of the same family (e.g., “cross-fund subsidization”).
Indeed, one other concern may be the fact that our analysis does not directly control for some
unobserved fund/family characteristics. However, matching within fund family may not deliver the
best match in terms of fund size, liquidity or other characteristics (Chen et. al., 2004). Matching
across fund families (within the same fund type), instead, allows us to better control for size-related
fund characteristics. We therefore jointly use these two matches as a way of guaranteeing the
robustness of our results.
The “matching within fund family” sample is constructed as follows. For each multi-hierarchy
fund, we first select another single hierarchy fund from the same fund family and most similar in
terms of fund size, and then we combine the matched single hierarchy funds with the original multihierarchy funds. The “matching across fund family” sample is constructed similarly except that the
matched single-hierarchy fund is chosen from different fund families but belonging to the same
fund type (mutual funds, insurance companies, etc.).
We start with some univariate analysis. We report the results in Table II, Panel A. We break
down the sample into 4 different levels of hierarchy: from the lowest (1 layer) to the highest (4
layers). We then report the sample mean of fund portfolio distance at different types of structures.
The number of observations appears in parenthesis. We report the results for the main sample and
the two matching samples described above. We also provide univariate tests of fund portfolio
distance regarding to single vs. multi- fund hierarchy. Multi-hierarchy (multi-specialty) means the
number of fund hierarchies (specialties) is greater than 1. The results show a monotonic increase in
fund portfolio distance as the number of layers increases. This holds regardless of the sample and
fund type. A three-layer mutual fund tends to invest in bonds of firms on average 140 (148 and 141)
3
Information on bond issuer locations is from Compustat and SDC global new issue database. Since Lipper only
provides county information of the managing firm, we use the location of the managing firm as the fund location. The
county level coordinates (latitude, longitude) are from the Gazetteer Files of Census 2000.
11
km further away than a one-layer mutual fund in the case of the main sample (sample based on
matching within family and sample based on matching across families). A three-layer insuranceowned fund tends to invest in bonds of firms on average 194 (104 and 175) km further away than a
one-layer insurance-owned fund in the case of the main sample (sample based on matching within
family and sample based on matching across families). We now move on to the multivariate
analysis. We estimate:
Disi ,t = α + β × Structurei ,t + δ × X i ,t −1 + ε i ,t ,
(1)
where Disi ,t represents the fund-bond distance of fund i at quarter t, Structurei ,t is the vector
containing both fund hierarchy and degree of specialty. X i ,t −1 is the vector made of the other control
variables defined above. We add fund type dummies in all specifications.
We report the results in Table II, Panel B for the main sample and in Panel C for the teammanaged subsample. In Panels D and E, the sample is based on the matching sample within and
across fund families respectively. Column (1) reports the results from an OLS regression with
standard errors clustered at fund level.
To address the potential endogeneity of fund structures, in Column (2), we implement an IV
regression, where family level structures are chosen as instruments. We instrument fund structure
using the following variables: family hierarchy (median of fund hierarchy within a family), family
employee specialty (median of employee specialty within a family), family team (median of team
dummy within a family), the interaction of family hierarchy with a financial center dummy and the
interaction of family employee specialty with a financial center dummy. This latter variable
resembles the instrument used by Chen et al. (2005) for the degree of outsourcing. The intuition is
the following. We know that the location in a financial center will have a direct impact on proximity
investment and performance, while the interaction with the structure of the family needs not be so.
At the same time, the interaction with the structure of the family helps to explain the structure of the
fund. Indeed, the incentive of the management family to include many layers or many different
areas of specialty depends on the availability of people. Availability is higher in financial centers
than in rural areas. Therefore, the desire of the family to set up a specific structure is constrained by
the location of the fund.
(Unreported) results show that the instruments help explain the organizational structure. Also,
they do not affect the dependent variable in the second stage through a channel different from the
impact on the instrumented variable. At the bottom of each IV specification, we report the Hansen’s
J statistic (p-value). We see that it always fails to reject the null, providing evidence for the quality
of our instruments.
12
Additional specifications based on changes in hierarchy are provided in the last section. They
are based in changes in the organizational structures induced by mergers and restructurings.
As additional robustness check, in Column (3), we provide Fama-Macbeth (1973) estimates
performed at the fund level, while in Column (4), we provide the results of Fama-Macbeth
estimates performed at the family level, based on family averages of all the variables. Column (5)
and (6) are estimated in the same way as in Column (3), but the sample is based on the funds owned
by mutual fund families and insurance companies.
The results indicate that there is a strong positive relation between the average distance of the
firms in which the fund invests and fund hierarchy. This holds across the different specifications as
well as for different sub-samples. The results are not only statistically significant but also
economically relevant. An increase of one layer in hierarchy raises the average distance (holding
weighted distance) of the firms in which the fund invests by 8.1% for the main sample, 7.7% for the
team-managed sample, 6.5% for the sample matched within family and 7.6% for the sample
matched across families. This shows that proximity investment is directly affected by the type of
structure of the fund. If we consider the other variables, we see that being managed by a team or by
a sole manager does not affect the decision to invest in closer firms. This suggests that our structure
variables do not just proxy for the mere fact that a fund is team-managed. There is in contrast no
relationship between proximity investment and the degree of specialty.
The other control variables are consistent with intuition. Being located in a financial center
increases the investment in closer firms. The same is true in the case the fund is more risk-conscious
and restricts itself to high-grade bonds. In the latter case, high risk prudence causes funds to shorten
their investment distances.
It is also interesting to note that funds that rotate their portfolio a lot (i.e., “high-turnover”
funds) are more likely to invest further away. This can be explained with the higher liquidity need
of these funds, not easy to meet in a more limited local area. However there is scarce evidence in
favor of an impact of the degree of employee specialty.
As a further robustness check, for each fund, no matter whether it is single-hierarchy or multihierarchy, we find a match with some other fund similar in type, geographical location and size, but
different in terms of fund family and hierarchy. Then, we run regressions based on the differences
between the original fund and the matched fund. The idea is that closely located funds are more
likely to face a homogenous information set, and by using differences, we can effectively cancel the
unobservable factors away. All the variables, excluding the financial center dummy, but including
both the dependent and the independent variables, are defined as differences between the value of
such a variable for the fund and the value of such a variable for its matched peer.
13
The matching procedure is as follows: for each fund-quarter, we first choose all the other funds
of the same fund type but from different fund families and having different fund hierarchy. Then,
we pick 20 funds located most closely and narrow them down to 10 according to similarity in fund
size. From those 10 funds we select the final one with the smallest geographical distance to the
original fund. If there is more than one matched fund located in the same place we choose the most
similar one in terms of fund size. The results confirm the previous ones, showing a strong positive
relation between the difference in portfolio distances and the difference in fund hierarchy. The
results are not included in the tables but available upon request.
One possible criticism of our measure of distance is that it measures the total distance between
the fund and the bond issuers represented in the portfolio. However, it may be that that it is not the
total distance that matters but the number of issuers that are located within a certain radius from this
investor. We therefore also use an alternative approach in which we fix a radius (300 km) and
define “close bonds” (“distant bonds”) the bonds of the firms located within (outside) the radius.
We then assign a value of 0 to the close bonds and 1 to the distant ones. The (unreported) results
based on this methodology are consistent with the reported ones.
Overall our findings hold across different specifications as well as for different sub-samples.
They also hold in the specification based on “differenced” variables with another similar fund
located close-by. These findings show that proximity investment is directly affected by the type of
structure of the fund. We now move on to see if fund hierarchy affects other portfolio strategies
such as portfolio concentration and herding.
B. Organizational Structure and Portfolio Concentration
We now consider the impact of the organizational structure on portfolio concentration. We know
that the availability of more information induces the manager to tilt the portfolio towards the set of
assets over which he has more information (e.g., Kacperczyk et al., 2005). We therefore expect that
a more vertical hierarchy, by reducing the incentive to collect soft information, lowers the degree of
portfolio concentration. We expect that funds characterized by a more hierarchical structure should
display a lower degree of portfolio concentration.
As in the previous analysis, we start with some univariate statistics. We define our measure of
portfolio concentration: Herfini ,t . It captures the degree of portfolio concentration in bonds of fund i
at quarter t. If we denote the set of bond issues held by fund i as Q and the fraction invested in bond
issue j as wi , j ,t , the fund herfindahl is defined as Herfini ,t = ∑ wi2, j ,t . In Table III, Panel A, we
j∈Q
report the sample mean of fund portfolio concentration overall and breaking down the sample into 4
different levels of hierarchy. We report the results for the main sample, for the team-managed
14
subsample and the two matching samples described earlier as well as tests of fund portfolio
differences in portfolio concentration between single and multi- fund hierarchy.
The results show a monotonic decrease in portfolio concentration as the number of layers
increases. This holds regardless of the sample and fund type. A three-layer mutual fund tends to
have a degree of concentration equal to just 46.9% (60.0% and 51.7%) of the concentration of a
one-layer mutual fund in the case of the main sample (matching within family and matching across
families). A three-layer insurance-owned fund tends to have a degree of concentration equal to just
19.8% (23.9% and 23.9%) of the concentration of a one-layer insurance-owned fund in the case of
the main sample (matching within family and matching across families). We then employ a
multivariate specification and estimate:
Herfini ,t = α + β × Structurei ,t + δ × X i ,t −1 + ε i ,t , (2)
where Herfini ,t is the degree of portfolio concentration in bonds of fund i at quarter t, and the other
variables are defined as in the previous specification.
We report the results in Table III, Panel B for the main sample and Panel C for the teammanaged subsample. In Panels D and E, the sample is based on the matching sample within and
across fund families respectively and with the same specifications as in Panel B. Column (1) reports
the results from an OLS regression with standard errors clustered at the fund level. In Column (2),
we report the results of an IV regression, where family level structures are chosen as instruments.4
As for the previous specification, additional tests based on changes in hierarchy are provided in the
last section.
The standard errors are clustered at the fund level. In Column (3), we provide Fama-Macbeth
(1973) estimates performed at the fund level, while in Column (4), we provide the results of FamaMacbeth estimates performed at the family level. Column (5) and (6) are estimated in the same way
as in Column (3), but they are restricted to the funds managed by mutual fund families and
insurance companies.
The results show a strong and consistent negative correlation between hierarchy and portfolio
concentration. This holds across the different specifications (OLS, IV and Fama-MacBeth) as well
as for the different sub-samples. An increase of one layer in the hierarchy reduces concentration by
50% (49% for the sample with more than one employee, 43% for the sample matched within family
and 46% for the sample matched across families). No relationship is found between the degree of
specialty and portfolio concentration.
4
We instrument the proxies for fund structure using the following variables: family hierarchy (median of fund hierarchy
within a family), family employee specialty (median of employee specialty within a family), family team (median of
team dummy within a family), the interaction of family hierarchy with financial center dummy and the interaction of
family employee specialty with financial center dummy.
15
The analysis based on the matching sample delivers consistent results. As in the previous
specifications, for further robustness checks, we run the regressions based on the difference in fund
concentration and the difference in fund hierarchy between the original fund and another similar
fund located close-by. The matching procedure is the same as described above. We also find
consistent results. These findings are consistent with the previous ones on proximity investment.
They confirm an overall picture in which a more hierarchical structure reduces soft information
collection. We now move on to herding.
C. Organizational Structure and Herding
We argued that a higher hierarchy would stifle fund managers’ incentive to collect soft information
and would induce them to invest more in line with their peers. To address this issue, we study the
relation between managerial herding and fund structure.
