The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Opinions expressed in the Economic Review do not necessarily reflect the views of the
management of the Federal Reserve Bank of San Francisco, or of the Board of Governors of
the Federal Reserve System.
The Federal ReServe Bank of San Francisco's Economic Review is published quarterly by the Bank's
Research and Public Information Department under the supervision ofJack H. Beebe, Senior Vice President
and Director of Research. The publication is edited by Gregory 1. Tong, with the assistance of Karen Rusk
(editorial) and William Rosenthal (graphics).
For free copies of this and other Federal Reserve publications, write or phone the Public Information
Department, Federal Reserve Bank of San Francisco, P.O. Box 7702, San Francisco, California 94120.
Phone (415) 974-3234.
2
I.
Major Trends in the U.S. Financial System:
Implications and I s su e s...................................................................... 5
Robert T. Parry
II.
Bank Capital Regulation and Asset Risk ..................................... 20
Frederick T. Furlong and M ichael C. Keeley
III.
Competition in the Semiconductor Industry.................................41
Randall Johnston Pozdena
IV.
Tax Revolt or Tax Reform? The Effects of Local Government
Limitation Measures in California ................................................ 57
Carolyn Sherwood-Call
Editorial Committee:
Jack Beebe, Bharat Trehan, Michael Keeley, Barbara Bennett,
Brian Motley, Carolyn Sherwood-Call.
3
4
Robert T. Parry*
The major trends shaping the current evolution of the U.S. financial
system conflict with an outmoded legal and regulatory framework.
Reform of that framework is needed to promote an efficient and stable
financial system. Specifically, reforms are needed in the deposit insurance system, bank powers, and the large-dollar, electronic payments
system.
This paper was originally presented at a Federal Reserve System
Management Conference held April 1987.
loopholes rather than in a manner that ensures the
evolution of an efficient financial system. The proliferation of new instrum~nts, the transfer of traditional banking activity to nonbanks, and the staggering volume of daily payments activity, for
example, may be as much a result of efforts to avoid
regulation as a response to fundamental economic
needs.
Second, although partial integration of financial
activities and of financial and commercial activities
is occurring, the important issues of how to reform
the federal safety net and how far to extend its
coverage are not being resolved. Third, as activity
shifts to international financial centers and lessregulated nonbank firms, domestic banking firms
are left with a diminishing proportion of overall
financial activity.
The legal and regulatory framework should be
reformed to accommodate market-driven forces for
change. However, such reform also must be consistent with the goal of preserving financial stability.
These criteria imply that federal supervision, regulation, and protection of the financial system
should be structured and conducted in a way that
promotes stability while limiting the perverse incentives for risk-taking and the possibility of large
government expenditures that government intervention can create. In this paper, I presume that limited
government guarantees, directed at payments balances and savings balances held by depositories, are
Shaped by the interaction of economic, technological, legal, and regulatory forces, the U.S.
financial system is undergoing signficant change.
During the next five to ten years, it increasingly will
be characterized by:
• reliance on primary securities markets, with a
diminishing role for traditional bank-provided
intermediation;
• institutional realignment of functions in the
provision of financial services, including clearing and settlement;
• expanded access to the payments system; and
• geographic integration, including internationalization of financial activity, with around-theclock trading and settlement.
The present legal and regulatory structure often
conflicts with fundamental economic and technological forces. Moreover, the piecemeal efforts to
resolve these conflicts and to accommodate market
forces have resulted in several undesirable consequences. First, financial change is occurring
through the exploitation of legal and regulatory
* President, Federal Reserve Bank of San Francisco.
Staff contributors are Barbara Bennett, Economist,
Michael Keeley, Senior Economist, