Full text of Economic Review (Federal Reserve Bank of Atlanta) : Fourth Quarter 1997, Volume 82, Number 4

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Treasury Auctions: What
Do the Recent Models
and Results Tell Us?
S A I K AT N A N D I
The author is a senior economist in the financial
section of the Atlanta Fed’s research department. He
thanks Peter Abken, Jerry Dwyer, and Larry Wall for
helpful comments and Rob Bliss for useful discussions.

A

UCTIONS, AS SELLING MECHANISMS, HAVE EXISTED FOR WELL OVER TWO THOUSAND YEARS
AND HAVE BEEN USED TO SELL A WIDE SPECTRUM OF GOODS.

THE ANCIENT BABYLONIANS

SOLD WIVES THROUGH AUCTIONS, AND THE LEGIONS OF ANCIENT
BOOTY THROUGH AUCTIONS.

CURRENTLY,

ROME OFTEN SOLD THEIR

AUCTIONS ARE EMPLOYED TO SELL OBJECTS AS

DIVERSE AS ARTWORK, MINERAL RIGHTS, CUT FLOWERS, GOLD, TOBACCO, THOROUGHBRED HORSES,
AND CORPORATIONS.

One of the most important auction markets in the
world today is that of U.S. Treasury securities; approximately $2 trillion worth of Treasury securities was
auctioned in 1995. As in other auctions, a set of rules
determines how bids are used to determine prices; this
set of rules makes up the format of the Treasury
auction.
A long-standing debate (dating back to the early
1960s) has centered on the selection of an appropriate
auction format for various Treasury securities, one that
would be least subject to possible manipulations by any
individual trader or a cartel and also result in the highest possible revenues for the Treasury.1 This debate
received fresh impetus after the infamous May 1991
auction for two-year Treasury notes, which led to a
squeeze in the available supply of these securities in
the postauction, or secondary, market. At this auction
Salomon Brothers, one of the biggest primary dealers,
grossly violated the maximum amount of the note that
it could buy (U.S. Treasury 1992). As a result, the two-

4

year notes started trading at abnormal premiums in the
secondary market. In other words, following the auction, the secondary market for the two-year notes suddenly became very illiquid. An illiquid secondary
market could not only increase the Treasury’s cost of
financing in subsequent auctions but is also detrimental to the Federal Reserve’s ability to carry out its open
market operations in the most efficient manner. An
understanding of the various auction formats and what
they entail for the Treasury’s cost of financing,
squeezes, and market liquidity is important to all participants of the Treasury auction market and, to some
extent, everyone interested in Treasury securities.
The nature of Treasury auctions differs from many
other types of auctions. Most consist of the sale of a single good that is not traded before or after the auction.
These auctions have been the subject of considerable
study. In contrast, in a Treasury auction, multiple units
of the same good are auctioned, and the auction is preceded by trading in a forward market, in which one can

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

obtain the to-be-auctioned securities at a price fixed in
advance of the auction. The auction is followed by
active trading in a secondary market. It is therefore
informative to look at models that take into account
these unique features of Treasury auctions. The vulnerability of any auction format to collusion among bidders
and the revenue superiority of one auction format to
another could very well depend on these particular features of the Treasury auction market.
This article seeks to explain the current understanding of Treasury auctions in light of recent theoretical research that takes into account the distinguishing
features of Treasury auctions and ongoing empirical
research that looks at these issues. It also informally
explores, in the context of current research findings,
the effects of certain contemplated changes in existing
auction formats on collusion and squeezes and hence
on the Treasury’s borrowing costs. The article first provides a brief description of the different types of auctions and the structure of the market for Treasury
auctions. The discussion then turns to the current theoretical models that incorporate the unique features of
Treasury auctions—what they imply for the two formats
currently used to auction securities and related empirical evidence. The final section analyzes the possible
effects of some of the contemplated changes in the present
auction mechanism.

The Market for Treasury Auctions
his section briefly reviews some of the institutional details of the Treasury auction market. (The
Joint Report on the Government Securities Market
[U.S. Treasury 1992] and Stigum 1990 provide more
detailed coverage.) Box 1 describes some of the commonly used auction formats, namely, the English auction,
the Dutch auction, the first-price sealed-bid auction, and
the second-price sealed-bid auction. The two currently
used formats for Treasury auctions, the uniform-price
auction and the discriminatory auction, are related to but
somewhat different from the first-price and second-price
auctions.
Bidders in Treasury auction declare themselves to
be one of two types, competitive and noncompetitive.
Competitive bidders, the bulk of whom are the thirty-nine
dealers designated as primary dealers by the Federal

T

Reserve Bank of New York, submit sealed bids specifying
the dollar amounts of the security the bidder is willing to
buy at each yield.2 Noncompetitive bidders submit a
quantity bid, up to $1 million for bills and $5 million for
notes and bonds, but do not specify any yield; the total
amount of noncompetitive bids is subtracted from the
dollar amount of the security to be auctioned, and what
remains is available for competitive bidding.3
Box 2 illustrates how bidders are awarded securities
in a Treasury auction. A bid is a demand schedule that
specifies the dollar amount of the security the bidder is
willing to buy at each yield. In a uniform-price auction,
all bidders pay the same yield (the market-clearing
yield) for the entire quantity they are awarded. In a discriminatory auction, each bidder pays for a quantity of
the accompanying yield in the demand schedule, as discussed in Box 2. Currently, only the two- and five-year
notes and the inflation-indexed bonds are auctioned
through the uniform-price procedure.4 For all other
bills and bonds, the discriminatory format is used.
As soon as the Treasury announces the total dollar
amount of the particular security to be auctioned, trading begins in a forward market, called the when-issued
market. An investor, instead of bidding at the auction,
can lock in a yield in the when-issued market by buying
the when-issued contract. The seller of the forward contract is under contractual obligation to deliver the
Treasury security to the buyer of the contract at the
time the Treasury delivers the securities to the successful bidders.5 After the auction, an active secondary market develops for the newly auctioned security, often
called an on-the-run security. The secondary market
also trades close substitutes of the newly auctioned
security, namely, securities from a previous issue of different maturity (off-the-run securities) that have maturities very close to the newly issued security’s maturity.
In general, on-the-run securities tend to be more liquid
than off-the-run securities of comparable maturity.
There is also a repurchase (repo) market for
Treasury securities (see Syron and Tschinkel 1987). It
is a market for short-term debt for which a specific
Treasury security is held as collateral by a lender. In
general, borrowers who want funds from the repo market
place their securities as collateral and agree to repurchase these securities at a later date (often overnight) at

1. Note that the concept of the Treasury’s maximizing revenue is equivalent to its minimizing the cost of debt.
2. Treasury securities dealers with whom the Federal Reserve trades directly as part of its open market operations are called
primary dealers. Trading by the primary dealers accounts for the bulk of trading in the secondary market. To become a primary dealer, a firm must also be committed to bidding nontrivial amounts at the Treasury auctions.
3. Noncompetitive bids, on average, compose about 15–20 percent of the total dollar amount of an auction.
4. The Treasury conducted six uniform-price auctions of long-term bonds during 1973 and 1974 but discontinued those thereafter. In fact, in one of the auctions, the tendered amount did not exceed the intended dollar amount of the issue.
5. Currently, there is a limit on the amount of securities a bidder can get at a single auction. Inclusive of positions in the
futures, forwards, and the when-issued market, a bidder’s net position in an auction at any given yield may not exceed 35
percent of the total dollar amount of the auction.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

5

B O X

1

Commonly Used Auctions
In an English (ascending-price) auction, often employed
to sell artwork or antiques, prices are progressively raised,
either by the auctioneer or with the bidders placing their
bids directly, until the object is sold to the bidder who stays
at the last round.

Unlike English and Dutch auctions, in which each bidder
gets to observe the bids of all other bidders, in sealed-bid
auctions, as the name suggests, the bidders submit sealed
bids. In a first-price sealed-bid auction, the object is
awarded to the highest bidder at the bid he submits.

The Dutch (descending-price) auction, used to sell cut
flowers in Holland, among other things, is the direct opposite of the English auction—that is, prices are successively lowered.1

In a second-price sealed-bid auction, the object is awarded to the highest bidder, but the winner pays the bid of the
second-highest bidder.

1. “Dutch auction” in this article refers to auctions of this type only and not the uniform-price Treasury auction, as is the case
sometimes.

a predetermined price. The predetermined price is higher than the amount loaned, the difference being the
interest earned on the loan of short-term funds.

Winner’s Curse
or bidders in a Treasury auction, the value of a
security (that will be auctioned) is its resale price
in the secondary market after the auction. The
true value is an unobserved quantity (that is, a random
variable) that is common across all bidders. Auctions of
this type are known as common-value auctions.
Winning a bid award in a common-value auction is
often associated with a phenomenon known as the winner’s curse. In a first-price auction of this type, the winning bidder is the one who has the highest estimate of
the object’s true value. Having won the auction may not
be particularly good news as it implies that everyone
else’s estimate of the true value was lower. The winner
may well have overestimated the value and could suffer
a loss in trying to sell the object. This winner’s curse has
been noted in auctions for off-shore oil rights and markets for baseball players, for example.6 Realizing this
possibility when bidding, bidders are likely to shade
their bids below their estimates of the object’s true
value, and the result is a loss of potential revenue to the
seller. A second-price auction, in which the winner pays
the highest losing bid, mitigates the winner’s curse by
having the bidder pay the second-highest bid. Because
the extent of bid shading is likely to be lower, it can also
result in higher expected revenue for the seller than a
first-price auction.7
Although Treasury auctions are for multiple units of
the same good, they share certain common features with

F

6

auctions for a single unit of a good in which the winner’s
curse is an important phenomenon. A uniform-price
Treasury auction is similar to a second-price single-unit
auction because the winning bidders pay not necessarily their bid prices but a common market-clearing price,
which could often be lower than their bid prices. On the
other hand, a discriminatory Treasury auction is similar
to a first-price single-unit auction because all winning
bidders pay their bid prices. Given the similarity
between second-price and uniform-price auctions, it is
quite possible that the severity of bid shading is much
lower in a uniform-price auction than in a discriminatory auction, leading some to argue that a uniform-price
auction would be the better choice for the Treasury.

Uniform-Price or Discriminatory Auctions?
The Traditional View
he debate regarding the revenue superiority of
uniform-price auctions over discriminatory auctions was first initiated by Friedman (1960). He
argued that the possibility of the winner’s curse in discriminatory auctions discourages participation by relatively uninformed bidders, in turn leading to reduced
competition and the possible formation of a cartel consisting of a small number of bidders. Another argument
in favor of a uniform-price auction is based on the notion
that bidders bid more aggressively in uniform-price auctions than in discriminatory auctions. These points have
been elaborated in Bikhchandani and Huang (1989),
Chari and Weber (1992), and Smith (1992). However,
these researchers modeled Treasury auctions as singleunit auctions and ignored the fact that bidders in
Treasury auctions can submit demand schedules.

T

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

Although Chari and Weber do recognize that a multipleunit auction is different from a single-unit auction, the
distinction is minimized by their claim that the economic logic of the arguments for the single-unit auction
seems likely to carry over to a multiple-unit auction.
In the aftermath of the Salomon episode, the
Treasury switched from discriminatory to uniform-price
auctions for two- and five-year notes on an experimental
basis.8 In fact, then Undersecretary of the Treasury for
Finance Jeremy Powell stated that one of the primary
reasons for switching was the “very substantial academic
opinion that the single price auction could result in lower
financing costs.” Recent research suggests, however, that
a uniform-price auction could actually benefit the
Treasury less than a discriminatory auction because it
allows bidders to submit demand schedules instead of
bidding for the whole unit.

Uniform-Price or Discriminatory Auctions?
Current Perspective and Models
s mentioned before, in a Treasury auction the competitive bidders submit demand schedules. It
turns out that the ability to submit demand schedules conveys an important strategic advantage to bidders
in uniform-price auctions, and it is one of the primary
focuses of the recent research on Treasury auctions.
Extending an important result by Wilson (1979), Back
and Zender (1993) shows that uniform-price auctions
may actually encourage implicit collusion among bidders
and cost the Treasury money by awarding the auction at
too low a price.
The following simple example illustrates the basic
intuition of Back and Zender (1993). Assume that the
Treasury is going to auction $10 billion worth of a oneyear zero-coupon security and there are three risk-neutral
bidders and no noncompetitive demand; the expected
yield in the after-market will be 5 percent, and each
competitive bidder agrees on that amount.9 Suppose
each of the bidders agrees to submit two bids, one for
$3333 million at a yield of 6 percent and another for
$6666 million at a yield of 20 percent. Given these bids,
the uniform-price auction will clear at the higher yield
of 20 percent (or equivalently at a lower price), which
is bad news for the Treasury. Each bidder gets one-third
of the $10 billion ($33331⁄3 million) Treasury issue at 20

A

percent, a very lucrative outcome; implicitly each bidder is part of a cartel that divides the issue equally
among its members. The usual problem with such cartels is that each member has an incentive to deviate and
bid a slightly higher price than agreed by the cartel. The
deviating member gets more volume (perhaps all the volume) at only a small loss in profits. Since the face value
of the security that a bidder gets from this arrangement is
$33331⁄3 million and the associated yield is 20 percent, the
cost of the securities is $(3333.33/1.2) million '$2777.78
million. Similarly, the expected revenues from selling
the securities at a yield of 5 percent is $(3333.33/1.05)
million ' $3174.6 million. Therefore, each bidder’s
expected profit as a member of the cartel is $(3333.33) 3
(1/1.05 – 1/1.2) ' $396.82 million. If one of the bidders
deviates and submits a bid of 19.99 percent, then he
or she gets $3334 million of the issue, but the marketclearing yield drops to 19.99 percent.10 As a result, expected profit drops to $(3334.0) 3 (1/1.05 – 1/1.1999) '
$396.67 million, which makes the bidder worse off than
being part of the cartel. Similarly, cornering the whole
issue or submitting any quantity greater than $3334 million will cause the yield to fall at or below 6 percent, both
of which are less profitable than sticking to the cartel.
Now, consider a slightly different bidding arrangement in which each of the bidders changes the bid at
$3333 million to 15 percent and everything else remains
the same. This particular bidding arrangement will
encourage a bidder to deviate and corner the whole issue
by submitting just one bid for $10 billion at 14.99 percent.
On these terms the expected profits from deviating,
$(10000.0) 3 (1/1.05 – 1/1.1499) million ' $827.40 million, exceed the profits from sticking to the cartel. What
exactly is the difference between the two bidding scenarios? Chart 1 shows that the demand schedules
in the former arrangement are steeper than those in the
latter. The steepness of demand schedules increases the
cost of deviating from a cartel and sustains the collusion.
However, a collusive arrangement such as the one
discussed above is difficult to sustain in a discriminatory auction. Because each bidder pays her bid yield for
the quantity awarded out of her demand schedule, submitting steep demand schedules could turn out to be
costly as a bidder ends up paying for the low-yield (highprice)–low-quantity points in her demand schedule.

6. See McAfee and McMillan (1987, 721) for some references.
7. In general, the revenue comparison between the first-price and second-price auctions could depend on the bidders’ attitudes toward risk. See Milgrom and Weber (1982) for more details.
8. As of now, the experiment has been extended indefinitely.
9. Risk neutrality implies that, facing an uncertain yield in the after market, bidders care only about the expected/average
level of yield. A risk-averse bidder would also care about the dispersion of the possible yields in the after market.
10. Each of the three bidders gets $3333 million (of the $10 billion issue) as each of them submits a bid at the low yield of 6 percent. However, the deviating bidder also gets an additional $1 million because the high bid of 19.99 percent is lower than
those of the other two bidders at 20 percent, and the auction clears at 19.99 percent.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

7

B O X

2

Bid Allocation in Treasury Auctions
Consider an auction in which $50 million worth of a
Treasury security will be auctioned and there are two competitive bidders. Each bidder submits a demand schedule,
that is, the dollar amount of the security that he or she is
willing to buy at each particular yield as shown in Chart A.
If the amount of noncompetitive bids is assumed to be
$10 million, the quantity available to the two competitive
bidders is $40 million. Starting at the lowest yield bid, the
Treasury computes the total quantity demanded at each
yield by adding the individual quantity demanded at each

Chart A
Demand Schedules
Bidder 1

Yield

5.04

5.02

5.00
0

5

10

15

20

25

Quantity (million $)

yield by the two bidders and arrives at an aggregate
demand schedule, as in the third panel of Chart A; each
bidder gets the entire quantity on his or her demand
schedule at the accompanying yield, provided the total
quantity demanded at the particular yield is less than the
available supply. This process is repeated for an increasing
sequence of yields as the Treasury works its way up the
demand schedule of each bidder until the available supply
is exhausted; the yield at which this happens is called the
market-clearing or stop-out yield. At the market-clearing
yield, if the total quantity demanded exceeds the available
supply, each bidder is awarded a quantity prorated on the
basis of quantities bid at that yield. For example, in the
third panel of Chart A, the market-clearing yield is 5.03
percent, and each of the bidders is awarded $10 million.
The total quantity demanded at 5.03 percent is $30 million,
but the total quantity available is $20 million; since each of
the bidders demands $15 million, $20 million is equally
divided between them.
In a discriminatory auction each competitive bidder
pays for a requisite amount of securities at the accompanying yield, while in a uniform-price auction all competitive (and noncompetitive) bidders are awarded the entire
quantity at the stop-out yield.1 Also, in a discriminatory
auction noncompetitive bidders are awarded the securities
at the quantity weighted-average auction yield of the
accepted bids.

Bidder 2

Aggregate

5.04

Yield

Yield

5.04

5.02

5.02

Market-clearing yield

5.00

5.00
0

5

10

15

20

25

Quantity (million $)

0

10

20

30

40

50

Quantity (million $)

1. For example, in a discriminatory auction, bidder 1 gets $5 million at 5.01 percent, $10 million at 5.02 percent, and $10 million at
5.03 percent. In a uniform-price auction, both bidders get the entire amount of securities at 5.03 percent although they did bid lower
yields.
8

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

CHART 1

Collusive and Noncollusive
Demand Schedules

Noncollusive

20

Yield

Indeed, Back and Zender (1993) show through a
sophisticated model that a uniform-price auction could
result in a loss of revenue to the Treasury as compared
with the discriminatory auction if the bidders submit
steep enough demand schedules in the uniform-price
auction. The steepness of the demand schedules binds
the bidders to a self-enforcing collusive arrangement; if
anyone deviates from the cartel, then he or she earns
lower expected profits. The interesting aspects of the
Back-Zender model are that collusion takes place
despite the bidders not being able to observe the bids of
other bidders in a sealed-bid auction and the model
considers only one auction at a time.11 On this question
of the effects of the steepness of demand schedules,
Feldman and Reinhart (1995) document that the
demand schedules in uniform-price gold auctions (conducted by International Monetary Fund) are steeper
than those in discriminatory auctions

10

Collusive

0
3000

4000

5000

6000

7000

Quantity (million $)

Noncompetitive Bids
f there are noncompetitive bidders, competitive bidders do not know at the time they submit their bids
the net amount of the Treasury security that will be
available to them. The net amount available to competitive bidders is therefore a random quantity. The random
supply represents a source of risk to the bidders in formulating and implicitly coordinating their bidding
strategies to maintain collusion. In particular, some of
the low-yield (high-price)–low-quantity points in steep
demand schedules (if they wish to submit such schedules) that otherwise do not matter in terms of the clearing price in uniform-price auctions may actually clear
the auction, costing the colluding bidders dearly. In the
example discussed above, suppose the total noncompetitive bid is for $50 million. The net amount available to
the three competitive bidders would be $950 million,
and, given the bids, the auction would clear at 6 percent
instead of at 20 percent as in the collusive outcome.
However, the Back-Zender model assumes that bidders are risk neutral; by definition, risk-neutral bidders
do not care about the risk that noncompetitive bids may
cause the low-yield points in their steep demand schedules to be realized as auction-clearing yields. Because
they care only about their expected gain, competitive
bidders are willing to submit steep demand schedules.
Consequently, a uniform-price auction could be a worse
choice for the Treasury than a discriminatory auction
even with unpredictable noncompetitive demand.12
Although most of the primary dealers are large financial

I

institutions, many of them are not the very large, welldiversified corporations that seem to fit the risk-neutral
description. Also, many of the competitive dealers are
known to hedge interest rate risk, an activity not compatible with risk neutrality. It is therefore important to
look at models in which bidders are risk averse.

What If Bidders Are Risk Averse?
ang and Zender (1996) relax the assumption of
risk neutrality of bidders; otherwise they retain
the same assumptions as those of Back and
Zender. The primary finding relevant for this article is
that if the number of competitive bidders and the average level of the random competitive demand are sufficiently high, a uniform-price auction could yield higher
revenues to the Treasury than a discriminatory auction
despite the ability of the bidders to submit steep
demand schedules. This result runs contrary to the general argument of Back and Zender and underscores the
importance of risk neutrality in that model.
In the example used to illustrate the intuition of
the Back-Zender model, it is clear that the profit per
bidder in the collusive arrangement decreases with the
addition of bidders because the issue is equally divided
among the bidders. Therefore, an increase in the number of bidders would make such a collusive outcome
more difficult to sustain in a uniform-price auction and
may increase its desirability for the Treasury. The higher the amount of noncompetitive bids is, the lower the

W

11. In the parlance of game theory, the model is one of a single-period game. Normally, collusion among agents is easier to
sustain in multiple-period games because the deviating agent in any period can be punished in a subsequent period.
12. Back and Zender (1993) do not analyze all possible game-theoretic equilibria (outcomes given rational actions of agents)
and instead analyze only a tractable set of such equilibria. It remains possible that with risk-neutral bidders a uniformprice auction could actually be a better choice for the Treasury.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

9

available supply is for competitive bidders and the higher the chances are that the auction will clear at the lowyield (high-price) points of the bidders’ steep demand
schedules. Therefore, if the amount of expected noncompetitive bids is high, risk-averse bidders may not be
willing to take the risk of submitting steep demand
schedules because such schedules increase the chances
of unfavorable outcomes (to bidders). As a result the
collusive outcome may not be realized and the Treasury
would benefit.

Pre- and Postauction Markets
he preauction when-issued market and the
postauction secondary market are integral parts
of the entire auction process and may affect the
analysis of possible collusion under uniform-price and
discriminatory auctions. If bidders are committed to sell
the to-be-auctioned securities in the when-issued market and fail to obtain a sufficient amount of them at the
auction, they will have no alternative other than to buy
these securities in the postauction market either
through a repo or directly in the secondary market.
However, often newly auctioned on-the-run securities
trade “on special” in the repo market—that is, one has
to lend funds at below-the-market rate to get these
securities. They are also more expensive to buy in the
cash secondary market than seasoned securities of comparable maturities. Thus, in formulating bidding strategies for an auction, bidders have to take into account
their positions in the when-issued market and the possibility of buying the securities at a premium in the
postauction market. A more complete model of Treasury
auctions, therefore, would take into account the whenissued market, the auction itself, and the possibility of
trading in the secondary market. Two recent studies by
Wang and Vishwanathan (1996) and Chatterjea and
Jarrow (1995) take into account the preauction and
postauction markets in their models of Treasury auctions and are discussed below.
Wang and Vishwanathan (1996) model the entire
auction process as consisting of three distinct markets:
preauction when-issued trading, the auction itself, and
postauction when-issued trading. They assume that
competitive bidders can submit demand schedules at
the auction but are averse to holding a large number of
positions (long or short) at the end of the auction cycle.
The assumption of aversion to large positions is not
unrealistic because bidders do not want to be caught in
a squeeze if they have substantial short positions and
financing unsold long positions is costly. Consequently,
bidders optimally restrain the steepness of their demand
curves to avoid ending up with unwanted excess inventory (long or short). As in Wang and Zender (1996),
Wang and Vishwanathan find that because of the diminished ability of bidders to submit steep demand curves,

T

10

uniform-price auctions can generate more revenue for
the Treasury than discriminatory auctions when the
number of competitive bidders and the mean level of
noncompetitive demand are high. An important aspect
of the model is that it is able to address the issue of the
temporal pattern of price volatility in the postauction
when-issued market. The authors find that auction surprises elicit a higher response in discriminatory auctions
than in uniform-price auctions. This result is consistent
with empirical evidence of Belzer and Reinhart (1996),
in which the surprise is measured by the difference
between the average auction yield and the contemporary
when-issued yield.
Chatterjea and Jarrow (1995) consider the
preauction when-issued market and the postauction
secondary market but ignore noncompetitive bids and
do not allow for bidders (assumed to be risk neutral) to
submit demand schedules; instead the bidders are
allowed to submit bids only for the entire quantity to be
auctioned. Although the bidders could end up getting
less than the entire unit if there is a tie with other bidders, essentially the model is that of a single-unit auction. In a common-value single-unit auction with
risk-neutral bidders, it is well known that a second-price
auction is superior to a first-price auction because the
extent of bid shading due to the winner’s curse in the
former is less. In keeping with this result, the authors
find that uniform-price auctions (that are similar to
second-price auctions) yield higher revenues to the
Treasury than discriminatory auctions (that are similar
to first-price auctions).
The theoretical papers that have been discussed
so far take into account the strategic advantage that
comes with submitting demand schedules or the institutional setup of the Treasury market. However, they
have focused mainly on the revenue superiority of one
auction format over another. They do not directly
address the important issue of whether either of the
auction formats (uniform-price or discriminatory) is
more vulnerable to short squeezes that are often known
to develop in the repo or the secondary market following an auction and, in fact, prompted the Treasury to
consider alternative formats. Nor do they recognize the
fact that bidders do communicate before the auction.
These two issues are addressed next.

Communication among Bidders
t seems possible that competitive dealers indulge
in mutual communication before submitting bids
for an ensuing auction.13 The theoretical models
developed to date do not take into account such
preauction communication, as otherwise the models
are intractable. A feasible way to tackle the issue of
bidder communication is to perform a controlled
experiment that tries to replicate the actual auction

I

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

market. Experiments by Goswami, Noe, and Rebello
(1996) suggest that, in the absence of communication,
the uniform-price auction results in higher revenues
for the auctioneer.14 However, if bidders are allowed to
discuss their future strategies, the outcomes of previous auctions, and so forth, collusive behavior emerges
and the uniform-price auction generates a lower revenue than the discriminatory auction. If there is communication among bidders, one might conclude that
discriminatory auctions could be a better choice for the
Treasury in terms of revenue enhancement. Other
empirical research (discussed below), however, does not
find any significant differences in revenues between the
two auction formats.