We define a variable (Herdingi,t) which proxies for the tendency of a fund to follow the trading
behavior of its peers or to go against it. We employ the same methodology used by Lakonishok,
Shleifer and Vishny (1992) and Grinblatt, Titman and Wermers (1995). A detailed description of
the construction of this variable is reported in the section Variables Definitions at the end of the
paper. In Panel B of Table I, we report some descriptive statistics of the measure of herding. The
mean in our sample is 1.4%. It is higher than the findings of 0.84% by Grinblatt, Titman and
Wermers (1995). This can be seen as evidence that bond funds herd more with each other than
equity funds.
We start with some univariate analysis. We break down the sample into 4 different levels of our
organizational variables. We also provide univariate tests of fund herding regarding to single vs.
multi- fund hierarchy. The results are reported in Table IV, Panel A. The results show a monotonic
increase in fund herding propensity as the number of layers increases and a decrease with the degree
of fund specialty. This holds regardless of the sample and fund type. A three-layer mutual fund
tends to herd on average 14.1% (25.7%, 16.5%) more than a one-layer mutual fund in the case of
the main sample (matching within family, matching across families). A three-layer insuranceowned fund tends to herd on average 20.7% (26.0%, 30.0%) more than a one-layer insuranceowned fund in the case of the main sample (matching within family, matching across families). We
then move on to the multivariate analysis and estimate:
Herding i ,t = α + β × Structurei ,t + δ × X i ,t −1 + ε i ,t . (3)
Herdingi,t is the herding measure as defined above of fund i at quarter t, and the other variables are
as from the previous specifications. We include fund type dummies in all the specifications.
We report the results in Table IV, Panel B for the main sample and Panel C for the teammanaged subsample. In Panels D and E, the sample is based on the matching sample within and
16
across fund families respectively and with the same specifications as in Panel B. The layout of the
columns is the same as in the previous analysis.
The results show a positive relation between hierarchy and herding. This holds across the
different specifications as well as for different sub-samples. An increase of one layer in the
hierarchy raises herding by 14% for the main sample, 11% for the team-managed, 15% for the
sample matched within family and 14% for the sample matched across families. This is in line with
the intuition that hierarchy, by reducing the incentives to collect soft information, translates in more
herding.
In contrast, there is a negative relation between hierarchy and herding. This holds across the
different specifications as well as for different sub-samples. An increase of one degree of specialty
lowers herding by 10% for the main sample, 12% for the team-managed, 15% for the sample
matched within family and 10% for the sample matched across families. This is consistent with the
idea that various specialties induce a collective thinking whose outcome is more unique and
different from the one arising from single-specialty.
For further robustness checks the regressions based on the difference in fund concentration and
the difference in fund hierarchy between the original fund and another similar fund located close-by
deliver consistent results.
D. Organizational Structure and Performance
We now focus on performance. We study the relation between performance and fund structure. We
start by defining the measure of performance ( Alphai ,t ) of fund i in month t. It is constructed in the
following way. First, for each fund-month (i,t), we estimate the monthly factor loadings by running
the following regression:
ri ,s − rrf ,s = ai ,t −1 + bi ,t −1 ( rdj ,s − rrf ,s ) + t i ,t −1TS s + ri ,t −1 RS s +
ci ,t −1CVs + mi ,t −1 ( rm ,s − rrf ,s ) + si ,t −1 SMBs + hi ,t −1 HMLs + ui ,t −1UMDs + ε i ,s ,
(4)
where t − 30 < s ≤ t − 1 and we require a minimum of 25 observations for each regression. The
dependent variable is the monthly return of fund i in month s less the risk-free rate rf ,s . The
independent variables include 8 factors: the excess return of Dow Jones Corporate Bond Index over
the risk-free rate ( rdj ,s − r f ,s ), the yield difference between the twenty-year constant maturity
treasury bonds and the two-year constant maturity treasury bonds (gs20-gs2, term spread, TS)
(Colin-Dufresne et. al., 2001), the yield difference between Moody’s BAA corporate bond index
and the thirty year constant maturity treasury bonds (Baa-gs30, risk spread, RS) 5 , the yield
5
Moody's includes bonds with remaining maturities as close as possible to 30 years. Moody’s drops bonds if the
remaining life falls below 20 years, if the bond is susceptible to redemption, or if the rating changes.
17
difference between the five year constant maturity treasury bonds and the average yield of two year
and ten year constant maturity treasury bonds (gs5-(gs2+gs10)/2, curvature spread, CV), the excess
return of market return over the risk-free rate ( rm , s − rf , s ), the return difference between small and
large capitalization stocks (SMB), the return difference between high and low book-to-market
stocks (HML), the return difference between stocks with high and low past returns (UMD). The
return data on Dow Jones Corporate Bond Index is from Dow Jones’ website. The data on treasury
bond yields and Moody’s Baa corporate bond yields are from the FRED database at the Federal
Reserve Bank of Saint Louis. The data on risk-free rate, market return, SMB, HML and UMD are
from Kenneth French’s website.
Then, using the estimated loadings, we calculate fund alpha in month t by:
α i ,t = ri ,t − rrf ,t − b̂i ,t −1 ( rdj ,t − rrf ,t ) − t̂i ,t −1TS t − r̂i ,t −1 RS t −
ĉi ,t −1CVt − m̂i ,t −1 ( rm ,t − rrf ,t ) − ŝi ,t −1 SMBt − ĥi ,t −1 HMLt − û i ,t −1UMDt .
(5)
We start with some univariate analysis. We break down the sample into 4 different levels of
hierarchy/specialty. We then report the sample mean of fund alpha. We also provide univariate tests
of difference in performance between single and multi- fund hierarchy. The results are reported in
Table V, Panel E. They show a monotonic decrease in fund performance as the number of layers
increases. This holds regardless of the sample and fund type. A three-layer mutual fund has a
performance 55 bp (68 bp, 54 bp) lower than a one-layer fund in the case of the main sample
(matching within family, matching across families). A three-layer insurance-owned fund has a
performance 18 bp (5 bp, 14 bp) lower than a one-layer insurance-owned fund in the case of the
main sample (matching within family, matching across families). We then move on to the
multivariate analysis and estimate:
Alphai ,t = α + β × Structurei ,t + δ × X i ,t −1 + ε i ,t , (6)
where alphai ,t is fund performance and the other variables are defined as above.The results are
reported in Table V. The analysis in Panel A is based on the main sample, Panel B is for the teammanaged subsample, while Panel C and Panel E are based on the matching sample within and
across fund families respectively. The layout of the columns as well as the specifications is the
same as in the previous tests.
The results show a strong negative relation between performance and fund structure and no
relationship between performance and degree of specialty. Hierarchy reduces performance. This
holds across the different specifications (OLS, IV and Fama-MacBeth) as well as for the case of the
matched sample. One additional layer reduces fund alpha by 21 bp for the main sample, 20 bp for
the team-managed subsample, 18 bp for the sample matched within family and 16 bp for the sample
matched across families.
18
In the previous section, we found that hierarchy is negatively related to proximity investment.
We know that proximity investment is positively related to performance (Coval and Moskowitz,
1999 and 2001, Chen et al., 2004). Is it the case that the negative relationship between performance
and hierarchy is just due to the lower proximity investment? To address this issue, in Panel A
(column (8)-(10)) we include a dummy variable taking a value of 1 if the fund portfolio distance is
below the third quantile of the sample in each quarter and 0 otherwise, and we interact it with fund
hierarchy. The interaction term is always significantly negative for fund performance. This suggests
that for funds investing in close firms the negative impact of hierarchy is stronger and is consistent
with wasting hard-to-transfer soft information.
If we consider the other variables, we notice the strong positive relation between performance
and the fraction invested in high-quality bonds. This suggests that constraints on the ability to
choose risky assets do not hamper performance. Fund size and family size are in general negatively
related to fund performance, consistent with the findings of Chen et al. (2004). It shows that the
effects of fund organization structures on performance are not driven by fund size and family size.
Overall, these findings provide evidence of the “dark side” of the fund organizational structure:
a negative relation between performance and hierarchy. We have argued that this is due to the fact
that hierarchy slows the flow of information within the organization. If this is the case, we expect
that most of the impact of hierarchy is concentrated in the investment in low quality bonds. Indeed,
these are the ones in which the lack of diffused and widespread information about the firms in
which to invest makes soft information more relevant. For instance, soft information of the firm
which may result in a potential rating downgrade should matter more for low-quality bonds than for
high quality bonds.
We test this issue, by studying whether the impact of hierarchy is more pronounced for the
investment in high-quality bonds or for low-quality bonds. In particular, we indentify the “high
rated bonds” in the fund portfolio and compare the impact of fund hierarchy on the fund
performance of investing in high rated bonds with that of investing in low rated bonds. High-rated
bonds refer to bonds with Moody’s credit rating above A3. Low-rated bonds have Moody’s credit
rating from B3 to BBB1. For each fund, we estimate two portfolio alphas separately. The first is
based on the value-weighted return of investing in high-rated bonds, while the second is based on
the return of investing in low-rated bonds. Performance is estimated as above.
We report the results in Table V, Panel F. We stack the high and low rated alphas together and
create a rating category dummy which equals 1 if it is a low rated alpha and 0 otherwise. Our focus
is the interaction term of fund hierarchy and the rating category dummy. Standard errors are
clustered at the fund level. In Column (1)-Column (6), we run both OLS and Fama-Macbeth
regressions for mutual funds and insurance-owned funds. The interaction term of fund hierarchy
19
and the low rating dummy of bond quality is always significant and negative. It means that the
strong negative correlation between performance and fund hierarchy is mostly concentrated in the
low-rated bonds. This confirms our intuition that hierarchy mostly affects the flow of information
mostly for low quality bonds. One additional layer of hierarchy reduces performance by 17 bp more
in the case of low quality bonds than for high quality bonds for the main sample. The results on the
team-managed subsample and the matching samples are consistent.
E. Robustness Checks
We now consider further robustness checks. The previous specifications assess the relationship
between hierarchy and fund behavior (portfolio concentration, herding, proximity investment) and
performance. We now test whether the same specifications hold on changes. That is, we investigate
whether changes in fund behavior and performance is related to a change in the degree of hierarchy
of the fund. We therefore focus on the subsample (fund-quarter) where the fund changes its
hierarchical structure from quarter t-1 to quartet t. These changes are attributable to either internal
restructurings or M&As. We estimate the following regression:
∆FundManagementi ,t = α + β × ∆Structurei ,t + δ × ∆X i ,t + γ × X i ,t −1 + FundManagementi ,t −1 + ε i ,t , (7)
where ∆FundManagementi ,t represents the change in fund behavior (portfolio distance, herding, and
portfolio concentration) as well as the change of fund performance. We consider two measures of
performance: the raw return (cumulative, quarterly) of the fund and the change of fund alpha
(cumulative, quarterly) respectively from quarter t -1 to quarter t. ∆Structurei ,t is the change of the
fund structure and ∆X i ,t are the changes of other control variables from quarter t-1 to quarter t. A
more detailed definition of these variables is reported in the Appendix. FundManagementi ,t −1 is the
lagged dependent variable at quarter t-1. We cluster the errors at fund level and we always include
time dummies and fund type dummies.
