Intermediated Investment Management

Neal M. Stoughton Youchang Wu Josef Zechner∗

First Dra : April 2006

Final Version: June 2010

ABSTRACT

Intermediaries such as financial advisers serve as an interface between por olio managers and in-

vestors. A large frac on of their compensa on is o en provided through kickbacks from the por olio

manager. We provide an explana on for the widespread use of intermediaries and kickbacks. Depend-

ing on the degree of investor sophis ca on, kickbacks are used either for price discrimina on or ag-

gressive marke ng. We explore the effects of these arrangements on fund size, flows, performance and

investor welfare. Kickbacks allow higher management fees to be charged, thereby lowering net returns.

Compe on among ac ve por olio managers reduces kickbacks and increases the independence of

advisory services.

∗
Neal Stoughton is at the School of Banking and Finance at UNSW Sydney. Youchang Wu is at the Wisconsin School of Business
at University of Wisconsin-Madison. Josef Zechner is at the Department of Finance, Accoun ng and Sta s cs at Vienna University
of Economics and Business. We thank the Gutmann Center for Por olio Management at the University of Vienna for financial
support, as well as the Social Sciences and Humani es Research Council of Canada (SSHRC). We also appreciate helpful discussions
with Daniel Bergstresser, Ingolf Di mann, Alain Durre, Fei Ding, David Feldman, David Gallagher, Rick Green, Christopher Hennessy,
Adolf Hengstschläger, Ernst Maug, Judy Posnikoff, Sco Schaefer, William Sharpe, Keith Wong, and the helpful sugges ons of an
anonymous referee and Campbell Harvey (Editor). The paper has been presented at Western Finance Associa on mee ngs in
2007, European Finance Associa on mee ngs in 2007, Financial Ins tu on Research Society mee ngs in 2008, American Finance
Associa on Mee ngs in 2009, University of Hong Kong, Peking University, University of New South Wales, University of Utah,
Erasmus University Ro erdam, University of Mannheim, University of Cologne, Norwegian School of Management (BI), Ca' Forscari
(Venice), Claremont-McKenna College, Chapman University, UC Riverside, Nanyang Technological University, Na onal University
of Singapore, Singapore Management University, Hong Kong University of Science and Technology, Vienna Graduate School of
Finance and the University of Warwick.

1
The money management industry has been recognized as having substan al influence on financial markets.

The Investment Adviser Associa on has es mated the total amount of assets managed by SEC-registered

investment advisers at $42.3 trillion at its peak in April 2008.1 An important reason for the enormous size

of the money management business is that there are o en mul ple layers of advisory services between

investors and the ul mate por olio manager. In many cases, investors do not delegate their wealth directly

to money managers. They rely on intermediaries. Examples of such intermediaries include financial advisers

who manage separate accounts for their clients, full-service brokers who assist investors to select mutual

funds, pension fund consultants who assist trustees with the selec on of por olio managers, and funds of

funds or feeder funds that iden fy mutual and hedge funds for their investors. According to a survey by the

Investment Company Ins tute, 80% of US households owning mutual funds (outside defined contribu on

plans) u lized professional advisory services in 2007.2 Further, Chen, Hong, and Kubik (2010) document that

the management of 27% of mutual funds in their sample are outsourced by fund management companies

to unaffiliated advisory firms. In the hedge fund industry, now more than one third of all assets under

management are through funds-of-funds according to Hedge Fund Research.3

The role of the intermediary in the money management industry has been the focus of considerable

a en on recently. In the notorious case of Bernard Madoff, the majority of assets directed to his scheme

came through ``feeder funds" (e.g., Fairfield Greenwich). The ``pay to play" nature of public pension plan

inves ng through placement agents was uncovered in the cases of the New York State pension plan as well

as CalPERS.4 Legisla on is currently under considera on that would extend the no on of client fiduciary

duty to the role of brokerage advice in order to mi gate conflicts of interest.

In all these cases, the central issue concerns the way in which intermediaries are compensated. O en

intermediaries are primarily compensated by rebates from the fund management company. This may result

in favorable treatment towards funds with high rebates. A typical example of such arrangements in the case

of mutual funds is that brokers receive direct compensa on by sharing front-end loads, back-end loads

and 12b-1 fees with the management company. Another common prac ce is that intermediaries receive

marke ng and sales support from fund management companies.5 Many other forms of kickbacks exist in

the industry under various revenue sharing agreements. While the rebate-based compensa on scheme is

s ll dominant in mutual fund distribu on, recent years have also seen a trend towards asset-based advisory

2
fees, where financial advisers charge their clients directly. Star ng in 1990, mutual fund advisory programs,

such as the ``wrap account", have become popular. Wrap account managers help the clients select mutual

funds and charge a percentage fee based on assets in the account. According to Strategic Insight, the annual

inflows into such accounts increased from less than $20 billion to an es mated $85-$90 billion in 2007.6

At the end of the compensa on spectrum are financial advisers who are ``fee-only", i.e., they are solely

compensated by their clients and receive no rebates.

Despite their prominent role, intermediaries in the investment management industry have largely been

ignored by the exis ng literature. Most previous studies on delegated por olio management consider only

the bilateral rela onship between investors and por olio managers.7 By contrast, this paper models the

intermediary as a dis nct agent and focuses precisely on the economic role that intermediaries play. We

analyze several ques ons related to investment management intermedia on. Why is intermedia on so

prevalent in the investment management industry? Why is it common prac ce for the intermediary to be

compensated by the por olio manager instead of directly by the client? How do intermedia on and kick-

backs affect fund performance and investor welfare and how do these results accord with empirical facts?

How does compe on among por olio managers affect the compensa on scheme of the intermediary?

Our model consists of investors, a representa ve financial adviser, a passively held pool of assets (e.g.,

index fund) and a pool of assets (ac ve fund) run by a por olio manager. The ac ve fund can involve

trading tradi onal assets (stocks and bonds) or alterna ve assets, such as private equity, foreign currency,

real assets etc., that are not present in the passive fund. When investors are sophis cated they are able

to fully an cipate equilibrium outcomes, while alterna vely, if investors are unsophis cated they can be

persuaded to invest in lower returning assets due to promo onal ac vi es by the adviser. Investors have

heterogeneous wealth levels, and can go directly to the por olio manager or through the indirect channel

by using an adviser. In order to invest directly, investors must pay a fixed cost to iden fy an ac ve por olio

manager who does not underperform the passive fund. As a result, only high networth individuals invest

directly. Por olio managers have market power and op mally select their fees. But as the ac ve fund has

diminishing returns to size, there is an op mal amount of assets invested ac vely. Financial advisers also

charge a fee, which compensates them for their costs of providing asset alloca on services to their clients.

We first derive an equilibrium assuming that the financial advisers are independent and must charge

3
the investors their full costs in order to break even. We then extend this model by allowing the por olio

manager to provide kickbacks to the adviser. These kickbacks can be used by the adviser to cover part of

the costs of opera ons or as marke ng support. We solve for the op mal amount of rebates preferred by

the por olio manager and the impact on management fees, fund sizes and flows. We then extend the anal-

ysis to the case of compe on between ac ve por olio managers. Finally we also derive the equilibrium

without an adviser and compare all the scenarios.

Our major findings are as follows. First, we ra onalize the widespread use of financial advisers. Ad-

visers exist in our model to facilitate the par cipa on of small investors in ac vely managed por olios by

economizing on informa on costs. As long as investors are ra onal and the advisory industry is compe -

ve, the existence of the adviser increases the elas city of investor demand and reduces management fees.

Therefore investor's welfare can be improved by the presence of financial advisers. However this result is

only true if kickbacks from the por olio manager to the adviser do not exist. When kickbacks exist, investor

welfare is always lowered but total welfare, including the benefit to por olio managers, is higher than

without advisers. Second, we explain the widespread use of side payments as a method for compensa ng

advisers. When investors are sophis cated, kickbacks serve as a price discrimina on mechanism, effec-

vely subsidizing the cost of advice to smaller investors. Alterna vely when investors are unsophis cated,

kickbacks support aggressive marke ng of the ac ve fund by the adviser. Surprisingly, when advisers are in-

fluenced by kickbacks from the por olio manager, the use of advisory services increases. Third, kickbacks

are always associated with higher por olio management fees and nega vely impact fund performance,

regardless of investor sophis ca on. When investors are sophis cated, kickbacks only affect the high net-

worth investors; when they are unsophis cated, all investors are nega vely affected. Fourth, the variety

of distribu on channels by which a fund is sold is related to its performance. Underperforming funds are

only sold indirectly. Ac ve funds with performance equal to or above passive funds are sold simultaneously

through direct and indirect channels. Fi h, we find that compe on among ac ve por olio managers re-

duces the use of kickbacks. Therefore the recent trend toward more independent advisory services can be

ra onalized as a consequence of an increasingly compe ve environment. Finally our results point to some

poten al policy implica ons on the regula on of kickbacks. Our model suggests that be er disclosure of

kickbacks and their uses would be beneficial to investors. In fact, if rebates are allowed it is be er for them

4
to be paid in the form of transparent monetary assistance to financial advisers rather than fund specific

promo onal ac vi es (e.g., sales seminars, marke ng materials, etc.).

Several recent studies have tried to measure empirically the economic impact of intermedia on in in-

vestment management. Bergstresser, Chalmers, and Tufano (2009) look at the performance of mutual

funds offered through the brokerage channel as compared to those offered directly to investors.8 They find

that, even before marke ng fees are deducted, risk-adjusted returns are lower for funds offered through

the brokerage channel as compared to those offered directly. Chen et al. (2010) document that mutual

funds managed externally significantly underperform those run internally. Ang, Rhodes-Kropf, and Zhao

(2008) and Brown, Goetzmann, and Liang (2004) have found that funds-of-funds underperform average

hedge funds. In all these cases the use of intermediaries does not appear to bring economic benefits to

investors.

There are also a few empirical papers that examine the poten al conflicts of interest in the mutual fund

distribu on channels more explicitly. Edelen, Evans, and Kadlec (2008) find that ac vely managed funds

improve fund distribu ons by compensa ng their brokers with abnormally high commissions and this leads

to lower fund returns. Christoffersen, Evans, and Musto (2007) find that higher revenue sharing with un-

affiliated brokers leads to more fund inflows, and higher revenue sharing with cap ve brokers mi gates

ou lows. Chen, Yao, and Yu (2007) show that mutual funds managed by insurance companies underper-

form their non-insurance counterparts by more than 1% per year. The authors find that this has to do

with the fact that insurance funds are o en cross-sold through the extensive broker/agent network of their

parent firms.

A seminal paper on the subject of investment management is the AFA Presiden al address of Sharpe

(1981). He analyzes the coordina on failure in the presence of mul ple por olio managers. Recently, this

model has been extended by Bindsbergen, Brandt, and Koijen (2008), who derive a linear performance

benchmark to be er align incen ves. By contrast, we consider the role of an intermediary between the

client and the por olio manager.9

An interes ng paper exploring the role of kickbacks in the medical field is that of Pauly (1979). He con-

siders a medical prac oner who is able to engage in `fee-spli ng' prac ces with a specialist. He finds that

there is no point in prohibi ng such prac ces in a fully compe ve environment because services are pro-

5
vided at marginal cost. However, when there are market imperfec ons such as monopoly or incomplete

pricing of insurance, fee-spli ng can actually improve client welfare. The role of rebates has been exten-

sively studied in the marke ng literature. However, most of the studies, e.g., Gerstner and Hess (1991),

focus on rebates that are provided to end consumers instead of to the intermediary (retailer). One excep-

on is Taylor (2002), who shows that rebates to the retailer can be used as a coordina on mechanism to

align the interests of the manufacturer and the retailer. In a recent paper, Inderst and O aviani (2009)

also address the issue of kickbacks to intermediaries. Their model structure is very different and does not

allow for simultaneous access to the product (the ac ve fund in our paper) through mul ple channels: all

customers must go through an adviser.