Short Squeezes
uite often, repo rates on specific on-the-run
issues are lower than overnight lending rates collateralized by similar securities. The difference
between that overnight lending rate and the repo rate
(collateralized by the specific security) measures the
degree of “specialness” of that specific security; owners
of securities on special can obtain overnight loans at a
lower rate than those of other comparable securities.
The occurrence of specials in the repo market is a common phenomenon and could be an outcome of the auction process itself. Dealers who have short positions in
the when-issued market and fail to obtain the desired
amount of securities at the auction have to acquire the
securities either from the secondary cash market or
the repo market. Sufficiently high demand for a specific security can increase the price and decrease the
repo rate on the specific security. Additionally, deliberate acts of cornering an auction by an individual or a
cartel could result in short squeezes and specials, as
was the case with Salomon Brothers in the May 1991
two-year note auction.
Repeated short squeezes are potential threats to
the integrity and liquidity of the Treasury market and
eventually could drive up the Treasury’s borrowing costs.
Can the Treasury undertake credible measures to prevent
and alleviate short squeezes? Are alternative auction formats more or less susceptible to short squeezes?
One possible way to alleviate short squeezes is to
reopen the squeezed security through an auction to provide additional supply. However, it is often difficult to differentiate between specials developing from deliberate
manipulation and those developing due to the dealers’
misjudging demands in the when-issued market or other

Q

phenomenon, such as the dealers’ selling the off-the-run
security and rolling into the when-issued market (going
long) for the soon-to-be on-the-run security.15 An additional problem with reopening is that, with a commitment to reopen, the future supply of Treasury securities
is essentially an uncertain quantity and poses a source of
risk to bidders in formulating their bidding strategies. If
bidders are risk averse, a risk premium can appear,
resulting in higher
average auction yields
and higher borrowing
The preauction whencosts for the Treasury.
issued market and the
Other specific measures
that have been suggestpostauction secondary
ed to increase the supmarket are integral parts
ply of squeezed secuof the entire auction
rities include selling
these securities directly
process and may affect the
through the New York
analysis of possible colluFed’s open market desk
sion under uniform and
or facilitating “synthetic
reopenings” by lifting
discriminatory auctions.
certain restrictions
on reconstituting the
coupon-bearing security
through Treasury Separate Trading of Registered Interest
and Principal (STRIPs) (whenever applicable).
Chari and Weber (1992) argue that bidders’ incentive to substantially affect the price they pay by submitting low-enough yield bids (high-enough prices) is less
in uniform-price auctions because the yield that bidders
are awarded is the stop-out yield, which could often be
higher than the yield that they had bid for. Chari and
Weber overlook the fact that bidders in uniform-price
auctions can increase the quantity awarded to them by
tendering low-enough yields. Provided there are other
bidders (who may be part of a cartel) whose bids would
make the auction clear at higher yields, the aggressive
bidders end up cornering a substantial fraction of the
uniform-price issue, acquired at lower prices than their
bid prices. However, bidders who attempt to corner a discriminatory auction by bidding low-enough yields will
have to pay high-enough prices for such bids. Thus, from
this perspective the incentive to submit low-enough
yields and corner the market may actually be higher in
uniform-price auctions than in discriminatory auctions.
In contrast, Nyborg and Sundaresan (1996) argue
that short squeezes are more likely to develop under discriminatory auctions if some traders are better informed

13. Anecdotal evidence suggests that they do, but there is no formal documentation.
14. The experiments, however, do not take into account the when-issued and secondary markets.
15. In one case a short squeeze had developed in the thirty-year bond (maturing February 2016) surrounding the auction of
the on-the-run thirty-year bond in May 1986; dealers had sold short the seasoned bond to take positions in the May 2016
bond, and suddenly there was a dearth of the seasoned thirty-year bond.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

11

about the value of the to-be-auctioned security than others. Theoretical models of securities markets that take
into account such asymmetry of information predict that
periods of higher information dissemination are also
periods of higher volatility. Nyborg and Sundaresan
(1996) find that when-issued yields in a discriminatory
auction display increasing volatility through time on the
day of the auction while under a uniform-price auction the when-issued
yields display decreasing volatility. As a result, the authors conclude that to the extent
Repeated short squeezes
the existence of higher
are potential threats to
preauction information
the integrity and liquidity
helps bidders to structure their bids better
of the Treasury market and
bidders face a lower
eventually could drive
probability of being
up the Treasury’s borrowsqueezed by others who
submit unanticipated
ing costs.
low yields.16 However,
the documented differences in temporal patterns of volatility
across the two auction formats are also consistent with
the predictions of Wang and Vishwanathan’s (1996)
model, in which all bidders are equally informed about
the value of the Treasury security and the differences in
volatility patterns are due to differences in hedging
behavior across auction formats; this result calls into
question Nyborg and Sundaresan’s interpretation.

Empirical Evidence
heoretical models, while insightful, cannot
model the auction market in its full complexity.
Assumptions often have to be made to keep a
model tractable. Therefore, theoretical predictions
regarding revenue superiority and susceptibility to
short squeezes of alternative auction formats are subject to question unless confirmed by empirical
research. Empirical research on Treasury auctions has
looked at evidence regarding the existence of the winner’s curse, has compared alternative auction formats
in terms of their potential savings to the Treasury, and
has explored the possible existence of collusion in
auctions.
In terms of the winner’s curse, Cammack (1991)
and Spindt and Stolz (1992) find that it is cheaper to
buy three-month Treasury bills in auctions than in the
postauction and preauction secondary markets, respectively. The difference in yields is on the order of 1 1⁄2 to
3 and 4 basis points. Assuming that the secondary market reflects the true value of the security, the authors
conclude that bidders do shade their bids in auctions,

T

12

a consequence of the winner’s curse. Although informative, the comparison of the auction yields and the
secondary market yields are not direct comparisons
because the secondary market securities are quoted for
a different delivery day than the auctioned securities. A
better approach is to compare the auction yields with
the when-issued yields that are for delivery on the same
business day. Using proprietary when-issued data from
competitive dealers, Simon (1994) and Nyborg and
Sunderasan (1996) find that the markup of the average
(quantity-weighted) auction yield over the contemporaneous when-issued bid yield (bid-side yield represents the rate that one can lock in to sell) for bills and
notes tends to be less than a basis point. Given the possibility of errors in measuring these yields, however, it
is not clear that there is any significant economic difference between the two yields and therefore any convincing evidence of winner’s curse. Also, for Treasury
bills the markup comparisons are compounded by the
fact that the required minimum difference between any
two yields, often called the tick size of the security, is
different in auctions and the when-issued market, as
noted by Cohen and McBeth (1994).
Nyborg and Sundaresan (1996) compare the
markups in uniform-price and discriminatory formats
to investigate whether the switch to the uniform-price
format in two- and five-year notes has resulted in higher revenues for the Treasury. They find that the differences in the markups of the average auction yield over
contemporaneous when-issued yields (a measure of
the Treasury’s possible savings) between the two formats depend on the time of the day the when-issued
yield is quoted and the maturity of the note. In short,
no definite conclusion can be reached regarding the
revenue superiority of the uniform-price auction over
the discriminatory auction by comparing these
markups.17 However, the data set used is relatively
small, and, furthermore, the comparison between the
two formats is not entirely controlled because the twoand five-year uniform-price auctions were held at different times and hence in a different interest rate
environment than the discriminatory auctions. Given
these conditions, small differences in markups would
be difficult to identify.
In terms of foreign auction markets, Tenorio (1993)
finds that average revenues to the Zambian Treasury
decreased after a switch from the uniform-price to the
discriminatory format due to lower bidder participation,
a result consistent with the assertions of Friedman
(1960). Similarly, Umlauf (1993), in a study of bidding in
Mexican Treasury auctions, finds that bidder profits, as
measured by the difference between quantity-weighted
average auction yield and yield in the immediate
postauction secondary market, dropped substantially
after the Mexican Treasury switched from the discrimi-

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

natory to the uniform-price format. However, collusion
among bidders in the Mexican Treasury auctions is often
thought to be a distinct possibility (Back and Zender
1993; Wang and Vishwanathan 1996), and caution is warranted in extrapolating this conclusion to U.S. Treasury
auctions.
From a somewhat different perspective, Gordy
(1996) examines discriminatory Treasury auctions in
Portugal and finds that the use of multiple bids per bidder and the dispersion in the bids of each bidder
increase with the volatility of the interest rates. Since
the possibility of the winner’s curse increases with the
uncertainty (volatility) of the value of the underlying
object, this evidence can be interpreted as suggesting
that the use of multiple bids in Treasury auctions acts
as a natural hedge against the winner’s curse. The
increased dispersion of bids in Swedish discriminatory
auctions as well has been found by Nyborg, Rydqvist,
and Sundaresan (1997).

Alternative Auction Formats
urrently a few approaches are being contemplated to change the format of auctions. One of these
is whether the Treasury should switch to an
ascending-price open-outcry auction. Another is
whether the Federal Reserve should preannounce its
noncompetitive bids. It is insightful to examine these
alternatives in light of the research that has been discussed.
Ascending-Price Open-Outcry Auctions. Extant
empirical evidence indicates that there is no significant
difference in the Treasury’s financing cost from selling
Treasury securities under either the uniform-price or
discriminatory format. One common feature of these two
formats is that they are sealed-bid auctions. As the auction procedure becomes more automated, it may be possible to hold electronic open-outcry auctions. In an
electronic open-outcry Treasury auction, bidders located
in diverse geographical regions of the country would
have access to a central computer at the Treasury and
would enter bids into their terminals. In fact, the Joint
Report (U.S. Treasury 1992) suggests that the Treasury
consider experimenting with an ascending-price/
descending-yield electronic open-outcry auction.
A descending-yield auction would start with the
Treasury announcing a yield, perhaps the contemporary
yield in the when-issued market or marginally higher, for
the opening round with bidders submitting bids at the
particular yield. After receiving the bids, the total

C

amount of bids would be announced publicly. The expectation is that the high yield available at the opening
round would lead to oversubscription. Thereafter, the
auctioneer would decrease the yield at each round gradually until the volume bid is less than the available supply. The bidders who remain until the last round would
get the securities at the next-highest yield (that is, the
yield of the previous round) while the bidders who
dropped out at the next-to-last round would be awarded
prorated quantities at that round’s yield. In this way, the
auction would get cleared at a single yield.
The open format seems to offer several positive features. It allows for the release of more information
through the bidding process as the bidders learn the
total volume at each yield and possibly the bids of other
bidders. The release of more information is favorable
from a winner’s curse point of view to the extent that
access to the additional information attenuates the winner’s curse. The availability of more information could
also increase bidder participation, thereby making the
auction more competitive. In addition, the open-outcry
format could be favorable from the perspective of a short
squeeze if the auction is structured such that the individual bids at the time they are entered are public knowledge. Bidders would be able to gauge the extent of
demand at low yields and revise their bids. For example,
those short in the when-issued market would have the
option of matching any abnormally high quantities at
low-yield bids (as and when they show up) and could
avoid being cornered, an option not feasible under the
sealed-bid format.
However, there are a few issues about ascendingprice Treasury auctions that deserve further study by the
Treasury. In particular, the ascending-price/descendingyield format could encourage a type of collusion as follows. If every bidder is part of a cartel and the cartel bids
conservatively enough that the net demand is less than
the available supply at the first or second round, then
the auction would get cleared at the high yields of the
initial rounds without providing the Treasury the opportunity to test demand at the lower yields. Is this type of
collusion sustainable? If a particular bidder defects from
this cartel and bids a much higher quantity at the initial
round, then the rest of the cartel members would bid
higher quantities at the successive rounds and drive the
auction yield successively lower (and prices higher)
such that it eventually becomes unprofitable for the
defector to get any quantity at the very low marketclearing yield. Realizing ex ante the cartel’s response to

16. Some supporting evidence is documented in Bikhchandani and Huang (1992), who find that in a majority of the bill auctions in their data set at least one yield in the auction was lower than the corresponding when-issued ask yield.
17. Wang and Vishwanathan (1996) question the validity of comparing markups across different auction formats because
their model predicts that prices in uniform-price auctions are much more variable than in discriminatory auctions, thus
introducing more noise in the uniform-price markups.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

13

defection, no cartel member would possibly defect, and
the auction could get cleared at the very initial rounds.
However, the Treasury may be able to deter this type of
collusion either by specifying a minimum number of
rounds or the minimum quantity that each bidder has to
tender at each round or at least during the first few
rounds. Another type of manipulation could take place
in the when-issued market itself if the Treasury were to
precommit to starting
the auction with a yield
that is always either
somewhat higher or
lower than the contemEmpirically . . . there
porary when-issued
yield and perfectly preseems to be no discernible
dictable by the bidders.
difference between discrimIn such a case, it is posinatory and uniform-price
sible that the whenissued yield could be
auctions in terms of
collectively manipulatrevenue to the Treasury.
ed to be much higher
than it would be otherwise and the auction
would clear at a much
higher average yield
even with multiple rounds, costing the Treasury revenue. To avoid this type of potential manipulation, the
Treasury could avoid precommitting to any opening
yield. Instead the Treasury could choose the opening
yield with the addition of a random component—for
example, by making it a little higher or little lower than
the when-issued yield in a way that is not predictable.
Even with these potential vulnerabilities, the openoutcry auction descending-yield/ascending-price format
remains a promising alternative. It allows for greater
information dissemination and, with a few extra features added to the contemplated design, perhaps could
deter collusion.
Preannouncement of Noncompetitive Bids.
Bikhchandani and Huang (1993) suggest that the Federal
Reserve should disclose the amount of securities it will
tender as noncompetitive bids on behalf of foreign central banks. The Federal Reserve’s tender normally constitutes a nontrivial part of the pool of noncompetitive bids.
Imperfectly predictable noncompetitive bids are a source
of uncertainty to competitive bidders, and decreasing
such uncertainty through the Federal Reserve’s disclosure might tend to diminish winner’s curse and increase
auction revenue. One potential disadvantage of such
disclosures in uniform-price auctions is that hiding noncompetitive bids could be a deterrent to the type of selfenforcing collusion that arises through the submission

14

of steep demand curves in such auctions. It is possible,
however, that the disclosures by the Federal Reserve
about the intent of its noncompetitive bids could be
beneficial to the Treasury in discriminatory auctions.

Conclusion
he U.S. Treasury is currently experimenting with a
uniform-price format for auctioning two- and fiveyear Treasury notes. All other Treasury securities
are still auctioned through the discriminatory format.
This experiment was begun with the notion that the
attenuation of the winner’s curse in uniform-price auctions would lead to increased auction revenues. However,
current theoretical research shows that the ability to submit demand schedules in Treasury auctions conveys a
strategic advantage to bidders under the uniform-price
format. As a result the reduction in winner’s curse in
uniform-price auctions could be outweighed by the bidders’ submitting steep demand schedules that beget a
self-enforcing collusion and cause the Treasury’s financing cost to actually increase in such auctions. The strategic advantage of demand schedules declines as the level
of noncompetitive demand and the number of competitive bidders rise. The existence of preauction and
postauction trading in Treasury securities dilutes the
strategic advantage that bidders have in uniform-price
auctions.
Empirically, however, there seems to be no discernible difference between discriminatory and uniformprice auctions in terms of revenue to the Treasury. While
this result may indicate that the theoretical models are
too stylized, it may also be the case that it is a little too
early to draw any robust conclusion; the data sets used
in the empirical tests do not span a sufficiently long time
period, and the comparison between the two auction formats is not entirely controlled to account for the different time periods and hence the different interest rate
environments in which these auctions were held.
The proposal to switch to electronic ascendingprice open-outcry auctions with an implied uniform
price may be more important than just switching to a
uniform-price auction. Although collusive behavior may
emerge in ascending-price auctions and the whenissued market may be manipulated to have an auction
clear at a high yield, this article has pointed to some
safeguards that the Treasury could adopt to preempt
such collusion. These take the form of imposing a lower
bound on the amount that competitive bidders need to
bid, at least during the early rounds of these auctions,
and the Treasury’s not precommitting to any opening
yield that is predictably related to the contemporary
when-issued yield.

T

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

REFERENCES
BACK, KERRY, AND JAIME ZENDER. 1993. “Auctions of Divisible
Goods: On the Rationale for the Treasury Experiment.”
Review of Financial Studies 6:733–64.

MILGROM, PAUL R., AND ROBERT J. WEBER. 1982. “A Theory of
Auctions and Competitive Bidding.” Econometrica
50:1089–1121.

BELZER, GREGORY, AND VINCENT REINHART. 1996. “Some
Evidence on Bid Shading and the Use of Information in the
U.S. Treasury’s Auction Experiment.” Carnegie Mellon
University, Graduate School of Industrial Administration,
Working Paper.

NYBORG, KJELL G., KRISTIAN RYDQVIST, AND SURESH SUNDARESAN.
1997. “Bidder Behavior in Multiple Unit Auctions: Evidence
from Swedish Treasury Auctions.” Columbia University
Working Paper.

BIKHCHANDANI, SUSHIL, AND CHI-FU HUANG. 1989. “Auctions with
Resale Markets: A Model of Treasury Bill Auctions.” Review
of Financial Studies 2:311–40.

NYBORG, KJELL G., AND SURESH SUNDARESAN. 1996. “Discriminatory versus Uniform Treasury Auctions: Evidence from WhenIssued Transactions.” Journal of Financial Economics
42:63–104.

———. 1992. “The Treasury Bill Auction and the WhenIssued Market: Some Evidence.” Massachusetts Institute of
Technology Working Paper.

SIMON, DAVID. 1994. “Markups, Quantity Risk, and Bidding
Strategies at Treasury Coupon Auctions.” Journal of
Financial Economics 35:43–62.

———. 1993. “The Economics of Treasury Securities
Markets.” Journal of Economic Perspectives 7, no. 3:117–34.

SMITH, CLIFFORD. 1992. “Economics and Ethics: The Case of
Salomon Brothers.” Journal of Applied Corporate Finance
5:23–8.

CAMMACK, ELIZABETH. 1991. “Evidence of Bidding Strategies
and the Information in Treasury Bill Auctions.” Journal of
Political Economy 99:100–30.

SPINDT, PAUL, AND RICHARD STOLZ. 1992. “Are U.S. Treasury Bills
Underpriced in the Primary Market?” Journal of Banking
and Finance 16:891–908.

CHARI, V.V., AND ROBERT WEBER. 1992. “How the U.S. Treasury
Should Auction Its Debt.” Federal Reserve Bank of Minneapolis Quarterly Review (Fall): 3–12.

STIGUM, MARCIA. 1990. The Money Market. Homewood, Ill.:
Dow-Jones Irwin.

CHATTERJEA, ARKADEV, AND ROBERT A. JARROW. 1995. “Market
Manipulation and a Model of the United States Treasury
Securities Auction Market.” University of Colorado Working
Paper.

SYRON, RICHARD, AND SHEILA TSCHINKEL. 1987. “The Government
Securities Market: Playing Field for Repos.” In Current
Readings on Money, Banking, and Financial Markets, edited by James A. Wilcox. Boston: Little, Brown & Co.

COHEN, HUGH, AND DOUGLAS MCBETH. 1994. “The Effect of Tick
Size on Treasury Auctions.” Federal Reserve Bank of Atlanta
Working Paper 94-9, September.

TENORIO, RAFAEL. 1993. “Revenue Equivalence and Bidding
Behavior in a Multi-unit Auction Market: An Empirical
Analysis.” Review of Economics and Statistics 75:302–14.

FELDMAN, ROBERT A., AND VINCENT R. REINHART. 1995. “Flexible
Estimation of Demand Schedules under Different Auction
Formats.” International Monetary Fund Working Paper
No. 95-116.

UMLAUF, STEPHEN. 1993. “An Empirical Study of the Mexican
Treasury Bill Auction.” Journal of Financial Economics
33:313–40.

FRIEDMAN, MILTON. 1960. A Program for Monetary Stability.
New York: Fordham University Press.
GORDY, MICHAEL. 1996. “Hedging Winner’s Curse with Multiple
Bids: Evidence from the Portuguese Treasury Bill Auction.”
Banco De Portugal Working Paper No. 21-96.

U.S. TREASURY DEPARTMENT, SECURITIES AND EXCHANGE COMMISSION, AND BOARD OF GOVERNORS OF THE FEDERAL RESERVE SYSTEM.
1992. Joint Report on the Government Securities Markets.
Washington, D.C: U.S. Government Printing Office.
WANG, JAMES J.D., AND S. VISHWANATHAN. 1996. “Auctions with
When-Issued Trading: A Model of the U.S. Treasury Markets.”
Duke University, Fuqua School of Business, Working Paper.

GOSWAMI, GAUTAM, THOMAS H. NOE, AND MICHAEL J. REBELLO.
1996. “Collusion in Uniform-Price Auctions: Experimental
Evidence and Implications for Treasury Auctions.” Review of
Financial Studies 9:757–85.

WANG, JAMES J.D., AND JAIME ZENDER. 1996.“Auctioning Divisible Goods.” Duke University, Fuqua School of Business,
Working Paper.

MCAFEE, R. PRESTON, AND JOHN MCMILLAN. 1987. “Auctions and
Bidding.” Journal of Economic Literature 30:699–738.

WILSON, ROBERT. 1979. “Auctions of Shares.” Quarterly
Journal of Economics 93:675–98.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

15

Movements in the
Term Structure of
Interest Rates
R O B E R T R . B L I S S
The author is a senior economist in the financial
section of the Atlanta Fed’s research department. He
thanks Peter Abken, Jerry Dwyer, Ehud Ronn, Mary
Rosenbaum, Steve Smith, and Tao Zha for helpful
comments and Janet Colter for editorial assistance.

B

OND PRICES TEND TO MOVE TOGETHER.

STOCKS TEND TO GO THEIR OWN WAY. THIS DISTINCTION

HAS IMPORTANT IMPLICATIONS FOR MANAGING THE RISKS ASSOCIATED WITH HOLDING THESE
SECURITIES.1

Because so much of the movement in stock prices
is idiosyncratic, or security-specific, it is impossible to
use one stock, or even a portfolio of stocks, to hedge the
price movements in any other stock. For this reason,
diversification, which helps minimize idiosyncratic risk,
plays an important role in modern equity portfolio management. Only the relatively small proportion of movement in stock prices that is systematic, or common,
across all stocks can be hedged using contracts whose
payoffs are tied to the value of a stock market index,
such as the S&P 500 futures contract traded on the
Chicago Mercantile Exchange. 2
In contrast, because so much of the movement in
bond prices is systematic, bond portfolio management
focuses on techniques for eliminating the common factor of interest rate risk by balancing short and long
exposures to fluctuations in interest rates. Doing so
does not require that portfolios be well diversified. 3 It
is possible to use a few bonds of differing maturities to
hedge the price fluctuations in any single bond or portfolio of bonds. Alternatively, interest rate derivatives,
such as the ten-year Treasury note futures contract
traded on the Chicago Board of Trade, whose payoff is
tied to the value of the ten-year Treasury note at the

16

expiration of the contract, may be used to hedge interest rate risk. Using derivatives rather than other bonds
for hedging is generally more cost efficient. The
futures markets are usually more liquid than the
underlying bond markets, and short positions may be
easily taken without the complications and costs of
shorting securities. The enormous volume of trading in
interest rate futures attests to the widespread use of
such techniques.
Hedging to reduce or eliminate the common factors influencing the value of an interest rate–sensitive
portfolio requires a model of how interest rates behave.
This model may be a formal equilibrium- or arbitragebased model, or it may be an ad hoc statistical model.4
The most widely used method for hedging bond portfolios is duration immunization, which matches the
Macaulay duration of assets and liabilities. Macaulay
duration is predicated on the assumption that interest
rates for all maturities move up and down in parallel.
Clearly, they do not do so. Nonetheless, Macaulay duration hedging is still widely used. Numerous studies have
developed more advanced hedging models aimed at
capturing changes in the shape of the term structure as
well as changes in the overall level of interest rates.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

This article first reviews these earlier studies,
including those showing that term structure movements can be decomposed into three components,
called factors. The empirical analysis then shows that
the nature of this decomposition has been consistent
since 1970 and that the structure of the factors has not
changed appreciably even though interest rate volatility has. The investigation then turns to the time series
behavior of the factors. Following this analysis, the
implications for hedging and interest rate modeling are
discussed. A numerical example demonstrates that hedging based on factor decomposition is superior to hedging
based on traditional methods.

A Review of the Literature
umerous articles compare various duration
measures, including Macaulay duration. For
example, a collection by Kaufman, Bierwag, and
Toevs (1983) presents a number of studies comparing
various duration specifications. These, in general, find
that Macaulay duration performs as well as other linear
models relating price and yield changes, although simple maturity did nearly as well in some cases. Ingersoll
(1983) developed a measure based on the single-factor
Cox-Ingersoll-Ross model and found it promising.
However, Gultekin and Rogalski, using actual bond data
rather than fitted yield changes, concluded that “the
data are not consistent with the hypothesis that price
and volatility of Treasury securities is adequately measured by simple duration” (1984, 252–53). They found
that a multifactor duration hedge, based on a factor
decomposition such as this article presents, outperformed the single-factor duration measures previously
proposed whether based on a theoretical model or on

N

ad hoc assumptions. Ilmanen’s (1992) finding that the
performance of duration as a measure of interest rate
risk varied through time helps to explain the differences in previous studies.
Formal interest rate models are based on assumptions about the number of sources of uncertainty and
their structure. It is possible to turn the process around
and first ask how many
factors underlie movements in the term structure, without specifying
the exact nature of the
Hedging to reduce or
relation between the
eliminate the common
factors and movements
factors influencing the
in bond prices beforehand. Litterman and
value of an interest rate
Scheinkman (1991)
sensitive portfolio requires
used factor analysis
a model of how interest
(discussed below) to
determine the number
rates behave.
of the factors underlying movements in interest rates and their
economic interpretation. They determined that three factors explain the
majority of movements in interest rates for various maturities. Knez, Litterman, and Scheinkman (1994) used the
same technique to examine short-term (less than one
year) interest rates across a variety of money market
instruments and found, surprisingly, that four factors are
important. This anomalous result is explained in part by
the mix of security types—Treasury bills, repurchase
agreements, commercial paper, and bankers’ acceptances—which may not all have the same risk.5

1. For instance, the draft for the recently implemented Basle Accord on the use of internal models for risk-based capital assessment proposed requiring that “[e]ach yield curve in a major currency must be modeled using at least six risk factors . . .”
(Joint Notice 1995, 38086; emphasis in the original). The evidence in this and numerous other studies is that there are only
three or four factors in some markets. Requiring models to incorporate six or more factors may well create the interesting
problem of identifying the extraneous factors to be included.
2. Dynamic hedging of individual stock price movements using stock-specific options, if they exist, is theoretically possible if
volatility is constant. However, under most pricing theories stock prices respond to changes in the systematic component of
stock-price risks while options prices respond to changes in total risk, which is the sum of systematic and idiosyncratic risk.
Thus, changes in the volatility of the idiosyncratic component of a stock’s return will not affect stock price but will affect the
value of options written on the stock. If volatility changes cannot be hedged, then the effectiveness of the stock/option hedge
will be reduced.
3. Although diversification has little role in managing interest rate risk, it is still important in managing credit or default risk
if the bond portfolio consists of risky debt.
4. Examples of equilibrium-based interest rate models include the Cox, Ingersoll, and Ross (1985), or CIR, model and the widely used Vasicek (1977) model—both usually implemented as single-factor models. The Longstaff and Schwartz (1992) model
is an example of a multifactor equilibrium model. These models all begin by modeling the process for the instantaneous
interest rate and then using equilibrium arguments to derive the implied structure and evolution of the entire term structure. Arbitrage-based models, such as the Heath, Jarrow, and Morton (1990, 1992) model, take the term structure as given
and model its evolution. These are naturally multidimensional, though single-factor restricted versions can be constructed.
5. Duffee (1996) has pointed out that Treasury bills of one month or less to maturity appear to show price movements that are
idiosyncratic, that is, unrelated to changes in other interest rates, either those of longer-maturity Treasury bills or
similar-maturity money market rates.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

17

The studies by Litterman and Scheinkman and by
Knez, Litterman, and Scheinkman both examine
changes in interest rates rather than the levels of interest rates or changes in bond prices. For hedging purposes, it is not the levels of interest rates that are
important but the changes, which in turn produce
changes in bond prices. Most bonds are coupon bonds
and are therefore portfolios of many different individual
cash flows, each responding to a different zero-coupon
bond yield change. Once the movements in zero-coupon
yields are understood, the movements in coupon bond
prices may be easily expressed as a function of these.6
Of Litterman and
Scheinkman’s three
factors, the first one
accounted for an average of 89.5 percent of
Factor analysis assumes
the observed variation
that changes, in this case
in yield changes across
maturities. This factor,
in interest rates, are driven
which they identified
by a few sources of variaas a “level change” faction that affect all interest
tor, helps to explain
why Macaulay duration
rates to varying degrees.
is so successful. While
changes in levels are
not the whole story,
they are such a large
part of what goes on in
interest rate movements that the assumption underlying
Macaulay duration (that is, parallel movements up and
down in interest rates) is a good first approximation.
Nonetheless, Litterman and Scheinkman show that
hedging based on three factors will improve hedge performance relative to Macaulay duration–based hedging
by 28 percent on average and in some cases much more.