The results are reported in Table VI. They are consistent with the previous ones. An increase in
the degree of hierarchy reduces proximity investment and portfolio concentration, while it increases
fund herding. Overall, this implies a lower performance. These findings are not only statistically
significant, but also economically relevant. An additional layer lowers the average distance from
firms whose bonds it invests in by 25 km, reduces portfolio concentration by 5% and raises fund
herding by 12%. It reduces fund performance by 17 bp per month. These results provide an
additional robustness check. They also show that changes in the fund structure quickly find their
way into the behavior of the fund managers.
20
Conclusion
We study how the internal organizational structure affects fund’s strategies and performance. We
focus on the trade-off between vertical structure and degree of specialization. We argue that a more
hierarchical structure reduces the incentives to collect “soft” information and proximity investment.
This lowers the incentive to concentrate the investment in few bonds and makes the manager more
likely to herd. The net result is lower performance. In contrast, a higher degree of specialization
should produce a more unique “thinking”, reducing herding with other asset managers.
And indeed, we show that that funds with more hierarchical structures tend to invest less in
firms located close to the funds. This has a direct negative effect on fund performance: more
vertical structures are characterized by worse performance. Funds with a more vertical structure
tend to herd more with the other funds and to hold less concentrated portfolios. In contrast, funds
characterized by a higher degree of specialties tend to herd less with each other. These findings
provide a first evidence of the trade-off between vertical hierarchy and degree of specialization
within organizations. While the effect of vertical structure seems to prevail, still the degree of
specialization plays a very important role in the provision of “unique” managerial solutions.
References
Adams, Renee B. and Ferreira, Daniel, 2007, A Theory of Friendly Boards. Journal of Finance, 62, 217-250.
Aghion, P. and Tirole, J., 1997, Formal and Real Authority in Organizations, Journal of Political Economy, 105, 1-29.
Bär, M., A. Kempf, and S.Ruenzi, 2005, Team Management and Mutual Funds, Working Paper.
Berger, A.N. N.H., Miller, M.A. Petersen, R.G. Rajan, and J.C. Stein, 2005, Does Function Follow Organizational
Form? Evidence from the Lending Practices of Large and Small Banks Journal of Financial Economics, 76(2), 237-69.
Blau, P.M. 1977. Inequality and Heterogeneity. New York: The Free Press.
Brown,S.J., and W.N.Goetzmann, 1995, Performance Persistence, Journal of Finance, 50, 679-698.
Carhart, M., 1997, On Persistence in Mutual Fund Performance, Journal of Finance, 52, 57-82.
Carter, D., Simkins,B., and Simpson, W. , 2003, Corporate Governance, Board Diversity, and Firm Value, The
Financial Review 38 (1), 33-53.
Chen, J., H. Hong, M. Huang, and J. Kubik, 2004, Does Fund Size Erode Mutual Fund Performance? The Role of
Liquidity and Organization, American Economic Review, 94, 1276-1302.
Chen, J., H. Hong, and J. Kubik, 2005, Outsourcing Mutual Fund Management: Firm Boundaries, Incentives and
Performance, Working Paper.
Chevalier, J., and G. Ellison, 1997, Risk Taking by Mutual Funds as a Response to Incentives, Journal of Political
Economy, 105, 1167-1200.
Colin-Dufresne,P.,R.Goldstein, andJ.S. Martin, 2001. The Determinants of Credit Spread Changes. Journal of Finance
56, 2177–2207.
Coval, J.D. and T. Moskowitz, 1999, Home bias at home: Local equity preference in domestic portfolios, Journal of
Finance 54, 2045-2073.
Coval, J.D. and T. Moskowitz, 2001, The geography of investment: Informed trading and asset prices, Journal of
Political Economy, 109, 811-841.
Crawford, V. and J. Sobel, 1982. Strategic Information Transmission. Econometrica, 50, 1431-145.
Cremer, J., 1980, A Partial Theory of the Optimal Organization of a Bureaucracy,
21
Bell Journal of Economics, 11(2): 683-693.
Cremer, J. 1993, Corporate culture and shared knowledge. Indust. Corporate Change 2(3) 351-386.
Del Guercio, D., J. Reuter, and P.A. Tkac, 2007, Why Do Some Mutual Fund Families Pick Stocks for the
Competition? Working Paper.
Dessein, W., 2002, Authority and Communication in Organizations. Review of Economic Studies, 69, 811-838.
Dessein, W. and Tano Santos, Adaptive Organizations, Journal of Political Economy, 2006, 114(5): 956-995.
Elton, E.J., M.J. Gruber, and C.R. Blake, 1996, The Persistence of Risk-Adjusted Mutual Fund Performance, Journal of
Business, 69, 133-157.
Gaspar, J.M., M. Massa, and P. Matos, 2006, Favoritism in Mutual Fund Families? Evidence on Strategic Cross-Fund
Subsidization, Journal of Finance, 61, 73-104.
Geanakoplos, John and Paul Milgrom, 1991, A Theory of Hierarchies Based on Limited Managerial Attention, Journal
of the Japanese and International Economy, 5(3): 205-25.
Graham, J.R., 1999, Herding Among Investment Newsletters: Theory and Evidence. Journal of Finance, 54, 237-268.
Grinblatt, M., Titman, S., and R. Wermers, 1995, Momentum Investment Strategies, Portfolio Performance, and
Herding: A Study of Mutual Fund Behavior, American Economic Review, 85, 1088-1105.
Guedj, I., and J. Papastaikoudi, 2004, Can Mutual Fund Families Affect the Performance of Their Funds?, Working
Paper.
Hong, H., Kubik, J.D. and J.C. Stein, 2004, Thy neighbor’s portfolio: word-of-mouth effects in the holdings and trades
of money managers, Journal of Finance, 60, 2801-24.
Hong, H., Kubik, J.D. and J.C. Stein, 2008, The only game in town. Stock-price consequences of local bias. Journal of
Financial Economics. Vol. 90. 20-37.
Kacperczyk, M., C. Sialm, and L. Zheng, 2006, Unobserved Actions of Mutual Funds, Review of Financial Studies,
forthcoming.
Kempf, A. and S.Ruenzi, 2004, Tournaments in Mutual Fund Families, Working Paper.
Khorana, A and H. Servaes, 1999, Determinants of Mutual Fund Starts, Review of Financial Studies 12, 1043–1074.
Kuhnen, C., 2009, Business networks, corporate governance and contracting in the mutual fund industry, Journal of
Finance, 64(5): 2185-2220
Kuhnen, C., 2004, Dynamic contracting in the mutual fund industry, Working Paper,
Lakonishok, J, A. Shleifer, and R.W.Vishny, 1992, The Impact of Institutional Trading on Stock Prices. Journal of
Financial Economics. Vol. 32. 23-43.
Mamaysky, H., M. Spiegel, 2001, A Theory of Mutual Funds: Optimal Fund Objectives and Industry Organization,
Working Paper.
Marshack, J. and R. Radner, 1972, Economic Theory of Teams. New Heaven, Yale University Press.
Massa M., J. Reuter and E. Zitzewitz, 2005, The Rise of Anonymous Teams in Asset Management, Journal of
Financial Economics, forthcoming.
Massa M., A. Yasuda and L. Zhang, 2007, Institutional Investors, Credit Supply Uncertainty, and the Leverage of the
Firm, Working Paper.
Nanda, V., Wang, J.Z., and L. Zheng, 2004, Family Values and the Star Phenomenon: Strategies of Mutual Fund
Families, Review of Financial Studies, 17, 667-698.
Stein, J.C., 2002, Information Production and Capital Allocation: Decentralized versus Hierarchical Firms, Journal of
Finance, 57, 1891-1921.
Van den Steen, E., 2005, Organizational Beliefs and Managerial Vision, The Journal of Law, Economics, and
Organization 21, 256 – 83.
Vayanos, Dimitri, 2003, The Decentralization of Information Processing in the Presence of Interactions, Review of
Economic Studies, 70: 667-695.
Wermers, R., 1999, Mutual Fund Herding and the Impact on Stock Prices, Journal of Finance, 54, 581-622.
Westphal, J. D., and E. J. Zajac, 1995, Who shall govern? CEO/board power, demographic similarity, and new director
selection." Administrative Science Quarterly, 40, 60-83.
22
Variable Definitions
Fund Hierarchy
For each fund we first define fund vertical layer according to the employee’s job title in the following way:
Vertical Layer
1
1
1
Job Name
Chairman
President
Chief Executive Officer
Job Code
CHR
PRE
CEO
2
2
Chief Financial Officer
Chief Investment Officer
CFO
CIO
3
3
Bond Department Head
Fixed-Income Head
BDH
HFI
4
4
4
4
4
4
4
4
Portfolio
Portfolio
Portfolio
Portfolio
Portfolio
Portfolio
Portfolio
Portfolio
PMG
PMB
PMV
PME
PMH
PMJ
PMP
PMT
5
Trader-General
TBG
6
Credit/Research Analyst-General
RAG
Manger-General
Manger-Balanced
Manger-Convertibles
Manger-Eq/Pref Stock
Manger-High Yield
Manger-Inv Grade Corp Bd
Manger-Pvt Placements
Manger-Short-Term/MM
Then fund hierarchy is counted as the distinct number of vertical layers. It is possible that a person can be entitled
with multiple job functions, so in the cases where the distinct number of vertical layers exceeds the number of
employees, we use the number of employees as the fund hierarchy.
Employee Specialty
For each fund we first determine the number of specialties according to the employee’s area of focus. We mainly
look at the following sectors:
1: Market Sector
Asset Backed
Corporate
Government
Mortgage backed
Local/Regional
US firms investing
non-domestically
Combination of Above
2: Credit Sector
3: Geographical Focus
Any Corporate Sector
Any Municipal Sector
Any Specialty
Canadian Dollar
Euro
U.S. Dollar
Any Country
Canada
Emerging Markets
Euroland
Any State/Terr.
United States
Combination of Above
Non-domestic
Combination of Above
4: Job Title
Chairman
President
Chief Executive Officer
Chief Financial Officer
Chief Investment Officer
Bond Department Head
Fixed-Income Head
Portfolio Manger-General
Portfolio Manger-Balanced
Portfolio Manger-Convertibles
Portfolio Manger-Eq/Pref Stock
Portfolio Manger-High Yield
Portfolio Manger-Inv Grade Corp Bd
Portfolio Manger-Pvt Placements
Portfolio Manger-Short-Term/MM
The number of specialties is counted as the number of total combinations of the above four sectors. Then we define
employee specialty as the number of specialties divided by the number of employees.
Team Dummy
We define team as a dummy variable which equals 1 if the fund has more than 1 employee and 0 otherwise.
Fund Portfolio Distance
Fund portfolio distance measures the distance between the fund and its bond portfolios. If we denote the set of bond
issues held by fund i by Q and wi , j ,t be the fraction of fund i invested in bond issue j, the fund portfolio distance
is defined as:
Disi ,t = log[∑ wi , j ,t 3963 * arccos( sin(lat i ) sin(lat j ) + cos(lat i ) cos(lat j ) cos(loni − lon j ))] ,
j∈Q
23
where ( lati , loni ), ( lat j , lon j ) are the (latitude, longitude) for fund i and bond issue j in radian degrees.
Information on locations of bond issuers is obtained from Compustat and SDC global new issue database. Since
Lipper only provides county information of the managing firm, we utilize location of the managing firm as the
location of the fund. County level coordinates (latitude, longitude) are from the Gazetteer Files of Census 2000.