In the next sec on we set up the basic model, with behavioral assump ons on the three sets of par c-

ipants in the game: the investors, the financial adviser and the por olio manager. In sec on II we derive

the equilibrium without kickbacks from the por olio manager. Sec on III derives the impact of kickbacks

from the por olio manager to the adviser, considering both the cases of sophis cated and unsophis cated

investors. Our model is generalized to imperfect compe on between por olio managers in sec on IV.

Sec on V considers the equilibrium in a situa on where advisory services are not available. Sec on VI

compares the outcome and welfare in the four alterna ve scenarios. Sec on VII concludes the paper.

I. Model Setup

In this sec on we describe the agents, their behavior and how they interact. There are three classes of

agents in the model: (1) the ac ve por olio manager; (2) the set of financial advisers, modeled as a repre-

senta ve agent; (3) the pool of investors in the economy.

A. Assets and Por olio Managers

There are two types of assets in which investors can invest. First, there is a passive fund, such as an index

fund, with an expected risk-adjusted gross return Rm (i.e., one plus the rate of return). Both investors and

advisers can invest in the passive fund without cost.

The second type of asset is an ac ve fund, whose expected gross return (once again risk-adjusted) is

equal to Rp . The ac ve por olio manager u lizes her exper se in managing the fund. However, because of

6
market impact or limited applicability of the por olio manager's exper se, we assume decreasing returns

to scale in the amount of investment. Specifically, we assume that

Rp = α − γA, (1)

where α represents the expected return on the first dollar of capital invested in the ac ve fund (assumed to

be greater than the passive return) and γ is a coefficient represen ng the rate at which returns decline with

respect to the aggregate amount of funds, A, that are placed with the por olio manager. For a discussion

of this assump on see Berk and Green (2004) and Dangl, Wu, and Zechner (2008).10

In addi on to inves ng in the passive asset and obtaining returns equal to Rm , the investors can choose

to delegate their por olio decisions to a financial adviser who advises mul ple clients, or they can decide

to invest directly with the por olio manager. Because there are poten ally many ac ve managers with

α < Rm , i.e., whose expected return does not exceed the return of the passive fund irrespec ve of fund

size, investors who want to invest directly must pay a fixed screening cost, C0 to avoid them.11 Therefore,

the only method for a direct investor to iden fy a por olio manager who is poten ally superior, α > Rm , is

to pay the cost C0 . In our single period model, it is impossible for some investors to ``free-ride" by wai ng on

the choices of others. In a more general model, it will always pay for some investors to pay such a screening

cost in order to obtain superior returns on ac vely managed funds, before their returns are depreciated by

the diseconomies of scale. Investors can avoid the fixed screening cost if they delegate their funds to the

adviser.

The financial adviser has the exper se required to ascertain poten ally superior por olio managers.

Any fixed cost of informa on that he might incur can be distributed across many individual investors and is

therefore negligible on a per client basis. On the other hand, the adviser is likely to face increased variable

costs related to the amount of assets he allocates to the ac ve por olio manager. As the amount allocated

for a given client increases this is likely to require more detailed and frequent communica on between

adviser and client, poten al conflicts of interest between adviser and client become more severe and the

adviser’s legal risks increase. Typically also, as the adviser handles more assets, he has to deal with a greater

number of clients. As a result, his compliance costs of monitoring client qualifica ons and dealing with

regulatory repor ng requirements increases with assets under management. Furthermore, more staff must

7
be retained to deal with larger numbers of clients and more branch offices must be opened. For simplicity

we therefore assume the adviser incurs a constant marginal cost, cA , propor onal to the amount of capital

allocated to the ac ve fund, and a zero fixed cost. The adviser can also allocate capital costlessly to the

passive fund.12

Finally we describe the nature of the fees that are charged by the por olio manager and the financial

adviser. We assume that the adviser charges a propor onal fee, fA , based on the end-of-period value of

ac vely managed assets. This fee is determined endogenously in the model due to compe on among

advisers. In fact we assume perfectly compe ve behavior so that the fee sa sfies a zero-profit condi on.

This also implies that the adviser is not able to charge for the funds allocated to the passive fund, since the

investors obtain no benefit to using him in this case.13 The por olio manager also charges a propor onal

fee, fP , on the end-of-period value of the assets managed. Using propor onal fees for advisory and por olio

management services on the basis of end-of period wealth links the fee to the por olio performance, and

therefore aligns the interests of the client, adviser and por olio manager. It is also common prac ce in the

advisory industry as well as for mutual funds. Introducing incen ve fees would only create distor ons in

risk-taking and would be a detriment unless the effort choices by the manager and the adviser need to be

mo vated.

The investors therefore have to decide amongst three inves ng strategies. If they invest in the passive

fund themselves, they get a return of Rm . If they invest directly with the por olio manager, they pay their

fixed search cost as well as the por olio management fee. If they delegate their decision to the adviser,

they have to pay two fees for the ac vely invested por on of their holdings: both the advisory fee and the

por olio management fee. The por olio manager can receive funds directly from the investors (the direct

channel) or indirectly through the financial adviser (the indirect channel).

B. Investors Behavior

Assume that each investor has wealth x + C0 , where x follows a Pareto distribu on with the following

probability density func on:

kAkm
f(x) = , k > 1, (2)
xk+1

8
where Am > 0 denotes the minimum wealth level (net of the search cost C0 ).14 The Pareto distribu on

has been widely used to describe the distribu on of wealth among individuals. Empirical studies have

found that this distribu on characterizes actual wealth distribu ons fairly well, except for its proper es

at the lower end.15 An important feature of this distribu on is that the probability density f(x) decreases

monotonically in wealth, implying that the frac on of wealthy investors is rela vely small while the frac on

of investors with low levels of wealth is rela vely large. The parameter k characterizes the extent of wealth

equality. Complete equality of wealth is characterized by k → ∞, while k → 1 corresponds to complete

inequality.

We standardize the popula on to be 1. Therefore the total wealth available for investment is

∫ ∞
kAm
W= xf(x)dx + C0 = + C0 , k > 1. (3)
Am k−1

Based on their wealth level, investors can choose whether to invest directly or indirectly. Let AD and

AI denote the amounts of direct and indirect investment to the ac ve fund. Therefore the total amount of

money under ac ve management is A = AD + AI and the expected return of the ac vely managed por olio

is Rp = α − γ(AD + AI ).

Investors take returns as given and have no market power, i.e., they are atomis c and do not take into

account the diseconomy of scale in ac ve por olio management when they decide where to channel their

funds. Therefore, the amount of capital invested in the ac vely managed fund via the adviser will adjust

un l investors earn their reserva on rate, Rm . Thus, inves ng through the adviser is iden cal to inves ng

in the passive fund. An investor with wealth A∗ + C0 will be indifferent between contrac ng directly with

the por olio manager and ge ng a net return Rp (1 − fp ) and inves ng via the adviser, where A∗ sa sfies

the following condi on:

[α − γ(AD + AI )](1 − fP )A∗ = Rm (A∗ + C0 ),

i.e.,

C0 Rm
A∗ = . (4)
(1 − fP )[α − γ(AD + AI )] − Rm

9
 It is obvious that all investors whose wealth is smaller than A∗ + C0 will prefer to invest via the adviser

who charges a propor onal fee whereas those with wealth greater than A∗ + C0 will prefer to contract

directly with the por olio manager (or equivalently hire a personal adviser with a fixed fee). We therefore

refer to this la er set of investors as ``high networth individuals".

Given that all investors with wealth levels greater than A∗ + C0 invest directly, we can solve for the

amount of money channeled directly to the por olio manager:

∫ ∞
kAkm
AD = xf(x)dx = . (5)
A∗ (k − 1)(A∗ )k−1

Note that if A∗ = Am , then AD = W − C0 , and all investors would contract with the por olio manager

directly. To make our analysis interes ng, we assume that if all wealth net of the search cost C0 is invested

in the ac ve por olio, the return of the ac ve por olio will be lower than that of the passive fund, i.e.,

α − γ(W − C0 ) < Rm . (6)

II. Independent Adviser Equilibrium

We now solve for the equilibrium in our model of investment management. First we discuss the behavior of

the adviser and subsequently the behavior of the por olio manager. In this sec on, we assume that there

are no kickbacks to the adviser so that his decision-making is uninfluenced in that it is based en rely on the

respec ve asset returns. We refer to such advisers as independent advisers. In the next sec on, we allow

for kickbacks from the por olio manager to the advisers.

A. Advisers Behavior

Since the representa ve fund adviser charges a propor onal advisory fee, fA , based on the end-of-period

value of the ac ve fund investment in his clients' accounts, he solves the following problem:

max w{fA [α − γ(AD + AI )](1 − fP ) − cA }. (7)

w

10
where w represents the por olio weight of the ac ve asset. When solving this problem, the adviser takes

the returns on the ac ve fund as given as well as the propor onal advisory fee.

In order for the adviser's op miza on problem to have a solu on that supports posi ve funds channeled

to both ac ve and passive funds, the coefficient on w in (7) must be zero, or

fA [α − γ(AD + AI )](1 − fP ) − cA = 0. (8)

In addi on, the fact that investors are indifferent to using the adviser implies

[α − γ(AD + AI )](1 − fP )(1 − fA ) − Rm = 0. (9)

Subs tu ng (9) into (8) gives the following equilibrium condi on:

[α − γ(AD + AI )](1 − fP ) = Rm + cA . (10)

No ce that (10) implies that the net-return of the ac vely managed por olio exceeds the return on the

passive asset by exactly the marginal cost of advisory services. If the net-return of the ac vely managed

por olio would be below this threshold value, no ra onal investor would invest ac vely using the adviser.

If it would be above, the advisor would invest everything in the ac vely managed por olio which would

depreciate its expected return.16

Equa ons (8) and (9) can also be used to determine the equilibrium fee charged by the adviser. Subs -

tu ng and solving for fA yields

cA cA
fA = = . (11)
Rp (1 − fP ) Rm + cA

In other words, the fee compensates the advisor for the cost incurred.17

In order to determine which investors will invest via the direct channel, we subs tute equa on (10) into

equa on (4) and get

C0 Rm
A∗ = . (12)
cA

11
No ce that this threshold level of wealth does not depend on α. Therefore the investors can op mally de-

cide whether or not to collect informa on in equilibrium without knowing the por olio managers' poten al

ability, α.

Another important property of this result is that the threshold level of wealth, which determines the

amount of money invested directly through equa on (5), is independent of the fees of the por olio man-

ager. This is because of the compe ve nature of the adviser. If the por olio manager a empts to increase

her fees, for the same gross return, investors would reduce the amount of capital that is allocated to ac ve

management via the adviser. Therefore the same net return is achieved by ac ve inves ng and thus there is

no effect on the marginal direct investor or the aggregate amount of money invested directly. This property

is cri cal to understanding the model and will be exploited below.

