be. But we do know his fees. The manager must add value at least equal to
his fees just to equal an index fund. So we must take fees seriously.
Letā™s recall a few facts of life. Great investment managers command
high compensationā”perhaps higher than warranted by their contribution
to society in general. Thatā™s true also of star athletes, popular actors, and
top corporate executives. Compensation is controlled by supply and de-
mand, which means charging what the market will bear. If we place a low
limit on our fee schedules for active management, we are likely to get no
more than what we are paying for.
On the other hand, it doesnā™t work the other way. The only thing we
know for sure is the fees weā™ll be paying. And paying high fees does not as-
sure us of better long-term performance. In the end, what counts is only
what we can spendā”performance net of fees.
In the selection of mutual funds, I also recommend sticking with funds that do not
charge so-called ā12(b)(1) fees.ā These fees equal up to 0.25%/year of assets and
are used by a mutual fund for advertising and promotional purposes. The fees are
permitted by the SEC in what I consider an inappropriate action by the SEC, be-
cause the fees clearly do not promote the interests of mutual fund investors. The
fees enable a mutual fund to become larger, which in due course reduces the ļ¬‚exi-
bility of its fund manager to perform. There is an ample supply of good mutual
funds with the integrity not to charge 12(b)(1) fees, and I would stick with them.
118 SELECTING AND MONITORING INVESTMENT MANAGERS
āManaging investment managers is easy,ā asserted a chief ļ¬nancial ofļ¬cer I
met back in the 1970s. āEach year we simply ļ¬re the managers with the
worst two records over the last three years.ā
Oh, if only it were that simple!
At least once each year, our adviser should review each individual
manager with the investment committee The review should cover not only
performance but also why the manager still ļ¬ts our criteria for hiring and
retaining managers. In particular, the adviser should explain why he still
considers the manager the best we can get in its asset class.
When to Take Action
Adding or withdrawing assets to or from a managerā™s account should not be
a sign that we have made a positive or negative evaluation of a manager.
We may increase an accountā™s assets as the result of a new contribution to
the plan, or a transfer of assets from another account that had become over-
weighted relative to its target. We may periodically have to withdraw assets
from an account to raise money to make payments to our sponsor or to
pensioners. More often, such actions simply reļ¬‚ect efforts to adjust the allo-
cation of total plan assets closer to the Policy Asset Allocation.
I do not advocate withdrawing a portion of a managerā™s account sim-
ply because we have given him a poor evaluation. Investment managers do
not need a wake-up call. Any manager worth his salt is going all-out con-
tinuously and knows when he is not performing well. He cannot perform
better simply by working harder. If heā™s performing poorly, he is probably
hurting even more than we are. We canā™t possibly ļ¬‚agellate him into better
performance. If we have lost conļ¬dence in a manager, we should terminate
On one or two occasions, I have heard sponsors consider giving a
manager a warning, something like, āWeā™ll give you X quarters to
straighten out your performance, or weā™ll have to terminate the account.ā
While itā™s essential to be honest with our managers, I believe such a warn-
ing is never appropriate. First of all, we can never know whether his (short-
term) performance during the warning period has any predictive value.
And second, the manager gets an absolutely wrong motivation: āWe must
do something, anything, because if we do nothing we will lose the account.
And if we do something, maybe weā™ll get lucky.ā We never want to moti-
vate our manager to roll the dice.
When to Terminate a Managerā™s Account
Reasons for termination may ļ¬t under ļ¬ve overlapping headings:
1. We lose trust in the manager. If we believe a manager is being less than
honest with us, or if he fails to honor agreements with us, it is time to
part company. Trust is a sine qua non of our relationship with any
2. We lose conļ¬dence that a manager can add much value to his bench-
mark. What would cause us to lose conļ¬dence?
I Performance is below benchmark
ā“ for a meaningful interval
ā“ by a sufļ¬cient magnitude, and
ā“ for reasons not explainable by investment style such as market
cap, or growth vs. value
so we can no longer objectively expect that the manager is likely to
exceed his benchmark materially in the future.