Fund Portfolio Concentration
Fund portfolio concentration represents the fund’s concentration ratio (herfindal) of its bond portfolio. If we denote
the set of bond issues held by fund i by Q and wi , j ,t be the fraction invested in bond issue j, fund portfolio
concentration is defined as:
Herfini ,t = ∑ wi2, j ,t .
j∈Q
Fund Herding
Fund herding represents the tendency of a fund following the crowd or going against it. We follow the same
methodology used by Lakonishok, Shleifer and Vishny (1992) and Grinblatt, Titman and Wermers (1995). The first
Bk ,t ( S k ,t ) be the number of funds buying
step is to define a measure of investor herding at the bond level. Let
(selling)
bond
issue
k
at
quarter
t,
then
the
herding
measure
(UHM)
is
expressed
as:
UHM k ,t =| p k ,t − E[ p k ,t ] | − E | p k ,t − E[ p k ,t ] | , where p k ,t = Bk ,t /( Bk ,t + S k ,t ) is the proportion of funds
trading issue-quarter k, t which are buyers. We use the proportion of all trades by funds that are purchases during
quarter t to proxy for E[ p k ,t ] . The first part represents the “extra” number of funds trading a bond issue during a
given quarter as the proportion of the total number of funds buying that issue-quarter minus the expected
proportion of buyers. The second term is an adjustment factor allowing for random variation around the expected
proportion of buyers under the null hypothesis of cross-sectional independence among fund trades. The expectation
in the second term is calculated by assuming
Bk ,t follows a binomial distribution with parameter ( Bk ,t + S k ,t )
and E[ pk ,t ] . The second step is to define the signed herding measure (SHM) which indicates the tendency of
whether fund i is following the crowd or going against it in trading bond k. It is calculated as:
SHM i ,k ,t = I i ,k ,t × UHM k ,t − E[ I i ,k ,t × UHM k ,t ] , where
I i , k ,t
is
an
indicator
variable:
I i ,k ,t =0 if
| p k ,t − E[ p k ,t ] |< E | p k ,t − E[ p k ,t ] | ; I i ,k ,t =1 if pk ,t − E[ pk ,t ] > E | pk ,t − E[ pk ,t ] | and fund i is a buyer of
− ( pk ,t − E[ pk ,t ]) > E | pk ,t − E[ pk ,t ] | and fund i is a seller of bond k; I i ,k ,t =-1 if
bond k, or if
pk ,t − E[ pk ,t ] > E | pk ,t − E[ pk ,t ] | and fund i is a seller of bond k, or if − ( pk ,t − E[ pk ,t ]) > E | pk ,t − E[ pk ,t ] |
and fund i is a buyer of bond k. Additionally, we impose the restriction that SHM i ,k ,t = 0 is there are fewer than
5 funds traded bond k during quarter t. Under the assumption that the number of buyers of bond k is binomially
distributed,
the
E[ I × UHM ] =
expectation
term
∑ (2 p − 1)UHM
p: p − pk , t > E [| p − pk , t |]
E[ I × UHM ]
k ,t
can
be
calculated
∑ (2 p − 1)UHM
( p ) Pr( p ) −
p:− ( p − pk , t ) > E [| p − pk , t |]
by
k ,t
the
following
formula:
( p ) Pr( p ) , where Pr(p) is the
probability of ( Bk ,t + S k ,t )p occurrences assuming a binomial distribution with parameter ( Bk ,t + S k ,t )
and E[ pk ,t ] . Finally, if we denote the set of bond issues held by fund i by Q and the fraction invested in bond k
by wi ,k ,t , the herding measure of fund i at quarter t is: Herding i ,t = 100 *
∑ (w
k∈Q
i ,k ,t
− wi ,k ,t −1 ) SHM i ,k ,t .
Fund Return Volatility: For each fund at month t, we calculate return volatility as the standard deviation of
monthly returns during the past 12 months. Then we use quarterly median as the fund return volatility.
Fund Portfolio Maturity: the value-weighted average maturity of all the bonds held by each fund.
Fund Size: logarithm of the par-amount of bond holdings held by each fund
Family Size: logarithm of the par-amount of bond holdings held by each fund family (managing firm)
24
Table I
Summary Statistics
This table presents summary statistics of the main variables used in the subsequent analysis. Our primary database is Lipper’s
eMAXX fixed income database. It contains information on quarterly bond holdings of major U.S. bond mutual funds, insurance
companies (life and property), annuities and pension funds from the first quarter of 1998 to the second quarter of 2005. It also
provides information on the fund employee’s job title, market sector, credit sector and geographical focus which enables us to
characterize the fund organizational structures.
Panel A: Summary Statistics of Fund Hierarchy
Panel A1 reports the summary statistics of fund hierarchy for the main sample. We report the number of observations by the
level of fund hierarchy. Panel A2 is for the sub-sample where the fund reports more than one employee (team-managed) in the
data. The detailed definitions can be found in the appendix. We separately report the results for funds owned by mutual fund
families, insurance companies and other institutions (annuities and pension funds).
Panel A1: Main Sample
No. of Obs. By Fund Hierarchy
1
2
3
4
Mean Hierarchy
All Institutions
(12592)
(6957)
(1374)
(226)
Mutual
(9259)
(3034)
(367)
(51)
Insurance
(2889)
(3335)
(896)
(140)
Others
(444)
(588)
(111)
(35)
1.491
(21149)
1.308
(12711)
1.764
(7260)
1.776
(1178)
Panel A2: Team-managed Subsample
No. of Obs. By Fund Hierarchy
1
2
3
4
Mean Hierarchy
All Institutions
(5279)
(6957)
(1374)
(226)
Mutual
(2515)
(3034)
(367)
(51)
Insurance
(2406)
(3335)
(896)
(140)
Others
(358)
(588)
(111)
(35)
1.750
(13836)
1.657
(5967)
1.818
(6777)
1.837
(1092)
Panel B: Summary Statistics of Fund Characteristics
In Panel B we report the sample mean of fund characteristics with the number of observations (fund-quarter) given in the
parentheses. We separately report our results for funds owned by mutual fund families, insurance companies and other
institutions (annuities and pension funds). The detailed definition of each variable can be found in the appendix.
Fund Characteristics
Employee Specialty
Team Dummy
Fund Size
Family Size
Fund Turnover
Fund Return Volatility
Fund Return (Quarterly)
Fraction: Investment-grade Bond
Financial Center Dummy
Fund Portfolio Maturity
Fund Portfolio Distance
Fund Portfolio Concentration
Fund Herding
Fund Performance (Monthly)
All Institutions
1.449
(21149)
0.654
(21149)
10.722
(21149)
14.212
(21149)
0.336
(21149)
0.014
(21149)
0.015
(21149)
0.677
(21149)
0.307
(21149)
8.622
(19990)
6.723
(20549)
0.056
(21149)
1.422
(18984)
-0.112
(30332)
25
Mutual
1.577
(12711)
0.469
(12711)
10.890
(12,11)
13.926
(12711)
0.395
(12711)
0.014
(12711)
0.015
(12711)
0.600
(12711)
0.356
(12711)
8.604
(12004)
6.795
(12297)
0.059
(12711)
1.517
(11690)
-0.225
(17738)
Insurance
1.245
(7260)
0.933
(7260)
10.402
(7260)
14.617
(7260)
0.236
(7260)
0.012
(7260)
0.015
(7260)
0.830
(7260)
0.214
(7260)
8.452
(6911)
6.580
(7076)
0.054
(7260)
1.306
(6150)
0.102
(11747)
Others
1.333
(1178)
0.926
(1178)
10.897
(1178)
14.802
(1178)
0.318
(1178)
0.015
(1178)
0.017
(1178)
0.563
(1178)
0.337
(1178)
9.917
(1075)
6.723
(1176)
0.036
(1178)
1.070
(1144)
-0.730
(847)
Table II
Fund Hierarchy and Portfolio Distance
This table presents the regression results of fund portfolio distance on fund hierarchical structure. We estimate the following
equation:
Disi ,t = α + β × Hierarchyi ,t + δ × X i ,t −1 + ε i ,t ,
where
Disi ,t
represents fund portfolio distance of fund i at quarter t, Hierachy i ,t is fund hierarchy and X i ,t −1 are other
control variables. The definitions are detailed in the appendix.
Panel A summarizes the sample mean of fund portfolio distance at different levels of fund hierarchy. We also provide
univariate tests of fund portfolio distance regarding to single vs. multi- fund hierarchy. Multi-hierarchy means the number of
fund hierachies to be greater than 1. We always report mutual funds and insurance companies separately. For robustness checks
we construct two matching samples, one within fund family and one across fund families. For each multi-hierarchy fund, we first
select another single hierarchy fund from the same fund family and most similar in terms of fund size, then combine the matched
single hierarchy funds with the original multi-hierarchy funds as our “matching within fund family” sample. The “matching
aross fund family” sample is constructed similarly except that the matched single-hierarchy fund is chosen from different fund
families but belonging to the same fund type (mutual funds, insurance companies, pension funds etc). We report the results for
the main sample as well as the two matching samples. Both two tailed T-test and Wilconxon rank-sum test are performed to test
the differences. The number of observations (fund-quarter) is given in the parentheses.
The analysis in Panel B is based on the main sample including funds owned by mutual fund families, insurance companies
and other institutions (annunities and pension funds). We add the fund type dummies across all specifications. Column (1) is
OLS regression with standard errors clustering at fund level. To address the possible engogeneity issue of fund-specific structures,
we implement an IV regression (2SLS) in Column (2) where family level structures are chosen as instruments. Specifically, we
instrument fund structure variables using the following variables: family hierarchy (mean of fund hierarchy within a family),
family employee specialty (mean of employee specialty within a family), family team (mean of team dummy within a family), the
interaction of family hierarchy with financial center dummy and the interaction of family employee specialty with financial
center dummy. Hansen’s J statistic (p-value) is reported to examine the quality of instruments. The standard errors are clustered
at fund level. Column (3) is estimated under a Fama-Mecbeth (1973) framework where the reported coefficients are time-series
averages of 29 quarterly cross-sectional slope estimates. Column (4) is done by first calculating family averages of all the
variables and then running Fama-Mecbeth regressions based on the family averages. Column (5) and (6) are estimated in the
same way as in Column (3) but only based on funds owned by mutual fund families and insurance companies.