B. Por olio Manager Behavior

The por olio manager op mizes the management fee fP to maximize her profit:

max ΠP = [α − γ(AD + AI )](AD + AI )fP , (13)

fP

taking into account the equilibrium condi on (10).

It is easy to solve this for the op mal por olio manager fee, which we record as a proposi on.

P 1. The op mal por olio manager fee in the investment management equilibrium without fee

rebates is

α − Rm − cA
f∗P = . (14)
α + Rm + cA

Using this result, it is straigh orward to see that the op mal management fee is increasing in the man-

agerial ability α. Further we can solve for the por olio manager's profit:

(α − Rm − cA )2
ΠP = , (15)
4γ

12
and the total assets allocated to the por olio manager:

α − Rm − cA
AI + AD = . (16)
2γ

We assume that C0 is big enough, or γ is small enough, to ensure that AI implied by equa ons (5), (12) and

(16) is posi ve, i.e., not all investment into the ac ve por olio comes from the direct channel.18

In summary the investment management equilibrium features posi ve profits of the por olio manager

and zero profit of the investment adviser. Returns on the ac vely managed por olio net of management

fees are greater than those of the passive fund. Net returns earned by direct investors in the ac ve fund

exceed those earned by indirect investors. Nevertheless only the high net worth individuals find it op mal

to invest directly. Furthermore, since the total fund size is increasing in managerial ability, while the amount

of money invested through the direct channel is not, our model implies that the importance of indirect sales

through financial advisers increases with fund size. Ceteris paribus, large funds sell a larger frac on through

advisers whereas small funds feature propor onally more direct investors.

III. Fee Rebates

We now extend the model to allow for rebates or ``kickbacks" from the por olio manager to the financial

adviser. The idea here is that the por olio manager desires to influence the decisions of the financial adviser,

so that the fund is accessed by small investors to a greater extent. The purpose of this sec on is to derive

the equilibrium amount of kickbacks and evaluate the impact of such ac vi es on asset returns and fund

flows, as well as the fees charged by advisers and the por olio manager.

We begin by se ng up the general model in which the rebate can be used for two purposes by the com-

peti ve advisers: (1) as a subsidy to cover opera ng business costs; (2) to support the adviser's promo onal

efforts in aggressively selling the ac ve fund to investors. Our analysis then considers two scenarios. First,

investors are sophis cated in the sense that they fully an cipate the impact of rebates on the equilibrium

outcomes of net asset returns and their decisions cannot be easily influenced by aggressive marke ng ef-

forts. Second, investors are assumed to be unsophis cated which means that they are suscep ble to selling

pressures by the adviser and are ex ante unable to fully an cipate the extent to which their judgments are

13
compromised.

A. General Model with Rebates

Suppose that the por olio manager provides a rebate of δ for each dollar directed to her por olio by the

adviser. The kickback is specified ex ante and assumed without loss of generality to be paid at the end of the

period. Further, suppose that the adviser can spend some frac on, e ∈ [0, 1], of the rebate in promo onal

effort, and retain the rest, (1 − e)δ, to assist with his opera ng business costs, cA . The promo onal effort

helps to embellish the returns to the ac ve fund and make it look more a rac ve. Alterna vely, we can

suppose that the promo onal ac vi es provide a nonpecuniary benefit to investors.19 In order to model

the inflated demand for ac ve funds, we assume that indirect investors' reserva on expected returns can

be lowered by the amount ηeδ, where η is a parameter characterizing how suscep ble investors are to

promo onal effort. Higher values of η imply higher levels of investor susceptability.

In contrast to the previous case without rebates, now the adviser has to consider the net rebate a er

effort expenditure, (1 − e)δ. Therefore his objec ve becomes:

max w{fA [α − γ(AD + AI )](1 − fP ) − cA + (1 − e)δ}.

w,e∈[0,1]

Since the adviser's promo onal effort reduces the indirect investors' reserva on return by ηeδ, the

appropriate modifica on of (9) becomes

[α − γ(AD + AI )](1 − fP )(1 − fA ) = Rm − ηeδ. (17)

Subs tu ng this par cipa on constraint into the adviser's op miza on problem gives the following objec-

ve:

max w{[α − γ(AD + AI )](1 − fP ) − Rm − cA + [1 + (η − 1)e]δ}.

w,e∈[0,1]

It is obvious that if η < 1, the op mal e∗ is 0, and the adviser keeps 100% of the rebate. If η = 1, e is

irrelevant for the adviser's objec ve func on, so it is assumed to be 0 without loss of generality. If η > 1,

then the op mal e∗ is 1, and all rebates are employed in promo onal ac vi es by the adviser. As before, in

14
order for the adviser's op miza on problem to have a solu on that supports posi ve funds channeled to

both the ac ve and passive funds, the coefficient on w must be zero, or

[α − γ(AD + AI )](1 − fP ) = Rm + cA − (1 + (η − 1)e∗ )δ.

Subs tu ng the expressions for the op mal e∗ into the above equa on, we see the net expected return of

the ac ve por olio (before advisory fee) is



 R m − δ + cA if η ≤ 1
[α − γ(AD + AI )](1 − fP ) = (18)

 Rm − ηδ + cA if η > 1.

Combining this result with equa on (17) and taking account of the op mal e∗ , we find the equilibrium

advisory fee to be:



 cA −δ
Rm −δ+cA if η ≤ 1,
fA = (19)

 cA
Rm −ηδ+cA if η > 1.

One can see that there are essen ally two cases of interest. First, when promo onal efforts are not

very efficient in lowering investors' reserva on return (η ≤ 1), then advisers choose not to engage in such

ac vi es by se ng e∗ = 0. From equa on (17), we have a situa on in which indirect investors achieve

the same reserva on return as in the case without rebates, Rm . As a result of compe on among advisers,

their fees, fA , are decreasing in the amount of the rebate, δ. This indicates that the rebate is effec vely

``passed on" to the investors.

The situa on is very different when η > 1. In this case indirect investors hold por olios with returns

below the passive asset and the advisory fee increases with the rebate. The intui on here is as follows.

Since the rebate is en rely devoted to promo ng the ac ve fund (e∗ = 1), the fee of a compe ve adviser

must be set to exactly cover his advisory cost. But now because the returns of the adviser's por olio are

nega vely impacted by the marke ng expenditure, the fee expressed as a percent of end-of-period por olio

value must increase to cover the advisory costs.

These two cases therefore lead to very different outcomes and we therefore analyze them separately

15
in the following two subsec ons. In the first case, η ≤ 1, indirect investors achieve the same return as

the alterna ve (passive) asset. We refer to this as the sophis cated investor scenario. In the second case,

η > 1, indirect investors are manipulated by the adviser's marke ng ac vi es. We therefore refer to this

as the unsophis cated investor scenario.

B. Sophis cated Investors

In our model with sophis cated investors we assume that investors not only an cipate that aggressive

marke ng will not be employed, but also ra onally an cipate the amount of fee rebates and the effect

that this will have on equilibrium net returns. From (18) the expected net return of the ac ve por olio is

Rm + cA − δ for the case of η ≤ 1. As in the case without rebates, we can subs tute this equilibrium net

return into (4) and get the threshold wealth level:

C0 Rm
A∗ = . (20)
cA − δ

We see first by comparing respec vely equa ons (20) with (12) that the rebate changes the marginal in-

vestor who is indifferent between inves ng directly and indirectly to one with a higher wealth level. Conse-

quently the amount of funds invested directly decreases. Hence a key result is that kickbacks shi investors

from the direct investment channel to the indirect investment channel. This occurs because kickbacks lead

to a lower equilibrium return of the ac ve por olio. As a result, less investors are willing to pay the fixed

search cost. As investors are shi ed into the indirect channel, they suffer a welfare loss because they are

now pushed down to their reserva on return. As before, though, the por olio manager's fee is constrained

by the asset alloca on decision of the adviser and the amount of direct investment.

We now endogenize the rebate by allowing the por olio manager to choose her op mal δ. The total

amount of rebate equals the amount of indirect investment in the ac ve por olio mes the rebate for each

dollar invested: AI δ. The por olio manager maximizes her profit net of the kickback payments. Therefore,

her problem is

max Πp = [α − γ(AD + AI )](AD + AI )fP − AI δ

fP ,δ

16
subject to the constraints (18) (for the case η ≤ 1), (5), and (20). We can now solve for the op mal por olio

manager fees and kickback payments imparted to the financial adviser.

P 2. In the sophis cated investor equilibrium with rebates, the op mal fee charged by the port-

folio manager is

α − Rm − cA + 2cA /k
f∗P = . (21)
α + Rm + cA

The op mal rebate from the por olio manager to the financial adviser is

cA
δ∗ = . (22)
k

Proof. See Appendix A.

Proposi on 2 shows that there is an interior op mal rebate. The intui on for this result is as follows.

The rebate allows the por olio manager to increase her fees, as embodied in equa on (21). However, for

fund shares which are sold to indirect investors the por olio manager does not benefit from higher fees

since they are fully offset by the rebate to the advisers. Therefore the por olio manager simply op mizes

the rebate to maximize the surplus she extracts from the direct investors. As the rebate increases, the sur-

plus she extracts per dollar of directly invested wealth increases; however, the amount of directly invested

wealth decreases because some investors will switch to the indirect channel. The la er effect dominates

for sufficiently large rebates.

Subs tu ng the op mal fee given by equa on (A.1) back into the objec ve func on and using again

constraint (18) for the case of η ≤ 1, we get

(α − Rm − cA )2
ΠP = + AD δ. (23)
4γ

Comparing this expression with equa on (15), we see that the por olio manager's profit in the new equi-

librium is simply her profit in the equilibrium without kickbacks plus the loss of remaining direct investors

due to lower fund return.

Effec vely subsidizing the adviser permits the por olio manager to price discriminate between large

17
and small investors, while charging the same management fee. Since high networth investors enjoy some

surplus, they have a lower elas city of demand compared to the indirect investors, who only get their

reserva on return. Therefore the por olio manager would op mally like to charge higher fees to the high

networth investors, without adversely affec ng the demand of small investors.20 Rebates, therefore, allow

the por olio manager to extract some surplus of the large investors.

From equa on (22), we can easily see that the op mal rebate is increasing in the advisory cost cA and

decreases in k, which measures the degree of equality of the wealth distribu on. The advisory cost cA rep-

resents the maximum rebate that can be provided before all investors leave the direct channel. Therefore,

not surprisingly, the op mal rebate is increasing in cA . The rela on between the op mal rebate and k is

also intui vely appealing. When k is large, there are fewer high networth investors, therefore, the por olio

manager does not extract much surplus by providing a rebate. By contrast, when k is close to 1, the frac on

of high networth investors is rela vely large. As a result, the por olio manager has a stronger incen ve to

subsidize to be able to extract their surplus. In this case, the adviser's fee, from equa on (19), approaches

zero. Advisory fees are always strictly posi ve since k > 1. Explicit examples in prac ce include the fees

charged by ``wrap account'' managers and funds of hedge funds advisers.