I The managerā™s performance has become inexplicably erratic.
3. Even though a managerā™s performance remains satisfactory, its predic-
tive value declines materially. This judgment is rarely based on a single
factor. It is inļ¬‚uenced by factors such as the following:
I A key person (or persons) left our account.
I The manager embarked on a different management approach than
that on which his track record is based.
I The manager is now managing much more money than that on
which his track record is based, and we believe this added money
will impair his future performance.
4. We ļ¬nd another manager we believe would add materially more value
than an existing manager in the same niche, even though the existing
manager has done well for us.
5. Here are two reasons related to our diversiļ¬cation needs:
I We perceive that two of our managers in the same asset class are
pursuing the same investment style.
I We have reduced our Policy Allocation to a particular asset class to
an extent where we no longer ļ¬nd it important to have as many
managers in that asset class.
Much of this comes down to assessing the managerā™s likely perfor-
mance in the years ahead. Assessing that requires tough objectivity of all
120 SELECTING AND MONITORING INVESTMENT MANAGERS
relevant facts. The key word is relevant. Figuring out whatā™s relevant and
whatā™s not for a particular manager is a major challenge, and if we get it
wrong, we are likely to take the wrong action.
Once we have gotten the actual asset allocation of our portfolio to be the
same as our Policy Asset Allocation, it wonā™t stay that way. One asset class
will perform better than another, and weā™ll soon be off target. We should
rebalance periodically to our Policy Allocation. This periodic rebalancing
forces us to do something that is not intuitively comfortableā”sell from as-
set classes that have performed best and reinvest the proceeds in those that
have performed worst.
David Swensen has articulated clearly the case for rebalancing: āFar
too many investors spend enormous amounts of time and energy con-
structing policy portfolios, only to allow the allocations they established to
drift with the whims of the market. . . . Without a disciplined approach to
maintain policy targets, ļ¬duciaries fail to achieve the desired characteristics
for the institutionā™s portfolio.ā13
Many plan sponsors set ranges for their Policy Asset Allocationsā”
such as 20% plus or minus 5%. The market could drive such an asset class
mighty far from its 20% target before the plan sponsor would be moti-
vated to take some action.
My preference is for a pinpoint target for each asset classā”X% of the
portfolio. No range. And when the market drives the asset class away from
that target, letā™s rebalance to bring it back. Why?
If the outperformance of a particular asset class gave any valid pre-
diction that the same asset class would outperform in the next inter-
val of time, that would be a valid reason for utilizing a range. But
that is not the case. Outperformance by Asset Class A in Interval 1
gives almost zero information about its performance in Interval 2ā”
with one exception. One of the pervasive dynamics in investments is
reversion to the mean, and sooner or later Asset Class A will begin to
Swensen, Pioneering, p. 4.
Hence, over the long term, rebalancing to a target may add a tiny in-
crement of return by forcing us, on average, to buy low and sell high.
That can be a counterintuitive discipline, of courseā”adding money to
an asset class (and therefore to managers) who have been less success-
ful lately, and taking it away from stellar performers. But it makes
sense, provided we retain high conļ¬dence in all our managers.
If we were conļ¬dent we could predict with reasonable accuracy which
asset class would outperform or underperform others in Interval 2,
then we should take advantage of tactical asset allocation insights. I,
for one, have no such conļ¬dence, and I donā™t tend to have much conļ¬-
dence in such insights of others. Unless we justiļ¬ably have that conļ¬-
dence, rebalancing is the way to go.
We can spend a lot of time agonizing over where to take that with-
drawal we need, or where to place our latest contribution. A rebalanc-
ing discipline removes a good deal of the agonizing and also makes
Presumably we established our Policy Asset Allocation in order to earn
the best expected return for a given level of aggregate portfolio risk. To
the extent we stray from our Policy Allocation, we are probably stray-
ing from our target portfolio risk and from the Efļ¬cient Frontier.