Empirical Analysis
hat Are “Factors?” Factor analysis assumes
that changes, in this case in interest rates, are
driven by a few sources of variation that affect
all interest rates to varying degrees. These sources of
variation, in turn, summarize changes in the economy.
Economic variables that may or may not affect interest
rates include (along with innumerable others) the supply and demand for loans, announcements of unemployment and inflation, and changes in market participants’
risk aversion arising from perceived changes in the
prospects for continued economic growth. The key
assumption of factor analysis is that this multitude of
influences, which change interest rates continually, can
be compactly summarized by a few variables, called factors, that capture the changes in the underlying determinants of interest rates. That thousands of influences
can be boiled down to a few inputs into a compact, or

W

18

parsimonious, model is a very strong assumption. Part
of the process of performing a factor analysis is to examine just how reasonable that assumption is.
As this article demonstrates, the assumption is
quite reasonable for interest rate changes, but it is less
so for stock returns. The relation among the underlying
changes in the economy, economic agents’ reactions to
these changes, and the factors extracted by a factor
analysis of changes in interest rates are not necessarily
explicit. Factor analysis is a purely statistical description of the data. The factors obtained by such an analysis summarize the changes in interest rates (or stock
returns or other variables) compactly. Interpreting the
extracted factors and relating them to possible causal
economic events is a separate challenge.
Because it is impossible to predict these underlying economic fluctuations completely, and hence the
factors that summarize them, from one period to the
next, the factors may be thought of as “shocks” to the
term structure.7 By construction, over the sample period used to estimate the factor decomposition, each factor has an expected value of zero each month and a
standard deviation of one. In any given period, the factor can take on any positive or negative value. Also, by
construction, the factors are not correlated with each
other in each period. Each factor has an impact on
changes in interest rates, but the degree of that impact
may vary across the term structure. Factor analysis
describes the way each factor affects (or “loads on”)
each interest rate. The relations between interest rate
changes and the factors are called factor loadings.8
Data and Methodology. The raw data used in this
study are the prices of bills, notes, and noncallable bonds
found in the monthly Government Bond Files produced by
the Center for Research in Securities Prices (CRSP) at
the University of Chicago. The analysis began with computation of a zero-coupon, or discount rate, term structure from the prices of bills, notes, and bonds (excluding
those with embedded options) using a term structure
estimation method developed in Fama and Bliss (1987).9
Data used are for the period from January 1970
through December 1995. Prior to 1970 insufficient numbers of eligible long-maturity bonds were available for
computing usable term structures. Over time, there is
more variation in the shape of the term structure at
shorter maturities than at longer maturities. For this
reason, ten unevenly spaced maturities are used in this
study, namely, three and six months and one, two, three,
five, seven, ten, fifteen, and twenty years to maturity.
The Fama-Bliss yields each month at these horizons
were differenced to compute the month-to-month yield
changes used in this study.
Historical Performance of Factor Models. Before
analyzing the factors themselves, it is useful to examine
their ability to explain interest rate movements through

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

time and to compare this with the ability of a similar number of factors to explain stock movements. For comparison purposes the stock returns were gathered from the
CRSP Monthly Stock Returns File. The stocks selected
were required to have no missing data during the 1970–95
period. Of the stocks meeting this requirement, ten were
selected to cover a broad range of market values.
Beginning with the 1970–71 period a three-factor
model was fitted to the twenty-four months of yieldchange data. The cumulative percentage of variation in
the observed data for one, two, and then three factors for
that period was computed and plotted. The twenty-fourmonth window was then advanced one month and the
process repeated until the last window covered the period
from 1994 through 1995. The same procedure was applied
to the ten stock returns. The results appear in Chart 1.
There is a clear difference in the ability of a few factors to explain changes in interest rates versus stock
returns. A single factor, as yet unidentified, explains at
least 60 percent of the observed variation in interest rate
changes since 1970 and at least 78 percent since 1978. For
stocks, however, the maximum variation in returns
explained by a single factor is only 58 percent. Including
two additional factors raises the interest rate variation
explained to a minimum of 86 percent overall and 95 percent since 1978. Adding two factors raises the stock return
variation explained to a maximum of only 84 percent.
The ability of a few factors to explain changes in
interest rates is remarkably constant. During the period
from 1971 through 1978, explanatory power was somewhat

lower than in subsequent periods.10 Since 1978 the yieldchange variation explained by three factors has remained
between 95 and 98 percent. In the same period the stock
return–related figures varied from 63 to 80 percent.
The relatively moderate ability of a few linear factors, even though not tied to any particular theory, to
explain stock returns underscores the poor performance of specific stock return models. For example,
the capital asset pricing model (CAPM) developed in
Sharpe (1964) and Linter (1965) hypothesizes that
there is a single factor underlying stock returns and
identifies that factor as “the market,” usually measured using a broad index of stocks. The CAPM, or the
closely related market model, is able to explain
between 1 and 60 percent for individual stocks.11
Another theory of stock returns, the arbitrage pricing
theory (APT) developed by Ross (1976), does not specify ex ante the nature or number of factors underlying
these returns and so is similar to factor analysis in its
application. Studies that have applied the APT, for
example, Dhrymes, Friend, and Gultekin (1984), have
found an ambiguous number of factors underlying
stock returns, with the number of factors tending to
increase with the number of stocks being analyzed.
Because of the poor ability of stock return models to
explain individual stock performance, most studies of
stock return models are performed using portfolios of
stocks. Portfolios tend to reduce the idiosyncratic component in returns and thus increase the importance of
the common factors.

6. Bond prices are not solely a function of the term structure of interest rates. Bliss (1997) shows that, regardless of the term
structure estimation used, there is a discrepancy between actual bond prices and the prices fitted using a term structure.
However, Bliss also shows that these errors tend to be persistent, as one would expect if they resulted from nonpresent value
factors such as liquidity or tax effects. Because pricing errors persist, they have an even smaller effect on bond returns than
on bond prices. To check for the impact of bond-specific, versus term structure-specific, components in bond returns, the
actual one-month returns for all bonds used in this study were regressed against the returns predicted solely by changes in
the term structure. The resulting regressions show that changes in the term structure of zero-coupon yields explained 99.9
percent of the variation in actual returns. Thus, hedging based on changes in the term structure of zero-coupon yields is,
for all practical purposes, sufficient to hedge actual bills and coupon-bearing notes and bonds.
7. This usage is heuristic. The factor shocks are not the fundamental causes of changes in the term structure; rather, they are
sufficient statistics for fully capturing the underlying economic shocks that do cause the changes. Furthermore, statisticians
use the term shocks for strictly independent events. Subsequent analysis will show that the interest rate factors are reasonably close to being, but are not precisely, serially uncorrelated.
8. The factor decompositions produced by a factor analysis are not unique. Factors can be recombined with each other in any
manner so long as they remain uncorrelated and of unit variance. For example, one could let New Factor 1 = (Factor 1 +
Factor 2)/√2; New Factor 2 = (Factor 1 – Factor 2)/√2; New Factor 3 = Factor 3. This process is called rotating the factors.
Rotating the factors simultaneously rotates the factor loadings. The new factors are just as valid a summary of the underlying economic influences as are the original factors. However, some rotations may be more easily interpreted than the original factors produced by the factor analysis. It is common practice to first run a factor analysis and then search for
rotations that make economic interpretations clear.
9. Bliss (1997) extends the Fama-Bliss method to longer maturities. Bliss also compares various term structure estimation techniques on the basis of the ability of the estimated term structures to price out-of-estimation-sample bills, notes, and bonds.
Using this criterion, the Fama-Bliss method is superior to the alternatives tested.
10. Interestingly, this period does not coincide with the 1979–82 period, when interest rates were extraordinarily volatile.
11. This calculation is usually done by regressing stock returns on a market-proxy index. The resulting R2s are measures of
explanatory power in regressions, comparable to the “percent of variation explained” in the factor analysis results. See
Brealey and Myers (1991, Table 9-2).
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

19

In contrast to the weak stock return results, three
factors explain almost all the variation in interest rate
changes. While there is much debate about the modeling
of interest rate movements, empirical studies of these
models have no need for first reducing idiosyncratic
variation through aggregation of individual securities
into portfolios and thus are conducted using individual
bonds or maturities of zero-coupon yields.
The top panel of Chart 1 is suggestive of some variation through time in the factor model: during the mid- to
late 1970s Factor 1 made a smaller, and Factor 3 (which
shows up as the difference between the top two lines) a
larger contribution to explaining movements in interest
rates than they did in the post-1982 period. However, by
itself, this evidence is not conclusive. If a single threefactor model is fitted to the entire 1970–95 period, the
percentages of variation explained by the first, first and
second, and then all three factors are 80.6, 92.2, and 95.3
percent, respectively. The performance of the three-factor
model, when estimated over the entire sample period,
approximately equals the average performance when different factor models are estimated over shorter, twentyfour-month subperiods. This finding indicates that the
factor model’s performance is robust to constraining the
factor loadings to be constant over long periods.
Another method for analyzing differences in models
through time is to look at the estimated factors themselves. Chart 2 plots the values of the three factors estimated using a single model over the entire period. What
stands out is the higher volatility of all three factors during the 1979–82 period, which corresponds to the period
during which the Federal Reserve focused primarily on
reducing the rate of growth of monetary aggregates,
rather than targeting interest rates, in an effort to
reduce inflation. Numerous studies of interest rate
behavior, including Bliss and Smith (1997), have found
an apparent structural shift in the process governing
interest rates in this period. Chart 2, which shows a
change in the volatility of the factors driving interest
rates, is consistent with this result.
Analysis of Interest Rate Changes. The preceding analysis showed that three factors could explain

most of the variation in changes in interest rates, particularly since 1978. The next step is to determine, if
possible, what these factors are and how to interpret
them by looking at the factor loadings or the impact of
each factor on each maturity. For instance, if Factor 1
changes, what happens to the three-month interest
rate, to the one-year rate, and so forth? From these
responses it is possible, in this case, to give an economic interpretation of the factors themselves.
The sample period from 1970 through 1995 is first
divided into three subperiods suggested by the top panel
of Chart 1 and by Chart 2 and observation of the changes
in interest rate behavior in the early 1980s. The first
period is from 1970 through September 1979, the second,
from October 1979 through October 1982, and the third,
from November 1982 through 1995. Table 1 presents the
cumulative explanatory power of the factors over the
entire period and in each of the three subperiods.
For each of the three subperiods the factor loadings are shown in Chart 3. Although they vary in the
details, the factor loadings show a consistent pattern
across the different periods.
The first factor loading is very close to being constant at between 0.80 and 0.85 across all maturities. This
result is true for all three subperiods. Thus, while Chart 2
shows an increase in the volatility of Factor 1 during the
October 1979–October 1982 period relative to other periods, the responsiveness of interest rate changes to this
factor is unchanged. The result is, of course, that interest rate changes themselves became more volatile. Since
Factor 1 has roughly equal effects on all maturities, a
change in Factor 1 will produce a parallel movement in
all interest rates. For this reason Factor 1 can be interpreted as a “level” factor, producing changes in the overall level of interest rates. The loadings on Factor 1 are
also larger in magnitude than the loadings on the other
two factors. Since the variances of the factors themselves are equal by construction, this result means that
more of the changes in interest rates come from the first
factor than from the others.
The loadings on the second factor increase uniformly
from a relatively large negative value at short maturities to

TA B L E 1

Explanatory Power of the Three-Factor Model
in Various Periods
Factor 1
(percent)

Factors 1 and 2
(percent)

All three factors
(percent)

January 1970–December 1995

80.6

92.2

95.3

January 1970–September 1979

74.9

87.9

92.2

October 1979–October 1982

84.0

92.6

95.9

November 1982–December 1995

83.0

94.0

97.0

Estimated Period

20

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

C H A R T 1 Factor Analysis

Zer o-Coupon Y ield Changes

P e r c e n t a g e o f Va r i a t i o n E x p l a i n e d

100

80

All Three Factors

60
Factors 1 and 2
Factor 1

40

20

0
1972

1980

1990

1996

Note: Ten maturities of three and six months and one, two, three, five, seven, ten, fifteen, and twenty
years; twenty-four month moving window

Stock Retur ns

P e r c e n t a g e o f Va r i a t i o n E x p l a i n e d

100

All Three Factors

80

60

40

Factors 1 and 2

20
Factor 1

0
1972

1980

1990

1996

Note: Ten randomly selected stocks; twenty-four month moving window

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

21

C H A R T 2 Rotated Factor Values for Zero-Coupon Yield Changes

Factor 1

5

Level

3

1

–1

–3

–5
1970

1980

1990

1996

1990

1996

1990

1996

Factor 2

5

Slope

3

1

–1

–3

–5
1970

1980

Factor 3

5

Curvature

3

1

–1

–3

–5
1970

22

1980

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

C H A R T 3 Rotated Factor Loadings for Zero-Coupon Yield Changes

Januar y 1970–September 1979
1.0

Factor 1

0.6

Factor 2

0.2

Factor 3

–0.2

–0.6
0

5

15

10

20

October 1979–October 1982
1.0

Factor 1

0.6

Factor 3

0.2

–0.2
Factor 2

–0.6
0

5

15

10

20

November 1982–December 1995
1.0

Factor 1

0.6

Factor 2

0.2
Factor 3

–0.2

–0.6
0

5

10

15

20

M a t u r i t y ( Ye a r s )

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

23

C H A R T 4 Decomposition of Yield Changes, February 15 – March 15, 1996

Zer o-Coupon Y ields
7
15 Mar 1996

Percent

6

15 Feb 1996

5

4
0

2

4

6

8

12

10

14

16

18

20

16

18

20

Y ield Changes a
1.2

Yield Changes

1.0
Resulting from factor shocks

0.8

0.6
Actual

0.4

0.2
0
a

2

4

6

8

12

10

14

Difference reflects idiosyncratic shocks.
Factor Contributions

0.6

Factor Contributions

Yield changes resulting from Factor 1 = 0.4526

0.4
Yield changes resulting from Factor 3 = 1.0305

0.2

0
Yield changes resulting from Factor 2 = 0.2350

–0.2

–0.4
0

2

4

6

8

10

M a t u r i t y ( Ye a r s )

24

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

12

14

16

18

20

a moderate positive value at the longest maturities.
While the values of the short and long ends of these
curves are approximately the same across subperiods,
the shapes of the curve clearly vary. This pattern of
increasing loadings is consistent with interpreting
Factor 2 as a “slope” factor, affecting the slope of the
term structure but not the average level of interest
rates. Factor 2 produces movements in the long and
short ends of the term structure in opposite directions
(twisting the yield curve), with commensurate smaller
changes at intermediate maturities.
During the period from October 1979 through
October 1982, the Factor 2 loadings increase in an approximately linear fashion, centered at approximately ten
years’ maturity. A change in Factor 2 then would have produced no change in the ten-year rate (since the ten-year
loading on Factor 2 was then zero), shorter maturity rates
would have decreased, and longer maturity rates would
have risen. In both cases, the size of changes would have
increased as the maturity involved was further from ten
years. In contrast, in the period from November 1982
through December 1995 the Factor 2 loadings were centered at approximately two years’ maturity, and for maturities longer than ten years the effects of a change in
Factor 2 are approximately constant.
Factor 3 may be interpreted as a “hump” or “curvature” factor. The loadings are zero at the shortest maturities, indicating the short rates are unaffected by
Factor 3, positive for intermediate maturities, and then
decline to become negative for the longest maturities.
Thus a positive change in Factor 3 would tend to
increase intermediate rates and decrease long rates,
altering the curvature of the term structure. In the first
and last subperiods the loadings peak at three to five
years and decline fairly uniformly thereafter. In the
intermediate period the peak is around seven years, and
the loadings do not decline markedly until around fifteen years’ maturity.
A classic example of a sharp change in the shape of
the term structure occurred in February and March of
1996. At the beginning of the period the market, as evidenced by federal funds futures prices, expected the
Federal Reserve to continue to lower short-term rates
after two consecutive 25 basis point declines in the federal funds target rate in December and January.12 At the
same time, the term structure was declining out to
about two years before it sloped upward, as shown by
the heavier line in the top panel of Chart 4. This “inverted hump” shape is unusual. It probably reflects concerns that the weak 0.9 percent increase in real gross

domestic product in the fourth quarter of 1995 might
foreshadow a recession and that the Federal Reserve
would have to cut interest rates to offset this slowdown.
Early in March it became clear that budget impasses,
which had shut down the government repeatedly (and
particularly hampered the reporting of economic data),
had been resolved, and it was becoming increasingly evident that the weak fourth quarter economic results were a
temporary aberration. This positive macroeconomic news
led market participants to reverse their expectations
of the Federal Reserve’s
near-term policy actions,
and the term structure
became sharply upwardly sloped out to two
The ability of a few factors
years before continuing
to explain changes in interto increase at a more
est rates is remarkably
moderate rate. The thinner line in the middle
constant. Three factors
panel shows the resultcan explain most of the
ing yield curve changes.
variation, particularly
Applying factor analysis
to these changes, using
since 1978.
the 1982–95 decomposition, one can compute
the shocks in terms of
the three factors; the
estimates for Factors 1–3 are 0.4526, 0.2350, and 1.0305,
respectively. Overall, there was an increase in the level
of interest rates (Factor 1) of about 0.4 percent, substantially offset at the shortest maturities by an increase
in the slope (Factor 2). The increase in the long-term
rate is the sum of the slope change, which is positive at
long maturities, and the level. The change in the curvature of the term structure from convex to concave is
entirely due to the large Factor 3 shock, the curvature
factor. Adding the effects of the three factors, shown
individually in the bottom panel, produces the total
changes in the yield curve resulting from common factors, shown as the heavier line in the middle panel. The
differences between the actual yield changes and the
changes due to factor shocks reflect the idiosyncratic, or
maturity-specific, component in the yield changes. It
can be seen that this idiosyncratic component is of second-order importance.
The Time-Series Behavior of Factor Changes.
By construction, the factors produced by a factor
analysis are uncorrelated with each other and have
identical unit variances. Nothing in the estimation
procedure, however, constrains how the factors

12. Federal funds futures contracts trade on the Chicago Board of Trade. Each contract’s terminal value is tied to the average
effective federal funds rate over its expiration month. Because contracts for various expiration months trade simultaneously, and because the Federal Reserve directly targets the federal funds rate in its open market operations, the market’s
expectations of Federal Reserve actions can be inferred from the term structure of federal funds futures prices.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

25

C H A R T 5 Impulse Responses, Six Lags

Response of
Factor 1

Factor 2

Factor 3

Factor 1

Shock to

1.00

0.50

0.00

–0.50
0

5

10

15

0

5

10

15

0

5

10

15

0

5

10

15

0

5

10

15

0

5

10

15

0

5

10

15

0

5

10

15

0

5

10

15

Factor 2

Shock to

1.00

0.50

0.00

–0.50

Factor 3

Shock to

1.00

0.50

0.00

–0.50

26

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

behave through time. Interest rate levels may persist
(that is, positive Factor 1 shocks may tend to be followed by additional positive Factor 1 shocks). Slope
increases in one period (a positive shock to Factor 2)
may increase the likelihood that rates will increase in
the next period (a positive shock in the next period to
Factor 1). The time-series behavior of the three factors was examined using a vector autoregressive
model, or VAR (see the appendix). Impulse-response
functions based on the VAR are generated to show how
a hypothesized single shock to one factor leads to subsequent changes in that factor and to current and subsequent changes for other factors. The estimated
response functions, along with the 95 percent confidence bands, are shown in Chart 5. Thus, a positive
shock of one standard deviation (+1.00) to Factor 1 at
time 0 will be followed on average by a +0.25 change in
Factor 1 at time 1. The error bands indicate that this
follow-up response is barely significant. After time 1
the effects on subsequent outcomes of Factor 1 are
negligible. The same time 0 shock to Factor 1 is associated with a slight, marginally significant, negative
change in Factor 2 at time 1, after which the responses die out.13 Factor 3 shows a negative time 1 response
and a positive time 2 response to the Factor 1 shock,
both marginally significant. Thus, changes in levels of
interest rates show a slight tendency to persist and to
have a slight effect on changes in slope and curvature
in the subsequent period. None of these effects persists beyond two periods.
Shocks to Factor 2 show the same slight tendency
to persist for a single period, after which the responses
die out. There is, however, no evidence of effects on
Factors 1 or 3 of changes in Factor 2, either contemporaneously or subsequently. Shocks to Factor 3 do not
persist even for a single period. Factor 1 shows no
response to Factor 3 shocks. Factor 2 shows negative,
though only marginally significant, responses at lags of
four and six months.

Implications of the Factor Model
he factor analysis approach used in this paper is
primarily exploratory in nature. It reveals that
three factors account for a large portion of the
underlying interest rate changes and hence for bond
price movements. The analysis also provides a clue to
the economic interpretation of those factors, suggesting
how they might be related to broader economic factors.14
The factor analysis provides an ad hoc, though very
effective, approach to hedging. However, the factor analysis does not constitute a coherent theoretical model of

T

interest rate movements. It provides only a set of stylized
facts that any such model should be able to capture.
The first, and most obvious, implication of this analysis is that single-factor-based interest rate models, such
as the single-factor CIR or Vasicek models, are incomplete, notwithstanding tests of these models that have
failed to reject them. Single-factor models may be “good
enough” for some applications such as managing portfolios of similar-maturity bonds, but they will result in hedging error when applied
to complex securities,
such as spread derivatives, for example. The
The factor analysis probox illustrates the relavides an ad hoc, though
tive performance of
Macaulay duration and
very effective, approach
factor durations hedgto hedging. However, the
ing for several hypofactor analysis does not
thetical portfolios.
The factor analysis
constitute a coherent
also explains why a simtheoretical model of
ple procedure such as
interest rate movements.
Macaulay duration
immunization works as
well as it does, despite
being based on the false
premise that interest rates for all maturities always
change by the same amount. While interest rates do not
always move in parallel, the largest single factor in interest rate movements is a parallel shift, accounting for
about 80 percent of the variation. This result helps
explain why other single-factor-based hedging approaches have not done much better than Macaulay duration.
Any single-factor-based model would result in changes in
the term structure at all maturities being perfectly correlated, which is not in fact the case. The factor analysis
shows that at least a portion of the non-parallel-shift component in changes in interest rates is uncorrelated with
the parallel-shift component. Thus, it is difficult for any
single-factor-based hedging technique, even one that
attempts to capture nonparallel shifts, to improve on
Macaulay duration.
A multifactor-based hedging strategy incorporating
the results developed in this article would begin by constructing an interest rate model using the factor decomposition developed above and the evidence of the
behavior of the factors themselves taken from the VAR
analysis—a simple three-factor model with lags of no
more than two periods would be sufficient. As the laggedvariable coefficients are only marginally significant and
thus likely to be spurious, consideration should also be

13. Apparent significant contemporaneous correlation between Factors 1 and 2 is an artifact of the estimation method.
14. Litterman, Scheinkman, and Weiss (1991) show how term structure curvature may be related to volatility in a simple singlefactor interest rate model (there factor has a different meaning from the linear factors in a factor analysis decomposition).
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

27

B O X

A Comparison of Macaulay Duration Hedging
and Factor Durations Hedging
he month from February 15 to March 15, 1996, provides a vivid example of the comparative advantage of
factor durations hedging over Macaulay duration hedging.
Three portfolios of bonds were constructed on February 15
as follows:
(1) Portfolio 1: A single twenty-year 8 percent coupon
bond (paying coupons semiannually, as is usual)
(2) Portfolio 2: Equal numbers of one-year and twentyyear 8 percent coupon bonds
(3) Portfolio 3: Long positions in one unit each of oneyear and twenty-year zero-coupon bonds, together
with a short position in one unit of a ten-year zerocoupon bond
The face values of the bonds in the portfolios were adjusted
so that the initial price of each portfolio was $100. The portfolios were priced using the February 15 term structure and
their Macaulay and factor durations, the latter computed
using the November 1982–December 1995 factor loadings.
See Table A.
Portfolio 1 loads heavily on the level factor. This
result suggests that Macaulay duration will provide a reasonable basis for hedging. However, since the slope- and
level-factor durations are not zero, hedges based on all
three factors should do somewhat better. Portfolio 2 is a
portfolio of coupon bonds of widely divergent maturities.
Such a portfolio is likely to be sensitive to both changes in
levels and changes in the slope of the term structure.
Macaulay duration hedging is expected to do even less
well in this instance. Portfolio 3 is a portfolio of mixed
long and short positions. Its level-factor duration is approximately equal to its Macaulay duration. However, as shown
by the curvature-factor duration, Portfolio 3 is particularly
sensitive to changes in the curvature of the term structure. It is unlikely that Macaulay duration can capture the
interest rate sensitivity of such a portfolio.
Two hedge portfolios were constructed for each portfolio. The first “Macaulay duration–matched” portfolio consisted of two zero-coupon bonds of adjacent (six months
apart) maturity, chosen to match both the price and the
Macaulay duration of the portfolio being hedged.1 The second “factor durations–matched” portfolio consisted of zero-

T

TA B L E A

Macaulay Duration versus Factor Durations
Macaulay
Duration

Factor Durations
Level

Slope

Curvature

4.45

1.29

2.33

0.89

0.27

–2.07

Portfolio 1
10.98

9.47
Portfolio 2

6.40

5.53
Portfolio 3

1.58

1.65

coupon bonds of one, five, ten, and twenty years’ maturity, in
amounts chosen to match the price and all three factor
durations of the portfolio being hedged.
Each portfolio and the associated two hedge portfolios
were then repriced on March 15, 1996. An ideal hedge portfolio would have the same return over the period from February
15 to March 15 as the portfolio it is hedging. See Table B.
These results clearly show that hedges based on the
three factor durations outperform hedging based on Macaulay
duration. Even for the “plain vanilla” long bond (Portfolio 1),
the hedging error of the Macaulay duration hedge portfolio,
while small (only 0.27 percent), is more than twice that of the
factor durations hedge portfolio. The Macaulay duration
hedge portfolio for Portfolio 2 shows a large hedging error,
greater than 1 percent, while the factor durations hedge portfolio shows a small error of only 1/20 of 1 percent. Portfolio 3’s
duration is short, only 1.6 years, and thus the portfolio has less
price sensitivity to interest rate swings than do Portfolios 1
and 2. Nonetheless, even though price changes were small,
Macaulay duration hedge portfolio missed the mark by more
than 1 percent while the hedging error produced by the factor
durations hedge portfolio was insignificant.
These three sample portfolios illustrate how in times
of unusual interest rate movements, when the slope and
curvature of the term structure are changing significantly,
Macaulay duration is unable to provide a sound basis for
hedging a wide variety of cash flows. It is in these cases that
factor durations hedging becomes more valuable.
Litterman and Scheinkman (1991) examined the ability of

1. This “near bullet” bond approximates a zero coupon of the desired duration while avoiding the need to interpolate between coupon
payment dates.
28

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

B O X

( C O N T I N U E D )

TA B L E B

Macaulay Duration versus Factor Durations
Macaulay Duration
Hedge Portfolio
Actual
Portfolio
Return
(Percent)

Return
(Percent)

Hedging Error
(Percent)

Factor Durations
Hedge Portfolio
Return
(Percent)

Hedging Error
(Percent)

0.29

–7.31

–0.10

1.03

–4.03

–0.06

1.07

0.27

0.0

Portfolio 1
–7.41

–7.70

–4.08

–5.12

0.27

–0.79

Portfolio 2

Portfolio 3

factor durations hedging and Macaulay duration hedging to
hedge weekly returns for thirteen Treasury bonds of maturities ranging from six months to nearly thirty years over
the period from February 1984 through August 1988. Their

results confirm that, over a wide range of term structure
movements and portfolios, factor durations hedging is significantly better than simple Macaulay duration hedging.

given to a simple three-factor model with no lags. The
hedge would then construct a portfolio of short and
long exposures to the three factors so that the net
exposure of the portfolio to each factor is minimized.15

evidence of increased volatility of the factors during the
Federal Reserve experiment period of October 1979
through October 1982, which is to be expected given the
rise in interest rate volatility in that period.
The success of a parsimonious factor model
involving interest rates contrasts with the moderate,
at best, success of the same approach to modeling
stock returns. This dichotomy underlies the completely different approaches to risk management used for
equities and interest rate–sensitive securities. In the
former case, the emphasis is on idiosyncratic risk
reduction through portfolio diversification, perhaps
with futures used to hedge the small systematic component of stock returns. For interest rate–sensitive
securities the emphasis is on precisely balancing a
portfolio to achieve the desired exposure to systematic risk factors. There is little use for portfolio diversification in managing interest rate–sensitive portfolios,

Conclusion
he three-factor decomposition of movements in
interest rates, first uncovered by Litterman and
Scheinkman (1991), is robust. The ability of the
three factors to explain observed changes in interest
rates is high in all subperiods studied and particularly
since 1978, when they explained virtually all interest rate
movements. Furthermore, the nature of the movements—level, slope, and curvature—has not changed,
although the cross-sectional loadings have varied slightly. The factors themselves appear to be well behaved.
There is only slight evidence of time-series or cross-factor
interactions that would complicate modeling. There is

T

15. Two key principles govern how to hedge. The first is that, in general, risks cannot be hedged piecemeal. If stock prices and
bond prices tend to move together, then it is incorrect to hedge the market risk in the stock portfolio independently of the
interest rate risks in the bond portfolio. Hedging correlated risks separately results in costly redundancy in the hedges since
the correlations tend to reduce the total risk of the combined portfolio below that obtained by summing the risks of the components. The exception to this rule is when the risks are uncorrelated, as are the risks associated with the factors obtained
by factor analysis. In this article, the factors are the sources of interest rate risk and are not correlated with each other.
Therefore, they may be hedged individually, one factor at a time, though for each factor hedging should be done across all
bonds in the portfolio. This approach may not work if the portfolio contains other types of securities, such as stocks, subject
to additional sources of risk.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

29

and when it is used it is primarily to address the separate issue of credit risk where it exists.
The factor analysis shows that parsimonious interest
rate models are adequate for hedging. The analysis also
indicates that interest rate models should not be so parsimonious as to have only a single underlying source of
uncertainty. Applying this information to hedging is fairly

straightforward. As shown in the appendix, factor durations are analogous to Macaulay duration and can be easily computed. Furthermore, the factor durations of a
portfolio are the weighted averages of the portfolio components’ factor durations. Thus, factor immunization is
straightforward. Translating the factor structure into a rigorous interest rate model is apt to be more problematic.

A P P E N D I X

Mathematical Details


∑ ≡ E ( Xt − µ )( Xt − µ )′  = LL′ + Ψ.

Factor Analysis
At each period, t = 1,…, T, we observe a p vector of
variables, Xt . Factor analysis assumes that these observations are linearly related to m, m < p, underlying unobserved factors by the following relation:
Xt − µ =
p×1

L

+ εt

Ft

p× m m×1

p×1

(A1)

where
E(Xt ) = µ
E(Ft ) = 0
E(εt ) = 0
cov(Ft ) = E(Ft Ft 9) = I
m3 m
cov(ε t ) = E(ε tε t ′) = Ψ
p× p
Ψ1

0
=
M

 0

0
Ψ2
M
0

L 0

L 0
.
O M 

L Ψp 

The Ft s are called factors and the Ls are called factor
loadings.
In this article p = 10, corresponding to the maturities
of the yield changes being analyzed, and t indexes the
months for which observations are made. Based on the evidence in previous studies, in this article m = 3. Johnson
and Wichern (1982, chap. 9) discuss techniques for determining the appropriate numbers of factors where these are
not known ex ante.
Everything on the right-hand side of the first equation
is unknown; however, the structure implies that

30





Note that the factors, Ft , do not appear, nor do the individual residual errors, εt . Because these variables do not
appear in structure of ∑, the number of unknown parameters is reduced so that it then becomes possible to compute
the factor loadings, L, and idiosyncratic variances, Ψ.
Two methods are widely used for estimating the elements of L and Ψ. The first uses the first m principal components of the estimated variance-covariance matrix, ∑,
to construct L. Then Ψ is constructed from ∑ – LL′ by
setting the off-diagonal elements to zero.1 The second
method is to assume that the residuals are multivariatenormal and to estimate the model parameters using maximum likelihood. Both methods are explained in detail in
Johnson and Wichern (1982) and other standard texts on
factor analysis. In this article, principal components estimation is used.
The solution, L, is not unique. If T is any orthogonal
matrix (that is, TT′ = T′T = I), then L* = LT is also a
solution:

∑ = L*L*′ + Ψ = LTT′L′ + Ψ = LIL′ + Ψ = LL′ + Ψ.
Rotation also affects the factors, Ft . When L in equation (A1)
is replaced with L* = LT, the factors become Ft * = T′Ft so that
Xt – µ = L*Ft * + εt = LTT′Ft + εt = LFt + εt .
Rotation of the factor loadings has no effect on Ψ. This
indeterminacy permits rotating the original solution until
factor loadings that have meaningful economic interpretations are obtained. For instance, for this article the load-

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

A P P E N D I X

( C O N T I N U E D )

ings were first rotated so that yield changes at all maturities had approximately the same loading on the first factor.
Since any perturbation to the first factor then affects all
maturities equally, this factor is interpreted as a “level”
shift. The interpretation of the remaining factors was obvious without further rotations.
To construct the rotation matrix, T, the following
algorithm was used. Since T is 3 3 3, it has nine elements.
However, the requirement that it be orthogonal reduces
the effective number of free parameters to three. T was
built up from three orthogonal rotation matrices, each
leaving one column of L (and Ft ) unchanged, and recombining the two remaining columns while preserving the
orthogonal structure of the resulting L*.
 cosθ1 sinθ1 0


T1 = − sinθ1 cosθ1 0.
 0
0
1
 cosθ 2 0 sinθ 2 


1
0 .
T2 =  0
− sinθ 2 0 cosθ 2 
1
0
0 


T3 = 0 cosθ 3 sinθ 3 .
0 − sinθ 3 cosθ 3 
Then T = T1T2 T3 is an orthogonal matrix with three free
parameters θ1, θ2, and θ3. A nonlinear optimizer was used
to search over feasible values of θ1, θ2, and θ3 (in the range
–π to +π) to find the value that minimized the variance of
the first column L*.
Once the appropriate estimated factor loadings have
been obtained, the estimated factors themselves can be
computed by
Ft = (L′Ψ –1L)–1L′Ψ –1(Xt – µ).

time-series structure of the factors, a vector-autoregressive
(or VAR) model is used. The model relates current values of
each of the factors to past values of all the factors.
Ft = A1
m×1 m× m

Ft –1 + K + Aj
m×1

Ft – j + ηt

m× m m×1

m×1

where E(ηt) = 0 and E(ηt′ηt) = I. Normally it is advisable
that this assumed restriction on the residuals be tested.
However, in this case the restriction holds by construction
since the factors are orthogonal. This model can be estimated using maximum likelihood methods as outlined in
Hamilton (1994).
The VAR analysis shows whether there are any linear
time-series relations among the factors—that is, whether
a shock to one factor in this period may have an influence
on another factor the next period. These relations are evidenced by nonzero elements in Aj . Equivalently, the
effects through time of individual shocks on each of the
factors can be examined using impulse response functions.
An impulse response function uses the estimated VAR
model to estimate the impact of a single, hypothetical,
one-time, one-standard-deviation shock to each factor on
the other factors in subsequent periods. The confidence
intervals for these response functions are computed using
the techniques developed in Sims and Zha (1995).

Hedging Portfolios against Factor Risks
Before beginning the discussion of hedging interest
rate movements using the factor model, it will be useful to
consider the simpler Macaulay duration hedging.
Macaulay duration is based on an assumed single interest
rate (that is, flat term structure). The price of any stream
of cash flows, CFm, m = 1,…, M, is thus2
M

P = ∑ CFm e – my.
m=1

The properties of the estimated factors themselves can
then be studied. For example, the time-series properties of
the factors can be investigated, as is done in this article.

Vector Autogression
The three factors estimated through factor analysis are,
by construction, orthogonal and of unit variance (that is, Ft′Ft
= I). However, there is no structure imposed on the timeseries behavior of the factors. To investigate the possible

For a normal coupon bond, the cash flows would correspond to the individual, usually equal, coupon payments
each period prior to maturity and the final coupon plus
principal repayment at maturity. For a portfolio of interest
rate–sensitive securities, the cash flows each period would
be aggregated across securities in the portfolio. These cash
flows may be either positive (for assets or long positions)
or negative (for liabilities or short positions).

1. An examination of the residual matrix, ∑ – LL9, to see if the off-diagonal elements are “small” before setting them to zero, is
advised. If some off-diagonal elements are large, there may be additional omitted factors, and a larger value for m may be warranted. In this article, this problem did not occur.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

31

A P P E N D I X

( C O N T I N U E D )

M
∂P
dym = ∑ ( − m)CFm e− my m dym
m =1 ∂ ym
m =1

The continuously compounded Macaulay duration is
found by taking the derivative of price, P, with respect to
yield, y:
dP M
= ∑ (– m) CFm e– my,
dy m =1

M

dP = ∑

m
3


= − ∑  mCFm e− mym ∑ Li ,m Fi .
m=1 
i =1


Dividing through by P and rearranging yields

then dividing both sides by P and expressing changes in P
in terms of changes in y:
 M − mCFm e− my 
dP

M
= − ∑
 dy = − ∑ mwm  dy = − Ddy.
P
P

 m=1
 m=1

Since CFm e–my is the present value of the cash flow at time
m, the ratio wm ≡ CFm e–my/P is the fraction of the bond’s or
portfolio’s value due to the cash flow at time m. Thus,
Macaulay duration, D, is the weighted average of the time
to each cash flow, m, where the weights, wm, are the fractions each cash flow currently contributes to the value of
the total portfolio.
When securities are combined into portfolios, the
duration of the portfolio is the weighted average of the
durations of the assets (either individual securities or
portfolios) in the portfolio, where the weights are the fractions invested in each asset. These weights may be positive
or negative but must sum to unity.
Suppose two portfolios have durations D1 and D2 and
the objective is to construct a portfolio that is immunized
to changes in yields, that is, Dp = 0. The portfolio weights,
or fractions invested in each of portfolios 1 and 2, are
w1 ≡ D2 /(D2 – D1) and w2 ≡ (1 – w1) = –D1/(D2 – D1). If
both durations are positive (or negative), then one of the
basic portfolios will need to be sold short, resulting in a
negative weight.
The same ideas apply to the factor structure of yield
changes, although the mathematics is slightly more complicated. The change in any given zero-coupon interest
rate, ym, for maturity m, is related to the factor shocks, Fi ,
i = 1, 2, 3 (recall that factor outcomes each period are the
sources of yield changes), by the factor loading, Li,m, appropriate to maturity m, for that zero coupon interest rate:
3

dym = ∑ Li ,m Fi .
i =1

The change in value of a portfolio of interest rate–sensitive
cash flows, P, is a function of the changes in each interest
rate indicated by

3  M
3

dP
CF e− mym

M
= − ∑ ∑ m m
Li ,m  Fi = − ∑  ∑ mwm Li ,m  Fi
P
P
1
1
=
=
i =1  m=1
i
m



3

= − ∑ Di Fi .
i =1

To summarize, Macaulay duration, D, is a measure of
the portfolio’s or bond’s sensitivity to changes in the (single) interest rate, y. Macaulay duration is computed by
taking the weighted average time-to-cash flow where the
weights are the present values of the cash flows divided by
the total value of the portfolio. Factor durations, Di, are
analogous. They measure the sensitivity of a portfolio’s
value to each of the factors. Factor durations are the
weighted average of the time-to-cash flows multiplied by
the factor loadings appropriate to the horizon of the cash
flow. As in the case of Macaulay duration, the weights are
the present value of each cash flow divided by the value of
the portfolio.
Factor durations combine linearly. If two portfolios
have prices P1 and P2 and associated factor durations, D1i
and D2i , i = 1, 2, 3, then the factor durations of the portfolio will be
DiP =

P1
P
Di1 + 2 Di2 ,
P1 + P2
P1 + P2

or more generally
N

DiP = ∑ x j Dij ,
j =1

where the xj are the fractions of total portfolio value represented by the value of each component (which may be
individually positive or negative but must sum to unity).
Immunization in this context requires selecting portfolio
weights, xj , so that all the combined portfolio factor sensitivities are zero. The three-factor interest rate decomposition presented here requires four bonds or bond portfolios
with sufficiently different (that is, linearly independent)
factor durations to serve as building blocks. Then four
simultaneous equations, one for each factor and one to guarantee that the weights sum to unity, are solved to arrive at
the weights for each element of the portfolio.

2. For notational simplicity the t denoting time of observation, Pt or yt , or arrival of shocks, Ft , is omitted in the following.
32

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

REFERENCES
BLISS, ROBERT R. 1997. “Testing Term Structure Estimation
Methods.” Advances in Futures and Options Research
9:197–231.

JOHNSON, RICHARD A., AND DEAN W. WICHERN. 1982. Applied
Multivariate Analysis. Englewood Cliffs, N.J.: Prentice-Hall.

BLISS, ROBERT R., AND DAVID C. SMITH. 1997. “The Stability of
Interest Rate Processes.” Federal Reserve Bank of Atlanta
Working Paper 97-13, November.

JOINT NOTICE OF PROPOSED RULEMAKING. 1995. Risk-Based
Capital Standards: Market Risk. Federal Register 60, no. 142
(July 25): 38082ff.
KAUFMAN, GEORGE G., GERALD O. BIERWAG, AND ALDEN TOEVS,
1983. Innovations in Bond Portfolio Management.
Greenwich, Conn.: JAI Press.

BREALEY, RICHARD A., AND STEWART C. MYERS. 1991. Principals
of Corporate Finance. New York: McGraw-Hill.

EDS.

COX, JOHN C., JONATHAN E. INGERSOLL JR., AND STEPHEN A. ROSS.
1985. “A Theory of the Term Structure of Interest Rates.”
Econometrica 53 (March): 385–407.

KNEZ, PETER J., ROBERT LITTERMAN, AND JOSÉ SCHEINKMAN.
1994. “Explorations into Factors Explaining Money Market
Returns.” Journal of Finance 49 (December): 1861–82.

DHRYMES, PHOEBUS J., IRWIN FRIEND, AND N. BULENT GULTEKIN.
1984. “A Critical Reexamination of the Empirical Evidence
on the Arbitrage Pricing Theory.” Journal of Finance 39
(June): 323–46.

LINTNER, JOHN. 1965. “The Valuation of Risk Assets and the
Selection of Risky Investments in Stock Portfolios and
Capital Budgets.” Review of Economics and Statistics 47
(February): 13–37.

DUFFEE, GREGORY R. 1996. “Idiosyncratic Variation of Treasury
Bill Yields.” Journal of Finance 51 (June): 527–51.

LITTERMAN, ROBERT, AND JOSÉ SCHEINKMAN. 1991. “Common
Factors Affecting Bond Returns.” Journal of Fixed Income
1 (June): 54–61.

FAMA, EUGENE F., AND ROBERT R. BLISS. 1987. “The Information
in Long-Maturity Forward Rates.” American Economic
Review 77 (September): 680–92.
GULTEKIN, N. BULENT, AND RICHARD J. ROGALSKI. 1984. “Alternative Duration Specification Measurement of Basis Risk:
Empirical Tests.” Journal of Business 57:241–64.

LITTERMAN, ROBERT, JOSÉ SCHEINKMAN, AND LAURENCE WEISS.
1991. “Volatility and the Yield Curve.” Journal of Fixed
Income 1 (June): 49–53.
LONGSTAFF, FRANCIS A., AND EDUARDO S. SCHWARTZ. 1992. “A
Two-Factor Interest Rate Model and Contingent Claims
Valuation.” Journal of Fixed Income 2 (December): 16–23.

HAMILTON, JAMES D. 1994. Time Series Analysis. Princeton,
N.J.: Princeton University Press.

ROSS, STEPHEN A. 1976. “The Arbitrage Theory of Capital Asset
Pricing.” Journal of Economic Theory 13 (December):
341–60.

HEATH, DAVID, ROBERT JARROW, AND ANDREW MORTON. 1990.
“Bond Pricing and the Term Structure of Interest Rates:
A Discrete Time Approach.” Journal of Financial and
Quantitative Analysis 25 (December): 419–40.

SHARPE, WILLIAM F. 1964. “Capital Asset Prices: A Theory of
Market Equilibrium under Conditions of Risk.” Journal of
Finance 19 (September): 425–42.

———. 1992. “Bond Pricing and the Term Structure of
Interest Rates: A New Methodology for Contingent Claims
Valuation.” Econometrica 60 (January): 77–105.

SIMS, CHRISTOPHER A., AND TAO ZHA. 1995. “Error Bands for
Impulse Responses.” Federal Reserve Bank of Atlanta
Working Paper 95-6, September.

ILMANEN, ANTTI. 1992. “How Well Does Duration Measure
Interest Rate Risk?” Journal of Fixed Income 1 (March):
43–51.
INGERSOLL, JONATHAN E., JR. 1983. “Is Immunization Feasible?
Evidence from the CRSP Data.” In Innovations in Bond
Portfolio Management, edited by George G. Kaufman, Gerald
O. Bierwag, and Alden Toevs. Greenwich, Conn.: JAI Press.

VASICEK, OLDRICH. 1977. “An Equilibrium Characterization
of the Term Structure.” Journal of Financial Economics
5:177–88.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

33

The Insider Trading
Debate
J I E H U A N D T H O M A S H . N O E
Hu is a senior economist in the financial section of
the Atlanta Fed’s research department. Noe, a former
Atlanta Fed visiting scholar, is the A.B. Freeman
Professor of Finance at the A.B. Freeman School of
Business, Tulane University. They thank Peter Abken,
Robert Bliss, Jerry Dwyer, and Joe Whitt for helpful
comments.

S
LATE

ECURITIES TRADING BY OFFICERS, DIRECTORS, AND OTHER KEY EMPLOYEES OF CORPORATIONS WHO HAVE ACCESS TO PRIVATE INFORMATION HAS GENERATED SOME OF THE MOST
SENSATIONAL SCANDALS IN THE POPULAR BUSINESS PRESS.
CASES OF INSIDER TRADING IS THAT OF IVAN

F. BOESKY

ONE

AND

OF THE MOST PUBLICIZED

MICHAEL R. MILKEN

IN THE

1980S. MILKEN WAS SENTENCED TO TEN YEARS IN PRISON, THE LONGEST SENTENCE IN U.S. HISTO-

RY METED OUT FOR VIOLATION OF INSIDER TRADING CODES.
MENT,

BOESKY

FOR

HIS COOPERATION WITH THE GOVERN-

RECEIVED A MORE LENIENT THREE-YEAR SENTENCE.

BOTH

WERE ORDERED TO PAY

HUNDREDS OF MILLIONS OF DOLLARS IN DAMAGE ASSESSMENTS AND PUNITIVE PENALTIES. IN ADDITION,
SEVERAL OTHER INVESTMENT BANKERS AND TRADERS WERE IMPLICATED AND PUNISHED IN THE CASE.1

ANOTHER CLASSIC CASE OF ILLEGAL INSIDER TRADING, THE CASE OF TEXAS GULF SULPHUR COMPANY, IS
BRIEFLY DESCRIBED IN

BOX 1.

Unlike other illegal activities, insider trading
remains, at least among economists and legal scholars,
one of the most controversial economic transactions. A
substantial body of academic and legal scholarship questions whether insider trading is even harmful, much less
worthy of legal action. The views on insider trading
range from moral revulsion to positive evaluations of its
economic benefits. In turn, many scholars support the
current restrictions placed on insider trading while others advocate a laissez-faire government policy. Why are
there such sharply contrasting views? What different
rationales are advanced for permitting and prohibiting

34

insider trading? This article explores the sources of the
insider trading controversy and suggests a road map for
blending the divergent scholarly opinions into a policy
framework for regulating insider trading.

Legislating Insider Trading
publicly listed firm’s “insiders” include its directors, officers, and other key employees. While the
legal definition of who the insiders are may extend
further (see Box 2), this article is concerned with this
classic sort of insider. Trading by a firm’s insiders on the
firm’s stock or derivative assets is not illegal unless,

A

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

loosely speaking, it is determined that the trading activity takes advantage of their confidential information
regarding the firm’s future prospects.2 The essence of the
existing laws on insider trading is that insiders must
either abstain from trading on such information or release
it to the public before they trade. A quick consideration of
some particulars of the relevant laws may be helpful.
The governmental body in charge of regulating
insider trading is the Securities and Exchange
Commission (SEC), which was established by the
Securities and Exchange Act of 1934 (referred to subsequently as the 1934 Act). The 1934 Act, its amendments,
and additional legislation passed in subsequent decades
formulate the legal bounds on insider trading. Among
all the code sections, the broad language in Section
10(b)-5 banning any “manipulative or deceptive device”
used “in connection with the purchase or sale of any
security” is the most often cited legal basis for banning
insider trading. The courts have interpreted this section
of the law, in cases such as Speed v Transamerica
Corporation (99 FSupp. 808, 828-32 [D. Del. 151]), as a
broad prohibition of insider trading that takes advantage of confidential information. Sections 10(b) and
17(a) of the 1934 Act are interpreted as more generally
prohibiting insider trading on material, nonpublic information about the firm. Section 16(b) requires the
returning of short-swing profits by insiders to the corporation, with a short swing defined as “round-trip”
transactions (a purchase and a sale or a sale and a purchase) within six months; and Section 16(c) prohibits
short sales by insiders. As noted earlier, insiders are
permitted to trade as long as the trading does not take
advantage of confidential information; Section 16(a),
however, requires that all trading by insiders be reported to the SEC within the first ten days of the month following the month in which the trading is executed.3 The
SEC publishes such transactions in its monthly Official
Summary of Insider Transactions on the assumption
that making insider trading transparent helps expose
any illegal trades and thus serves as a deterrent.
Prosecution of insider trading was not very common until the second half of this century. Since 1961,
insider trading regulations have become more restrictive
through a number of cases and interpretations. In 1975
Section 32 of the 1934 Act was amended to increase the
maximum criminal penalty fines to $10,000 and the
maximum prison sentence to five years. Vigorous
enforcement of the stiffer penalties did not happen

until the 1980s though. Between 1966 and 1980 the SEC
filed only thirty-seven cases of insider trading, and twentyfive of them were settled out of court; that is an average of
only 2.6 cases per year, and the SEC sought or obtained disgorgement of profits in twelve of these (Seyhun 1992).
From 1982 to 1986, according to Haddock and Macey
(1986, 1987), the SEC initiated seventy-nine cases based
on Section 10(b)-5, an average of 17.2 cases per year.
Meulbroek (1992) reported that between 1980 and 1989
there were at least 464 defendants in the insider trading
cases pursued by the SEC.
In 1984 Congress passed the Insider Trading
Sanctions Act of 1984 (ITSA), which provides for up
to three times the insiders’ illegal profits in civil
penalties and a tenfold
increase in criminal
penalties (from $10,000
A substantial body of
to $100,000). As enforcement became more vigacademic and legal
orous, the courts began
scholarship questions
imposing prison senwhether insider trading
tences in 1985, whereas
none of the cases proseis even harmful, much
cuted before 1980 ended
less worthy of legal action.
with jail sentences. In
1988, Congress passed the Insider Trading and Securities
Fraud Enforcement Act
(ITSFEA), which creates a bounty program for insider
trading informants and holds the top management of a
firm responsible for its employees’ illegal insider trading
activities. Moreover, ITSFEA increased the maximum
criminal penalties to $1 million and the maximum jail
sentence to ten years. Trading partners who suffer losses
because of insiders’ illegal activities have the right to
recover their losses under ITSFEA.
Despite the SEC’s efforts to curb insider trading that
is based on confidential information, there is evidence
that insider trading is active and insiders’ trading profits
are excessive relative to the average market return.
Seyhun documents that “corporate insiders earned an
average of 5.1 percent abnormal profits over a one-year
holding period between 1980 and 1984, increasing further
to 7.0 percent after 1984, compared with 3.5 percent before
1980. During the 1980s, insiders increasingly sold stock
before bad news. Moreover, after increases in regulations,

1. For details about this particular insider trading event, see Stewart (1991).
2. A derivative asset is a security for which the payoff at a future time depends on the price of another security or the prices of
several other securities.
3. Those who hold more than 10 percent of any equity class must also report their trading activity to the SEC. Whether they are
covered by the other insider trading rules depends on whether they have actual access to corporate inside information (see
Box 2).
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