Panel C has the exact same specifications as in Panel B but for the team-managed subsample where the fund has more than
one employee. Panel D and Panel E are based on the matching sample within and across fund families respectively with the same
specifications as in Panel B. For the sake of brevity we only report the coefficients of interested variables from Panel C to Panel
E. ***, ** and * represent significance levels at 1%, 5% and 10% respectively with t-statistics given in parentheses.
26
Table II (Cont’d)
Panel A: Univariate Results
Fund Portfolio Distance
by Fund Hierarchy
Main Sample
Mutual
Insurance
Matching Sample
(within family)
Mutual
Insurance
Matching Sample
(across family)
Mutual
Insurance
6.781
(8978)
6.826
(2923)
6.871
(347)
6.934
(49)
6.492
(2773)
6.621
(3280)
6.654
(888)
6.708
(135)
6.796
(2090)
6.843
(1728)
6.890
(219)
6.973
(42)
6.544
(867)
6.609
(1704)
6.629
(603)
6.717
(90)
6.781
(4206)
6.827
(2851)
6.872
(339)
6.946
(48)
6.505
(1893)
6.622
(3215)
6.650
(878)
6.708
(135)
T-test
Multiple vs. Single Hierarchy
5.19***
11.64***
3.60***
3.92***
4.45***
9.48***
Wilconxon Test
Multiple vs. Single Hierachy
2.83***
13.79***
2.86***
3.70***
2.69***
11.58***
FM Family
(4)
0.067***
(7.56)
-0.001
(-0.24)
0.055***
(3.61)
0.018***
(4.96)
-0.006*
(-1.90)
0.006***
(4.13)
0.043
(1.42)
-8.560***
(-2.90)
1.072
(1.58)
-0.486***
(-10.65)
-0.071***
(-7.44)
6.958***
(87.83)
Y
0.1502
7838
Mutual
FM
(5)
0.041***
(3.42)
0.006
(1.07)
0.031*
(1.87)
0.000
(0.00)
0.009***
(3.07)
-0.002*
(-1.90)
0.010
(0.40)
-3.914
(-1.61)
0.576
(0.87)
-0.364***
(-10.36)
-0.078***
(-7.68)
6.899***
(90.32)
Y
0.1156
11635
Insurance
FM
(6)
0.070***
(8.15)
0.001
(0.10)
0.070
(1.31)
0.030***
(9.89)
-0.039***
(-7.22)
0.010***
(3.99)
0.061
(1.13)
-0.180
(-0.03)
-1.433
(-1.46)
-0.423***
(-8.01)
0.077***
(3.21)
6.915***
(50.80)
Y
0.1544
6737
1
2
3
4
Panel B: Main Sample
Fund Hierarchy
Employee Specialty
Team Dummy
Fund Size
Family Size
Portfolio Maturity
Fund Turnover
Fund Return Volatility
Fund Return
Fraction in Investment-grade Bonds
Financial Center Dummy
Const
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
0.059***
(4.35)
-0.001
(-0.21)
0.015
(0.65)
0.014***
(2.75)
-0.007
(-1.14)
0.002
(0.99)
0.020
(1.31)
-3.580*
(-1.90)
-0.001
(-0.01)
-0.374***
(-11.36)
-0.061***
(-2.69)
6.923***
(61.91)
Y
Y
Fund
0.1031
19450
All Institutions
IV
FM
(2)
(3)
0.081***
0.066***
(4.61)
(8.85)
-0.040
-0.001
(-1.36)
(-0.20)
-0.023
0.024
(-0.71)
(1.49)
0.016***
0.015***
(2.91)
(7.32)
-0.014
-0.009***
(-1.63)
(-5.44)
0.002
0.002**
(0.96)
(2.03)
0.017
0.026
(1.09)
(1.55)
-3.675*
-2.483
(-1.95)
(-1.11)
-0.006
0.313
(-0.04)
(0.77)
-0.381***
-0.400***
(-11.41)
(-16.12)
-0.065***
-0.049***
(-2.85)
(-6.62)
7.050***
6.848***
(44.14)
(102.99)
Y
Y
Y
0.98
Fund
0.0960
0.1325
19450
19450
27
Table II (Cont’d)
Panel C: Team-managed Subsample
Fund Hierarchy
Employee Specialty
Control Variables
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
0.054***
(3.89)
-0.010
(-0.77)
Y
Y
Y
Fund
0.1183
12733
All Institutions
IV
FM
(2)
(3)
0.077***
0.060***
(3.94)
(7.76)
-0.039
-0.009**
(-0.76)
(-2.56)
Y
Y
Y
Y
Y
0.55
Fund
0.1158
0.1546
12733
12733
Mutual
FM
(5)
0.034***
(2.85)
-0.012*
(-1.81)
Y
Y
0.1304
5437
Insurance
FM
(6)
0.071***
(8.58)
0.002
(0.29)
Y
Y
0.1473
6295
FM Family
(4)
0.059***
(5.57)
-0.041***
(-3.84)
0.036
(0.95)
Y
Y
0.2233
3398
Mutual
FM
(5)
0.038***
(2.83)
-0.028***
(-4.28)
0.056**
(2.20)
Y
Y
0.1602
3858
Insurance
FM
(6)
0.063***
(4.29)
-0.035*
(-1.86)
-0.069
(-0.82)
Y
Y
0.1979
3210
FM Family
(4)
0.060***
(5.73)
-0.001
(-0.12)
0.014
(0.78)
Y
Y
0.1365
6763
Mutual
FM
(5)
0.035**
(2.55)
0.001
(0.21)
0.014
(0.70)
Y
Y
0.1056
7167
Insurance
FM
(6)
0.075***
(8.66)
-0.004
(-0.37)
-0.035
(-0.61)
Y
Y
0.1472
6054
FM Family
(4)
0.062***
(8.45)
-0.011**
(-2.59)
Y
Y
0.1701
5714
Panel D: Matching Within Family
Fund Hierarchy
Employee Specialty
Team Dummy
Control Variables
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
0.047***
(3.17)
-0.024**
(-2.04)
0.033
(1.11)
Y
Y
Y
Fund
0.1233
7352
All Institutions
IV
FM
(2)
(3)
0.065***
0.046***
(3.42)
(4.97)
-0.002
-0.027***
(-0.05)
(-4.21)
0.024
0.030
(0.54)
(1.20)
Y
Y
Y
Y
Y
0.88
Fund
0.1206
0.1754
7352
7352
Panel E: Matching Across Family
Fund Hierarchy
Employee Specialty
Team Dummy
Control Variables
Time Dummies
Fund Type Dummies
Clustering
Hansen J (p-value)
(Average) R-squared
Number of observations
OLS
(1)
0.058***
(4.11)
0.001
(0.10)
-0.003
(-0.10)
Y
Y
Y
Fund
0.1009
14172
All Institutions
IV
FM
(2)
(3)
0.076***
0.057***
(4.43)
(6.90)
-0.014
-0.002
(-0.62)
(-0.45)
-0.034
-0.006
(-0.99)
(-0.37)
Y
Y
Y
Y
Y
Fund
0.83
0.0987
0.1291
14172
14172
28
Table III
Fund Hierarchy and Portfolio Concentration
This table presents the regression results of fund portfolio concentration on fund hierarchical structure. We estimate the
following equation:
Herfini ,t = α + β × Hierarchyi ,t + δ × X i ,t −1 + ε i ,t ,
where
Herfini ,t
hierarchy and
represents the portfolio concentration (herfindal index) of fund i at quarter t,
X i ,t −1
Hierachyi ,t
is fund
are other control variables. The definitions are detailed in the appendix.
Panel A summarizes the sample mean of fund portfolio concentration at different levels of fund hierarchy. We also provide
univariate tests of fund concentration regarding to single vs. multi- fund hierarchy. Multi-hierarchy means the number of fund
hierarchies to be greater than 1. We always report mutual funds and insurance companies separately. For robustness checks we
construct two matching samples, one within fund family and one across fund families. For each multi-hierarchy fund, we first
select another single hierarchy fund from the same fund family and most similar in terms of fund size, then combine the matched
single hierarchy funds with the original multi-hierarchy funds as our “matching within fund family” sample. The “matching
aross fund family” sample is constructed similarly except that the matched single-hierarchy fund is chosen from different fund
families but belonging to the same fund type (mutual funds, insurance companies, pension funds etc). We report the results for
the main sample as well as the two matching samples. Both two tailed T-test and Wilconxon rank-sum test are performed to test
the differences. The number of observations (fund-quarter) is given in the parentheses.
The analysis in Panel B is based on the main sample including funds owned by mutual fund families, insurance companies
and other institutions (annunities and pension funds). We add the fund type dummies across all specifications. Column (1) is
OLS regression with standard errors clustering at fund level. To address the possible engogeneity issue of fund-specific structures,
we implement an IV regression (2SLS) in Column (2) where family level structures are chosen as instruments. Specifically, we
instrument fund structure variables using the following variables: family hierarchy (mean of fund hierarchy within a family),
family employee specialty (mean of employee specialty within a family), family team (mean of team dummy within a family), the
interaction of family hierarchy with financial center dummy and the interaction of family employee specialty with financial
center dummy. Hansen’s J statistic (p-value) is reported to examine the quality of instruments. The standard errors are clustered
at fund level. Column (3) is estimated under a Fama-Mecbeth (1973) framework where the reported coefficients are time-series
averages of 29 quarterly cross-sectional slope estimates. Column (4) is done by first calculating family averages of all the
variables and then running Fama-Mecbeth regressions based on the family averages. Column (5) and (6) are estimated in the
same way as in Column (3) but only based on funds owned by mutual fund families and insurance companies.
Panel C has the exact same specifications as in Panel B but for the team-managed subsample where the fund has more than
one employee. Panel D and Panel E are based on the matching sample within and across fund families respectively with the same
specifications as in Panel B. For the sake of brevity we only report the coefficients of interested variables from Panel C to Panel