Doesnā™t rebalancing incur unnecessary transaction costs? It doesnā™t
have to. If we have sizable amounts of contributions to or withdrawals
from our fund, we can probably rebalance without any incremental
costsā”simply by using those cash ļ¬‚ows to rebalance. Or if we invest in no-
load mutual funds or commingled funds, we can usually rebalance without
cost to us. Or if we use index funds, we can keep part of our assets in index
futures, which we can rebalance at little cost.
But what if rebalancing will necessitate additional transactions by
some of our investment managers?
We can actually execute such transactions with minimal incremental
cost if we give our manager enough notice. If we tell a manager, āwe will
need $10 million from our $100 million account any time in the next three
months,ā he can often raise much of that $10 million through his normal
transactions during that quarter, simply by his not reinvesting proceeds
from his routine sales.
Even if we can rebalance entirely from the withdrawals we must take
from our fund periodically to pay pension beneļ¬ts or endowment income,
we should still forecast our cash needs well in advance and try to give our
122 SELECTING AND MONITORING INVESTMENT MANAGERS
managers a few months, if possible, to raise the cash. All for the purpose of
minimizing transaction costs.
How often should we rebalance?
There have been some good academic studies on rebalancing. Some
suggest we might be ahead by adopting a quarterly discipline. Others seem
to suggest that there is little difference between doing it once a quarter or
once a year. Still others suggest not more than once a year. Yet others indi-
cate that use of a target range is best. Differences in results of the studies
depend on the particular time intervals of the studies. I favor doing it con-
tinuously with cash ļ¬‚ow and then taking action once a year if we are still
materially off our Policy Allocation.
How about rebalancing managers within the same asset class? If two
managers are in two different subclasses, such as large-cap growth and
large-cap value, I think rebalancing makes sense. If two managers are in
the very same asset class, well, I havenā™t seen scholarly studies on that, and
I think thatā™s up to our qualitative judgment on a case-by-case basis, but re-
balancing between them may make the best sense.
The most important point is to have a rebalancing plan and stick to it,
with exceptions rarely more frequent than once in 10 years.
Once we have established our portfolioā™s Policy Asset Allocation, we
then want the best managers we can get in each individual asset class.
This typically means a relatively large number of managers.
As committee members, we should know the criteria for hiring and re-
taining each manager and be prepared to question our adviser on the
basis of these criteria.
We should know at least as much about our existing managers as
about managers we are considering to hire, and we should ask each
year if our existing managers are still the best we can get.
o matter how small, an investment fund with multiple managers should
N have a custodian. Why?
A custodian safeguards the assets. It sees that every penny is always in-
vested, at least in a money market fund if not otherwise directed. It exe-
cutes all transactions as directed and provides regular transaction reports
and asset statements.
Yes, it is possible for a tiny investment fund to own shares in multiple
mutual fund accounts without a custodian, but it places a burden on the
sponsorā™s small staff or a volunteer board member. And even if they can do
the job well, their successors may not be as competent in that task.
If the investment fund has one or more separately managed accounts
(as opposed to mutual funds), a custodian is a must. For one thing, it sepa-
rates the function of custody from that of investment management. The in-
vestment manager doesnā™t get its hands on any assets. Nobody at the
managerā™s ļ¬rm can abscond with any assets as long as custody is with a
different party. Nor can it lie to us about the assets it is managing for us.
For most kinds of skullduggery, the trustee and investment manager (or
their staff people) would have to colludeā”a less likely occurrence. Some
notorious cases of fraud could hardly have happened if the client had used
The custodian is usually a bank, although the task might feasibly be
done by a broker or even a consultant. One advantage of the bank is that
the assets never belong to the bank. If the bank ever went bankrupt, its
creditors could not get their hands on any of our assets. This is not true of a
broker who holds client assets in a custody account, because the assets in
that custody account are in the brokerā™s name. If the broker ever went
bankrupt, our assets would be in jeopardy.