35

B O X

1

Insider Trading at Texas Gulf Sulphur Company
nsider trading not only concerns scholars and regulators
but also attracts the attention of the general public. To
get a practical idea of insider trading, consider the famous
Texas Gulf Sulphur Company case (Manne 1966, 49).
Texas Gulf Sulphur Company was established in 1909.
In 1959 its exploratory prospecting with magnetic surveying equipment produced some evidence that valuable
deposits of copper, zinc, and silver might exist in an area of
Ontario. In 1963 the first drilling confirmed the possibility,
and the commercial value of the find proved to be enormous. The company instituted tight control of the drilling
project so as not to leak the information to outsiders.
Meanwhile, various officers, directors, and employees of
the company, knowing this information and the fact that it
was not released to the public, bought shares of, and call
options on, Texas Gulf Sulphur Company or were given
stock options by the company and tipped other people to
purchase the stock or options of the company. These activities happened between November 12, 1963, and April 16,
1964, a period when the stock prices of Texas Gulf Sulphur
Company were relatively low due to its lackluster performance in business.
Rumors about the company’s discovery surfaced and
became rife in mid-April 1964. By then the stock price had
risen to $29.375 from $17.375 on November 10, 1963. On
April 12, 1964, the company made an announcement,
which the Security Exchange Commission (SEC) later
accused of misleading the public, that the company’s

I

drilling had “not been conclusive” and “the rumors about
the discovery were unreliable . . . premature and possibly
misleading,” and originated with speculators not connected with the company. Four days later, on April 16, 1964,
however, the company announced “a major ore discovery”
of about 25 million tons of copper, zinc, and silver. The
stock price jumped to $71 on April 19, 1964. Those who had
purchased or acquired stocks and options before this date
reaped substantial financial gains.
In April 1965 the SEC filed a suit in the United States
District Court for the southern district of New York against
a number of individual defendants who were directors,
managers, and employees of Texas Gulf Sulphur Company.
The charges were based on the defendants’ violation of
Rule 10(b)-5 of the Securities Exchange Act of 1934 for
“engaging in the purchase and sale of securities on the
basis of information with respect to material facts relating
to Texas Gulf acquired by said defendants in the course of
their corporate duties or employment with Texas Gulf
which information had not been made available to Texas
Gulf, its stockholders and other public investors; (b) making available such information, directly or indirectly, to
other persons for the purpose of permitting or allowing
such other persons to benefit from the receipt of such
information through the purchase and sale of securities;
and (c) engaging in other conduct of similar purport and
object.” The SEC won the case.

data indicate that a larger volume of insider trading
activity was followed by greater favorable abnormal
price movements. Also, top executives appear to have
traded on more valuable private information in the
1980s” (1992, 176). Keown and Pinkerton (1981) note
that 40–50 percent of the price increase of an acquisition target firm occurs before the acquisition announcement, suggesting that some people have taken
advantage of the information before it is available to the
public. Meulbroek (1992), in a study of a pool of illegal
insider trading cases, has made a more specific statement that 43 percent of the stock price increase for an
acquisition target firm happened on the days when illegal trading occurred. Although insider trading is not prohibited if it technically does not violate the rules of the

SEC, the fact that insiders’ trading profits are higher
than others’ may indicate that the rules currently in place
are not serving their intended purposes.
Trading by insiders, whether legal or illegal, is substantial and increasing. Seyhun reports that “the number of shares traded by insiders went up by four times
from the pre-1980 period to the post-1984 period” (1992,
176) and the frequency of high-volume insider trading
also increased after the ITSA became law in 1984. These
developments raise questions about the effectiveness of
enforcement and the existence of loopholes in insider
trading laws,4 which are not the topics of this article.
This discussion focuses instead on the fundamental
question of whether it is desirable to restrict insider
trading in the first place.5

36

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

Who Is Affected by Insider Trading?
or understanding the effects of insider trading, it
is helpful to categorize the agents involved or
affected into several groups. Economic analysis of
insider trading typically considers the following parties:
insiders, market professionals, liquidity traders, and
investors, who are defined as follows.
Insiders, as defined earlier, are the officers, directors, and other key employees of a firm who, by the
nature of their employment, obtain or possess confidential information regarding the firm’s prospects. An
example of an insider is the chief executive officer or
the chief engineer of the firm.
Market professionals are informed noninsiders,
including securities analysts, brokers, or arbitrageurs,
who have acquired private information regarding the
firm’s prospects by spending their own resources and
who do not have any fiduciary relationship with the
firm. For example, a security analyst may have called
the firm’s major customers and learned that they are
not interested in buying its new product line.
Liquidity traders, sometimes referred to as “noise”
traders, are short-term stock market participants who
have some, usually negligible, holdings of the firm’s shares
and trade in order to hedge risk or balance their portfolios
without consideration of a firm’s prospects. An example of
a liquidity trader is a large pension fund that buys and
sells the firm’s shares from time to time in order to meet
the investment and redemption needs of its clients.
Investors may be small or large shareholders who
have a long-term investment objective such that they “buy
and hold.” While not privy to management’s private information, investors have a significant beneficial interest in
the firm’s actual performance. For instance, the heir to a
substantial holding of the firm’s stock who does not take
an active role in its daily management is an investor.
Insider trading involves and affects each of the
above classes of agents. If insiders were allowed to trade
on their privileged information, they would of course
reap trading profits. At the same time, insiders who are
professional managers (see Box 2) may receive reduced
compensation from investors to reflect the profits managers can earn from trading.
Insider trading also affects liquidity traders, who face
the prospects of incurring losses when trading with agents

F

possessing superior information. On the other hand, if they
avoid trading, they will lose the diversification/hedging
benefits that prompt them to trade in the first place.
In addition, insider trading implies that informed
noninsiders or market professionals face informed competitors in the financial marketplace. The rivalry
between informed insiders and informed noninsiders
may drive the latter out of the market, making prices less
informative, or, by furthering competition, increase the
speed with which information is released to uninformed
traders. Insider trading has an impact on investors
through its effects on both investors’ trading profits
(when they buy and sell holdings for liquidity reasons)
and managerial incentives to create value. If insider
trading were not prohibited by law, investors, especially
large shareholders, would need to decide their firm’s
policy toward insider trading.
The legal and economic literature on insider trading attempts to weigh the trade-offs discussed above to
formulate optimal policies. Different authors focus on
different classes of actors and different types of effects.
Given the number of classes of actors involved in and
affected by insider trading and the multiplicity of
effects, differences in focus have led to rather discordant assessments of insider trading and conflicting policy recommendations. The following sections review
this literature. Because the bodies of legal and economic literature on insider trading have evolved somewhat
independently, each is discussed separately.6 A tentative synthesis of the arguments presented in the literature concludes the discussion.

Legal Scholarship
or the most part the legal literature on insider trading attempts either to support the existing scope of
10(b)-5 or argue for its elimination. A number of
rationales have been advanced within the legal community for prohibiting insider trading. These rationales fall
into three broad categories—fraud theories, fiduciaryduty theories, and information-access theories. The earliest rationalization for restricting insider trading
identified it as a fraudulent or exploitative business practice. In fact, such a perception appears to be the basis for
the court interpretation of 10(b)-5 as a prohibition on
insider trading.

F

4. For example, “passive” insider trading is something difficult to detect and not punishable. When there is favorable inside
information, insiders may continue holding on to the firm’s stock, which they would have sold otherwise. Conversely, they
may refrain from buying (more of) the firm’s stock when there is adverse inside information about the firm (Fried 1996).
5. There is some evidence that there is not a strong interest on the shareholders’ part to restrict insider trading (Seyhun 1992)
in the firms’ code of ethics. But it should also be noted that this position is against the backdrop of existing restrictions from
the SEC.
6. The division is along the line of the research but not the professional identity of the authors. Some legal literature authors
are actually accomplished economists and vice versa.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

37

B O X

2

Who Is Covered by Insider Trading Laws?
o one disagrees that directors, officers, and key
employees are corporate insiders, and there is no
legal uncertainty about whether they are covered by the
existing laws restricting insider trading. Because the purpose of this article is to look into the fundamental logic of
whether or not insider trading should be banned instead of
how to define insiders, focusing the discussion on the
clearly defined “hard core” insiders seems appropriate.
The definition of who insiders are in legal practice is
wider than the one used here. The major extension of the
definition is based on the idea of fiduciary duty and information misappropriation. Directors, officers, and key
employees of a firm bear a fiduciary duty to the firm, and any
trading based on the confidential information obtained
when they perform their corporate duties may be viewed as
(1) breaching their fiduciary duties and (2) misappropriating information that belongs to the firm. With this rationale as the essential basis for banning insider trading,
agents who are not directors, officers, or key employees of
a firm but who bear fiduciary duties to the firm, such as the
firm’s contracted lawyers, consultants, and investment
bankers, would also be banned from trading on any information about the firm they have obtained when performing
their duties.
This argument may be pushed further. If the information obtained from a firm by someone with a fiduciary duty
to the firm is not about the firm itself but about some other
firm(s), and the individual trades on such information, is
he or she liable for breach of fiduciary duty or misappropriating information?
These are debatable questions regarding the legal
definition of insiders, which future court cases are likely to
gradually clarify. Some borderline rulings hint about the
direction in which the courts are leaning. A famous case is
Chiarella v United States. Vincent Chiarella was a worker
in a financial printing company who figured out the names

N

of the acquisition target firms of the printing company’s
clients and bought the stocks of the target firms prior to
the public announcements. He was charged by the SEC
with committing illegal insider trading. In 1980 the
Supreme Court found him not guilty since he bore no fiduciary duty to the acquisition target firms and was therefore
not an insider. A more recent case in point is James H.
O’Hagan v United States. O’Hagan was a lawyer who made
$4.3 million by trading in stock options of the Pillsbury
Company after learning that a client of his law firm was
planning a takeover of Pillsbury. The SEC prosecuted
O’Hagan with insider trading based on the misappropriation theory, which argues that though he owed no fiduciary
duty to Pillsbury, his violation lay in his deceitful acquisition and misuse of information that properly belonged to
those to whom he did owe a duty: his law firm and its
clients. A federal appeals court rejected the SEC approach
as unauthorized by Congress. But when the case went
before the Supreme Court this year, the ruling was 6-3 in
favor of the SEC.1
Another related front of development is that the SEC
in 1980 began interpreting Section 14(e)-3 of the Securities
and Exchange Act of 1934 as part of its anti–insider trading
rules, making illegal the purchase or sale of a security by
anyone who is in possession of material information relating to a tender offer if such information is acquired directly or indirectly from the issuer, an officer, or any person
acting on the issuer’s behalf (Seyhun 1992). Based on this
rule, no fiduciary duty relationship or intent to defraud is
needed to convict someone who trades on the information
of a tender offer. The SEC, however, experienced a setback
in 1990 when its interpretation of Section 14(e)-3 of the
1934 Act was ruled by the court of appeals, in the Chestman
v United States case, as having exceeded its rule-making
authority.

1. The Fourth and Eighth Circuit Courts in the cases of United States v Bryan (58 F3d 933, 949 [1995]) and United States v O’Hagan
(92 F3d 612 [1996]), respectively, have rejected the extension of insider trading restrictions under the “misappropriation theory” to
agents who do not have a fiduciary relationship to the firm. But other circuits accept the misappropriation theory. The Supreme
Court split 4-4 in its consideration of the misappropriation doctrine (Carpenter v United States, 484 US 19, 24 [1987]).
38

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

The evolution of legal scholarship leading to this
view of insider trading is interesting. The common law
does not, in general, impose any duty to disclose confidential information regarding material facts on the part
of participants in voluntary transactions (Carlton and
Fischel 1983). All that is required is that agents not make
untruthful or deceptive statements regarding such facts.
Because insider trading does not require affirmative misrepresentation, it is not surprising that it was legal before
the turn of the century. At that time, concern for the
excesses of the capitalism of the Gilded Age led to disparagement of insider trading. Further, some state securities commissions attempted to restrict the practice.
Following the first prohibitions on insider trading
with the enactment of the 1934 Act, legal interpretations evolved in the 1940s through the early 1960s
through a series of court decisions arguing that insider
trading amounts to a dishonest and fraudulent practice
used by informed investors to enrich themselves at the
expense of uninformed investors and thus violates the
general prohibition of manipulative and deceptive practices under Section 10(b)-5. In these accounts, insider
trading restrictions protect “those who do not know
market conditions from those who do” (Charles Hughes
& Co. v the SEC, 139 F.2d 434, 437, 2d Cir. 1943). This
line of argument emphasizes the adverse effect of insider trading on liquidity traders who incur trading losses.
The extent of outrage felt by some scholars regarding
these losses can be gauged by the fact that, when Manne
(1966) raised economic argument favoring elimination
of insider trading restrictions, some opponents of insider trading argued that the wrong inflicted by insider
trading on uninformed investors is so great that, even if
permitting trading increased economic efficiency, the
ethical questions raised by the exploitation of uninformed investors would still weigh heavy enough to
rationalize its prohibition because the gains are due to
“unfair” behavior (Schotland 1967).
A basic question not addressed by the fraud rationale for prohibiting insider trading concerns the
assumption that exploiting informational advantages for
the purposes of security trading is unethical. All sorts of
economic agents profit from informational advantages in
a market economy, and such exploitation is not in gen-

eral viewed as unethical. Why then is exploiting an informational advantage in securities trading unethical?
This question brings the discussion to the second
set of rationales the legal literature raises for prohibiting
insider trading—those based on fiduciary duty. These
arguments emphasize the effect of insider trading on the
insider-investor relationship. The basic argument is that
the agents engaging in prohibited insider trades obtain
their information via fiduciary relationships, and trading
on this information for personal gain represents a breach
of duty. This position takes two forms. The first is a narrow variant that restricts the scope of fiduciary duty to
corporate insiders. A second, more liberal version of the
theory, frequently termed the “misappropriation theory”
(Merwin 1996), interprets prohibitions against violating
fiduciary duty as including third parties in possession of
confidential firm-specific information (for example,
contracted lawyers and accountants).
A fairly straightforward counterargument, even to
the narrow variant of the fiduciary-duty theory, can be
made based on the Coase (1960) theorem: If investors
have a property right to inside information that is violated by the expropriation of this information by insiders
for personal profit, why can investors not legally sell this
right to corporate insiders? The Coase theorem implies
that, absent regulation, any property right, including
that to inside information, will be allocated to the party,
investors or insiders, who values the right the most.7
Thus, if the harm to investors from insider trading
exceeds the profits to insiders from engaging in such
trade, the firm will voluntarily prohibit insider trading.
On the other hand, if the gains to insiders from trading
exceed the costs to the firm, trading will be permitted
since investors can sell to insiders the right to trade and
both can profit. From this perspective, contractual
arrangements within the firm, rather than government
fiat, should determine insider trading policies.
A rejoinder can be made to this Coasian argument
on the basis of transactions costs. The regulation of
insider trading by contract is costly on two accounts.
First, firms, lacking the enforcement technology available to the state (for example, surveillance of wire transfers), may find enforcement of contractual restrictions
on insider trading very expensive. They may therefore

7. The Coase theorem is a classical argument in economics. It argues that regardless of how the legal system initially assigns
property rights to agents, if trade is allowed and transactions costs are zero, agents will trade to efficient allocations. The
classic example, offered originally by Coase (1960), is that of the allocation of pollution rights between farmers and railroads. Consider two different allocations of pollution rights: railroads may or may not have the right to emit soot that damages the crops of farmers. The Coase theorem argues that, if the economic losses from emitting soot to farmers do not exceed
the economic gains to railroads from emission, emission will occur regardless of how the legal system allocates the right to
emit soot. If railroads are granted the right to emit soot, they will simply retain this right. If railroads are prohibited from
emitting soot, they will be willing to pay farmers enough to buy the right to emit soot. Conversely, if the economic losses to
farmers from the emission of soot exceed the gains to railroads, soot will not be emitted; if railroads initially have the right
to emit soot, farmers will be willing to pay enough to buy this right from them. Thus, the allocation of the property right to
emit soot affects the wealth of farmers and railroad owners but does not affect the amount of soot emitted.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

39

avoid prohibiting trading even if the costs of trading to
the firm exceed the private benefits to insiders. Second,
the parties affected by insider trading are numerous,
including both current and future owners of the firms’
shares. Even the interests of the shareholders on record
at a given point in time are not the same regarding insider trading. Investors following buy-and-hold policies
know that any losses from insider trading incurred when
selling their shares will occur in the distant future and
thus be fairly insignificant in present-value terms. Thus
they may prefer compensating managers by offering them insider trading
A basic question not
profits. Shareholders
addressed by the fraud
who plan to trade actively in the near term, with
rationale for prohibiting
some liquidity traders in
insider trading concerns
this category, may feel
the assumption that
differently. If the cost of
case-by-case contracting
exploiting informational
is high on average, a flat
advantages for the purposprohibition on such trades of security trading is
ing may improve welfare.
The third category
unethical.
of legal theories rationalizing insider trading
prohibitions moves away
from considering the effect of trading on managers and
investors and returns the focus to its impact on liquidity
traders. However, in contrast to the arguments cited
above that stress the harm to the individual uninformed
trader, scholars in this camp stress the aggregate effect of
trading on the market. These “information access” theories argue that the information possessed by insiders is
special because it either (a) cannot be legally obtained by
other investors (Brudney 1979) or (b) is residual, that is,
not produced for its own sake but rather as a by-product
of other managerial activities (ten Oeuvre 1997). Because
insiders have the only legal access to certain kinds of
information, they have an “insurmountable” advantage
when trading with other market participants. The presence of investors with such advantages renders financial
markets unfair (Brudney 1979). If information possessed
by insiders is residual, prohibiting them from profitably
trading on their information will not reduce information
acquisition activity (ten Oeuvre 1997)—even if not
compensated via insider trading profits, managers will
still produce information in the course of administering
the firm. Banning insider trading, however, by increasing
the confidence of uninformed investors may lower the
premium they require to transact and in turn lead to
more stable and liquid markets. Proponents of informational rationales argue that the primary congressional
motivation for Section 10(b)-5 of the 1934 Act was to
increase the stability and fairness of the securities mar40

ket (Brudney 1979). (The arguments against insider
trading based on its effect on market liquidity will be
revisited in a later section.)
These arguments rationalizing the prohibition on
insider trading have been countered by arguments from
legal scholars representing the “law and economics”
school of thought at the University of Chicago. Manne
(1966) provides the classic exposition of this viewpoint
in favor of insider trading. Manne advances two defenses. The first is that trading by insiders allows information to be rapidly impounded in the prices of securities.
As a result, the efficiency of capital markets increases.
Because firms use securities prices in making investment and capital budgeting decisions, increases in price
efficiency will lead to higher levels of economic output.
This argument points to the social gains from insider
trading as reasons it should not be prohibited by the
state; however, it does not explain how investors, liquidity traders, and market professionals gain from
insider trading and thus why investors should permit
insider trading. Carlton and Fischel (1983) address this
issue by noting that increased price efficiency can benefit firms by reducing investor uncertainty. They also
point out that price efficiency established by insider
trading, as opposed to direct disclosure, may better protect confidential corporate information.
Manne’s second argument in favor of permitting
insider trading holds that security trading can improve
the alignment of interests between outside claimants
and management by allowing managers to profit from
the appreciation in firm value their efforts engendered.
Of course, the salience of this argument is somewhat
muted by the obvious rejoinder, offered by opponents of
insider trading, that managers may as easily profit by
taking short positions and engendering corporate failures. Manne argues, however, that, although the security market profits may be the same for success as they are
for failure, almost all non-trade-related incentives, such
as compensation and reputation, favor engendering success and thus, given the neutrality of the trade-related
incentives, insiders would never produce “bad news”
solely in order to trade on such news. Further, as Macey
(1991) points out, even if managers had an incentive to
engage in such short trades, it would always be possible
to place broad restrictions on the direction of their trading activities, precluding trading on “bad news” (that is,
short sales) but permitting trading on “good news” and
thereby eliminating this incentive problem.
Manne’s arguments on the incentive effects have
been extended and clarified by a number of others.
Easterbrook (1985) argues that insider trading may
increase the managers’ willingness to take on risk.8 This
bias toward risk may actually be beneficial because
other factors that affect managerial proclivities toward
risk taking, such as firm-specific human capital, will bias

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

managers against risk taking.9 For example, insider trading opportunities represent an antidote to the propensity toward conservatism and excessive caution generated
by managers’ desire to protect their jobs by avoiding
risky projects. Easterbrook and Fischel (1991) argue
that in fact the current prohibition on realizing shortswing profits, even if private information is disclosed as
required by 16(b) of the 1934 Act, actually exacerbates
the conservatism of managers by forcing them to hold
large portions of their wealth in corporate stock rendered illiquid by insider trading restrictions.
Yet another extension of Manne’s argument for the
incentive-alignment effects of insider trading is provided by Carlton and Fischel (1983). They point out that
firms, when hiring managers, frequently have difficulties assessing both the talents of managers and their
willingness to take risks necessary to create economic
value. Offering contracts for compensation via insider
trading rather than fixed salaries would help distinguish the kind of high-quality managers sought and
resolve this uncertainty. In other words, by accepting
lower levels of explicit compensation in exchange for
more opportunities to engage in insider trading on personal accounts, managers would have the opportunity
to demonstrate superior abilities and a high level of risk
tolerance. Thus, making insider trading opportunities
part of the menu of contracts available when hiring
mangers could resolve some uncertainty regarding managerial attributes and abilities to the benefit of the firm.

Economic Models
icroeconomic theorists have also addressed the
issues surrounding insider trading restrictions
by formulating models of the insider trading
process. Such models, for reasons of tractability, cannot
each include all the groups of agents involved but necessarily concentrate selectively on a limited number of
the many effects of insider trading. Their analysis may
therefore lead to varied, even opposing, conclusions
about those effects, depending on the specification and
parameters of the models. On the other hand, a formal
analysis forces the researcher to make assumptions
explicit and to trace out all the latent causal effects rigorously. The possible costs and benefits to insider trading, some of which are covered in the largely informal
and heuristic treatment in the legal literature, may
hence be logically confirmed or refuted in these clearly

M

defined models, providing additional insights into insider trading and a partial foundation for rationally weighing the question of its regulation.
The effects of insider trading considered in economic models may be usefully classified into two categories—effects on aggregate economic performance and
effects on the relative welfare of different market participants. Aggregate variables considered include the liquidity of the firm’s stock, the firm’s capital cost, the
information content of the stock price, and the level of
investment. Such variables are significant only in certain
economic contexts. Liquidity, for example, is an indicator
of the ability to sell quickly, which can influence attractiveness to potential buyers. The cost of a firm’s capital
may affect its future development as it measures its capability to raise new capital. The informativeness of a firm’s
stock price is relevant to the risk of investing in the firm,
which is an important factor in determining demand for
it. The investment level indicates how many resources
are devoted to expanding the firm’s, and possibly therefore the economy’s, production capacity.
The second category of economic variables concerns the relative impact of insider trading on the different agents—that is, on who gains and who loses if
insider trading is restricted. When discussing relative
impact on welfare, any simple categorization of agents
involved in or affected by insider trading may not be perfect. In fact, each model used in the economic research
may be slightly different in this aspect. Despite this
shortcoming, the taxonomy of insiders, market professionals, liquidity traders, and investors serves as a unified if rough framework for summarizing and presenting
the existing economic literature.
Effect on Aggregate Economic Variables. Several
models imply that allowing insider trading of a firm’s
stock would reduce its liquidity but improve the informational efficiency of its price. Moreover, insider trading would raise the firm’s cost of capital. For example,
Manove (1989) assumes that a firm’s insiders have private information regarding future corporate cash flows.
Permitting insider trading increases the trading losses
that liquidity traders incur when they sell to meet liquidity needs, thereby discouraging their trading activity
and lowering the liquidity of the firm’s stock. Insiders and
investors, facing the reduced liquidity, will attach a liquidity discount when buying the firm’s stock, which drives up the firm’s capital cost. At the same time, however,

8. The argument for insider trading increasing risk-taking incentives is as follows: Prices represent unbiased forecasts of true
firm value conditioned on all publicly available information. Insiders have better information than the market and thus
can earn trading profits by buying if and only if market prices are too low given their private information. Since market
prices are not biased, the likelihood of a large difference between true value and market value is proportional to the market’s uncertainty regarding future firm value. Thus, insider trading profits will be positively related to the variability of firm
cash flows.
9. Bebchuk and Fershtman (1994) formalize the argument.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

41

insiders’ private information will be incorporated into
the stock price through their trading activity, thus
enhancing the informativeness of the stock price and
reducing the risk of owning the stock.
Most models also show that the price of a firm’s
stock will be more responsive to random changes in order
flow on stock exchanges if insider trading is permitted
because people will infer that such changes are most
likely due to insiders’ activity based on their superior
information. The increased responsiveness further
reduces the liquidity of the stock. These results are
found, for example, in models by Leland (1992), Noe
(1997), Hu and Noe (1997), Dye (1984), and Shin (1996)
but partially refuted by
Ausubel (1990) and
Fishman and Hagerty
The effects of insider trad(1992), among others.
Shin’s argument
ing considered in economic
considers the interacmodels may be usefully
tion between insiders
classified into two cateand market professionals. He points out that
gories—effects on aggremarket professionals
gate economic performance
may have also acquired
and effects on the relative
private information
regarding the firm.
welfare of different market
Competition between
participants.
market professionals
and insiders in using
their information will
influence the stock price. Allowing some insider trading
may improve the informational efficiency of the stock
price while it may also alleviate trading losses by liquidity
traders. The key consideration in his conclusion is the
“appropriate amount” of insider trading to match market
professionals’ trading. If the balance is not achieved,
then insider trading may not be as salubrious.
Although Shin (1996) considers the interaction
between insiders and market professionals, his model
does not feature market professionals’ decisions about
spending resources on information acquisition. Fishman
and Hagerty (1992) consider this factor and present a
scenario in which insider trading leads to a stock price
that is less efficient in providing information. Insider
trading may deter market professionals from acquiring
information and trading the stock and, consequently,
reduce the total amount of information impounded into
the stock price. Market professionals have the option of
becoming informed by spending resources to investigate
the firm’s prospects. Insiders have costless access to
such valuable information. When insiders’ information
is too good and is used in their trading, it may not be
worthwhile for market professionals to expend
resources. Thus under certain conditions, the presence
of insiders may discourage market professionals or
42

crowd them out, leading to less efficient stock prices.
Fishman and Hagerty also show that, when insider trading is allowed, the better the information that insiders
have, the less efficient the stock price is. For a given
aggregate amount of information acquired by all traders,
the stock price efficiency increases with the number of
market professionals. Finally, for a fixed number of market professionals, stock price efficiency is maximized
when information is evenly distributed between professionals and insiders.
Ausubel (1990) offers a contradicting point. He
argues that banning insider trading actually takes away
the incentive of insiders to hide their private information and may result in its earlier revelation, promoting
the informational efficiency of stock prices.
A firm’s investment can also be increased or
decreased by insider trading. Ausubel (1990), for example, argues that insider trading reduces investment.
Because investors expect insiders to take advantage of
them, investors may price the investment capital
accordingly, resulting in a reduction of their capital
injection. Insiders may reduce their contribution to the
capital as well because they know that the quality of the
investment deteriorates as the capital from outsiders is
lower. Banning insider trading therefore provides a
greater return on investment, which induces greater
investment by investors and in turn improves returns to
insiders and encourages greater investment by the
insiders as well.
Manove (1989) has concluded that insider trading
alters investment levels in two ways. Using an argument
similar to Ausubel’s, he suggests that insider trading
may depress investment activity below the socially optimal level. On the other hand, it may lead to wasteful
increase in investment; in his model, investors overinvest to eliminate uncertainty (this idea is discussed in
more detail in the next section). Leland’s (1992) model
also predicts that a firm’s investment will rise if insider
trading is allowed, but he concludes that such increased
investment may have some benefits.
Effect on Relative Well-Being of Agents. The welfare effects of insider trading on different agents vary.
Economic models help to identify them and serve as
guides for weighing the benefits and costs of regulation to
each group.
Economists generally agree that any informed trading, including insider trading, hurts liquidity traders, who
may be forced to trade in order to balance their portfolios
or hedge their positions but are at an informational disadvantage and inevitably lose money to insiders and
other informed traders.10 This argument is confirmed in
the analysis by, for example, Manove (1989), Noe (1997),
Fishman and Hagerty (1992), and Leland (1992). A
clever counter argument is found in Shin (1996), who
points out that insiders are not the only ones who profit

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

from information asymmetry at the expense of liquidity
traders; market professionals gain as well. According to
Shin, allowing a certain amount of insider trading may
actually alleviate liquidity traders’ losses because insiders and market professionals have to compete with each
other in exploiting their informational advantage. This
competition results not only in faster and more thorough revelation of insiders’ and market professionals’
information but also in smaller losses for liquidity
traders.
Some studies show that market professionals may
be placed in a worse position if insider trading is permitted. Unlike insiders who acquire private information
about a firm without incurring any cost other than performing their corporate duties, market professionals
obtain this information by deliberately expending effort
and money. Allowing insider trading would mean smaller returns on these expenditures and would diminish
the value of market professionals’ information (Shin
1996; Fishman and Hagerty 1992). In some cases market professionals may drop out of the market because
they cannot compete against insiders, who acquire
information at no cost (Fishman and Hagerty 1992).
Would insiders always be better served by being able
to trade on their privileged information? The answer is
not straightforward because of the strategic interaction
between insiders and investors. Insiders could be expected to trade on their unique information only if they were
able to reap trading profits. But at the same time, the
compensation they receive for performing their corporate
duties may be adjusted by investors to reflect the profits
they can earn from such trading. Investors, if in complete
or partial control of the firm’s key policies, may also take
other actions, such as adjusting the firm’s investment to
advance their own interests, that can change the welfare
of insiders. Therefore the welfare effects of insider trading
on investors and insiders are not obvious.
Leland (1992) suggests that insiders are most likely to have net welfare gains if insider trading is permitted, but the welfare effects on investors are mixed.
Among investors, those who own the firm from its inception, whom Leland calls “project owners,” will gain from
allowing insider trading while those who invest later—
“outside investors”—will be hurt. Insiders gain because
of their trading profits at the expense of liquidity traders
and outside investors. Part of the gain for project owners
is from the higher stock prices due to insider trading,
which give their original investment a higher market
value, especially if outside investors respond favorably
to a more informative stock price.