To further analyze the impact of op mal fee rebates, we compute the equilibrium size of the fund. Sub-

s tu ng the op mal fee given by equa on (A.1) into (18) (for the case of η ≤ 1), we find that the fund size

in the equilibrium with kickbacks is exactly the same as the fund size in the equilibrium without kickbacks,

i.e., equa on (16). Intui vely this occurs because in both cases, the indirect investors are marginal investors

in the sense that a slight decrease in net returns would lead them to switch to the passive por olio. Their

reserva on return is Rm . Since the advisory services market is compe ve, the por olio manager has to

cover the cost of advisory services and therefore the marginal cost of obtaining one dollar from the indirect

investor is Rm + cA , which is independent of kickbacks. When the por olio manager op mizes the fund

size, she equates this marginal cost with the marginal increase in the end-of-period por olio value resul ng

from an addi onal dollar of indirect investment, which is also independent of the kickback. Since both the

marginal benefit and the marginal cost are independent of the kickback, the fund sizes are iden cal in both

cases.

We summarize the impacts of kickbacks in Proposi on 3.

18
P 3. In the sophis cated investor equilibrium with kickbacks, the ac ve fund size is the same

as when advisers are independent. However, more investors u lize advisory services by inves ng indirectly

and the net return (a er the management fee) of the ac ve fund is lower. The por olio manager charges

a higher fee, while advisers charge a lower fee but receive a compensatory kickback from the por olio

manager.

Since we have shown that fund size is constant, while management fees are increased because of re-

bates, our model predicts lower performance for ac vely managed mutual funds that use greater levels

of rebates. This conforms to some recent evidence from mutual funds, which shows that rebates in the

form of excess commissions paid to brokerage firms(Edelen et al., 2008) are associated with poor fund per-

formance. Also our results predict that the por olio management fee has a one-for-one nega ve impact

on the fund's net return, which is consistent with the finding of Carhart (1997). Our model shows these

empirical pa erns can result from conflicts of interest in the distribu on channel for funds.

A striking result of our analysis is that despite the poten al conflicts of interest associated with the re-

bates, financial advisers are actually used to a greater extent in equilibrium than when there are no rebates.

The reason is that the por olio manager op mally raises her fees, which makes direct investment less at-

trac ve. The adviser is forced to lower his own fees in order to remain compe ve with the alterna ve

asset. Hence, the volume of the adviser's asset management business increases.

C. Unsophis cated Investors

We now turn to the analysis where η > 1 and financial advisers use the rebate for promo onal ac vi es,

which results in a reduc on of the reserva on return of indirect investors below the alterna ve (passive)

asset. Obviously if this were an cipated ex ante, nobody would u lize a financial adviser and we would

wind up in an equilibrium without advisers (see sec on V). Since investors are unsophis cated, they may

be unable to an cipate this outcome. To analyze the equilibrium outcome under such scenarios, we now

assume that at the me when the investors have to decide whether to pay the search cost, they believe

(incorrectly) that they are in the equilibrium without rebates. As a result, equa on (12) for the threshold

value A∗ applies.21 However ex post, a er an investor pays the search cost, she ra onally invests passively if

the expected net return of the ac ve fund is below the passive asset. Our unsophis cated investors model

19
therefore assumes that the ini al unsophis ca on is overcome a er paying the search cost. Of course, this

cost is sunk in the sense that direct investors may regret having paid it once they find out that the ac ve

fund may underperform.

From equa on (18), we know that the expected return of the ac ve por olio is not less than Rm if and

only if δ ≤ cA /η for the case η > 1. Therefore the amount of money invested through the direct channel

is the same as in the independent adviser equilibrium only if δ ≤ cA /η. Otherwise the amount of money

invested directly is AD = 0. Subs tu ng out fP in the por olio manager's objec ve func on using the

second case in equa on (18) gives

ΠP = [α − γ(AD + AI ) − (Rm + cA − ηδ)](AD + AI ) − AI δ

= −γ(AD + AI )2 + (α − Rm − cA )(AD + AI ) + [ηAD + AI (η − 1)]δ, (24)

kAkm ck−1
where AD = A
(k−1)(C0 Rm )k−1
over the range δ ≤ cA /η and is zero otherwise. Since η > 1, this objec ve

func on is linearly increasing in δ except at δ = cA /η, where AD has a discrete jump toward zero. It is

easy to see that in this case, if rebates are unbounded, the op mal δ goes to infinity, since the por olio

manager can essen ally expropriate unlimited amounts of wealth from indirect investors. To account for

the fact that there are natural limits to the degree to which advisers' marke ng efforts can exploit indirect

investors, we assume that the rebate is bounded above by δ̄ > cA /η. 22 Therefore, there are two poten al

levels of δ that may maximize the por olio manager's profit: δ = cA /η, or δ = δ̄.

We record the equilibrium solu on in the case of unsophis cated investors in the following proposi on.

P 4. In the unsophis cated investor equilibrium (η > 1), there are two possible op mal rebates.

In one case δ = cA /η and the op mal management fee is equal to

α − Rm + cA /η
f∗P = .
α + Rm + cA /η

In this case the ac ve fund and passive fund have iden cal net returns. In another case the op mal rebate

is equal to its upper bound, δ = δ̄, the op mal management fee is

′ α − Rm − cA + (η + 1)δ̄
f∗P = ,
α + Rm + cA − (η − 1)δ̄

20
and the ac ve fund underperforms the passive fund.

Proof. See the internet appendix.

The compara ve sta cs that determine which of the two possibili es occurs is given by the following.

C 1. The unsophis cated investors equilibrium is more likely to imply underperformance by the

ac ve fund if (1) the upper limit on feasible rebates, δ̄, is high; (2) or the frac on of high networth investors

in the economy is rela vely low, i.e., if k is high; (3) or the fixed search cost, C0 , is high; (4) or the managerial

ability, α, is high; (5) or investors are more vulnerable to marke ng ac vity, i.e., if η is high.

Proof. See the internet appendix.

The economic intui on for these results is as follows. The por olio manager faces a tradeoff between

the benefit of extrac ng more from the indirect investors and the cost of losing the direct investors. When

the restric on on maximum rebate δ̄ is less binding, the poten al gain for aggressively marke ng the fund

and serving only the indirect investors is high, which induces the por olio manager toward the underper-

formance equilibrium. When C0 or k is high, fewer investors will pay the search cost, so the cost of losing

those investors is rela vely low to the por olio manager, which also makes the por olio manager favor

the equilibrium of serving only the indirect investors. When managerial ability α is high, the poten al gain

from ac vely marke ng the fund is bigger because the por olio manager can capitalize on a bigger fund

size. Finally, when η is high, the marke ng effort required to influence the investors is lower, which again

favors the equilibrium with aggressive marke ng and underperformance of the ac ve fund.

Looking at different scenarios of investor sophis ca on, we can derive several implica ons of the in-

s tu onal nature of intermediated investment management. First, fund performance net of management

fees is nega vely related to rebates to financial advisers, irrespec ve of the degree of investor sophis -

ca on. When investors are ra onal, the rebates are passed on to the investors through lower advisory

fees, the performance of the ac ve fund is compromised, but is s ll above that of the passive fund. If in-

vestors are unsophis cated, then the rebates are used for aggressive marke ng, and the ac ve fund will

either underperform or perform equally well as the passive fund. Second, the diversity with which a fund

is channeled to investors is related to its performance. Underperforming funds are only sold indirectly.

Ac ve funds with performance equal to or above passive funds are sold simultaneously through direct and

21
indirect channels. Third, when investors are ra onal, disclosure of rebates is not relevant because they

are an cipated. When investors are unsophis cated, disclosure of the rebates can be valuable. Through

disclosure, investors will be alerted to the presence of excessive promo onal ac vi es, and will therefore

make a more informed decision about using advisory services. Fourth, when investors are sophis cated

the existence of rebates affects the welfare of direct investors only. When they are unsophis cated, then

rebates make both direct and indirect investors worse off.

IV. Compe on in Ac ve Por olio Management

Now we introduce an environment where there are more than one ac ve por olio managers. For

simplicity in this sec on we consider two managers compe ng for the same pool of investors as before.

Our results can easily be generalized to the case with a finite number of por olio managers. We focus on

the case of sophis cated investors in this sec on (η ≤ 1).

A central issue in the case of compe on is whether it increases or decreases the rebate imparted to

the asset alloca on choices of advisers. One hypothesis is that compe on disciplines por olio managers

and induces lower kickbacks; the opposite hypothesis holds that compe on creates stronger incen ves

for the por olio managers to provide kickbacks to advisers. In this case they are in a race to outdo each

other.

A. Independent Advisers

We assume that the two por olio managers are pursuing similar ac ve strategies. The por olio managers

are symmetric with respect to ability and therefore both have poten al abnormal returns, α. Because the

por olio managers use similar strategies, the expected return of each fund is not only related to her own

fund size, but also to the size of the other fund, i.e., there is a nega ve size externality within each fund

sector. More specifically, we assume that the gross return, RPi , of por olio manager i is given by

RPi = α − γ(Ai + ρAj ), i = 1, 2; j = 1, 2; i ̸= j

22
where Ai ≡ ADi + AIi is the size of the fund managed by manager i (both direct and indirect investment);

ρ ∈ [0, 1] is a parameter characterizing the similarity between the investment strategies employed by the

two por olio managers. When ρ = 1 the two investment strategies are iden cal, so the compe on be-

tween por olio managers is most intense, and the same diseconomy of scale occurs regardless of whether

an increase in fund size occurs within the fund or with its rival. On the other hand, ρ = 0 indicates that the

strategies are uncorrelated, and as a result, the diseconomy of scale effect is confined to the individual fund

level. Consistent with the diseconomy of scale at the fund sector level, Naik, Ramadorai, and Stromqvist

(2007) find that for four out of eight hedge fund strategies, capital inflows have sta s cally preceded neg-

a ve movements in strategy alpha. Wahal and Wang (2010) measure the compe on between mutual

funds by the extent of overlapping security holdings. When new funds enter with similar holdings, they

find a significant nega ve impact on fees, returns and flows of the incumbent funds.

We employ the concept of Cournot-Nash compe ve strategies with respect to fund sizes. In this case,

each por olio manager op mizes her fund size taking as given the size the other fund:

max ΠPi = [α − γ(Ai + ρAj )]Ai fP , (25)

Ai

subject to essen ally the same condi on as equa on (10) considered earlier on the behavior of the financial

advisers:

[α − γ(Ai + ρAj )](1 − fP ) = Rm + cA . (26)

Subs tu ng out fP in the objec ve func on using the constraint, and considering the first order condi-

ons for both por olio managers simultaneously, we obtain the following proposi on.

P 5. The Cournot-Nash equilibrium involving compe on among por olio managers is unique

and symmetric. The equilibrium fund sizes are

α − Rm − cA
A∗i = , i = 1, 2. (27)
(2 + ρ)γ

23
The equilibrium management fee is

α − Rm − cA
f∗P = . (28)
α + (1 + ρ)(Rm + cA )

Proposi on 5 shows that as long as the investment strategies are posi vely correlated, i.e., ρ > 0,

the size of each individual ac ve fund is smaller than the fund size in the monopolist case (equa on 16).