A tiny investment fund will want to ļ¬nd the lowest-cost custodian,
124 THE CUSTODIAN
perhaps from its local bank. The fund will receive transaction reports and
asset statements, but probably little else. Without paying extra, it will not
receive performance reports or other analyses, and it may not need them if
it has a consultant who can provide those services.
At the end of each month, a sophisticated custodian provides the client the
following statements for each account and for the overall fund:
An asset statement showing the month-end book value and market
value, usually categorized into convenient asset groupings selected by
the client, with subtotals for each grouping.
A list of all transactions, including those that were accrued but not set-
tled at the end of the month. Accrued transactions are noted on the as-
set statement as āAccounts Payableā and āAccounts Receivable.ā
For each account, a daily log of all purchases and sales of securities
during the month, as well as all contributions and all withdrawals or
expenses paid at the direction of the fundā™s sponsor.
A list of all income (interest and dividend payments) received.
A list of all cash ļ¬‚ows into (contributions) and out of (disbursements)
Foreign investments are shown in terms of both local and U.S. currency.
Special reports requested by the client, such as aggregate brokerage
commission reports, or a listing of the largest 10 transactions during
For an investment fund with separately managed accounts, we might
require that the custodian send a copy of each accountā™s monthly statement
to the manager of that account, and that the manager, within two weeks of
receipt, review that statement and write a letter to the custodian (with a
copy to our adviser) saying either (1) the manager agrees entirely with the
custodianā™s statement, or (2) it agrees except for the following items (and
then lists each variance). The custodian and manager must then get to-
gether and resolve each variance, again with a copy to the adviser explain-
ing the resolution. This procedure serves to improve the accuracy of both
the custodianā™s and managerā™s records. The adviserā™s task is that of moni-
toring the mailā”to make sure each manager responds to the custodian
each month, and that each item of variance is resolved.
Why do I like this process? No one knows more than the manager
about his particular assets and transactions. He is aware of more nuances
than any outside auditor. Also, he has a vested interest in keeping the
custodianā™s records accurate, because he knows we are relying on the cus-
todianā™s statements, not his (the managerā™s). Moreover, any errors in the
custodianā™s records can affect the calculation of investment performance,
perhaps the most important management information of all.
The custodian will provide another important service: For an organiza-
tion with multiple restricted endowments or multiple pension plans that the
organization commingles for purposes of investment, a custodian can pro-
vide the unitized recordkeepingā”like the recordkeeping for a mutual fund.
The custodian also prepares and ļ¬les tax returns and special govern-
ment reports as necessary.
Custodianship today is a highly capital-intensive industryā”with the
capital all going into systems and software to carry out the trusteeā™s monu-
mental information-processing function. The needed capital can run into
hundreds of millions of dollars. Much of this capital investment is for the
provision of management information, which we will discuss next.
Given all the raw data that a custodian has about every investment ac-
count, no one is better situated to provide us with analyses of performance
and the composition of individual investment accounts and our overall in-
vestment account. Besides the usual performance charts that are regularly
presented at committee meetings, the sophisticated custodian can provide
our adviser with a range of analytics that is as broad as oneā™s imagination.
The more sophisticated master custodians, with all of the raw data in
their computers, have over the years developed reliable and ļ¬‚exible perfor-
mance measurement systems. A good trustee keeps track of some 2,000
different indexes and combinations of indexes for use as benchmarks, as
requested by its various clients. And it can compare our fundā™s perfor-
mance against a universe of other endowment or pension funds. The custo-
dian can then present this information in any number of graphic forms.
The more sophisticated custodian can slice and dice the composition of
any particular portfolio by virtually any measure one can imagineā”by in-
dustry, by country, by price/earnings ratio, by market capitalization, and
by innumerable other measures. And it can display these in a wide range of
126 THE CUSTODIAN
The convenience afforded by having all of our planā™s assets under one
roof gives us, the plan sponsor, remarkable ļ¬‚exibility. We can do things
easily that would be difļ¬cult or impossible to do without a custodian.
If we terminate a manager, we simply have to instruct our custodian not
to accept any more transactions from the terminated manager, and then