E. ***, ** and * represent significance levels at 1%, 5% and 10% respectively with t-statistics given in parentheses.
29
Table III (Cont’d)
Panel A: Univariate Results
Fund Portfolio Distance
by Fund Hierarchy
Main Sample
Mutual
Insurance
Matching Sample
(within family)
Mutual
Insurance
Matching Sample
(across family)
Mutual
Insurance
0.064
(9259)
0.052
(3034)
0.030
(367)
0.013
(51)
0.081
(2890)
0.042
(3342)
0.016
(898)
0.019
(140)
0.050
(2113)
0.049
(1783)
0.030
(237)
0.012
(42)
0.046
(847)
0.030
(1723)
0.011
(612)
0.022
(95)
0.058
(4646)
0.051
(2569)
0.030
(339)
0.013
(47)
0.067
(2227)
0.041
(3081)
0.016
(870)
0.020
(135)
T-test
Multiple vs. Single Hierarchy
-7.66***
-21.14***
-1.34
-8.93***
-5.18***
-15.59***
Wilconxon Test
Multiple vs. Single Hierachy
-12.75***
-26.22***
-8.27***
-10.99***
-10.80***
-21.32***
FM Family
(4)
-0.016***
(-13.34)
0.000
(-0.09)
-0.008***
(-4.02)
-0.012***
(-13.66)
-0.003***
(-5.86)
-0.001***
(-2.76)
-0.004
(-1.28)
-0.154
(-0.37)
0.051
(0.54)
0.000
(0.06)
0.014***
(9.32)
0.245***
(15.70)
Y
0.2876
8023
Mutual
FM
(5)
-0.006***
(-5.08)
-0.001
(-0.85)
-0.011***
(-4.37)
-0.018***
(-12.96)
0.002***
(3.41)
-0.002***
(-6.39)
0.001
(0.52)
0.288
(0.76)
-0.025
(-0.23)
0.018***
(2.66)
0.008***
(5.01)
0.246***
(14.45)
Y
0.2742
12004
Insurance
FM
(6)
-0.034***
(-13.89)
0.004***
(3.48)
0.009**
(2.28)
-0.020***
(-21.06)
0.006***
(9.35)
0.001***
(4.40)
-0.009
(-0.88)
0.239
(0.46)
0.065
(0.33)
-0.006
(-0.44)
-0.003
(-0.93)
0.200***
(10.09)
Y
0.3711
6921
1
2
3
4
Panel B: Main Sample
Fund Hierarchy
Employee Specialty
Team Dummy
Fund Size
Family Size
Portfolio Maturity
Fund Turnover
Fund Return Volatility
Fund Return
Fraction in Investment-grade Bonds
Financial Center Dummy
Const
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
-0.021***
(-8.46)
0.001
(0.62)
-0.004
(-0.91)
-0.019***
(-16.89)
0.003***
(2.72)
-0.001**
(-2.39)
0.001
(0.49)
-0.376
(-1.04)
0.036
(1.01)
0.018***
(2.85)
0.002
(0.53)
0.271***
(13.71)
Y
Y
Fund
0.2138
19990
All Institutions
IV
FM
(2)
(3)
-0.029***
-0.019***
(-9.28)
(-16.30)
0.000
0.001
(0.33)
(1.01)
-0.002
-0.003
(-0.47)
(-1.49)
-0.019***
-0.018***
(-17.04)
(-17.64)
0.003**
0.003***
(2.42)
(6.73)
-0.001**
-0.001***
(-2.24)
(-4.84)
0.002
0.000
(0.64)
(-0.06)
-0.355
0.179
(-0.98)
(0.51)
0.036
0.039
(0.99)
(0.34)
0.018***
0.016**
(2.84)
(2.48)
0.002
0.002*
(0.59)
(1.67)
0.287***
0.251***
(13.41)
(15.00)
Y
Y
Y
0.50
Fund
0.2110
0.2598
19990
19990
30
Table III (Cont’d)
Panel C: Team-managed Subsample
Fund Hierarchy
Employee Specialty
Control Variables
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
-0.021***
(-8.27)
0.003*
(1.93)
Y
Y
Y
Fund
0.2059
13063
All Institutions
IV
FM
(2)
(3)
-0.028***
-0.019***
(-9.05)
(-17.28)
0.003
0.003***
(1.60)
(3.74)
Y
Y
Y
Y
Y
0.41
Fund
0.2023
0.2665
13063
13063
Mutual
FM
(5)
-0.006***
(-5.63)
0.000
(-0.23)
Y
Y
0.2815
5607
Insurance
FM
(6)
-0.034***
(-13.91)
0.004***
(3.85)
Y
Y
0.3659
6452
FM Family
(4)
-0.011***
(-10.51)
0.001
(0.67)
0.002
(0.37)
Y
Y
0.3331
3333
Mutual
FM
(5)
-0.007***
(-3.81)
0.001
(0.73)
0.004*
(1.71)
Y
Y
0.3079
3808
Insurance
FM
(6)
-0.018***
(-8.24)
0.003*
(1.78)
-0.019
(-1.51)
Y
Y
0.3726
3067
FM Family
(4)
-0.012***
(-11.08)
0.000
(-0.41)
-0.011***
(-3.93)
Y
Y
0.2982
6596
Mutual
FM
(5)
-0.005***
(-3.95)
-0.001
(-0.79)
-0.008***
(-2.74)
Y
Y
0.2975
7006
Insurance
FM
(6)
-0.025***
(-16.35)
0.005***
(4.63)
-0.002
(-0.38)
Y
Y
0.3757
6000
FM Family
(4)
-0.014***
(-13.28)
0.002*
(1.80)
Y
Y
0.2520
5833
Panel D: Matching Within Family
Fund Hierarchy
Employee Specialty
Team Dummy
Control Variables
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
-0.014***
(-5.86)
0.002
(1.22)
0.006
(1.36)
Y
Y
Y
Fund
0.1680
7133
All Institutions
IV
FM
(2)
(3)
-0.020***
-0.012***
(-7.72)
(-9.58)
0.003
0.002**
(1.21)
(2.55)
0.004
0.007**
(0.57)
(2.53)
Y
Y
Y
Y
Y
0.29
Fund
0.1643
0.2729
7133
7133
Panel E: Matching Across Family
Fund Hierarchy
Employee Specialty
Team Dummy
Control Variables
Time Dummies
Fund Type Dummies
Clustering
Hansen J (p-value)
(Average) R-squared
Number of observations
OLS
(1)
-0.017***
(-8.43)
0.001
(1.08)
0.000
(-0.06)
Y
Y
Y
Fund
0.2070
13850
All Institutions
IV
FM
(2)
(3)
-0.023***
-0.016***
(-9.47)
(-16.54)
0.001
0.001**
(0.70)
(1.99)
0.002
-0.001
(0.48)
(-0.64)
Y
Y
Y
Y
Y
Fund
0.26
0.2052
0.2741
13850
13850
31
Table IV
Fund Hierarchy and Herding
This table presents the regression results of fund herding on fund hierarchical structure. We estimate the following equation:
Herding i ,t = α + β × Hierarchyi ,t + δ × X i ,t −1 + ε i ,t ,
where
Herding i ,t
represents fund herding tendency of fund i at quarter t,
Hierachyi ,t
is fund hierarchy and
X i ,t −1
are
other control variables. The definitions are detailed in the appendix.
Panel A summarizes the sample mean of fund herding at different levels of fund hierarchy. We also provide univariate
tests of fund herding regarding to single vs. multi- fund hierarchy. Multi-hierarchy means the number of fund hierarchies to be
greater than 1. We always report mutual funds and insurance companies separately. For robustness checks we construct two
matching samples, one within fund family and one across fund families. For each multi-hierarchy fund, we first select another
single hierarchy fund from the same fund family and most similar in terms of fund size, then combine the matched single
hierarchy funds with the original multi-hierarchy funds as our “matching within fund family” sample. The “matching aross fund
family” sample is constructed similarly except that the matched single-hierarchy fund is chosen from different fund families but
belonging to the same fund type (mutual funds, insurance companies, pension funds etc). We report the results for the main
sample as well as the two matching samples. Both two tailed T-test and Wilconxon rank-sum test are performed to test the
differences. The number of observations (fund-quarter) is given in the parentheses.
The analysis in Panel B is based on the main sample including funds owned by mutual fund families, insurance companies
and other institutions (annunities and pension funds). We add the fund type dummies across all specifications. Column (1) is
OLS regression with standard errors clustering at fund level. To address the possible engogeneity issue of fund-specific structures,
we implement an IV regression (2SLS) in Column (2) where family level structures are chosen as instruments. Specifically, we
instrument fund structure variables using the following variables: family hierarchy (mean of fund hierarchy within a family),
family employee specialty (mean of employee specialty within a family), family team (mean of team dummy within a family), the
interaction of family hierarchy with financial center dummy and the interaction of family employee specialty with financial
center dummy. Hansen’s J statistic (p-value) is reported to examine the quality of instruments. The standard errors are clustered
at fund level. Column (3) is estimated under a Fama-Mecbeth (1973) framework where the reported coefficients are time-series
averages of 29 quarterly cross-sectional slope estimates. Column (4) is done by first calculating family averages of all the
variables and then running Fama-Mecbeth regressions based on the family averages. Column (5) and (6) are estimated in the
same way as in Column (3) but only based on funds owned by mutual fund families and insurance companies.
Panel C has the exact same specifications as in Panel B but for the team-managed subsample where the fund has more than
one employee. Panel D and Panel E are based on the matching sample within and across fund families respectively with the same
specifications as in Panel B. For the sake of brevity we only report the coefficients of interested variables from Panel C to Panel
E. ***, ** and * represent significance levels at 1%, 5% and 10% respectively with t-statistics given in parentheses.
32
Table IV (Cont’d)
Panel A: Univariate Results
Fund Portfolio Distance
by Fund Hierarchy
Main Sample
Mutual
Insurance
Matching Sample
(within family)
Mutual
Insurance
Matching Sample
(across family)
Mutual
Insurance
1.420
(8416)
1.532
(2869)
1.620
(355)
1.743
(50)
1.179
(2308)
1.232
(2944)
1.423
(775)
2.454
(123)
1.525
(1811)
1.682
(1544)
1.917
(209)
1.933
(38)
1.375
(731)
1.321
(1467)
1.732
(474)
2.642
(78)
1.430
(3953)
1.559
(2240)
1.666
(308)
1.917
(42)
1.162
(1738)
1.236
(2599)
1.511
(706)
2.521
(113)
T-test
Multiple vs. Single Hierarchy
2.19**
2.20**
2.01**
0.98
2.18**
2.67***
Wilconxon Test
Multiple vs. Single Hierachy
2.92***
3.03***
2.54**
0.93
2.49**
3.35***
FM Family
(4)
0.216***
(5.19)
-0.085***
(-4.32)
-0.097
(-1.02)
0.022
(0.82)
0.072***
(3.90)
0.013*
(1.69)
1.164***
(7.12)
50.083***
(5.12)
1.573
(0.48)
0.206
(1.01)
0.207***
(3.52)
-1.498***
(-3.70)
Y
0.1485
7370
Mutual
FM
(5)
0.152**
(2.46)
-0.084***
(-3.28)
-0.072
(-1.18)
0.010
(0.41)
0.026
(1.40)
0.027***
(2.89)
0.733***
(5.70)
31.796***
(4.18)
3.407
(1.15)
0.348**
(2.31)
0.290***
(3.68)
-0.745*
(-1.80)
Y
0.0878
11084
Insurance
FM
(6)
0.118*
(1.74)
-0.211***
(-4.73)
0.031
(0.16)
-0.024
(-0.99)
-0.001
(-0.03)
-0.012
(-1.25)
2.042***
(6.44)
35.922*
(1.82)
-1.717
(-0.36)
1.039***
(2.92)
0.063
(0.70)
-0.663
(-1.02)
Y
0.1739
5867
1
2
3
4
Panel B: Main Sample
Fund Hierarchy
Employee Specialty
Team Dummy
Fund Size
Family Size
Portfolio Maturity
Fund Turnover
Fund Return Volatility
Fund Return
Fraction in Investment-grade Bonds
Financial Center Dummy
Const
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
0.153***
(3.25)
-0.105***
(-3.44)
-0.069
(-0.87)
-0.013
(-0.74)
0.016
(0.89)
0.022***
(2.69)
0.825***
(10.10)
25.416***
(3.14)
2.567**
(2.41)
0.314**
(2.32)
0.284***
(3.68)
-1.114***
(-2.95)
Y
Y
Fund
0.0658
17942
All Institutions
IV
FM
(2)
(3)
0.186***
0.155***
(3.23)
(3.93)
-0.145***
-0.097***
(-3.99)
(-4.57)
-0.083
-0.054
(-0.86)
(-0.99)
-0.012
-0.013
(-0.68)
(-0.76)
0.010
0.019
(0.56)
(1.23)
0.020**
0.019***
(2.48)
(2.92)
0.817***
0.946***
(10.12)
(7.43)
27.157***
27.606***
(3.43)
(4.26)
2.563***
2.410
(2.75)
(1.03)
0.331**
0.341**
(2.48)
(1.99)
0.282***
0.276***
(3.66)
(5.10)
-0.646*
-0.351
(-1.67)
(-0.92)
Y
Y
Y
0.75
Fund
0.0652
0.0993
17942
17942
33
Table IV (Cont’d)
Panel C: Team-managed Subsample
Fund Hierarchy
Employee Specialty
Control Variables
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
0.139***
(2.87)
-0.101**
(-2.33)
Y
Y
Y
Fund
0.0718
11743
All Institutions
IV
FM
(2)
(3)
0.156***
0.161***
(2.66)
(4.10)
-0.175***
-0.089**
(-3.63)
(-2.59)
Y
Y
Y
Y
Y
0.47
Fund
0.0709
0.1204
11743
11743
Mutual
FM
(5)
0.144**
(2.15)
0.011
(0.21)
Y
Y
0.1222
5302
Insurance
FM
(6)
0.120*
(1.75)
-0.227***
(-4.68)
Y
Y
0.1792
5483
FM Family
(4)
0.290***
(3.19)
-0.150**
(-2.13)
0.239
(1.31)
Y
Y
0.2223
3164
Mutual
FM
(5)
0.196**
(2.17)
-0.018
(-0.24)
-0.016
(-0.10)
Y
Y
0.1763
3602
Insurance
FM
(6)
0.067
(0.63)
-0.168*
(-1.82)
0.154
(0.78)
Y
Y
0.3019
2750
FM Family
(4)
0.166***
(3.38)
-0.078**
(-2.31)
-0.140
(-1.47)
Y
Y
0.1626
6083
Mutual
FM
(5)
0.155**
(2.05)
-0.056
(-1.62)
-0.129*
(-1.73)
Y
Y
0.1156
6543
Insurance
FM
(6)
0.129*
(1.81)
-0.210***
(-4.44)
-0.074
(-0.34)
Y
Y
0.1891
5156
FM Family
(4)
0.168***
(4.54)
-0.062
(-1.65)
Y
Y
0.1682
5493
Panel D: Matching Within Family
Fund Hierarchy
Employee Specialty
Team Dummy
Control Variables
Time Dummies
Fund Type Dummies
Hansen J (p-value)
Clustering
(Average) R-squared
Number of observations
OLS
(1)
0.184***
(2.87)
-0.129**
(-2.01)
0.045
(0.34)
Y
Y
Y
Fund
0.0755
6605
All Institutions
IV
FM
(2)
(3)
0.231***
0.180***
(3.06)
(2.99)
-0.228***
-0.084
(-2.98)
(-1.23)
0.088
0.066
(0.55)
(0.49)
Y
Y
Y
Y
Y
0.81
Fund
0.0747
0.0757
6605
6605
Panel E: Matching Across Family
Fund Hierarchy
Employee Specialty
Team Dummy
Control Variables
Time Dummies
Fund Type Dummies
Clustering
Hansen J (p-value)
(Average) R-squared
Number of observations
OLS
(1)
0.174***
(3.49)
-0.092***
(-2.79)
-0.092
(-1.04)
Y
Y
Y
Fund
0.0670
12522
All Institutions
IV
FM
(2)
(3)
0.191***
0.179***
(3.24)
(3.61)
-0.140***
-0.085***
(-3.66)
(-2.91)
-0.089
-0.118*
(-0.88)
(-1.75)
Y
Y
Y
Y
Y
Fund
0.89
0.0663
0.1229
12522
12522
34
Table V
Fund Hierarchy and Performance
This table presents the regression results of fund performance on fund hierarchical structure. We estimate the following equation:
Alphai ,t = α + β × Hierarchyi ,t + δ × X i ,t −1 + ε i ,t ,
where
Alphai ,t
represents the alpha of fund i at month t,
Hierachyi ,t
is fund hierarchy and
X i ,t −1
are other control variables. The definitions are detailed in the appendix.