However, the aggregate amount of funds under ac ve management is always greater, since each individual

fund size is more than one half of the size of the monopolist fund. Further, management fees are lower

(equa on (28) as compared to (14)). As a result, even though the aggregate fund size is larger, net returns on

ac ve por olios are the same as in the monopolist case. If por olio managers employ completely different

strategies, i.e., ρ = 0, individual fund size and management fee will be the same as in the monopolist case,

and the aggregate amount of money under ac ve management will be doubled.

B. Subsidized Advisers

Now we consider the Cournot-Nash equilibrium in which each por olio manager provides a rebate, δ, to

the adviser. In this case the por olio managers' op miza on problem can be wri en as

max ΠPi = {α − γ[(ADi + AIi ) + ρ(ADj + AIj )]}(ADi + AIi )fP − AIi δ, (29)
ADi ,AIi

subject to the constraints

{α − γ[(ADi + AIi ) + ρ(ADj + AIj )]}(1 − fP ) = Rm + cA − δ, (30)

∑ 2
kAkm (cA − δ)k−1
AD ≡ ADi = . (31)
(k − 1)(C0 Rm )k−1
i=1

The solu on to this problem is provided in the following proposi on.

P 6. The solu on to the Cournot-Nash compe on game between two por olio managers who

can influence the financial advisers through kickbacks is unique and symmetric. The equilibrium features the

same total fund size as in the case without kickbacks. However the alloca on through the indirect channel

24
is larger and the direct channel is smaller. The op mal rebate is

cA
δ∗ = , (32)
2k − 1

and the op mal fee schedule for each por olio manager is

α − Rm − cA + (2 + ρ)δ∗
f∗P = . (33)
α + (1 + ρ)(Rm + cA )

Proof. See Appendix B.

This proposi on shows that our results on the impact of fee rebates are carried through to the case

of (imperfect) compe on between mul ple por olio managers. As before the amount of funds ac vely

managed is not affected by kickbacks. Not surprisingly we find that more funds are managed ac vely when

there is more compe on. However it is not true that there is a ``race in outdoing" the other por olio

manager in terms of rebates. In fact they op mally select a lower rebate to the advisers as compared to

the monopolis c case. The reason has to do with the direct investors. Increasing the rebate financed by

increasing fees implies that one fund will lose high networth investors to a compe tor that does not follow

suit. Therefore, even if the fund could gain access to more indirect investors, it winds up losing direct

investors and this effect dominates. As a result, compe ve forces actually counteract the tendency to

subsidize the advisers. Hence, recent trends towards more independent advisory services could be due to

a greater degree of compe ve pressures between ac ve por olio managers.

Our results point out that the nega ve size externality parameter, ρ, only impacts the fees charged by

the por olio manager, but not the extent of rebates. This occurs because for any given rebate, the threshold

wealth level that determines whether someone is a direct or indirect investor is independent of ρ. Rebates

are used as a means to extract surplus from direct investors and therefore are not affected by ρ either. This

also implies that net returns are unaffected by the size externality. On the other hand, when the nega ve

externality is more significant, por olio managers do compete more aggressively by lowering their fees,

and the aggregate ac ve inves ng is diminished.

25
 V. Equilibrium Without Advisers

In order to inves gate the role of financial advisers in delegated por olio management, we now exam-

ine an equilibrium in which financial advisers do not exist. When there is no investment adviser, the only

vehicle for ac ve inves ng is directly through the por olio manager. As a result, AI = 0; the fund size is de-

termined solely by AD . The por olio manager maximizes her profit by choosing an op mal fee. Therefore,

the por olio manager's problem can be wri en as:

max ΠP = (α − γAD )AD fP (34)

fP

subject to the constraints (4) and (5) with AI = 0.

Note that the informa onal assump ons made here are somewhat stronger than those needed pre-

viously with the financial adviser. Recall that the decision whether to invest directly did not depend in

equilibrium on the poten al value of ac ve management, α, when the financial adviser is present. Now we

must assume that the direct investor knows in advance which α will obtain, a er the cost, C0 , is expended.

This, of course does not violate our earlier jus fica on for the cost, since without paying it, there would be

an adverse selec on problem borne by the investor.

Problem (34) is now solved in the next proposi on.

P 7. In the por olio management equilibrium without advisers, there exists a unique interior

op mal fund size and management fee, which are the solu ons to the following set of equa ons:

λk 1/(k−1)
α − Rm − 2γAD − A = 0, (35)
k−1 D

and

1/(k−1)
α − γAD − (Rm + λAD )
fP = , (36)
α − γAD

where

k − 1 1/(k−1)
λ ≡ C0 Rm ( ) . (37)
kAkm

26
Proof. See the internet appendix.

Analy cal solu ons to the equa on system in Proposi on 7 can be obtained for some special values

of k. For example, k = 1.5 or k = 2. For general values of k, the solu ons can be found using numerical

methods.

In the following sec on we compare outcomes for the four scenarios previously considered. In par c-

ular we address the ques on whose interests financial advisers really serve: the investors' or the por olio

manager's. The key ques on is how investors are impacted by the presence of the adviser and the kickback.

We also consider the consequence of rebates on total welfare of investors and the por olio manager.

VI. Comparison of Equilibria

To illustrate the differences, we construct a numerical example and solve for the equilibria in the four

cases: (1) no adviser case; (2) independent adviser case; (3) rebates to the adviser with sophis cated in-

vestors -- referred to as the subsidized adviser case; (4) rebates to the adviser with unsophis cated investors

-- referred to as the unsophis cated investor case.

A. Calibra on

There are seven parameters in our model with sophis cated investors: α, Rm , γ, Am , C0 , cA , and k. For the

unsophis cated investor case, there are two addi onal parameters: η and δ̄. We now describe how we

calibrate these parameters.

We first set the expected gross return of the passive por olio, Rm , to be 1.04. According to Hung,

Clancy, Dominitz, Talley, Berrebi, and Suvankulov (2008), a typical fee charged to investors with $100,000 to

$1 million assets under management by investment advisers is 1.25% in the US. Considering the poten al

rebates received by some advisers, we set the advisory cost, cA , equal to 1.5%. For the diseconomy of scale

parameter, γ, we refer to the study by Chen, Hong, Huang, and Kubik (2004). In their table 1, they report a

difference of 8 basis points per month in market-adjusted returns between the second and fourth mutual

fund size quin les. The average size difference between these two groups is $143 million. Since bigger

mutual funds in reality may be run by managers with higher ability, thereby counterac ng the nega ve size

effect, this return difference ra o of 0.0008 · 12/(143 · 106 ) understates the true diseconomy of scale. We

27
therefore mul ply it by a factor of three to give an es mated γ ≈ 2 · 10−10 . We set α equal to 1.08, so that

the size of the ac ve fund in the equilibrium with sophis cated investors is $62.5 million, which matches

the fund size in the median quin le of Chen et al. (2004). For the wealth distribu on parameter k, we u lize

the interval, k ∈ [1.5, 2]. The midpoint of this range, 1.75, translates into a Gini coefficient equal to 0.4,

which matches the US income distribu on as published in the 2009 Human Development Report (United

Na ons Development Program). The minimum wealth level, Am , is set equal to 5·107 , so that the aggregate

wealth of the investors net of the search cost exceeds the ac ve fund size in all equilibria considered in our

compara ve study. The fixed search cost, C0 , is set equal to 5 · 106 . In the subsidized adviser equilibrium

with k=1.75, this implies that 23% of the investment into the ac ve fund comes from the direct channel.

This accords with the empirical data reported by the Investment Company Ins tute, that 74% of the mutual

fund assets held by US households (outside defined contribu on plans) are purchased through professional

financial advisers.23 Finally in order to parameterize the unsophis cated investor equilibrium, we set δ̄ to

be 0.015, which implies that the por olio manager cannot provide a rebate higher than the actual advisory

cost, cA . We also set η = 1.6 for illustra on purposes.

B. Outcomes

In this subsec on we analyze the equilibrium fund size, the amount of direct investment, the management

fee and net returns as a func on of k. In drawing cross-sec onal conclusions from these outcomes we inter-

pret these results as though they occur in segmented environments, rather than a common environment

where all distribu onal choices are endogenously determined. The results are illustrated in Figure 1. Insert Figure

We plot the fund size in the four equilibria in Panel A of Figure 1. Here the solid line represents the fund 1 here.

sizes in both the independent and subsidized adviser equilibria since they are the same. As one can see

from the graph, the size of the ac ve por olio is substan ally larger in the presence of financial advisers,

especially when k is large, i.e., when the frac on of high networth investors in the economy is small. This

is because financial advisers assist small investors to invest in the ac ve por olio. The unsophis cated

investor case has the highest fund size. This is because the marke ng efforts of the adviser generate more

demand from indirect investors. As discussed in subsec on III.C, when k is low both direct and indirect

investors hold the ac ve fund. As k increases, there is a discrete jump in fund size when the maximal rebate

28
 Panel A: Fund size Panel B: Investment through the direct channel
fund size AD

8. ´ 107
5. ´ 107

6. ´ 107 4. ´ 107

3. ´ 107
7
4. ´ 10
2. ´ 107

1. ´ 107
2. ´ 107
0

k k
1.6 1.7 1.8 1.9 2.0 1.6 1.7 1.8 1.9 2.0

Panel C: Fees charged by the por olio manager Panel D: A er-fee return of the ac ve por olio
fP H1 - f P L R P

0.035 1.06

0.030
1.05
0.025

0.020
1.04
0.015

0.010
1.03
0.005

k k
1.6 1.7 1.8 1.9 2.0 1.6 1.7 1.8 1.9 2.0

Figure 1: Comparison between equilibria. The solid lines correspond to the equilibrium with indepen-
dent advisers (in Panel A, it also corresponds to the equilibrium with subsidized advisers), dashed lines
correspond to the equilibrium with subsidized advisers, do ed lines correspond to the equilibrium without
advisers. Fund size and fund flow are in dollar terms. Management fee is a decimal frac on. A er-fee
return is one plus the rate of return. The parameter η is between zero and one in these three cases. The
values of parameters other than k are as follows: Rm = 1.04, α = 1.08, γ = 2 · 10−10 , Am = 5 · 107 , C0 =
5 · 106 , cA = 0.015. The dashed-dot lines correspond to the case of unsophis cated investors, in which two
addi onal parameters are specified: η=1.6, δ̄=0.015. In Panel B, the dashed-dot line is invisible over the
lower range of k because it overlaps with the solid line.

29
is used and the extreme amount of promo onal ac vity causes excessive investment by indirect investors,

even though direct investors are driven away.

Note that before-fee returns are inversely related to fund size. Panel A of Figure 1 therefore implies

that the before-fee performance is higher if the fund is sold only through the direct channel (i.e., in the no

adviser equilibrium) than in an environment where the fund is sold through both channels. By contrast, in

a situa on where a fund is sold solely through indirect channels (i.e., in the equilibrium with δ = δ̄), we

should expect to observe lower gross returns. This is supported by empirical evidence in Bergstresser et al.

(2009), who show that returns are higher for direct channel funds as compared to brokered funds even

before marke ng fees are deducted.

Panel B of Figure 1 compares the amount of investment through the direct channel. In the case without

advisers, this is equivalent to the total fund size. In all scenarios direct investment decreases as k increases,

i.e. when there are fewer high-networth individuals. Direct inves ng in the unsophis cated investor case

corresponds to the independent adviser case for lower values of k and disappears en rely for higher values

of k. Not surprisingly, the direct-channel investment is smaller when advisers are subsidized than when

they are independent because some investors are shi ed to the indirect channel due to the kickback.