Alternatively, as mentioned above, Manove (1989)
argues that, when insider trading is permitted, investors
may overinvest in the firm to diminish insiders’ informational advantage. Increased investment may reduce
uncertainty in the firm’s prospects by increasing the
chance of its success, and the reduced uncertainty in
turn renders insiders’ confidential information less valuable. Investors have an incentive to decrease the value of
insider information because they may ultimately liquidate their own positions, and informationally advantaged
insiders could then exploit investors’ trades. The overinvestment ultimately results in a reduced return on the
firm and an increased relative share of profits for
investors, which may harm both insiders and investors.
In the same model, Manove provides another scenario, in
which investors, understanding that they will lose to
insiders when they liquidate their positions, rationally
decrease their investment because of lower actual
returns. The underinvestment in the firm hurts both
investors and insiders. Whether over- or underinvestment occurs depends on the specific parameters of the
model. If the confidential information of insiders is
about some unusual event that may not be easily diminished by overinvestment, then underinvestment is more
likely. When the insiders’ confidential information is of
some general nature, investors may overinvest to diminish insiders’ information advantage.
Ausubel (1990) has provided a technically different
model supporting the underinvestment story by Manove.
In Ausubel’s model, two groups of investors invest in the
firm at inception, one of them to process private information while the other does not. Ausubel calls the former insiders and the latter outsiders. Consequently,
outsiders have only partial control of the firm or project.
With insider trading permitted, not only outsiders but
also insiders reduce their investment, decreasing the
value of the firm and hurting the welfare of both. In
other words, insiders’ losses from diminished investor
confidence may more than offset their trading gains.
Ausubel shows that if insider trading were considered in
isolation from its impact on the initial investments by
outsiders and insiders, its welfare effect would be merely redistributional—that is, the decreased welfare of
outsiders would equal the increased welfare of insiders.
When investors are in complete control of the firm,
the question arises about how insider trading opportunities alter the compensation packages of insiders. Noe
(1997) presents a model in which the trading profits of
insiders are an inexpensive substitute for insiders’ wages
that ensure the same effort from them. That is to say,

10. Informed trading by noninsiders is generally tolerated by those opposed to insider trading under the rationale that a noninsider’s information does not result from having a special position that allows access to the firm’s information. In theory,
any investor can become an informed trader if he is willing to invest the necessary resources. The option to become an insider is not generally available.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

43

investors get a bigger share of the firm’s profit if insider
trading is allowed and the welfare of investors is
increased while the welfare of insiders is decreased. Hu
and Noe (1997) show that the information revealed
through insider trading may be so valuable to investors
that their losses to insiders may be more than compensated for. Investors can use the information to adjust
their investment portfolios optimally and modify insiders’
compensation packages to their own advantage, probably
ending up with a net welfare gain. The welfare of insiders
may also be improved because the trading profits can
arguably exceed the losses in their direct compensation.
An appropriate measure of overall economic efficiency depends on the economic context as well as on
value judgments. It is therefore difficult to define a universal measure for deciding whether overall economic
efficiency is improved or impaired by allowing insider
trading, except in the case in which the welfare of all the
agents involved increases or decreases. As discussed
above, most studies have shown different welfare effects
of insider trading on different agents. Hu and Noe have
provided a possible scenario in which both investors and
insiders, who are the only agents in their model, gain
from permitting insider trading. On the other hand,
Ausubel has provided an example of just the opposite—
that is, all agents lose when insider trading is allowed.

Policy Implications: An Illustration with
Hypothetical Cases
wo hypothetical situations illustrate how theory,
with the help of empirical research, can be translated into policy. First, consider an economy that
empirical research has identified as fast-developing,
characterized by numerous positive net-present-value
projects, a lack of experienced outside analysts, and
insiders who tend to have major stakes in firms’ ownership. In such an economy, the theoretical consensus indicates that permitting insider trading may be optimal.
Ensuring maximal price informativeness and thus optimal allocation of capital across sectors is especially
important. Given the lack of other information sources,
insider trading will have a strong positive impact on price
informativeness and thus will strongly further this goal.
Because of the abundance of good projects, an increase
in the costs of capital will have little adverse effect on
investment. Further, because insiders have major ownership stakes, their interests are closely aligned with the
other owners’ interests and thus any adverse effects of
trading in terms of agency costs can be minimal.
On the other hand, consider an economy characterized by a separation of ownership and management, a
sophisticated system of security analysis, and a mature
investment climate in which most projects return the
average market rate. In this economy prohibiting insider
trading may be optimal. The separation between owner-

T

44

ship and management implies that investors will have an
incentive to substitute cheaper compensation based on
insider trading for expensive salary packages designed to
ensure high performance. At the same time, managers
have an incentive to manipulate project returns to
increase risk. The adverse effects of insider trading on
market liquidity can decrease investment. The presence of
a sophisticated security analysis industry, at the same
time, can reduce the importance of insider trading for
market efficiency. Thus, in this case the costs of permitting
trading may be outweighed by the benefits of prohibiting it.

Conclusion
s the above discussion elucidates, the policy recommendations proposed by scholars regarding the
regulation of insider trading differ widely. The disparity can be traced to (a) differences in the criteria
used to evaluate insider trading and (b) differences in
assumptions regarding the importance of the distinct
effects of insider trading activity for overall economic
well-being. Differences in criteria for evaluating insider
trading, ethical versus economic, are apparent primarily
in the legal literature on insider trading. Some legal
scholars (for example, Schotland 1967) believe that
insider trading is immoral per se and that state regulation should thus prohibit such trading even if it is economically beneficial. It is difficult, if not impossible, to
resolve the conflicts in opinion about such issues.
Little if any of the divergence of opinion expressed in
the economic literature can be attributed to disagreements over the basic criteria for evaluating insider trading. Instead, the divergence can primarily be ascribed to
disagreement over which effects of insider trading will
have the most significant impacts on economic wellbeing. Despite these disagreements, there is a common
core of opinion regarding the effects of insider trading on
certain economic variables under some circumstances. In
fact, this common core of opinion can be summarized fairly simply in three points: (1) Whenever other informationally advantaged investors are absent or insignificant,
insider trading increases trading losses to investors and
liquidity traders and makes markets less liquid.
Otherwise, insider trading, by increasing competition
between informed investors, may assist investors and liquidity traders. (2) Unless other informed agents are
crowded out of the financial market, insider trading renders prices more informative, potentially increasing the
efficiency of investment and capital budgeting decisions.
(3) Insider trading opportunities provide low-cost, highpowered incentives for managerial performance. However,
the incentives provided are imperfect for two reasons:
insider trading encourages managers to undertake risky
activities and investors to underprovide more traditional
forms of compensation that may lead to increased managerial performance and reduced risk taking.

A

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

Policymakers’ decision to regulate insider trading
may depend on the structural characteristics of the economy. For any given economy, this issue is an empirical one
to be addressed by research. The results of this research
could then be effectively combined with the conclusions
of theory to produce practical policy guidelines.
In general, deriving policy from theory will not be as
easy as it was in the two hypothetical examples given. It
may be difficult to assess empirically the values of the
key variables, or the estimated value of the variables

considered in isolation may point toward conflicting
policies. However, this discussion does show that the
voluminous literature on insider trading identifies a
number of important variables to be considered when
formulating policy on insider trading. Designing such
policy requires a detailed assessment of the structure of
the economy, some sensitivity to cultural attitudes
toward the appropriateness of such trading activity, and
careful consideration of the enforcement costs associated with regulating trade.

REFERENCES
AUSUBEL, LAWRENCE. 1990. “Insider Trading in a Rational
Expectations Economy.” American Economic Review 80
(December): 1022–41.

KEOWN, ARTHUR J., AND JOHN M. PINKERTON. 1981. “Merger
Announcements and Insider Trading Activity: An Empirical
Investigation.” Journal of Finance 36 (September): 855–69.

BEBCHUK, LUCIAN ARYE, AND CHAIM FERSHTMAN. 1994. “Insider
Trading and the Managerial Choice among Risky Projects.”
Journal of Financial and Quantitative Analysis 29,
no. 1:1–14.

LELAND, HAYNE E. 1992. “Insider Trading: Should It Be
Prohibited?” Journal of Political Economy 100, no. 4:859–87.
MACEY, JONATHAN R. 1991. Insider Trading: Economics,
Politics, and Policy. Washington, D.C.: AEI Press.

BRUDNEY, VICTOR. 1979. “Insiders, Outsiders, and Informational Advantage under the Federal Securities Law.”
Harvard Law Review 93:322–76.

MANNE, HENRY. 1966. Insider Trading and the Stock Market.
New York: Free Press.

CARLTON, DENNIS, AND DANIEL FISCHEL. 1983. “The Regulation of
Insider Trading.” Stanford Law Review 33:857–95.

MANOVE, MICHAEL. 1989. “The Harm from Insider Trading and
Informed Speculation.” Quarterly Journal of Economics 104
(November): 823–46.

COASE, RONALD H. 1960. “The Problem of Social Cost.” Journal
of Law and Economics 3:1–44.
DYE, RONALD. 1984. “Insider Trading and Incentives.” Journal
of Business 57 (July): 295–313.
EASTERBROOK, FRANK. 1985. “Insider Trading as an Agency
Problem.” In Principals and Agents: The Structure of
Business, edited by John Pratt and Richard Zechhauser.
Cambridge, Mass.: Harvard University Press.

FISHMAN, MICHAEL, AND KATHLEEN HAGERTY. 1992. “Insider
Trading and the Efficiency of Stock Prices.” RAND Journal
of Economics 23 (Spring): 106–22.
FRIED, JESSE M. 1996. “Towards Reducing the Profitability of
Corporate Insider Trading.” Harvard Business Law School
Discussion Paper No. 195.

———. 1987. “Regulation on Demand: A Private Interest
Model with an Application to Insider Trading Regulation.”
Journal of Law and Economics 30 (October): 311–52.

MEULBROEK, LISA K. 1992. “An Empirical Analysis of Illegal
Insider Trading.” Journal of Finance 47, no. 5:1661–98.
NOE, THOMAS. 1997. “Insider Trading and the Problem of
Corporate Agency.” Journal of Law, Economics, and
Organization, forthcoming.

EASTERBROOK, FRANK, AND DANIEL FISCHEL. 1991. The Economic
Structure of Corporate Law. Cambridge, Mass.: Harvard
University Press.

HADDOCK, DAVID D., AND JONATHAN R. MACEY. 1986. “Coasian
Model of Insider Trading.” Northwestern University Law
Review 80:1449–72.

MERWIN, JAY G., JR. 1996. “Misappropriation Theory Liability
Awaits a Clear Signal.” The Business Lawyer 51 (May):
803–23.

SCHOTLAND, ROY A. 1967. “Unsafe at Any Price: A Reply to
Manne.” Virginia Law Review 53 (November): 1425–78.
SEYHUN, NEJAT H. 1992. “The Effectiveness of the InsiderTrading Sanctions.” Journal of Law and Economics 35
(April): 149–82.
SHIN, JHINYOUNG. 1996. “The Optimal Regulation of Insider
Trading.” Journal of Financial Intermediation 5 (January):
49–73.
STEWART, B. JAMES. 1991. Den of Thieves. New York: Simon
and Schuster.
TEN OEUVRE, J.E.A. 1997. “Insider Trading and the Dual Role
of Information.” Yale Law Review 106 (January): 1325–30.

HU, JIE, AND TOM NOE. 1997. “Insider Trading, Costly
Monitoring, and Managerial Incentives.” Federal Reserve
Bank of Atlanta Working Paper No. 97-2, May.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

45

Financial Development
and Growth
ZSOLT BECSI AND PING WANG
Becsi is an economist in the regional section of the
Atlanta Fed’s research department. Wang is a professor in the Department of Economics, Pennsylvania
State University. They are grateful for valuable comments and suggestions from Tom Cunningham,
Aruna Srinivasan, and Larry Wall.

F
THAN

10

INANCIAL INTERMEDIATION PLAYS AN IMPORTANT ROLE IN ECONOMIC ACTIVITY.
MANCE BY THE FINANCIAL SECTOR CAN BE VERY COSTLY FOR SOCIETY.
ING CRISES IN VARIOUS COUNTRIES IN THE

1970S

AND

1980S,

FOR

POOR PERFOR-

EXAMPLE, BANK-

PROVOKED BY DEREGULATION

OF BANKS, SOMETIMES EXACTED A HIGH COST, IN MANY CASES ESTIMATED TO BE GREATER

PERCENT OF GROSS DOMESTIC PRODUCT

(GDP).1 ON

THE OTHER HAND, A HEALTHY BANKING

SECTOR HAS BEEN ASSUMED TO CONTRIBUTE TO THE GROWTH OF THE ECONOMY.

BAGEHOT ([1873] 1991)

AND

SCHUMPETER ([1911] 1936)

ECONOMISTS

SINCE

HAVE THOUGHT SO, AND UNTIL RECENTLY

ECONOMISTS DID NOT SERIOUSLY QUESTION THE LINKAGES BETWEEN THE FINANCIAL AND REAL SIDES OF
THE ECONOMY.

THIS

PROPOSITION HAS BECOME FAR MORE CONTROVERSIAL THAN MIGHT BE EXPECTED,

THOUGH, AND LATELY BOTH THEORETICAL AND EMPIRICAL WORK HAS BEEN REDIRECTED AT THIS ISSUE.

THIS

ARTICLE PROVIDES A CRITICAL SURVEY OF SOME IMPORTANT THEMES AND STUDIES IN THE LITERA-

TURE OF FINANCIAL INTERMEDIATION AND GROWTH.2

In recent years with the renaissance of interest in
growth theory by economists there has been a reappraisal
of factors that matter for growth. Traditional growth theory says that as capital grows diminishing returns set in
and long-term growth is determined by factors other than
capital, such as technological progress, that are independent of policy intervention. Thus, growth theory at least
as it applied to policy analysis was effectively dead in the
water. But in the mid-1980s economists took a new look
at the determinants of growth by broadening the notion
of capital to include knowledge, technological progress,
and the like. Under this definition diminishing returns
46

can take a long time to materialize and growth is susceptible to many influences, including policy.3
The justification for this methodological shift was
twofold (Barro and Sala-i-Martin 1995). First, as an
empirical matter, traditional growth theories could not
explain the variety of countries’ long-term growth experiences. Second, since even small differences in growth
rates upheld over generations will cause appreciable differences in living standards, finding policies that mattered became crucial.
Economists have examined various explanations
for growth, including the role of financial intermedi-

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

aries. This emphasis has accompanied big strides in
understanding financial intermediaries. Presumptions
that financial intermediaries matter for real economic
activity were contradicted by theoretical models showing that they are inessential to a benchmark world of no
market frictions. Modern theories explicitly show how
financial intermediaries overcome market frictions and
lower the cost to society of transferring information or
wealth between households and firms. Many of the
arguments boil down to the idea that in one way or
another financial intermediaries give individuals or
firms access to economies of scale that they would not
have otherwise. Thus, intermediaries enhance economic efficiency and ultimately growth because they help
allocate capital to its best possible use. How well financial intermediaries perform their functions and
enhance efficiency helps explain differences across
countries’ economic performance.
After discussing frictions that give financial intermediaries an important efficiency-enhancing role (and
frictions such as government intervention that may constrain financial efficiency), this article shows how these
roles can be integrated into modern theories on growth.
Most importantly, this article provides an illustrative
model that is meant to capture current thinking about
the ways in which financial intermediaries affect growth.
This simple model shows how households, firms, and
financial intermediaries interact to determine equilibrium growth rates and various interest rates and rate
spreads. It is used to discuss the effects of financially
repressive policies such as reserve requirements, interest
rate controls, directed credit flows, and entry limitations
such as barriers to interstate banking, conventionally
termed financial repression (McKinnon 1973).
Finally, the discussion briefly surveys some of the
recent empirical literature on growth and financial intermediation. This literature has shown that different measures of financial development are positively correlated
with economic growth rates. Although there have been
some initial attempts to quantify the effect of financial
repression, more work needs to be accomplished before
precise policy recommendations can be made. Also, some
puzzles remain. For instance, it has been shown that the
effect of financial intermediation on growth becomes
weaker as countries become more developed, perhaps
because of problems with measuring financial develop-

ment or because financial intermediaries actually have
larger effects in less-developed countries than in moredeveloped ones. Other empirical work suggests that financial intermediaries differ across countries in their cost
efficiency and the degree of competition, all of which
might affect their roles in economic efficiency and growth.
It may be noteworthy to remark that for several
reasons this article uses the terms financial intermediation and banking interchangeably. As a practical matter financial development is usually measured by
banking variables because of a lack of alternative data.
In addition, banks are usually the most important form
of intermediation in both less-developed and developed
countries. Finally, as stressed by Stiglitz (1993), the
focus on banks emphasizes primary capital markets in
which new capital is raised as opposed to secondary
capital markets on which claims are traded.4 Primary
capital markets are directly linked to economic development, and most primary capital functions in developing countries are performed by banks.

Efficiency Arguments for Financial Intermediation
n a world of perfect competition, perfect information, and no market frictions, there would be no role
for financial intermediaries. Individuals could take
their savings and invest them in projects and firms with
payoffs that are optimal given individuals’ time horizons
and preferences. Even with uncertainty, financial intermediaries are unnecessary. Financial markets could be
created that would provide funds for firms at one point
in time in return for repayment at another. These markets could be specialized further to trade contracts that
exchange funds subject to all imaginable types of contingencies. Such markets would provide efficient diversification of risks. In this benchmark world the efficient
markets hypothesis, where prices reflect all available
information (Malkiel 1992), holds as well as the famous
Miller-Modigliani theorem (1961) that says real economic decisions are independent of the methods of
financing, thus leaving only a passive role for the financial sector, as shown in Fama (1980).
Presence of Market Frictions. This perfect world is
built upon unrealistic assumptions, of course. The key to
explaining why intermediaries exist is to introduce
imperfections or frictions into this world. Doing so means
relaxing the assumptions of perfect information, perfect

I

1. See Goldstein and Turner (1996) for a discussion of these banking crises.
2. This article has benefited from many excellent surveys that emphasize different aspects of the literature. In particular, see
Galetovic (1994), Pagano (1993), and Levine (1997) as well as Allen and Gale (1994), Arestis and Demetriades (1997),
Bhattacharya and Thakor (1992), Demetriades (1997), Gertler (1988), and Greenwood and Smith (1997).
3. This is the so-called endogenous growth theory, which is developed by Romer (1986) and Lucas (1988).
4. Of course, abstracting from securities markets (such as the important role of initial public offerings for the success of Silicon
Valley start-ups) simplifies things greatly. See Boot and Thakor (1997), Greenwood and Smith (1997), Levine (1997), and
Stiglitz (1996) for some current theoretical arguments on the relationship of financial intermediaries and financial markets.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

47

competition, or frictionless markets and showing how
intermediaries can improve on the outcome. When conditions are less than perfect, economic exchange is costly, and if it is sufficiently costly it may not occur at all.
Financial intermediaries make these exchanges affordable, thus offsetting the underlying market frictions.
Though no single general model explains why banks
exist, the fundamental frictions that give rise to financial
intermediaries can be classified into two categories, as
either a technological friction or an incentive friction, discussed in more detail
below. Technological
frictions prevent individuals from having
access to economies of
In a world of perfect
scale. In other words,
individuals are preventcompetition, perfect infored
from activities that
mation, and no market
would be cheaper per
frictions, there would
person if more people
participated in the
be no role for financial
activity. Incentive fricintermediaries.
tions occur because
information is costly and
individuals are differentially informed and act
in their self-interest,
and contracts are incomplete because not all contingencies
can be spelled out, not every action is accountable, and
because the specific legal environment matters.
Reducing Technological Frictions. The role of
financial intermediaries in overcoming technological
frictions was introduced by Gurley and Shaw (1960). In
their analysis financial intermediaries transform primary securities issued by firms into indirect financial securities desired by final investors. Financial intermediaries
transform bonds and stocks issued by firms into demand
or savings deposits for households. They help transform
savings into investments by repackaging wealth and
transferring capital and information.
One way in which intermediaries help individual
savers is by giving them access to large investment
projects via the so-called funds-pooling mechanism.
Individual investors might be too small to be able to afford
securities issued by firms, especially if these cannot be
divided into small affordable units. By pooling together
the funds of many small savers, financial intermediaries
overcome this indivisibility of firms’ securities. Because it
is less costly for financial intermediaries to transform
securities, gather funds and pool them, and invest those
funds on behalf of savers than it is for individuals to hold
securities issued by firms directly, financial economies of
scale arise. Thus, financial intermediaries improve the
efficiency of the economy by letting savers invest in large
projects and making more of these projects possible.
48

Another important way in which intermediaries benefit small savers is by making riskier investments available to them via what is called the risk-pooling
mechanism. Although riskier projects tend to yield higher returns than low-risk projects, individuals might not
want to take on much risk when their available funds are
too small to effectively insure themselves. Intermediaries
can provide the risk-reducing benefits of diversification
by holding a portfolio of loans to many entrepreneurs of
all different types of risk and giving depositors less risky
claims against it. The intermediary can offer this service
at lower cost than savers can manage individually. Savers
therefore have access to economies of scale not otherwise available to them.
An intermediary can also help investors by providing
access to long-term projects through liquidity management. Some projects require long-run commitment of
capital because they can take a long time before yielding
results, but savers have shorter and uncertain time horizons. They do not want their funds tied up and, as a precaution in case of unexpected demands, prefer to have
investments that are liquid. While long-term projects
often yield higher returns, they can be costly because
they can be illiquid, or hard to turn into quick funds.
Intermediaries can provide liquidity for these investments by pooling savers with different liquidity needs, in
essence diversifying across liquidity risks and thus giving
savers access to the higher returns.5 In other words, pooling provides financial economies of scale by reducing the
cost of illiquid investments; savers gain by the lower
costs, and efficiency is enhanced because more investment will occur in long-term projects.
A final way in which an intermediary can improve
investors’ access to worthwhile investments is by means
of the so-called screening mechanism. Savers by themselves usually have too little time and income to inform
themselves about all good and bad investment opportunities. Doing so would require searching, collecting, and
then processing information on firms, managers, economic conditions, and so on. Because of economies of
scale, financial intermediaries can collect large amounts
of information, provide expertise in evaluating, screening, and sorting prospects, and monitor firms’ actions at
a lower cost than individuals can. By economizing on
information acquisition costs, financial intermediaries
help capital move to its highest value, thus improving
allocative efficiency (see Diamond 1984 and Boyd and
Prescott 1986). They can provide information from not
only intermediaries who hold savers’ funds but also
investment analysts, credit rating agencies, auditing
firms, and other institutions.
In summary, scale economies imply intermediation
that improves allocative efficiency, or the degree to
which resources flow to the most productive investments.
Without intermediation, individuals generally would not

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

have the means to finance a single firm or exploit scale
economies in monitoring. Financial intermediaries provide individuals access to economies of scale by pooling
their funds and creating small-denomination securities
that allow households to hold diversified portfolios,
invest in firms with economically efficient scales, and
increase their asset liquidity. Without this pooling of capital from savers, many businesses would be constrained
to economically inefficient scales and time horizons.
Thus, financial intermediaries overcome frictions,
improve resource allocation, and bring markets closer to
the efficient-markets view of the world.6
Reducing Incentive Frictions. The foregoing arguments for financial intermediation are based on the
assumption that there are no conflicts of interest in the
behavior of savers and firms. However, it is not always in
firms’ best interest to reveal all, and investors typically
have less-than-complete information about firms. With
asymmetric information, conflicts of interest are possible. Thus, if financial contracts were to apply equally to
different types of firms, adverse selection might occur in
that only firms of lower quality would demand the contracts. Higher-quality firms would stay away from the
contracts because they would not get terms that reflected their higher quality. There are also issues of moral
hazard because it is not always in firms’ best interest to
behave honestly. For example, managers may not report
whether they are dillydallying or whether the firm is pursuing risky or questionable strategies, nor do they always
truthfully reveal how their projects turned out.
Considerable resources are required for monitoring
firms in order to ensure practices and decisions in
savers’ best interests. Financial intermediaries can help
reduce problems associated with asymmetric information or moral hazard by offering financial contracts that
are not available in markets and providing economies of
scale in monitoring and control.
Markets sometimes fail to resolve incentive problems efficiently because, as noted above, for individuals
the amount of time needed to monitor performance and
control behavior is too costly. Individuals may rely on
publicly available information transmitted though markets rather than gathering information themselves, and
as a result fewer individuals may make loans. In addition, firms may hold back from borrowing because it
would increase the likelihood of their being monitored
by lenders. Reconciling differing incentives is costly,
and when it is required, resources are diverted from
investment projects themselves. Financial intermedi-

aries perform an important role in mediating divergent
incentives between lenders and borrowers that arise
from imperfect information and incomplete contracts.
Information theories emphasize what is known as
the monitoring and control role of banks. In the monitoring function intermediaries collect information to verify
desirable behavior and compliance with covenants. They
also use this information in their control function to
improve performance under the contract terms by punishing undesirable behavior and collecting from borrowers who do not repay in full on time. Diamond (1984)
shows that households delegate financial intermediaries
as monitors to take an active role in firms’ activities to get
information and maintain discipline to avert incentive
problems. He argues that there are economies of scale for
monitoring and controlling firms. A single financial intermediary can perform these duties at least as effectively as
many individual lenders and more cheaply because effort
is not duplicated. Some information theories stress the
role of commitment and emphasize the role of banks in
offering financial contracts not available in competitive
markets. Mayer (1988) observes that intermediaries
make long-term relationships possible by devising contracts that ensure that firms fulfill their commitments.
To summarize, both asymmetric information and
incompleteness of contracts create incentive frictions
that potentially cause savers’ and firms’ behaviors to be
incompatible. Financial intermediaries provide contracts and discipline that enhance the economy’s ability
to achieve efficient risk sharing.