Panel E summarizes the sample mean of fund portfolio performance at different levels of fund hierarchy. We also provide univariate tests of fund performance regarding to single
vs. multi- fund hierarchy. Fund alpha is reported in percentage terms. Multi-hierarchy means the number of fund hierarchies to be greater than 1. We always report mutual funds and
insurance companies separately. For robustness checks we construct two matching samples, one within fund family and one across fund families. For each multi-hierarchy fund, we first
select another single hierarchy fund from the same fund family and most similar in terms of fund size, then combine the matched single hierarchy funds with the original
multi-hierarchy funds as our “matching within fund family” sample. The “matching aross fund family” sample is constructed similarly except that the matched single-hierarchy fund
is chosen from different fund families but belonging to the same fund type (mutual funds, insurance companies, pension funds etc). We report the results for the main sample as well
as the two matching samples. Both two tailed T-test and Wilconxon rank-sum test are performed to test the differences. The number of observations (fund-quarter) is given in the
parentheses.
The analysis in Panel A is based on the main sample including mutual funds, insurance companies and other institutions (annunities and pension funds). We put fund type
dummies in all specifications. Column (1) is OLS regression with standard errors clustering at fund level. Column (2) is OLS regression with standard errors clustering at fund family
level. To address the possible endogeneity issue of fund-specific structures, we implement an IV regression (2SLS) in Column (3) where family level structures are chosen as instruments.
Specifically, we instrument fund structure variables using the following variables: family hierarchy (median of fund hierarchy within a family), family employee specialty (median of
employee specialty within a family), family team (median of team dummy within a family), the interaction of family hierarchy with financial center dummy and the interaction of
family employee specialty with financial center dummy. Hansen’s J statistic (p-value) is reported to examine the quality of instruments. The standard errors are clustered at fund level.
Column (4) is estimated under a Fama-Mecbeth (1973) framework where the reported coefficients are time-series averages of 60 monthly cross-sectional slope estimates. Column (5)
is done by first calculating family averages of all the variables and then running Fama-Mecbeth regressions based on the family averages. Column (6) and (7) are for funds owned by
mutual fund families and insurance companies. From Column (8) to Column (10) we add the interaction term of fund hierarchy and a “close investment” dummy. It equals 1 if the
fund portfolio distance is below the third quantile of the sample quarter and 0 otherwise. As before, Panel B is for the team-managed sub-sample where the fund has more than one
employee. Panel C and Panel D are based on the matching sample within and across fund families respectively.
In Panel F we compare the impact of fund hierarchy on the fund performance of investing in high rated bonds with that of investing in low rated bonds. High rated bonds refer
to bonds with Moody’s credit rating above A3. Low rated bonds are bonds with Moody’s credit rating from B3 to BBB1. For each fund we estimate two portfolio alphas separately.
One is based on the value-wighted return of investing in high rated bonds while the other is based on the return of investing in low rated bonds. The estimation procedure is the same
as described in the appendix. We stack the high and low rated alphas together and create a rating category dummy which equals 1 if it is a low rated alpha and 0 otherwise. Our focus
is the interaction term of fund hierarchy and the rating category dummy. Standard errors are clustered at the fund level. ***, ** and * represent significance levels at 1%, 5% and
10% respectively with t-statistics given in parentheses.
35
Table V (Cont’d)
Panel A: Main Sample
All
Institutions
Fund Hierarchy
OLS
(1)
-0.164***
(-3.51)
OLS
(2)
-0.164***
(-3.84)
IV
(3)
-0.213***
(-3.61)
FM
(4)
-0.148***
(-6.53)
FM
(Family)
(5)
-0.087***
(-3.28)
Mutual
Insurance
All
Institution
Mutual
Insurance
FM
(6)
-0.140***
(-3.76)
FM
(7)
-0.133***
(-4.24)
FM
(8)
-0.102***
(-4.37)
-0.153***
FM
(9)
-0.093**
(-2.39)
-0.199***
FM
(10)
-0.107***
(-3.61)
-0.081**
(-5.49)
0.149***
(3.05)
0.016
(1.31)
0.052*
(1.67)
-0.012
(-1.23)
-0.016**
(-1.97)
-0.024*
(-1.84)
0.031
(0.63)
19.889
(1.62)
-4.433
(-1.00)
0.973***
(3.69)
-0.085***
(-3.52)
-0.403
(-1.12)
Y
0.3075
30332
(-3.74)
0.216***
(2.70)
0.056**
(2.26)
0.015
(0.35)
-0.011
(-1.04)
-0.024**
(-2.17)
-0.037***
(-3.02)
0.095
(1.47)
31.359***
(2.88)
-4.442
(-0.98)
1.038***
(3.85)
-0.003
(-0.13)
-0.502
(-1.33)
0.3217
17738
(-2.23)
0.057
(0.88)
-0.038*
(-1.81)
0.161
(1.56)
-0.022**
(-2.51)
-0.009
(-1.18)
0.006
(0.48)
-0.010
(-0.13)
-3.429
(-0.20)
-1.382
(-0.22)
1.074***
(3.67)
-0.142***
(-4.17)
-0.048
(-0.11)
0.3036
11747
Fund Hierarchy *
Close Investment Dummy
Close Investment Dummy
Employee Specialty
Team Dummy
Fund Size
Family Size
Portfolio Maturity
Fund Turnover
Fund Return Volatility
Fund Return
Fraction in Investment-grade Bonds
Financial Center Dummy
Const
Time Dummies
Fund Type Dummies
Clustering
Hansen J (p-value)
(Average) R-squared
Number of observations
0.012
(0.58)
-0.031
(-0.41)
-0.005
(-0.30)
-0.012
(-0.86)
-0.025***
(-3.51)
-0.041
(-0.69)
66.294***
(7.36)
-13.409***
(-10.57)
0.882***
(6.66)
-0.081
(-1.43)
-1.530***
(-4.37)
Y
Y
Fund
0.2722
30332
0.012
(0.56)
-0.031
(-0.39)
-0.005
(-0.28)
-0.012
(-0.77)
-0.025***
(-3.18)
-0.041
(-0.72)
66.294***
(6.26)
-13.409***
(-9.13)
0.882***
(6.33)
-0.081
(-1.19)
-1.530***
(-3.62)
Y
Y
Family
0.2722
30332
0.017
(0.62)
0.009
(0.09)
-0.006
(-0.38)
-0.012
(-0.88)
-0.025***
(-3.49)
-0.039
(-0.66)
66.389***
(7.40)
-13.413***
(-10.58)
0.884***
(6.68)
-0.075
(-1.34)
-1.801***
(-4.75)
Y
Fund
0.82
0.2720
30332
0.014
(1.19)
0.061*
(1.94)
-0.011
(-1.13)
-0.018**
(-2.21)
-0.024*
(-1.86)
0.027
(0.56)
19.595
(1.60)
-4.525
(-1.03)
0.924***
(3.44)
-0.089***
(-3.63)
-0.333
(-0.92)
0.3032
30332
36
-0.001
(-0.08)
-0.019
(-0.40)
0.008
(0.99)
-0.019**
(-2.08)
-0.021
(-1.44)
-0.033
(-0.44)
5.581
(0.41)
-5.103
(-1.01)
0.663**
(2.18)
-0.062**
(-2.32)
-0.031
(-0.08)
0.3436
12667
0.049**
(2.23)
0.017
(0.41)
-0.010
(-0.93)
-0.026**
(-2.50)
-0.038***
(-3.08)
0.086
(1.36)
33.234***
(3.09)
-4.706
(-1.05)
1.019***
(3.68)
-0.008
(-0.33)
-0.410
(-1.11)
0.3142
17738
-0.041*
(-1.94)
0.160
(1.58)
-0.020**
(-2.40)
-0.011
(-1.43)
0.008
(0.57)
-0.012
(-0.14)
-3.017
(-0.18)
-1.401
(-0.23)
0.991***
(3.37)
-0.134***
(-3.97)
0.024
(0.05)
Y
0.2972
11747
Table V (Cont’d)
Panel B: Team-managed Subsample
Fund Hierarchy
OLS
(1)
-0.160***
(-3.35)
OLS
(2)
-0.160***
(-3.79)
All Institutions
IV
(3)
-0.199***
(-3.31)
FM
(4)
-0.136***
(-6.08)
FM (Family)
(5)
-0.079***
(-2.82)
Mutual
FM
(6)
-0.102**
(-2.59)
Insurance
FM
(7)
-0.134***
(-4.36)
Fund Hierarchy *
Close Investment Dummy
Close Investment Dummy
Same Specifications as in Panel A
Number of Observations
Y
20447
Y
20447
Y
20447
OLS
(2)
-0.154***
(-3.71)
All Institutions
IV