We now turn to the effect of the existence of independent advisers on the amount of direct investment

in the ac ve fund. There are two effects to consider. First, for given management fees, the absence of

advisers increases the amount of direct investment, as there is no subs tute. However, there is a second

effect due to the endogeneity of the management fee. In the absence of advisers, it is op mal for the

por olio manager to charge higher fees since demand is less elas c. The magnitude of this second effect

increases when there are more high networth investors, which corresponds to a small k. The second effect

dominates the first in this region.

Panel C of Figure 1 illustrates the por olio manager's fees as a func on of k. In the case of independent

advisers the fee is constant. This is because the indirect investors are the marginal ones, and their reser-

va on return is independent of their wealth. Moreover, the fee charged in the independent adviser case

is by far the lowest of the four scenarios. The reason for the low fee is that the presence of the adviser

effec vely makes the demand more elas c and thus fee reduc ons are more profitable for the por olio

manager.

30
 If investors are sophis cated, rebates are used to price discriminate against high networth investors. As

k becomes larger, i.e. as the number of high networth investors becomes smaller, the poten al benefit of

price discrimina on decreases. Therefore the op mal fee is decreasing in k. By contrast, in the unsophis -

cated investor case the rebate is used for promo on instead of price discrimina on. Now, for higher values

of k, the por olio manager prefers to only sell indirectly and does not moderate her fees to a ract direct

investors. Thus fees are higher at the upper end of the range of k.

Panel D of Figure 1 compares the return of the ac ve por olio a er management fee. Consistent with

the results on the difference in fees, the net return is always lower in the cases with kickbacks as compared

to the situa on without kickbacks. The net returns in the absence of the adviser are related to the direct

investment decision of high networth investors. When there are many of them (k small) the por olio man-

ager can capitalize by increasing her fees to a greater extent. As a result the net returns are reduced below

the independent adviser case. When there are fewer of them (k large) the por olio manager is only able

to sell to a smaller number of direct investors at a lower fee, therefore the net returns are more a rac ve

than in the independent adviser case. Finally the net returns are lowest in the unsophis cated investor

case because the fund size is not only larger than in the other cases, but fees are higher as well.

C. Welfare Analysis

We now analyze how the aggregate welfare is affected by the presence of advisers, as well as by fee rebates.

Investment advisers (when in existence) always have zero surplus. Therefore we consider only the welfare of

the por olio manager and the investors. We measure the surplus of the investors rela ve to the default of

inves ng in the passive asset and earning net return Rm . Therefore, except for the unsophis cated investor

case, indirect investors earn zero surplus.

The effect of rebates on investor welfare is unambiguously nega ve. Recall that when investors are

sophis cated, fee rebates shi some investors from the direct channel to the indirect channel. These in-

vestors lose their surplus, while investors who remain in the direct channel get a lower net return as the

por olio manager raises her fee. Indirect investors are not affected because the higher por olio manage-

ment fee is offset by the lower advisory fee. In the unsophis cated investor case, the expected return of

the ac ve fund is either equal to or below the passive return. High networth investors get no benefit from

31
having paid the search cost, therefore they experience a welfare reduc on. Furthermore, since rebates are

used for marke ng instead of being passed through, indirect investors' welfare is also nega vely affected.

Obviously the por olio manager's profit increases when rebates are used, since zero rebate is feasible in

her op miza on problem.

Combining the profit for the por olio manager with the surplus earned by the investors we compute

the total welfare in the four scenarios. The details of these computa ons are carried out in appendix B.

Denote the total welfare and investor surplus by Ui and Si , respec vely, where the superscript i indicates

different equilibria: i = 0 (no adviser), i = 1 (independent advisers), i = 2 (subsidized advisers), i = 3

(unsophis cated investors). We are able to prove the following proposi on:

P 8. The levels of total welfare in the four equilibria are given respec vely by

U0 = (α − γA0D − Rm )A0D − θ 0 C0 Rm , (38)

U1 = A1D cA − θ 1 C0 Rm + Π1P , (39)

U2 = A2D (cA − δ) − θ 2 C0 Rm + Π2P , (40)

U3 = −A3I ηδ − θ 1 C0 Rm + Π3P , (41)

kAkm
where AiD = (k−1)(A∗i )k−1
denotes the amount of assets invested into the ac ve por olio directly, θ i ≡ ( AAm∗ )k
i

denotes the frac on of investors choosing the direct channel, ΠiP denotes the por olio manager's profit, A3I

denotes the amount of indirect investment in the unsophis cated investor equilibrium. Furthermore,

U1 > U2 , U1 > U3 , S1 > S2 > S3 . (42)

Proof. See the internet appendix.

Insert Figure
Panel A of Figure 2 plots the total welfare in the four equilibria under the same parameter values used 2 here.

to plot Figure 1. Consistent with Proposi on 8, the total welfare in the independent adviser equilibrium

(the solid line) is always higher than in the equilibrium with subsidized advisers (the dashed line). When

kickbacks are permi ed, the equilibrium features excessive use of investment advisers. This occurs as some

clients are being induced to use the propor onal cost technology of the advisers when the fixed-cost tech-

32
 Panel A: Fund size
U
1.5 ´ 106

1. ´ 106

500 000

-500 000

k
1.6 1.7 1.8 1.9 2.0

Panel B: Por olio manager's profit Panel C: Investor surplus

PP S
500 000
1.5 ´ 106
0

-500 000
1. ´ 106
-1. ´ 106

500 000 -1.5 ´ 106

-2. ´ 106

k k
1.6 1.7 1.8 1.9 2.0 1.6 1.7 1.8 1.9 2.0

Figure 2: Welfare comparison across equilibria. This figure compares welfare in four different equilibria.
The solid lines correspond to the equilibrium with independent advisers (in Panel A, it also corresponds
to the equilibrium with subsidized advisers), dashed lines correspond to the equilibrium with subsidized
advisers, do ed lines correspond to the equilibrium without advisers. The parameter η is smaller than one
in these three cases. The values of parameters other than k are as follows: Rm = 1.04, α = 1.08, γ =
2 · 10−10 , Am = 5 · 107 , C0 = 5 · 106 , cA = 0.015. The dashed-dot lines correspond to the case of unsophis-
cated investors, in which two addi onal parameters are specified: η=1.6, δ̄=0.015.

33
nology, i.e. the direct channel would be more efficient. Importantly, the figure also shows that both of

these equilibria dominate the no adviser equilibrium: the total welfare in the equilibrium without advisers

(the do ed line) is lower than in the other two equilibria for all different values of k we consider. This

provides a ra onale for the existence of advisory services in facilita ng capital investment through ac ve

funds.

The total welfare with unsophis cated investors is lowest among all the cases. First, the marke ng effort

is an addi onal deadweight loss. Second, investors who expend the search cost do so without obtaining

any benefits in equilibrium since the net return of the ac ve fund is either equal to or below the passive

fund. Within the unsophis cated investor case, the underperformance equilibrium, which occurs when k

is high, is substan ally worse than the equal performance equilibrium. This is because marke ng is more

aggressive in that scenario, and more indirect investors are affected.

Panel B of Figure 2 compares the profits of the por olio manager in four equilibria for various values of

k. In the independent adviser equilibrium, the profit of the por olio manager is independent of the wealth

distribu on parameter k, therefore it is a horizontal line in the diagram. The por olio manager is strictly

be er-off in the equilibria with kickbacks. She benefits more from paying kickbacks when k is smaller. This

is because she extracts more surplus from the high networth investors when there are many such investors

in the economy. In the unsophis cated investor case, the low networth investors are also expropriated,

thereby leading to the highest level of profit.

For most reasonable values of k, the por olio manager benefits from the presence of financial advisers,

even when kickbacks are forbidden. This is because the existence of financial advisers allows the por olio

manager to provide her services to small investors, who will otherwise not par cipate in the ac ve por olio.

Interes ngly, this is not always the case. When k is small, the por olio manager's profit is higher in the

no adviser equilibrium than in the independent adviser equilibrium, indica ng that when there are many

wealthy investors, the por olio manager may be be er off by declining investment through an indirect

channel.

Panel C of Figure 2 plots the investor surplus in different equilibria, derived in the internet appendix.

Consistent with our analy cal results, kickbacks always reduce investor welfare. Clearly investors are at a

severe disadvantage when they are unsophis cated, even more so when there are fewer high networth

34
investors and rebates are at the maximum level. The figure also shows that from the investors' point of

view, subsidized advisers are worse than no advisers at all. More subtly, when k is sufficiently large, even

independent advisers can reduce investor welfare. This is consistent with the shape of the net return of the

ac ve por olio plo ed in Panel D of Figure (1).

In summary, our analysis shows that as long as investors are sophis cated, the presence of financial

advisers improves total welfare with or without the use of rebates. Most of the benefit, however, accrues

to the por olio manager. The use of rebates is op mal from the por olio manager's perspec ve, whether

investors are sophis cated or unsophis cated. Investors are worse off from the use of rebates, especially

when they are suscep ble to marke ng efforts and are unable to an cipate this.

VII. Conclusions

The market for financial products and services is expanding rapidly as corpora ons and financial in-

s tu ons package cash flows and con ngent claims in increasingly sophis cated ways. As the number of

alterna ve investment opportuni es placed before investors increases, financial advisers play a more and

more prominent role in alloca ng assets. Investment advisory services are employed by many types and

categories of investors, including retail investors, corporate pension funds, university endowment commit-

tees and many other ins tu onal investors. The purpose of this paper has been to inves gate the effect

of such financial intermediaries and their compensa on schemes on investors' por olio decisions, fund

returns, management fees and welfare.

A unique feature of our model is that the decision to use an adviser is endogenously determined. Ad-

visory services provide an opportunity for smaller investors to par cipate in an ac vely managed por olio,

consis ng for instance of alterna ve investments which would not be economical without the use of an

adviser. As long as investors are ra onal and the advisory industry is compe ve, the presence of advisers

improves the total welfare of the por olio manager and investors even when they are subject to poten al

conflicts of interest. Investors’ welfare alone may increase or decrease due to the existence of financial

advisers. Consistent with the widespread usage of rebates as part of financial advisers' compensa on in

prac ce, our model shows that it is op mal for the por olio manager to subsidize advisers via kickbacks.

35
Depending on the degree of investor sophis ca on, rebates are used by the por olio manager either as

a price discrimina on mechanism or to support aggressive marke ng of the ac ve fund. In both cases

kickbacks strictly reduce investors’ welfare. Nevertheless, they increase the use of advisory services.

Consistent with the exis ng empirical finding that brokered funds underperform direct channel funds,

our model predicts that underperforming ac ve funds can only be sold via financial advisers to unsophis-

cated investors. By contrast, outperforming funds are generally sold simultaneously through direct and

indirect channels. Funds with highest gross performance are likely to be those sold directly and exclusively

to a small subset of high networth investors. We also predict that funds distributed by intermediaries that

are more heavily subsidized by the por olio managers, such as insurance mutual funds, high-load funds,

and funds paying abnormally high so -dollars to improve fund distribu on, underperform other funds.