Imperfect Efficiency
t has been argued that financial intermediaries
improve economic efficiency by overcoming frictions,
thereby giving households access to economies of
scale that they could not attain by themselves. How well
financial intermediaries perform this role affects economic performance. However, markets in which there are
economies of scale do not always work well because by
nature they tend to be imperfectly competitive. With
economies of scale larger firms tend to be more efficient
or have lower average costs than smaller firms, so there
is a tendency for firms to grow large relative to the size of
the market. Firms that achieve efficiencies ahead of their
rivals will gain a competitive advantage they can exploit
by continuing to stay ahead, growing ever larger until
only a small number are left to dominate the market.
In this way scale economies also imply market power
and noncompetitive pricing. Market power may also arise

I

5. Diamond and Dybvig (1983) show that individuals have different liquidity needs, but verifying them is prohibitively costly so that banks cannot write insurance contracts. Banks offer liquid deposits to savers and have the right mix of liquid and
illiquid investments that provides complete insurance to savers.
6. Allen and Gale (1994) show that institutions that arise as a result of the presence of various frictions do not bring conditions completely back to a frictionless Miller-Modigliani world.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

49

through specialization by financial intermediaries. For
example, private or inside information can allow financial
intermediaries to differentiate themselves from rivals.
Knowing more than their competitors, innovative firms
can adjust the quality of products or services or create
new products and markets and temporarily gain some
market-pricing power.
Markets with economies of scale and product differentiation may still work relatively efficiently. Baumol,
Panzar, and Willig (1982) show, for example, that when
barriers to entry are low,
markets are contestable
in the sense that the
threat of potential competition will force exFinancial intermediaries
isting firms to act competitively. However,
enhance economic efficienentry into the financial
cy and ultimately growth
services sector is made
because they help allocate
difficult by the costs of
required investments
capital to its best possible
in structures and
use.
equipment but perhaps
more importantly by
investments in obtaining proprietary information, developing or
hiring expertise in monitoring and making loans, and
establishing a good reputation (Vives 1991). Diamond’s
(1984) analysis suggests that new banks are implicitly
discouraged from entering the market by the fact that
larger banks can do a better job of diversifying risks
than can smaller banks. Sussman and Zeira (1995)
model another potential barrier in showing that banks
have local economies of scale with advantages for monitoring the closer they are to their clients, advantages
especially in making loans to smaller businesses. In
addition, Petersen and Rajan (1995) argue that commitment models imply efficiency in markets with economies
of scale but also that there is less competition because
banks gain an information monopoly over firms to whom
they have made prior loans, and they eventually exploit
their positions.
Perhaps more importantly than economies of scale,
legal constraints and government regulations affect the
efficiency of financial intermediaries. Governments
have historically and extensively intervened in the
banking sector. One rationale for intervening is that
markets may fail to work well due to the presence of
frictions, and another is that intervention promotes
financial stability by lowering the probability of bank
failure. Historically, through legal restrictions on entry
and allowing collusive activities, governments have
tended to make banking markets less contestable and
thus less competitive. For instance, the United States
50

until recently has had restrictions on entry (interstate
banking rules) and branching that made markets less
contestable. Other highly developed and less-developed
countries allow or have allowed collusion between
banks. Bingham (1985) shows that, except for Italy,
Switzerland, and the United Kingdom, industrial countries have regulated or allowed collusion that distorted
deposit rates. Such governmental actions affect the efficiency of credit allocation. In some Latin American
countries governmental controls on interest rates and
credit allocation have been very severe, and economists
since McKinnon (1973) and Shaw (1973) have worried
about the costs of repressed financial sectors on growth
and efficiency.7

Background to Modeling Growth and
Financial Intermediation
he previous section explained that efficient
financial intermediation helps channel resources
toward activities with high rates of return. More
financial resources go to profitable projects that are
larger in scale, longer in maturity, and with riskier
prospects than if intermediation were less efficient.
Likewise, efficient intermediation means that investment costs less so that savings transformed into investment go further. Finally, efficiency also means that
information is processed well, allowing good investment
opportunities to be identified and then helping ensure
that businesses act in ways that do not conflict with
savers’ interests.
Early Development of the Ideas. Some of these
ideas are not new but have been put forth by classical
economists to understand and compare the performance
of various financial intermediation systems. The roles
played by financial intermediaries and the services provided in enhancing efficiency have been used to explain
the economic growth of some countries. One of the earliest to connect finance and growth was Bagehot, who
argued that financial intermediation was critical for the
rapid industrialization of England in the early nineteenth century: “Political economists say that capital
sets towards the most profitable trades, and that it
rapidly leaves the less profitable and non-paying trades.
But in ordinary countries this is a slow process. . . . In
England, however, . . . capital runs as surely and instantly where it is most wanted, and where there is most to be
made of it” ([1873] 1991, 6). In other words, information
flowed rapidly and was used to divert funds from poorquality investments to high-quality investments, thus
enhancing the overall efficiency of investment.
Bagehot also stressed the importance for growth
and development of readily accessible pools of funds
that are sufficiently large to allow risky and large-scale
projects: “We have entirely lost the idea that any undertaking likely to pay, and seen to be likely, can perish for

T

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

want of money; yet no idea was more familiar to our
ancestors, or is more common now in most countries. A
citizen of Long in Queen Elizabeth’s time . . . would have
thought that it was no use inventing railways, . . . for you
would not have been able to collect the capital with
which to make them. At this moment, in colonies and all
rude countries, there is no large sum of transferable
money; there is no fund from which you can borrow, and
out of which you can make immense works” ([1873]
1991, 20). Bagehot’s argument, intriguingly, also implies
that financial development may be necessary for inventions that lead to growth.
Schumpeter is more explicit on this point, writing
that financial intermediaries promote growth by identifying and redirecting funds toward innovative projects.
“The banker . . . stands between those who wish to form
new combinations and the possessors of productive
means. . . . He makes possible the carrying out of new
combinations, authorizes people, in the name of society
as it were, to form them” ([1911] 1936, 74). He continues, “The essential function of credit . . . consists in
enabling the entrepreneur to withdraw the producers’
goods which he needs from their previous employments,
by exercising a demand for them, and thereby to force
the economic system into new channels” ([1911] 1936,
106). Schumpeter also stresses that “the relation . . .
between . . . credit creation by banks and innovation . . .
is fundamental to the understanding of the capitalist
engine” (1939, 111).
On the New Growth Theory. Modern theories of
growth build on the ideas provided by classical economists like Adam Smith and Joseph Schumpeter, such as
technological progress through specialization and innovation and the role of market power as an incentive for innovations.8 In traditional growth models, capital was
narrowly defined as physical capital only. Capital goods
are inputs into production that are themselves produced
goods or reproducible; other inputs such as land and labor
are not capital by this definition. In these models, as the
economy grows and capital accumulates, diminishing
returns to capital inevitably set in so that further increases of output can be achieved only by ever-larger investments. As the stock of capital rises over time the returns
to capital fall until investment is no longer profitable.
Thus, investment-led sustained growth is not possible.
Growth is determined by technology and demographics,
both of which were assumed to be exogenous, or forces
not determined in the economy but acting on the econo-

my, although it was generally recognized that this was a
simplifying assumption. In essence, growth was assumed
rather than explained. Furthermore, these models were
inadequate on empirical grounds, predicting global convergence of economies that is counter to the evidence
from cross-country studies.
Modern endogenous growth models avoid the problems of earlier theory by broadening the definition of
capital to include human capital, intangible capital
such as knowledge, and other things that yield quality
enhancements to inputs and output from within the system. The critical innovation is that for the expanded
notion of capital there need not be diminishing returns
(at least over long periods), although there may still be
diminishing returns to individual capital inputs. Thus,
as capital rises, returns will not fall to the point where
investment is unprofitable, so investment continues
and sustained growth is possible.
The endogenous growth literature has explored several forces that offset the propensity for diminishing
returns to capital. Explanations that have gained attention recently include knowledge and allow technological
progress to be endogenous. One explanation is that there
are spillover effects to investments in capital, broadly
defined, that prevent the returns to investments from
falling (Romer 1986). Firms learn from the process of
making investments. Workers’ learning on the job creates
knowledge that becomes publicly available and spills
over to other firms. Thus, “learning by doing” increases
the stock of knowledge and human capital, offsetting the
tendency for diminishing returns. Another explanation is
that imperfect competition in markets for innovative
goods allows firms temporarily to earn above-normal
profits, encouraging technological progress and sustained growth. Innovation shows up as increases in the
quality of goods and inputs or as specialization. With the
ability to differentiate their product, innovative firms
achieve market power over their prices. The incentive to
innovate is to gain market power in order to earn the
above-normal profits until competitors catch up. The
desire to get ahead and then stay ahead is self-reinforcing and leads to sustained growth. Simply put, it is the
battle over market share and profits through constant
innovation that propels economies forward.
Linking Financial Activity to Economic Growth.
While innovation and knowledge creation are the forces
behind capital accumulation and growth, financial
intermediaries can also affect the process.9 To the

7. See Espinosa and Hunter (1994) and Roubini and Sala-i-Martin (1992) for some recent views.
8. Schumpeter expresses this point succinctly in his statement that “without development there is no profit, without profit no
development” ([1911] 1936, 154). While he was referring to economic development, this argument applies as well to financial development.
9. The role of venture capitalists in Silicon Valley is a good example of how innovation can lead to growth and financial intermediaries can facilitate innovation.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

51

extent that innovation and knowledge creation are
financed with external funds, growth will depend on the
efficiency of the financial sector, which is directly related to the extent to which financial economies of scale
have been realized and on how developed the sector is.
Growth will also depend on how well financial intermediaries can increase the quality of aggregate investment
by enhancing profitable opportunities, which is accomplished partly through the information role of intermediation, the monitoring and control functions discussed
previously. But it also importantly depends on how well
technological frictions are overcome. For instance, better risk diversification means that more funds can be
shifted to long-term projects. However, even if few of
the important activities for growth are externally
financed, financial intermediaries can have an effect
because they transform the quality of investments and
also because the information side of the business also
adds to the common stock of knowledge and human
capital. Ultimately the spillover effects will lower the
costs and raise the quality of investments. Finally,
financial development is also driven by innovation, and
financial innovation develops from the same incentives
that drive technological innovation, namely, the possibility of temporary market power and above-normal
returns. In other words, noncompetitive forces are central for innovation in the financial sector.10
Recent theoretical work on financial activity and
growth emphasizes that the emergence of financial intermediation spurs higher growth. For instance, Greenwood
and Jovanovic (1990) highlight financial intermediaries’
risk-pooling and monitoring functions. By pooling savings
for diversified investment projects and by monitoring the
behavior of the borrowing firms, banks ensure higher
expected rates of returns to promote growth. Saint-Paul
(1992) considers similar portfolio diversification via the
stock market. In both models financial intermediation
costs are fixed or less than proportional to the volume of
intermediated funds, and economic growth and financial
development reinforce each other while raising welfare.
Bencivenga and Smith (1991) follow Diamond and
Dybvig (1983) to elaborate the liquidity management
role of banks. Financial intermediaries reduce lowreturn investment due to premature liquidation and redirect funds into longer-term, high-yield projects, leading
to faster growth. Levine (1991) incorporates both portfolio diversification and liquidity management aspects to
show the role of financial intermediaries in pooling consumers’ liquidity risks via the securities market, concluding that setting up a stock market is growth
enhancing. Chen, Chiang, and Wang (1996) generalize
Schumpeter’s view to show that by allowing for a more
sophisticated and specialized production process, financial intermediation results in more investment projects
and spurs economic growth. Thus, despite very different
52

channels through which the real and the financial sector
interact, a consensus that betterment in financial markets is associated with faster real growth has developed.

A Simple Model
o understand how financial intermediation can be
modeled in an endogenous growth framework, this
study sketches a simple model. The model shows
how decisions of households, firms, and financial intermediaries interact to determine growth rates and real
interest rates. Making several simplifying assumptions
allows focusing on critical components. For instance,
capital is broadly defined as including physical and
human capital, knowledge, quality, and so on. The model
also does not deal with uncertainty and international
capital markets, although these could easily be incorporated into it. Finally and possibly most critically, the
model analyzes only long-run effects and not short-term
and transition effects that also can be very important.
This highly simplified model is designed to show
(1) when there is a positive relationship between financial development and economic growth and (2) how
changes in tastes, technologies, and an array of financial and monetary policies affect the endogenous
growth rate, the loan and deposit rates of interest, and
the spread between the two rates. Results indicate that
betterment in financial services either by improving the
productivity of the real sector via a variety of channels
or by reducing the cost of financial intermediation
enables faster growth. A higher cost of financial intermediation causes the equilibrium growth rate to
decrease, the deposit rate to fall, and the loan-deposit
interest rate differential to widen. A financial innovation that increases the productivity of firms raises the
growth rate and both interest rates but reduces the
interest rate differential.
The model also suggests some useful policy implications. When a ceiling is imposed on loan rates, growth
tends to fall and both the deposit and loan rates are
lower, with the differential widened. A cap on deposit
rates or an increase in the reserve requirement ratio
reduces the growth rate and the deposit rate and leads
to a higher interest rate differential. A limit on the entry
of banking firms may have ambiguous effects on economic growth and deposit rates, but it surely raises loan
rates and the interest rate differential.
A Sketch of the Analytic Structure. For the analysis it is important to look at the equilibrium conditions
from the three important sectors: firms, households, and
financial intermediaries. The first sector discussed is the
production sector. The assumption is that firms have a
linear production function—that is, increases of inputs
cause output to increase proportionately. This assumption means that there are no diminishing returns to scale,
capturing a central element of endogenous growth theo-

T

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

ry. In other words, the marginal product of capital, designated by the term A, does not fall when capital inputs are
increased. It is also assumed that firms act competitively. They maximize profits by setting the marginal productivity of capital equal to the (marginal) cost of capital,
which is the real interest rate they must pay to the owners of capital, or the real loan rate, RL. Thus,
RL = A.

premium and denoted by the product G*P, where G is the
per capita growth rate of consumption and P is the rate
at which this premium affects households’ preferences
over time. With this notation, the condition for optimal
consumption growth and savings can be written,
RD = I + P*G,

(2)

D

(1)

To concentrate on the finance and growth nexus it is
assumed that the productivity of capital, A, is not affected by the firm’s stock of capital or by the economy’s
growth rate but can be enhanced by technological
progress and innovations. The financial sector and the
government can also affect the productivity of capital.
For example, as is argued earlier, financial intermediation can raise the quality of investments, meaning that
productivity, A, will rise. The government also can affect
firms’ choices by affecting productivity directly or indirectly through regulations on intermediaries. For instance,
taxes on capital will tend to lower the productivity of capital broadly defined while government investments in
infrastructure will tend to raise firms’ productivity.
The household sector is also very straightforward.
Households choose consumption and savings to maximize their lifetime well-being. Although households’
employment, education, and fertility choices are very
important, again the model abstracts from them for the
sake of simplicity. Households adjust their consumption
across time and their savings until they are indifferent
between consuming more today and saving more, that
is, until the market return on their savings is equal to
the return they require to sacrifice current consumption for future consumption.
The required return on savings is made up of two
parts. The first part depends on how impatient individuals are no matter what their consumption pattern is.
More impatient people require a higher return, designated by I, to give up current consumption in exchange for
savings. The second part is that individuals require a premium to save more than they normally would. In other
words, to get individuals to increase the rate at which
consumption grows through additional savings they have
to be compensated. This variable is called the growth

where R is the households’ gross rate of return on
deposits, which with banks is the market rate of return to
saving.11 Again for simplicity, the focus is limited to
deposit holdings. Other forms of savings can be easily
incorporated but do not add much insight at this level of
generality. Also, note that demographics may influence
preferences and government may also play a role in households’ willingness to save. For instance, an older society
may be more impatient to consume. Alternatively, taxes on
household income may raise the growth premium.
Adding financial intermediaries or a banking sector
is just a matter of including a condition that relates loan
rates to deposit rates. Profit-maximizing banks adjust the
volume of their loans and deposits until the rate differential is just equal to the unit cost of intermediation,
denoted by C. Thus the optimality condition is
RL = RD + C.

(3)

The determinants of the costs of financial intermediation are important in the analysis to come. Some of
these costs are normal costs of running a business, such
as administrative costs, and others are unique to financial intermediation, such as the cost of information
gathering. Regardless of whether the costs arise from
banks overcoming technological or incentive frictions,
they may depend on the scale of loan or deposit activities. Larger banks may be more efficient and better able
to diversify loan risks or the risks of early withdrawals,
both of which cause the loan-deposit rate differential to
fall. On the other hand, banks that are larger, especially
relative to a small or local market, may have some market power to set either loan or deposit rates. The result
is a rise in the loan differential, which can be proxied by
a rise in the cost of intermediation.12 The cost of intermediation may also depend on the growth rate of the
economy. One might think that when the economy grows
faster over long periods of time, more efficient and less

10. See Allen and Gale (1994) for a different perspective.
11. This condition is also known as the Euler equation, after the mathematician who came up with the mathematical procedure, or the Keynes-Ramsey rule, after the economists who saw how to apply the procedure to optimal consumption decisions.
12. Berger and Hannan (1994) find that after controlling for efficiency there is a significant positive correlation between loandeposit spreads and concentration in the United States. Note also that we highlight the possibility of imperfect competition
with financial intermediaries but not with producers. Extending the model in this direction will not affect the results at this
level of generality (Jones and Manuelli 1997). Most growth models, however, do not consider imperfect competition in both
product and financial markets. Recent attempts to do so include Sussman and Zeira (1995) and Becsi, Wang, and Wynne
(forthcoming).
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

53

CHART 1

CHART 2

Determination of Equilibrium

Increase in Impatience
R L, R D

R L, R D

LD' CS'
LD
CS

RL

LD
CS

FF

RD

FF
RD

C

I

G

G
G'

G*

costly intermediation is required.13 Thus, intermediation
depends on growth and growth depends on intermediation. That is to say, there are feedback effects from and
to growth.
All these equations can be graphed, and an equilibrium with financial intermediation can be determined. In Chart 1 the horizontal axis measures the
growth rate, G, and the vertical axis measures real
interest rates, RL and RD. The firms’ equilibrium condition, equation (1), is the line labeled FF, which is
drawn horizontally, meaning that productivity is independent of growth. For example, a line drawn with a
slight upward tilt to it initially that becomes horizontal
after some point would indicate that the effect of
growth on productivity is positive to a point but then
levels off, reflecting positive knowledge spillovers in
that individuals’ knowledge enhancement reinforces
that of others to generate an ever-higher level of aggregate knowledge (Romer 1986). The curve showing the
optimal consumption-savings rule is denoted CS and
graphs equation (2). It is drawn with a vertical intercept, I, for the impatience of the economy, below the
intercept of the FF curve and with an upward slope of P.
Finally, the real loan-rate curve, LD, graphs equation
(3) by adding the loan-deposit differential to the CS
curve, which captures only deposit rates. As drawn, the
differential narrows as growth increases, indicating that
scale effects on costs dominate, and financial efficiency
increases with growth. The equilibrium loan rate and
equilibrium growth rate are determined by the intersection of the LD and the FF curves, and then the equilibrium deposit rate is the rate consistent with the
equilibrium growth rate. Finally, the equilibrium rate of
consumption growth equals that of output (under conditions of long-term balanced growth in the absence of
externalities).14

54

G*

Economic Implications of the Model. To see the
properties of the equilibrium, consider an increase in
society’s impatience, shown by a leftward shift of the CS
and LD curves in Chart 2. This shift causes the equilibrium growth rate to decrease and the equilibrium loandeposit differential to widen. For this simple model, the
equilibrium loan rate is pegged to productivity and
remains unchanged. Impatience means that the deposit
rate falls or rises. A more impatient society cuts back on
savings and reduces the growth rate of consumption to be
able to consume sooner. Lower growth means that banks
have fewer projects to finance, so they must scale back
their lending activities. A reduction in scale implies that
the costs of intermediation increase; and because banks
cannot raise loan rates above the profitability of projects,
banks may reduce the rates they pay to savers. Notice
that if impatience rises until it exceeds productivity in
this model there will be no equilibrium growth, suggesting that for economies patience is a virtue.15
How do different financial innovations affect the
real economy? Suppose that the innovation reduced the
cost of intermediation only. This reduction could be
through new methods to reduce risks (for example, the
use of credit scoring for smaller business loans) or
cheaper means for pooling funds (such as replacing
branches with Internet banking). Because costs fall,
the loan-deposit differential tightens at all levels of
growth, and the LD curve shifts down and to the right as
in Chart 3. Thus, the equilibrium growth rate rises, but
there is no change in the loan rate. However, deposit
rates rise at the same time the loan-deposit rate spread
tightens. Deposit rates rise because in equilibrium the
scale of financial activities rises with the growth rate of
the economy.16
Alternatively, Chart 4 portrays a financial innovation that increases the average productivity of firms.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

CHART 3

CHART 4

Increase in Financial Intermediation Costs

Increase in Productivity
R L, R D

R L, R D

LD
LD'
CS

RL

LD
CS

FF

RD1
RD

R L'
R D'
RL

FF

RD

G

G
G*

G*

Such an innovation can be thought of as a permanent
shock that increases the monitoring and control functions of financial intermediaries and raises the quality of
investment outcomes. Banks are now able to be more
discriminating and on average fund better projects. Or, it
may be that banks can better influence the actions
taken by firms—say, because of legal reforms that make
contracts more enforceable—and thus achieve better
outcomes from investments. In either case, the FF curve
shifts up due to higher productivity. Again the equilibrium growth rate rises, but now the equilibrium loan rate
rises together with the deposit rate at the same time the
spread falls. Intuitively, an increase in productivity
means both that growth rises and financial intermediaries earn more from their loans. Since the scale of
financial activities also rises with increased growth, the
cost of intermediation falls. Thus, intermediaries can
pay out more to savers by raising deposit rates, which
also lures them to increase their rate of consumption.
Thus, no matter what the source of the financial
innovation is, it has the same effect on outcomes as a
technology shock. Technology shocks and financial
innovation shocks imply that long-run growth rates are

negatively correlated with the rate spread and positively correlated with deposit and loans rates individually.17
The positive correlation with loan rates is weaker the
flatter the FF curve is and the more the financial innovation serves to increase the quality of loans rather than
reduce the costs of intermediation. Notice, though, that
as long as increases in the scale of financial activities
increase efficiency and cause financial intermediation
costs to fall, one should expect to see a negative correlation of rate spreads and growth rates. A shock in the
economy’s patience will also cause the growth rate to be
positively correlated with loan rates, but all the other
correlations could go either way.
Evaluation of Financial and Monetary Policy. The
discussion next turns to several classic forms of government intervention and financial repression. First, consider a ceiling on loan rates typically thought to make credit
cheaper. When the loan rate ceiling is binding—that is,
below the unconstrained equilibrium rate—the ceiling
acts like a drop in productivity or a downward shift of the
FF curve, best seen by assuming that the FF curve is horizontal and reversing the changes in Chart 4. The equilibrium growth rate falls and the loan-deposit spread widens

13. Sussman and Zeira (1995) find that total bank costs per unit of extended credit have fallen with financial development
across U.S. states.
14. If there were no banks, equations (1) and (2), or EF and CS curves, would yield a solution for the basic endogenous growth
model, which is commonly known as the Ak-model, because output is linear in capital, k, with A the marginal productivity. One can easily find the long-term equilibrium. When financial intermediation is costless, the loan rate equals the deposit
rate. In this case, the economy’s equilibrium long-run growth rate is equal to (A – I)/P. Thus, anything that raises producers’ long-term productivity or lowers households’ impatience or smoothes consumption-savings trade-offs will raise longterm growth.
15. Also, if the FF curve is upward sloping to account for positive knowledge spillovers, the interest rate differential and the
deposit rate can either increase or decrease.
16. Note that if the FF curve sloped upward, then the equilibrium loan rate would rise too.
17. A productivity increase will raise loan rates while financial innovation will leave them unchanged. However, if the FF curve
is upward sloping, loan rates will increase in both cases.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

55

CHART 5

CHART 6

Deposit Rate Ceiling

Limit on Entry of Financial Intermediaries
R L, R D

R L, R D

LD
CS

RL

FF

RD

R'
RL
RD

LD
RD

RD

LD'
LD
CS

D

L

B

A

G

G
G*

G*

because the scale of financial activities has fallen, causing
costs to rise. Second, governments may choose to put a
cap on deposit rates. As shown in Chart 5, the effect is to
transform the CS locus into a horizontal line at the mandated deposit rate, which is assumed to be below the previous equilibrium deposit rate. In this case equilibrium
growth and loan rates are determined where the FF curve
intersects the now-downward-sloping LD curve (obtained
by adding C to the horizontal deposit rate ceiling). While
equilibrium growth rates fall, loan rates are unchanged
and the interest rate spread widens. In a sense banks are
profiting at the expense of growth, which may or may not
have been the desired outcome. The fact that the LD
curve is now downward-sloping produces some interesting effects on its own, implying a negative correlation
between loan and growth rates. Third, consider an
increase in bank reserve requirements. This move would
effectively raise the costs of intermediation and the loandeposit rate differential at all growth rates. Thus, the LD
curve would shift upward and to the left in Chart 3, causing growth and the scale of financial activity to fall, the
rate spread to widen, and deposit rates to fall.
Fourth, many governments have tried to direct
credit by picking and choosing recipient firms and sectors.18 For this policy to be successful, the government
must do a better job of evaluating investments and picking winners than the private sector does. Without the
requisite credit expertise or good luck, however, the
result will be a decline in the average quality of loans
and thus a downward shift in the FF curve, again a
reversal of the analysis in Chart 4.
Finally, Chart 6 illustrates a limit on the entry of
banks, which reduces the number of banks but increases
individual bank size. Such a policy may have the effect of
increasing the market power of financial intermediaries,
and thus the loan-deposit spread rises with an effect sim-