(3)
-0.178**
(-2.42)
Y
20447
Y
8877
Y
8180
Y
11478
Mutual
FM
(6)
-0.290***
(-5.62)
Insurance
FM
(7)
-0.117***
(-3.65)
All Institution
FM
(8)
-0.085***
(-4.10)
-0.163***
Mutual
FM
(9)
-0.032
(-0.71)
-0.252***
Insurance
FM
(10)
-0.102***
(-3.64)
-0.095**
(-4.05)
0.180***
(2.67)
Y
20447
(-3.22)
0.297**
(2.23)
Y
8180
(-2.54)
0.088
(1.29)
Y
11478
All Institution
FM
(8)
-0.118***
(-3.74)
-0.112***
Mutual
FM
(9)
-0.237***
(-3.69)
-0.183**
Insurance
FM
(10)
-0.056
(-1.56)
-0.124***
(-3.69)
0.152**
(2.59)
Y
11644
(-1.99)
0.233*
(1.67)
Y
5325
(-2.78)
0.213**
(2.34)
Y
6149
All Institution
FM
(8)
-0.104***
(-3.95)
-0.118***
Mutual
FM
(9)
-0.144***
(-3.05)
-0.174***
Insurance
FM
(10)
-0.092***
(-3.06)
-0.088**
(-4.53)
0.066
(1.56)
Y
20235
(-3.06)
0.100
(1.14)
Y
9581
(-2.53)
0.085
(1.29)
Y
9982
Panel C: Matching Within Family
Fund Hierarchy
OLS
(1)
-0.154**
(-2.61)
FM
(4)
-0.154***
(-5.31)
FM (Family)
(5)
-0.137***
(-3.29)
Fund Hierarchy *
Close Investment Dummy
Close Investment Dummy
Same Specifications as in Panel A
Number of Observations
Y
11644
Y
11644
Y
11644
OLS
(2)
-0.140***
(-3.12)
All Institutions
IV
(3)
-0.161**
(-2.60)
Y
11644
Y
5348
Y
5325
Y
6149
Mutual
FM
(6)
-0.180***
(-3.98)
Insurance
FM
(7)
-0.121***
(-3.53)
Panel D: Matching Across Family
Fund Hierarchy
OLS
(1)
-0.140***
(-2.79)
FM
(4)
-0.135***
(-5.17)
FM (Family)
(5)
-0.084**
(-2.62)
Fund Hierarchy *
Close Investment Dummy
Close Investment Dummy
Same Specifications as in Panel A
Number of Observations
Y
20235
Y
20235
Y
20235
Y
20235
37
Y
9973
Y
9581
Y
9982
Table V (Cont’d)
Panel E: Univariate Results
Fund Portfolio Distance
by Fund Hierarchy
1
2
3
4
T-test: Multiple vs. Single Hierarchy
Wilconxon Test: Multiple vs. Single
Hierachy
Main Sample
Mutual
Insurance
Matching Sample
(within family)
Mutual
Insurance
Matching Sample
(across family)
Mutual
Insurance
-0.21
(13811)
-0.28
(4278)
-0.76
(544)
-0.93
(89)
-3.84***
0.18
(5008)
0.01
(5533)
0.00
(1629)
0.05
(132)
-4.60***
-0.15
(3060)
-0.36
(2493)
-0.83
(338)
-0.94
(75)
-4.84***
0.072
(1520)
0.075
(3672)
0.027
(1301)
-0.116
(96)
-0.22
-0.28
(6453)
-0.29
(3634)
-0.82
(507)
-0.97
(86)
-2.11**
0.13
(3733)
-0.02
(5089)
-0.01
(1599)
0.05
(132)
-3.72***
-4.76***
-4.06***
-5.43***
-0.20
-2.94***
-3.47***
Panel F: Interaction with Rating Category
All Institutions
OLS
FM
(1)
(2)
Fund Hierarchy
Fund Hierarchy *
Rating Category
Rating Category
Employee Specialty
Team Dummy
Fund Size
Family Size
Portfolio Maturity
Fund Turnover
Fund Return Volatility
Fund Return
Fraction: Investment-grade Bonds
Financial Center Dummy
Const
Time Dummies
Fund Type Dummies
Clustering at
(Average) R-squared
Number of observations
Mutual
Insurance
OLS
(3)
FM
(4)
OLS
(5)
FM
(6)
0.046**
(2.07)
-0.218***
-0.006
(-0.67)
-0.104***
0.016
(0.46)
-0.161***
-0.002
(-0.10)
-0.081***
0.038
(1.26)
-0.212***
-0.016
(-1.08)
-0.100***
(-5.75)
-0.197***
(-2.93)
0.013
(1.34)
0.003
(0.09)
-0.004
(-0.58)
0.010
(1.57)
-0.004
(-1.13)
-0.010
(-0.35)
31.452***
(7.83)
-6.524***
(-10.77)
0.742***
(11.04)
-0.090***
(-3.55)
-0.749***
(-4.66)
Y
Y
Fund
0.2153
49097
(-5.56)
-0.406***
(-4.36)
0.014**
(2.21)
0.019
(1.38)
-0.010**
(-2.29)
0.002
(0.61)
-0.002
(-0.49)
0.034
(1.36)
9.214**
(2.56)
-0.768
(-0.38)
0.737***
(5.55)
-0.093***
(-6.91)
-0.588***
(-3.33)
Y
0.2910
49097
(-2.71)
-0.219**
(-2.44)
0.021*
(1.70)
0.002
(0.06)
-0.002
(-0.15)
0.010
(0.97)
-0.008**
(-2.16)
0.006
(0.16)
40.565***
(8.03)
-8.030***
(-10.53)
0.823***
(10.29)
-0.073**
(-1.99)
-2.022***
(-7.29)
Y
Y
Fund
0.1714
27432
(-3.56)
-0.410***
(-3.81)
0.010
(1.10)
-0.003
(-0.15)
-0.006
(-1.18)
0.003
(0.75)
-0.005
(-1.06)
0.078**
(2.02)
13.711***
(3.29)
-2.636
(-1.27)
0.850***
(5.63)
-0.094***
(-6.28)
-0.826***
(-4.39)
Y
0.2977
27432
(-3.76)
-0.246**
(-2.14)
-0.014
(-0.85)
0.217***
(2.94)
-0.008
(-0.76)
0.001
(0.17)
0.004
(0.65)
0.003
(0.05)
6.789
(1.08)
-1.391
(-1.46)
0.634***
(4.27)
-0.110***
(-3.16)
-0.963***
(-3.57)
Y
Y
Fund
0.2970
20252
(-2.98)
-0.419***
(-4.48)
-0.007
(-0.56)
0.100**
(2.51)
-0.015***
(-3.58)
-0.001
(-0.18)
-0.001
(-0.26)
-0.015
(-0.29)
7.728
(1.54)
7.116*
(1.83)
0.583***
(4.54)
-0.057***
(-2.98)
-0.198
(-0.83)
Y
0.2933
20252
38
Table VI
Changes in Fund Management with Changes in Fund Hierarchy
This table presents the regression results of changes in fund management on the changes in fund hierarchical structure. Here we
only focus on the subsample (fund-quarter) where there is a change in the fund hierarchical structure from quarter t-1 to quartet
t. We estimate the following regression:
∆FundManagementi ,t = α + β × ∆Hierarchyi ,t + δ × ∆X i ,t + γ × X i ,t −1 + FundManagementi ,t −1 + ε i ,t ,
where from Column (1) to Column (5)
∆FundManagement i ,t
represents the change of fund portfolio distance, the change
of fund herding, the change of fund portfolio concentration, the change of fund raw return (cumulative, quarterly) and the
change of fund alpha (cumulative, quarterly) respectively from quarter t -1 to quarter t.
hierarchy and
appendix.
∆X i ,t
∆Hierachyi ,t
is the change of fund
are the changes of other control variables from quarter t-1 to quarter t. The definitions are detailed in the
FundManagementi , −1
is the lagged dependent variable at quarter t-1. The standard errors across all
specifications are clustered at fund level and we always include time dummies and fund type dummies. ***, ** and * represent
significance levels at 1%, 5% and 10% respectively with t-statistics given in parentheses.
Change in Fund Hierarchy
Change in Employee Specialty
Change in Team Dummy
Change in Fund Size
Change in Family Size
Change in Portfolio Maturity
Change in Fraction in Investment-grade Bonds
Lag Fund Hierarchy
Lag Employee Specialty
Lag Team Dummy
Lag Fund Size
Lag Family Size
Lag Portfolio Maturity
Lag Fraction in Investment-grade Bonds
Lag Dependent Variable
Const
Time Dummies
Fund Type Dummies
R-squared
Number of observations
ΔFund
Portfolio
Distance
(1)
0.024***
(2.81)
-0.001
(-0.07)
0.060***
(2.80)
0.024
(1.38)
-0.026
(-1.55)
0.079
(1.58)
-0.392***
(-3.84)
0.026**
(2.23)
-0.001
(-0.09)
0.051
(1.38)
0.006*
(1.81)
-0.009**
(-2.42)
0.022
(1.52)
-0.136***
(-5.89)
-0.301***
(-12.75)
1.987***
(11.01)
Y
Y
0.1808
4036
39
ΔFund
Herding
(2)
0.191***
(3.03)
-0.044
(-0.67)
-0.119
(-0.56)
0.939***
(4.47)
-0.197
(-1.37)
0.937**
(2.03)
0.135
(0.17)
0.193**
(2.36)
0.043
(0.49)
-0.487
(-1.29)
-0.085***
(-3.15)
0.034
(0.98)
0.188*
(1.68)
0.255
(1.47)
-75.966***
(-34.68)
-0.538
(-0.63)
Y
Y
0.4166
3706
ΔFund
Portfolio
Concentration
(3)
-0.236**
(-2.06)
0.119
(0.70)
-0.113
(-0.21)
-3.737***
(-4.17)
0.442
(1.28)
-4.548***
(-3.47)
12.344***
(3.66)
-0.307**
(-2.16)
0.113
(0.59)
1.959**
(1.97)
-0.349***
(-4.39)
0.137
(1.57)
-1.149***
(-3.70)
0.751*
(1.73)
-15.675***
(-5.38)
3.721*
(1.80)
Y
Y
0.1892
4196
ΔFund
Raw
Return
(4)
-0.145***
(-2.62)
-0.071
(-0.94)
-0.020
(-0.05)
-0.247*
(-1.72)
-0.034
(-0.29)
0.152
(0.83)
1.995**
(2.20)
-0.100
(-1.40)
-0.045
(-0.45)
-0.052
(-0.11)
0.029
(1.35)
-0.075***
(-2.68)
0.417***
(4.06)
0.772***
(3.07)
-73.846***
(-20.69)
1.601
(1.59)
Y
Y
0.7773
1622
ΔFund
Alpha
(5)
-0.207**
(-2.11)
0.189*
(1.74)
0.768
(1.51)
-0.371
(-1.45)
-0.114
(-0.54)
-0.299
(-0.77)
0.049
(0.04)
-0.162
(-1.27)
0.258
(1.39)
0.450
(0.59)
0.039
(0.93)
-0.068
(-1.28)
0.122
(0.63)
0.432
(1.08)
-26.252***
(-14.77)
-1.138
(-0.74)
Y
Y
0.3308
1622