Furthermore, we show that compe on between ac ve por olio managers lowers the equilibrium rebate,

which results in a lower kickback-based component in the adviser's compensa on scheme. Therefore re-

cent trends toward more independent advisory services may be due to enhanced compe on among port-

folio managers. Our model also generates some empirical predic ons that have not been tested yet. For

example, we predict that the incen ve of the por olio manager to subsidize the adviser increases when

the frac on of large investors in the economy increases. Also, our model implies that the importance of

indirect sales through financial advisers increases with fund size. Ceteris paribus, large funds sell a larger

frac on through advisers whereas small funds feature propor onally more direct investors.

Several poten al policy implica ons emerge from our analysis. Investor educa on that decreases the

suscep bility of investors to marke ng ac vi es will imply less use of fee rebates for promo on and higher

investor welfare. Adequate disclosure of the magnitude of fee rebates and the extent to which this is passed

on to investors can also be important. Moreover, it would be be er to allow the por olio manager to sub-

sidize the adviser via general purpose monetary transfers rather than earmarking fund specific marke ng

support. It may be temp ng to draw a conclusion from our model that banning rebates en rely may be

op mal, however, such interpreta on must be made with cau on. Even though investors are worse off

with subsidized advisers, the por olio manager and investors, taken together, are be er off compared to

not having financial advisers, as long as investors are sophis cated. Furthermore, in a more general model

the value created by ac ve por olio managers would be endogenous. If their poten al profit is curtailed

36
by regula on, they are less likely to make the investment necessary to a ain high levels of exper se.

Appendix

A. Proof of Proposi on 2

Proof. Subs tu ng out AD and AI in the objec ve func on using the constraints, we have

( )
fP (Rm + c − δ)[α(1 − fP ) − (Rm + c − δ)] α(1 − fP ) − (Rm + c − δ) kAkm (c − δ)k−1
ΠP = − − δ.
(1 − fP )2 γ (1 − fP )γ (k − 1)(C0 Rm )k−1

The first order condi on for the op mal fee is

α − Rm − cA + 2δ
fP = . (A.1)
α + Rm + cA

Taking the par al deriva ve of ΠP with respect to δ, and subs tu ng out fP from the resul ng expression

using Equa on (A.1), we have

∂ΠP kAkm (cA − δ)k−2

= (cA − kδ).
∂δ (k − 1)(C0 Rm )k−1

Se ng this par al deriva ve equal to zero, we get the op mal rebate stated in Proposi on 2.24 Equa on

(21) is obtained by subs tu ng the op mal rebate into (A.1).

The second order condi ons can be verified in a straigh orward fashion.

B. Proof of Proposi on 6

Proof. Subs tu ng out fP and δ in the objec ve func on using the two constraints, we have

C0 Rm
ΠPi = {α − Rm − cA − γ[(ADi + AIi ) + ρ(ADj + AIj )]}(ADi + AIi ) + ADi (cA − ).
KAkm
[ (k−1)AD ]1/(k−1)

37
The first order condi ons are

∂ΠPi
= {α − Rm − cA − γ[(ADi + AIi ) + ρ(ADj + AIj )]} − γ(ADi + AIi ) = 0, (B.1)
∂AIi
∂ΠPi
= {α − Rm − cA − γ[(ADi + AIi ) + ρ(ADj + AIj )]} − γ(ADi + AIi )
∂ADi
C0 Rm ADi C0 Rm
+(cA − ) − . (B.2)
KAkm
[ (k−1)AD ]1/(k−1) AD (k − 1) [ KAkm
]1/(k−1)
(k−1)AD

Subs tu ng equa on (B.1) into (B.2), the la er reduces to

ADi C0 Rm
cA = [1 + ] , i = 1, 2. (B.3)
AD (k − 1) [ KAkm
]1/(k−1)
(k−1)AD

Note that the total size of each individual fund, Ai = ADi + AIi , is fully determined by equa on (B.1).

It is unique, symmetric, and independent of the kickback payments. Therefore equa on (27) remains to

hold. Equa on (B.3) shows that the direct channel choice is also unique and symmetric across por olio

managers. Therefore ADi = AD /2 in equa on (B.3). No ce further that from (31) we have

C0 Rm
= cA − δ,
KAkm
[ (k−1)AD ]1/(k−1)

thus we end up with

1
cA = (1 + )(cA − δ)
2(k − 1)

at op mum. Therefore the op mal rebate is given in equa on (32). Subs tu ng the op mal fund size and

op mal rebate into equa on (30), we obtain equa on (33).

38
References
Ang, Andrew, Ma hew Rhodes-Kropf, and Rui Zhao, 2008. Do funds-of-funds deserve their fees-on-fees?
Journal of Investment Management 6, 1--25.

Bergstresser, Daniel, John Chalmers, and Peter Tufano, 2009. Assessing the costs and benefits of brokers in
the mutual fund industry. Review of Financial Studies 22, 4129--4156.

Berk, Jonathan and Richard Green, 2004. Mutual fund flows and performance in ra onal markets. Journal
of Poli cal Economy 112, 1269--1295.

Bha acharya, Sudipto and Paul Pfleiderer, 1985. Delegated por olio management. Journal of Economic
Theory 36, 1--25.

Bindsbergen, Jules Van, Michael Brandt, and Ralph Koijen, 2008. Op mal decentralized investment man-
agement. Journal of Finance 63, 1849--1895.

Brown, Stephen, William Goetzmann, and Bing Liang, 2004. Fees-on-fees in funds-of-funds. Journal of
Investment Management 2, 39--56.

Carhart, Mark, 1997. On persistence in mutual fund performance. Journal of Finance 52, 57--82.

Chen, Joseph, Harrison Hong, Ming Huang, and Jeffrey Kubik, 2004. Does fund size erode performance?
Liquidity, organiza onal diseconomies and ac ve money management. American Economic Review 94,
1276--1302.

Chen, Joseph, Harrison Hong, and Jeffrey Kubik, 2010. Outsourcing mutual fund management: Firm bound-
aries incen ves and performance. University of Southern California Working Paper.

Chen, Xuejuan, Tong Yao, and Tong Yu, 2007. Prudent man or agency problem? On the performance of
insurance mutual funds. Journal of Financial Intermedia on 16, 175--203.

Christoffersen, Susan, Richard Evans, and David Musto, 2007. Fund flows vs. family flows: Evidence from
the cross sec on of brokers. Wharton School Working Paper.

Dangl, Thomas, Youchang Wu, and Josef Zechner, 2008. Market discipline and internal governance in the
mutual fund industry. Review of Financial Studies 21, 2307--2343.

Ding, Fei, 2008. Brokerage commissions, perquisites, and delegated por olio management. SSRN Working
Paper 1085963.

Edelen, Roger, Richard Evans, and Gregory Kadlec, 2008. What do so -dollars buy? Performance, expense
shi ing, agency costs. American Finance Associa on San Francisco Mee ngs.

Fung, William, David Hsieh, Narayan Naik, and Tarun Ramadorai, 2008. Hedge funds: Performance, risk,
and capital forma on. Journal of Finance 63, 1777--1803.

Gerstner, Eitan and James Hess, 1991. A theory of channel price promo ons. American Economic Review
81, 872--886.

Gervais, Simon, Anthony Lynch, and David Musto, 2005. Fund families as delegated monitors of money
managers. Review of Financial Studies 18, 1139--1169.

39
Grundy, Bruce, 2005. Combining skill and capital: Alternate mechanisms for achieving an op mal fund size.
University of Melbourne working paper.

Hung, Angela, Noreen Clancy, Jeff Dominitz, Eric Talley, Claude Berrebi, and Farrukh Suvankulov, 2008.
Investor and industry perspec ves on investment advisers and broker-dealers. Rand Ins tute For Civil
Jus ce Technical Report, Sponsored by the United States Securi es and Exchange Commission.

Inderst, Roman and Marco O aviani, 2009. How (not) to pay for advice: A framework for consumer pro-
tec on. Unpublished working paper.

Mamaysky, Harry and Ma hew Spiegel, 2002. A theory of mutual funds: Op mal fund objec ves and
industry organiza on. Yale School of Management Working Paper.

Massa, Massimo, 1997. Why so many mutual funds? Mutual funds, market segmenta on and financial
performance. INSEAD working paper.

Naik, Narayan, Tarun Ramadorai, and Maria Stromqvist, 2007. Capacity constraints and hedge fund strategy
returns. European Financial Management 13, 239--256.

Pauly, Mark, 1979. The ethics and economics of kickbacks and fee spli ng. The Bell Journal of Ecoomics
10, 344--352.

Persky, Joseph, 1992. Retrospec ves: Pareto's law. Journal of Economic Perspec ves 6, 181--192.

Sharpe, William, 1981. Decentralized investment management. The Journal of Finance 36, 217--234.

Stoughton, Neal, 1993. Moral hazard and the por olio management problem. Journal of Finance 48, 1009-
-2028.

Taylor, Terry, 2002. Supply chain coordina on under channel rebates with sales effort effects. Management
Science 48, 992--1007.

Tirole, Jean, 1988. The Theory of Industrial Organiza on. (The MIT Press).

Wahal, Sunil and Yan Wang, 2010. Compe on among mutual funds. Journal of Financial Economics Forth-
coming.

40
Notes
1
See Evolu on Revolu on: A Profile of the Investment Adviser Profession by the Investment Adviser Associa on (2008).
2
See Ownership of Mutual Funds Through Professional Financial Advisers, 2007 by the Investment Company Ins tute (2008).

According to the same study, 24% of all mutual fund assets are held through defined contribu on plans.
3
See HFR global hedge fund industry report: first quarter 2009 by Hedge Fund Research (2009).
4
In the la er case, one of these agents received more than $50 million from one por olio management firm for assistance in

obtaining CalPERS business.

5
For example, one of the major financial advisory firms, Ameriprise Financial, in its publica on Purchasing mutual funds through

Ameriprise Financial (2009), reported that it received more than $150 million marke ng support from the mutual fund companies

in 2008.
6
See Windows into the Mutual Fund Industry: 2007 in Review by Strategic Insight (2008).
7
E.g., Bha acharya and Pfleiderer (1985) and Stoughton (1993).
8
Some examples of mutual fund companies that u lize direct channels include Fidelity, Vanguard and Janus. Examples of

companies that offer their products through brokers include American Funds and Putnam.
9
Other papers concerning the structure of the investment management industry include Mamaysky and Spiegel (2002), Gervais,

Lynch, and Musto (2005), Massa (1997), Grundy (2005) and Ding (2008).
10
Chen et al. (2004) find large mutual funds underperform small ones. Fung, Hsieh, Naik, and Ramadorai (2008) find capital

inflows a enuate the ability of hedge fund managers to deliver alpha.

11
Note that ac ve funds with α < Rm do not exist on the equilibrium path. However, if off the equilibrium path an investor

chooses an ac ve fund without paying the screening cost, then inferior funds could be chosen.
12
Our simplified cost structure is representa ve of the more general situa on where the variable cost of inves ng is lower for

the direct channel and the fixed cost is lower for the indirect channel. In fact we can interpret the fixed cost of inves ng directly as

the cost of hiring a personal adviser and ensuring that he only works on behalf of the direct investor and is therefore not subject

to the same poten al conflict of interest or the same legal risks as in the case where the adviser works for many investors. The

fixed cost of retaining such a personal adviser is rela vely high, while the variable cost is rela vely low. Our results are essen ally

unchanged under this more general cost structure.