56

FF'
FF

ilar to a cost increase. This effect is shown as a move from
equilibrium A to B. However, increasing the scale of individual banks might arguably produce economies in lending and lead to improvements in loan quality, possibly
through better screening and control. In this case the FF
curve might shift up, and the combined effect of cost and
productivity increases is uncertain for growth and
deposit rates but surely leads to increases in loan rates
and the rate spread. This last effect is shown as a move
from B to D, illustrating a case where the net growth
effect is nil, but it could go either way.
Countries in which financial repression has
occurred, most notably in Latin America, have used a
combination of these policies, and the negative consequences for growth have been even more severe.
Financial sector reforms in Latin America have led to significant increases in real interest rates (Holden and
Rajapatirana 1995), an outcome consistent with the
model above. However, these effects were accompanied
by large spreads between lending and borrowing rates
that suggest increased inefficiency of the banking sector.
The analysis above also implies that there are only two
types of financial repression that might increase growth
rates.19 These policies are limits to entry—if demonstrated to give rise to very large productivity increases—and
directed credit, if pursued successfully. Park (1993)
argues that these two policies played an important role
initially in Korea’s strong growth performance.20

Empirical Review
he predictions from the model presented here
show that financial intermediation increases the
efficiency of investment by identifying and channeling resources toward high-return projects and by disciplining corporations. While innovation and knowledge
creation are the ultimate forces behind broad capital

T

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

accumulation and growth, financial intermediation will
enhance growth to the extent that the intermediaries
perform their functions efficiently. Thus, countries’
growth performance should vary with their level of
financial efficiency. Financial efficiency in turn depends
importantly on the extent to which financial economies
of scale have been realized and on how developed and
innovative the financial sector is. Financial development
can be measured by the size of the financial sector to the
extent that activities transform the quality of investments. Also, as the growth model with financial intermediation illustrated, financial development and
efficiency are reflected in lower loan-deposit rate
spreads. The growth model also predicts that growth
rates are positively correlated with real interest rates
and negatively correlated with loan spreads. In addition,
it has been shown that government intervention can
severely affect the efficiency of financial intermediaries
and economic growth and alter these correlations.
Importance of External Financing. In most theoretical models financial intermediaries lend to firms all
that is needed for investment. In the real world, however, businesses’ retained earnings finance a large part of
investment. This situation exists especially in lessdeveloped countries with financial markets that do not
operate well and operate at a higher cost to firms than
in countries with more-developed financial markets.
Mayer (1988) shows that firms in eight industrialized
countries over the period from 1970 to 1985 financed
most of their investment from retained earnings. He
finds that intermediated loans are the dominant source
of external funds for firms and that in all countries
(except Canada) intermediated loans contribute a
greater share of external financing than short-term
securities, bonds, and shares combined.21 There are
marked variations in external-financing percentages
across countries. Computed on a net (of accumulation
of equivalent financial assets) basis, external financing
shares ranged from lows of –2.3 percent in the United
Kingdom and 14.1 percent in the United States to highs
of 32.1 percent in Japan and 48.1 percent in Italy.
It is reasonable to suppose that the share of external
financing varies with the relative cost of external financ-

ing both overall and for particular forms of finance. The
costs of particular forms of financial intermediation
depend on technological and incentive frictions as well
as legal and regulatory costs and should fall as the efficiency of the sector increases. It is thus surprising that
the United States and the United Kingdom, countries
with arguably the most efficient financial sectors, have
the lowest external financing shares. This puzzle is further compounded by the fact that external financing
ratios in the United States were not any different in the
1970s from those in the first two decades of the century
(Taggart 1985) and in the United Kingdom were stable
over the postwar period (Mayer 1990). While it is
arguable that these countries are the most efficient for
all types of intermediation across countries (see Berger
and Humphrey 1997, who evaluate the few studies that
attempt cross-country comparisons of efficiency), it is
likely that their efficiency has increased over time.
There are several possibilities in accounting for the
relatively low external financing shares in the United
States and the United Kingdom.22 Mankiw (1988) argues
that times of rapid growth create needs for funds that
cannot easily be accommodated by retained earnings
alone, and more external financing occurs. Mayer (1988)
argues that competition may have increased the costs of
intermediation and external funds by making long-term
financial relationships less likely. Alternatively, one could
argue that while financial efficiency implies lower costs
of intermediation, the reduced costs are not necessarily
passed on to firms if intermediaries have sufficient market power over the price of their services. Another explanation that has not been explored is the effect on
financial intermediation costs from government intervention and regulation.
Does Financial Development Promote Growth?
While financial intermediaries may not finance a dominantly large share of investment, they still may be important for growth. As economists since Goldsmith (1969),
McKinnon (1973), and Shaw (1973) have shown, financial development and economic growth are positively correlated across countries. Goldsmith (1969), analyzing
data from thirty-five countries over the period from 1860
to 1963, finds that financial and economic development

18. Throughout this article, we abstract from cases in which funds move (either voluntarily by financial intermediaries or
because of government decree) toward government projects and households.
19. A rarely considered alternative is that the government could designate a fixed loan-deposit rate spread, which would cause
the LD curve to become linear, parallel to the CS curve. If the designated spread is above (or below) the unconstrained equilibrium spread, growth may rise (or fall). For initially low equilibrium growth rates with a large unconstrained spread,
the designated spread would be low and the unconstrained spread and growth would tend to rise. However, at high initial
growth rates this policy would slow growth.
20. For corroborating evidence, see also Demetriades and Luintel (1996).
21. Interestingly, equity markets have been negligible as a source of external funds, and the long-term trend has been downward. See also Corbett and Jenkinson (1994), who provide a more consistent data set for Germany, Japan, the United
Kingdom, and the United States over the period from 1970 to 1989.
22. See Gertler and Rose (1991) for a discussion on the factors that determine external financing costs.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

57

are positively correlated over periods as long as several
decades. Financial development is measured in his study
by the financial intermediation ratio, or the ratio of
financial intermediary assets divided by gross national
product. To the extent that these assets measure the provision of credit to firms (as opposed to households and
government), this measure captures the financial intermediaries’ role in overcoming frictions and enhancing
growth through quality enhancement. However, as
Goldsmith notes, it is an open question whether financial
development leads to economic development or vice
versa, because each has feedback effects on the other. In
a later study Goldsmith (1985) shows that financial
development largely occurs during the early stages of
economic development when countries have low levels of
income.23 However, even though financial development
occurs early and may precede economic growth, it is
unclear that it provides causality in an economic sense.
More recently, King and Levine in a series of studies (1992, 1993a, b) look at growth and financial development over various periods starting in 1960 for a
comprehensive cross section of countries. They expand
the set of financial development measures to better
capture the various services provided by financial intermediaries. For example, one measure approximates the
liquidity-providing role of financial intermediaries
through the ratio of liquid liabilities (currency plus
demand and interest-bearing deposits, or M2) to GDP.
Another measure, a ratio of credit provision to private
firms to GDP, captures monitoring, screening, and control activities as well as the pooling of funds and diversification of risks. The first measure approximates
intermediaries’ role in overcoming technological frictions while the second approximates their role in overcoming incentive frictions. However, these are very
crude measures of the specific roles of banks, such as
liquidity provision and firm-specific risk reduction or
overcoming divergent incentives. King and Levine find
that their measures are positively correlated with real
GDP growth rates even after controlling for initial conditions, education, government spending, inflation,
political stability, and some other policy measures. They
also show that subsequent growth rates are positively
correlated with initial liquidity ratios. This finding can
be taken as evidence that financial development causes
growth, but it may also be reflecting a buildup in anticipation of future growth. In any case, because measures
of financial intermediation and growth are endogenous
and respond in specific ways to shocks, studying this
question might be less informative than understanding
the sources of shocks that drive the correlations.
Lately, several studies have found that the correlations between financial intermediation and growth
depend critically on the sample of countries considered.
Fernandez and Galetovic (1994) split King and Levine’s
58

sample between Organisation for Economic Cooperation
and Development (OECD) and non-OECD countries and
show that the correlations fall and become insignificant
for OECD countries. DeGregorio and Guidotti (1992) add
more countries to King and Levine’s sample and show that,
when they divide the sample into three groups based on
per capita incomes at the start of the sample period, the
correlations rise and become more significant as initial
incomes fall.
In both studies it is argued that there might be
insufficient variation among developed countries with
mature financial sectors to determine their growth
effects and that the financial variables do not capture
intermediation through nonfinancial or nonbank intermediaries. Either the measures of financial development used are not broad enough or else they are too
broad to capture specific efficiency-enhancing roles of
financial intermediaries. Alternatively, financial intermediation may have stronger efficiency-enhancing
effects in less-developed countries than in developed
countries. DeGregorio and Guidotti also show that Latin
American countries had a significantly negative correlation. As shown by Diaz-Alejandro (1985), who analyzed
the liberalization in Latin America in the 1970s, insufficient regulation and expectations by banks of bailouts
resulted in overlending by banks. Thus, a large financial
intermediation sector reflected a very fragile system,
not financial efficiency. The lesson is that regulation
affects the behavior of banks and the system’s efficiency. One can interpret the negative correlation either as
a negative effect on growth of financial intermediaries
or more likely as a sign that insufficient provision of
financial services retards growth (Galetovic 1994).
In any case, these studies highlight a potential
problem with the post–World War II sample period considered in most of the cross-country studies. Barro and
Sala-i-Martin (1995) note that growth rates in the postwar period were atypical because in comparison with
rates over the last 100 years they were much above
trend. Also, starting in the 1970s, many policy experiments were conducted with more or less successful
attempts at deregulating financial sectors. As the previous section showed, different types of liberalization give
rise to different correlations with growth, suggesting
that empirical studies must control for the specifics of
initial policy regimes and policy switches.
Some studies of financial repression and growth have
attempted to control for these factors. Roubini and Sala-iMartin (1992), using dummy variables in standard growth
regressions to distinguish between repressed and other
countries, find that financial development and growth are
insignificantly correlated. However, using dummy variables that distinguish countries with annual real rates
less than –5 percent yields a significantly negative correlation. King and Levine (1992) also control for high nega-

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

tive real rates and policy distortions and find a significantly negative effect.
These studies show that governments can have a
negative effect on growth rates through financial intermediaries. To date, the empirical work has not provided
estimates for positive government interventions, nor are
there any estimates of the effect of specific governmental
policies on growth.24 However, the empirical literature
does show that the overall policy package matters for the
efficiency of financial intermediaries and for growth.25
As the illustrative growth model has demonstrated,
real interest rates and loan-deposit spreads reflect the
efficiency of financial intermediaries as well as the productivity of investment (and underlying policy experiments). There have been a few studies that analyze the
correlation of real interest rates and growth. For instance,
King and Levine (1992) find an insignificantly positive
association of real interest rates and growth for seventythree countries over the period from 1974 to 1989. But
including financially repressed countries is important for
these results. Another indicator of financial efficiency is
the difference between lending and borrowing rates.
Using this variable in their empirical work, King and
Levine (1992) find insignificant negative correlation with
per capita GDP growth, blaming the poor quality of interest rate data. Sussman (1993) finds a weak negative correlation between rate spreads and real per capita GDP for
1985 for eighty-one countries (omitting repressed countries with negative rate spreads). He shows that the
markup increases as incomes rise from poor countries to
middle-income countries and then falls again as incomes
rise to those in the group of rich countries.
Relevant Microeconomic Studies on Efficiency
and Competitiveness of Financial Intermediaries.
This article has argued that factors such as scale
economies and financial market structure may affect
the efficiency of the financial intermediaries and thus
ultimately growth. To date there are no studies that
explore the relationships among these factors and
growth. However, there are microeconomic studies that
look at some of these factors in isolation. One aspect
that deserves attention is the role of economies of scale
on efficiency and growth. The empirical evidence suggests that there may be returns to scale in at least U.S.
banking markets. Clark (1988) argues that economies
exist only for small depository institutions. Berger,
Hunter, and Timme (1993) conclude in their survey of

the literature that medium-scale financial intermediaries might be slightly more scale-efficient than large or
small firms. McAllister and McManus (1993) find that
managers at large financial intermediaries might be better able to control costs and that large firms have
approximately constant returns to scale. Berger and
Humphrey (1997) review the few studies that make
cross-country comparisons on the efficiency of the financial sector. These studies have methodological differences and do not control for regulatory differences, but
one finding is that U.S. banks are among the least efficient when comparing developed countries.
Economies of scale sometimes imply imperfect competition, and how competitive financial intermediaries
are may also affect how well they perform in overcoming
frictions and ultimately stimulating growth.
Unfortunately, only a very few studies make cross-country
comparisons. Shaffer (1995) estimates the degree of market power or contestability on the commercial banking
industry in each of fifteen industrialized nations over
multiyear periods. He finds that there is much variation in
the degree of competition across countries and that five
countries (Belgium, Denmark, France, Japan, and the
United States) had statistically significant evidence of
market power. Berger and Humphrey, in their survey of
the literature, conclude that “market power does seem to
affect the prices of some types of local deposits and loans
in the United States” (1997, 47). They caution, though,
that U.S. banking markets are an outlier because elsewhere markets tend to be more concentrated, sometimes
with explicit collusion, and usually they are national in
scope. Petersen and Rajan (1995) show that market
power varies geographically in the United States. They
find that young or small firms obtain more external
financing in concentrated markets than in competitive
markets because larger and more monopolistic banks
can extract future payments from them in an environment in which firms are less able to turn to other banks.

Conclusion
his article states that in one way or another financial intermediaries give individuals or firms access
to economies of scale that they would not have otherwise. Thus, intermediaries allocate capital to its best
possible use and economic efficiency is enhanced. How
well financial intermediaries carry out their functions
may explain differences across countries’ rates of growth.

T

23. See also Wachtel and Rousseau (1995), who provide asset ratios for various types of intermediaries in the United States, the
United Kingdom, and Canada over a period of 100 years.
24. One exception is Jayaratne and Strahan (1996), who find that when intrastate branching restrictions were relaxed in the
United States bank lending quality and real per capita state growth rose. They argue that lower barriers to entry improved
the average quality of surviving banks and that increases in the average size of banks led to economies of scale and scope.
25. See also Arestis and Demetriades (1997), who provide evidence that policy interventions matter if a study uses a time series
framework.
Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

59

A simple model illustrates this relationship and shows
how policies such as reserve requirements, interest rate
controls, entry limitations, and directed credit flows
affect growth and interest rates. While most of these
policies are inimical to growth, there are two exceptions: the quality of investments and growth may be
increased if entry restrictions cause financial intermediaries to operate at more efficient scales or if directed
credit policies are chosen well.
The article also briefly surveys some of the recent
empirical literature, which consistently finds a positive
relationship between growth and financial intermediation, and shows that because financially repression hurts
growth, financial intermediation matters if it is not effi-

cient. More work needs to be accomplished, however,
before policy recommendations can be made. This article argues that more attention needs to be paid to the
efficiency-enhancing role and imperfectly competitive
aspects of financial intermediation. The economies of
scale arguments used to motivate financial intermediation imply that markets may be imperfectly competitive
at the same time that the race for market power provides incentives for financial innovation and development. Empirical work suggests that financial
intermediaries differ across countries in the cost efficiency and the degree of competition they provide, a
finding that has implications for economic efficiency
and maybe for growth.

REFERENCES
ALLEN, FRANKLIN, AND DOUGLAS GALE. 1994. Financial
Innovation and Risk Sharing. Cambridge, Mass.: MIT Press.
ARESTIS, PHILIP, AND PANICOS DEMETRIADES. 1997. “Financial
Development and Economic Growth: Assessing the
Evidence.” Economic Journal 107 (May): 783–99.
BAGEHOT, WALTER. [1873] 1991. Lombard Street: A Description
of the Money Market. Philadelphia: Orion Editions.
BARRO, ROBERT, AND XAVIER SALA-I-MARTIN. 1995. Economic
Growth. New York: McGraw-Hill.
BAUMOL, WILLIAM, JOHN PANZAR, AND ROBERT WILLIG. 1982.
Contestable Markets and the Theory of Industry Structure.
New York: Harcourt Brace Jovanovich.
BECSI, ZSOLT, PING WANG, AND MARK WYNNE. Forthcoming.
“Endogenous Market Structure and Financial Development.”
Federal Reserve Bank of Atlanta Working Paper.
BENCIVENGA, VALERIE R., AND BRUCE D. SMITH. 1991. “Financial
Intermediation and Endogenous Growth.” Review of
Economic Studies 58:195–209.
BERGER, ALLEN, AND TIMOTHY H. HANNAN. 1994. “The Efficiency
Cost of Market Power in the Banking Industry: A Test of the
‘Quiet Life’ and Related Hypotheses.” Board of Governors of
the Federal Reserve System Finance and Economics
Discussion Paper 94-36, November.
BERGER, ALLEN, AND DAVID HUMPHREY. 1997. “Efficiency of
Financial Institutions: International Survey and Directions
for Future Research.” Board of Governors of the Federal
Reserve System Working Paper 97-11.
BERGER, ALLEN, WILLIAM HUNTER, AND STEPHEN TIMME. 1993.
“The Efficiency of Financial Institutions: A Review and
Preview of Research Past, Present, and Future.” Journal
of Banking and Finance 17 (April): 221–49.
BHATTACHARYA, SUDIPTO, AND ANJAN V. THAKOR. 1992.
“Contemporary Banking Theory.” Journal of Financial
Intermediation 3 (October): 2–50.

BOOT, ARNAUD W.A., AND ANJAN V. THAKOR. 1997. “Financial
System Architecture.” Review of Financial Studies
10:693–733.
BOYD, JOHN H., AND EDWARD C. PRESCOTT. 1986. “Financial
Intermediary Coalitions.” Journal of Economic Theory
38:211–32.
CHEN, BEEN-LON, YEONG-YUH CHIANG, AND PING WANG. 1996.
“A Schumpeterian Model of Financial Innovation and
Endogenous Growth.” Photocopy, November.
CLARK, JEFFREY A. 1988. “Economics of Scale and Scope
at Depository Financial Institutions: A Review of the
Literature.” Federal Reserve Bank of Kansas City Economic
Review 73 (September/October): 16–33.
CORBETT, JENNY, AND TIM JENKINSON. 1994. “The Financing of
Industry, 1970–89: An International Comparison.” Centre for
Economic Policy Research Discussion Paper No. 948.
DEGREGORIO, JOSE, AND PABLO E. GUIDOTTI. 1992. “Financial
Development and Economic Growth.” International Monetary
Fund Working Paper No. 92-101.
DEMETRIADES, PANICOS O. 1997. “Financial Markets and
Development.” Keele University Working Paper No. 97-03,
April.
DEMETRIADES, PANICOS O., AND KUL B. LUINTEL. 1996. “Financial
Repression in the South Korean Miracle.” Keele University
Working Paper No. 96-13, October.
DIAMOND, DOUGLAS W. 1984. “Financial Intermediation and
Delegated Monitoring.” Review of Economic Studies 51
(July): 393–414.
DIAMOND, DOUGLAS, AND PHIL DYBVIG. 1983. “Bank Runs,
Deposit Insurance, and Liquidity.” Journal of Political
Economy 85 (April ): 191–206.
DIAZ-ALEJANDRO, CARLOS. 1985. “Good-Bye Financial Repression, Hello Financial Crash.” Journal of Development
Economics 19 (September/October): 1–24.

BINGHAM, T. 1985. Banking and Monetary Policy. Paris:
Organisation for Economic Cooperation and Development.

60

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

ESPINOSA, MARCO, AND WILLIAM HUNTER. 1994. “Financial
Repression and Economic Development.” Federal Reserve
Bank of Atlanta Economic Review 79 (September/October):
1–11.
FAMA, EUGENE F. 1980. “Banking in the Theory of Finance.”
Journal of Monetary Economics 6:39–57.

———. 1993b. “Finance and Growth: Schumpeter Might
Be Right.” Quarterly Journal of Economics 108 (August):
716–37.
LEVINE, ROSS. 1991. “Stock Markets, Growth, and Tax Policy.”
Journal of Finance 46 (September): 1445–65.
———. 1997. “Financial Development and Economic
Growth: Views and Agenda.” Journal of Economic Literature
35 (June): 688–726.

FERNANDEZ, DAVID G., AND ALEXANDER GALETOVIC. 1994.
“Schumpeter Might Be Right—But Why? Explaining the
Relation between Finance, Development, and Growth.”
Johns Hopkins University SAIS Working Paper in International Economics No. 96-01.

LUCAS, ROBERT E., Jr. 1988. “On the Mechanics of Economic
Development.” Journal of Monetary Economics 22 (July):
3–42.

GALETOVIC, ALEXANDER. 1994. “Finance and Growth: A
Synthesis and Interpretation of the Evidence.” Board of
Governors of the Federal Reserve System International
Finance Discussion Paper No. 477.

MALKIEL, BURTON G. 1992. “Efficient Market Hypothesis.” In
New Palgrave Dictionary in Money and Finance, edited by
P. Newman, M. Milgate, and J. Eatwell. London: Macmillan.

GERTLER, MARK. 1988. “Financial Structure and Aggregate
Economic Activity: An Overview.” Journal of Money, Credit,
and Banking 20 (August): 559–96.

MANKIW, GREGORY. 1988. “Comments on New Issues in
Corporate Finance.” European Economic Review 32 (June):
1183–6.
MAYER, COLIN. 1988. “New Issues in Corporate Finance.”
European Economic Review 32 (June): 1167–83.

GERTLER, MARK, AND ANDREW ROSE. 1991. “Finance, Growth,
and Public Policy.” World Bank Working Paper No. 814,
December.
GOLDSMITH, RAYMOND. 1969. Financial Structure and
Development. New Haven and London: Yale University Press.
———. 1985. Comparative National Balance Sheets.
Chicago: University of Chicago Press.
GOLDSTEIN, MORRIS, AND PHILIP TURNER. 1996. “Banking Crises
in Emerging Economies: Origins and Policy Options.” Bank
for International Settlements Economic Paper No. 46,
October.

———. 1990. “Financial Systems, Corporate Finance,
and Economic Development.” In Asymmetric Information,
Corporate Finance, and Investment, edited by Glenn
Hubbard. Chicago: University of Chicago Press.
MCALLISTER, PATRICK, AND DOUGLAS MCMANUS. 1993. “Resolving
the Scale Efficiency Puzzle in Banking.” Journal of Banking
and Finance 17 (April): 389–405.
MCKINNON, RICHARD I. 1973. Money and Capital in Economic
Development. Washington, D.C.: Brookings Institution.
MILLER, MERTON H., AND FRANCO MODIGLIANI. 1961. “Dividend
Policy, Growth, and the Valuation of Shares.” Journal of
Business 34 (October): 411–33.

GREENWOOD, JEREMY, AND BOYAN JOVANOVIC. 1990. “Financial
Development, Growth, and the Distribution of Income.”
Journal of Political Economy 98 (October): 1076–107.
GREENWOOD, JEREMY, AND BRUCE SMITH. 1997. “Financial
Markets in Development and the Development of Financial
Markets.” Journal of Economic Dynamics and Control
21 (January): 145–81.
GURLEY, JOHN G., AND EDWARD S. SHAW. 1960. Money in a
Theory of Finance. Washington, D.C.: Brookings Institution.

PAGANO, MARCO. 1993. “Financial Markets and Growth.”
European Economic Review 37 (April): 613–22.
PARK, YUNG CHUL. 1993. “The Role of Finance in Economic
Development in South Korea and Taiwan.” In Finance and
Development: Issues and Experience, edited by Alberto
Giovannini. New York: Cambridge University Press.
PETERSEN, MITCHELL, AND RAGHURAM G. RAJAN. 1995. “The Effect
of Credit Market Competition on Lending Relationships.”
Quarterly Journal of Economics 110 (May): 407–43.

HOLDEN, PAUL, AND SARATH RAJAPATIRANA. 1995. Unshackling
the Private Sector. Washington, D.C.: World Bank.
JAYARATNE, JITH, AND PHILIP STRAHAN. 1996. “The FinanceGrowth Nexus: Evidence from Bank Branch Deregulation.”
Quarterly Journal of Economics 111 (August): 639–70.

ROMER, PAUL M. 1986. “Increasing Returns and Long-Run
Growth.” Journal of Political Economy 94 (October):
1002–37.

JONES, LARRY, AND RODOLFO MANUELLI. 1997. “The Sources of
Growth.” Journal of Economic Dynamics and Control 21
(January): 75–114.

ROUBINI, NOURIEL, AND XAVIER SALA-I-MARTIN. 1992. “Financial
Repression and Economic Growth.” Journal of Development
Economics 39 (July): 5–30.

KING, ROBERT G., AND ROSS LEVINE. 1992. “Financial Indicators
and Growth in a Cross Section of Countries.” World Bank
Working Paper No. 819.
———. 1993a. “Finance, Entrepreneurship, and Growth:
Theory and Evidence.” Journal of Monetary Economics 32
(December): 513–42.

SAINT-PAUL, GILLES. 1992. “Technological Choice, Financial
Markets, and Economic Development.” European Economic
Review 36:763–81.
SCHUMPETER, JOSEPH. [1911] 1936. The Theory of Economic
Development. Cambridge, Mass.: Harvard University Press.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997

61

———. 1939. Business Cycles: A Theoretical, Historical,
and Statistical Analysis of the Capitalist Process. New York:
McGraw-Hill.

SUSSMAN, OREN, AND JOSEPH ZEIRA. 1995. “Banking and
Development.” Center for Economic Policy Research
Discussion Paper No. 1127, February.

SHAFFER, SHERRILL. 1995. “Market Conduct and Aggregate
Excess Capacity in Banking: A Cross-Country Comparison.”
Federal Reserve Bank of Philadelphia Working Paper No.
93-28/R.

TAGGART, ROBERT A. 1985. “Secular Patterns in the Financing
of U.S. Corporations.” In Corporate Capital Structures in the
United States, edited by Benjamin Friedman. Chicago:
Chicago University Press.

SHAW, E.S. 1973. Financial Deepening in Economic
Development. New York: Oxford University Press.

VIVES, XAVIER. 1991. “Banking Competition and European
Integration.” In European Financial Integration, edited by
Alberto Giovannini and Colin Mayer. Cambridge: Cambridge
University Press.

STIGLITZ, JOSEPH. 1993. “Overview.” In Finance and
Development: Issues and Experience, edited by Alberto
Giovannini. New York: Cambridge University Press.
———. 1996. “Financial Markets and Development.” Oxford
Review of Economic Policy 5:55–68.
SUSSMAN, OREN. 1993. “A Theory of Financial Development.”
In Finance and Development: Issues and Experience, edited
by Alberto Giovannini. New York: Cambridge University
Press.

62

WACHTEL, PAUL, AND PETER ROUSSEAU. 1995. “Financial
Intermediation and Economic Growth: A Historical
Comparison of the United States, United Kingdom, and
Canada.” In Anglo-American Finance Systems: Institutions
and Markets in the Twentieth Century, edited by M. Bordo
and R. Sylla. Burr Ridge, Ill.: Irwin Professional Publishing.

Federal Reserve Bank of Atlanta E C O N O M I C R E V I E W Fourth Quarter 1997