13
In a model in which individual investors incur addi onal costs of inves ng in the passive fund on their own, the adviser could

charge a fee on total assets under management.

14
Our general conclusions do not depend on this specific assump on about the wealth distribu on. In previous versions, we

have u lized several different assump ons about the wealth distribu on; our results are qualita vely similar. Including C0 in the

ini al wealth simplifies the subsequent nota on.

15
See Persky (1992) for a brief review of this literature.
16
Note w in the solu on to (7) is indeterminate because it does not ma er in our model whether the funds invested passively are

held on account with the adviser or by the investors themselves. However, the total amount of wealth invested ac vely through

41
the adviser is determinate and solved for below.
17
Note that cA must be ``discounted" by Rp (1 − fP ), since it is propor onal to the assets allocated at the beginning of the period

whereas the advisory fee is propor onal to the end-of-period value of the por olio.
18
If this condi on would not hold, the adviser is not u lized at all. In this case, the analysis in sec on V would apply.
19
One actual scenario described to us by a well-known finance professor is that his mother kept her money with an adviser

simply because the adviser always remembered her birthday by sending a floral bouquet.
20
This is consistent with the inverse-elas city rule of monopolist pricing. See, for example, Tirole (1988).
21
Our results are qualita vely unaffected if instead the investors were to assume they were in an equilibrium in which rebates

are provided but not used for promo onal ac vi es.

22
Assuming an upper bound for the rebate in the case of η > 1 is equivalent to assuming that there is a limit to the degree that

marke ng effort is effec ve in lowering the investor’s reserva on return. This would arise, for example, if the marke ng effort

required is a convex func on of the reduc on of the investor’s reserva on return below Rm . The upper bound may also reflect

some regulatory restric ons on the maximum rebate.

23
Note that because the popula on has been normalized to one, Am and C0 are aggregate quan es rather than measured at

the individual level.

24
If k > 2, δ = cA also sa sfies the first order condi on. However, in this case, AD = 0; the por olio manager's profit is not

maximized. Therefore δ = cA is not op mal.

42
 Internet Appendix for ``Intermediated Investment Management''*

Proof of Proposi on 4 and Corollary 1

Proof. Equa on (24) shows that the por olio manager's profit is maximized either at δ = cA /η or δ = δ̄ in

the case η > 1. Consider first the scenario δ = cA /η, in which the ac ve fund and the passive fund have

the same expected return. From equa on (24) we see ΠP is maximized at

1 1
AI = [α − Rm − cA + (η − 1)δ] − AD = [α − Rm − cA /η] − AD ,
2γ 2γ

kAkm ck−1
where AD = A
(k−1)(C0 Rm )k−1
. Subs tu ng this result and δ = cA /η into the second case of equa on (18)

yields the op mal management fee f∗P stated in the proposi on.

In the second scenario δ = δ, the ac ve fund underperforms the passive fund, and AD = 0. ΠP is

maximized at

′ 1
AI = [α − Rm − cA + (η − 1)δ̄],
2γ

Subs tu ng this back into the second case of equa on (18) and no ng that AD = 0 and δ = δ, we have
′
the op mal management fee, f∗P , for this scenario.

To prove Corollary 1, note that the por olio manager's profit in the first scenario is

(α − Rm − cA /η)2 AD cA
ΠP = + ,
4γ η

and in the second scenario it is

′ [α − Rm − cA + (η − 1)δ̄]2
ΠP = .
4γ
0
Cita on format: Stoughton, Neal M., Youchang Wu and Josef Zechner, [year], Internet Appendix to “Intermediated Invest-
ment Management,” Journal of Finance [vol #], [pages], h p://www.afajof.org/IA/[year].asp. Please note: Wiley-Blackwell is not
responsible for the content or func onality of any suppor ng informa on supplied by the authors. Any queries (other than missing
material) should be directed to the authors of the ar cle.

1
The difference between the por olio manager's profits under these two scenarios is then

′
ΔΠP = ΠP − ΠP
c2A /η2 − [cA − (η − 1)δ̄]2 − 2(α − Rm )[cA /η − cA + (η − 1)δ̄]
= + AD cA /η.
4γ

If ΔΠP > 0, the por olio manager chooses the first (equal performance) equilibrium with δ = cA /η. Oth-

erwise she chooses the second (underperformance) equilibrium with δ = δ̄. To see which equilibrium is

more likely to occur, we take the par al deriva ve of ΔΠP with respect to various model parameters. A

nega ve par al deriva ve means the second equilibrium is more likely to occur as the parameter value

increases.

First, note that

∂ΔΠP 1
= − [α − Rm − cA + (η − 1)δ̄](η − 1) < 0,
∂ δ̄ 2γ

where the inequality from our assump ons α > Rm + cA , η > 1, and δ̄ > cA /η. Therefore, when δ̄ is high,

it is more likely that the por olio manager prefers the underperformance equilibrium.

Second, note that C0 and k affect ΔΠP only through AD . From the expression of AD , it is easy to see that

AD is decreasing in both C0 and k:

∂log(AD )
= −(k − 1)/C0 < 0,
∂C0

∂log(AD ) 1 1 Am cA 1 1 Am
=( − ) + log( )=( − ) + log( ∗ ) < 0.
∂k k k−1 C0 Rm k k−1 A

Since ΔΠP increases in AD , it follows that ΔΠP decreases in both C0 and k.

Third, note that

∂ΔΠP 1
= − [cA /η − cA + (η − 1)δ̄] < 0,
∂α 2γ

where the inequality follows from our assump ons η > 1 and δ̄ > cA /η.

2
 Finally, we have

∂ΔΠP 1
= {(α − Rm − cA /η)cA /η2 − [α − Rm − cA + (η − 1)δ̄]δ̄} − AD cA /η2
∂η 2γ
1
< {(α − Rm − cA /η)δ̄ − [α − Rm − cA + (η − 1)δ̄]δ̄} − AD cA /η2
2γ
1
= [(−cA /η + cA − (η − 1)δ̄)]δ̄ − AD cA /η2
2γ
< −AD cA /η2

< 0,

where the first inequality follows from the assump ons α > Rm + cA , η > 1, and δ̄ > cA /η, and the second

inequality follows from the assump ons η > 1 and δ̄ > cA /η.

Proof of Proposi on 7

Proof. Combining the two constraints in problem (34) we immediately obtain equa on (36). Subs tu ng

this expression back into the objec ve func on and differen a ng, we have

∂ΠP λk 1/(k−1)
= α − Rm − 2γAD − A ,
∂AD k−1 D
∂ 2 ΠP λk (2−k)/(k−1)
= −2γ − A < 0.
∂A2D (k − 1)2 D

Equa on (35) is given by se ng the first order condi on above equal to zero. Since ΠP is strictly concave

when AD > 0, the first order condi on is both necessary and sufficient condi on for the solu on to this

maximiza on problem; furthermore, the op mal AD is unique. To prove the existence of an interior solu on,

0 < AD < W − C0 , to the first order condi on, note that ∂ΠP
∂AD > 0 if AD = 0. Due to the monotonicity of the

first deriva ve, it suffices to show this deriva ve becomes nega ve as AD → W −C0 , i.e., as AD converges to

the aggregate wealth of the economy net of the search cost C0 . This is guaranteed by the condi on (6).

Proof of Proposi on 8

Proof. In the case without financial advisers, the number of direct investors is the same as the number of

investors inves ng in the ac ve por olio. Denote the total surplus of the (direct) investors, rela ve to the

3
default of passive investment, by S0 , we have
∫ +∞
0
S = [x(α − γA0D )(1 − fP ) − (x + C0 )Rm ]f(x)dx
A∗0

= [(α − γA0D )(1 − fP ) − Rm ]A0D − θ 0 C0 Rm ,

where A∗0 is the threshold wealth level (net of C0 ) that makes the marginal investor indifferent between the
∫ +∞ kAkm ∫ +∞
passive fund and the ac ve por olio, A0D = A∗ xf(x)dx = (k−1)(A ∗ )k−1 , θ ≡ A∗
0
f(x)dx = ( AAm∗ )k .
0 0 0 0

In the equilibrium with independent advisers, the net return of the ac ve fund is equal to Rm + cA .

Therefore the (direct) investor's surplus, S1 , is given by

∫ +∞
S1 = [x(Rm + cA ) − (x + C0 )Rm )]f(x)dx
C0 Rm
cA

= A1D cA − θ 1 C0 Rm ,

kAkm ck−1
where A1D = A
(k−1)(C0 Rm )k−1
, and θ 1 ≡ ( CcA0ARmm )k .

Similarly, since the net return of the ac ve por olio in the case with subsidized advisers equals Rm +

cA − δ, the total surplus of the (direct) investors in the subsidized adviser equilibrium is given by

S2 = A2D (cA − δ) − θ 2 C0 Rm ,

kAkm (cA −δ)k−1

where A2D = (k−1)(C0 Rm )k−1
, and θ 2 ≡ ( (cAC−δ)A
0 Rm
m k
).

In the unsophis cated investor case, high networth investors have a deadweight loss of C0 . The frac on

of investors who pay this cost is the same as in the case without rebate, i.e., θ 1 . The indirect investors earn

an expected return lower than the passive return by ηδ, where δ equals either cA /η or δ̄. Therefore, the

total investor surplus in this case is

S3 = −A3I ∗ ηδ − θ 1 C0 Rm < 0.

Note that investor surplus S0 , S1 and S2 must all be strictly posi ve, otherwise no ra onal investors will

pay the search cost. Therefore S3 < 0 is lowest among all the four equilibria. To prove S1 > S2 , note that
∫ C 0 Rm
1 2
cA −δ C0 Rm
S −S = A2D δ + (A1D − A2D )cA 1 2
− (θ − θ )C0 Rm = A2D δ + xf(x)[cA − ]dx > 0.
C0 Rm x
cA

This equa on indicates that investors' welfare loss due to the existence of the kickback can be decomposed

4
into two parts: investors who remain in the direct channel lose A2D δ; investors who would originally choose

the direct channel, but are forced to switch to the indirect channel because of the kickback lose (A1D −

A2D )cA − (θ 1 − θ 2 )C0 Rm . Both components are strictly posi ve.

Adding the por olio manager's profit to the investor surplus, we get the total welfare U0 , U1 , U2 and U3

in Proposi on 8. To prove U1 > U2 , recall that allowing kickbacks increases the por olio manager's profit

by A2D δ (equa on (23)), therefore the first component of the investor welfare loss described above is exactly

offset by the gain of the por olio manager. However, the second component is a deadweight loss.

To prove U1 > U3 , we first compare the independent adviser equilibrium with the unsophis cated

investor equilibrium with δ = cA /η. Using the expressions for investor surplus and the por olio manager's

profit for both cases and a er some algebra, we have

c2A 1 A1 cA
U1 − U3 = (1 − 2 ) + I > 0,
4γ η η

where A1I denotes the amount of indirect investment in the independent adviser equilibrium. Similarly,

comparing the independent adviser equilibrium with the unsophis cated investor equilibrium with δ = δ̄,

we have

2
(η − 1)(η + 1)δ̄
U1 − U3 = A1D cA + (A1D + A1I )δ̄ + > 0.
4γ

This completes our proof of Proposi on 8.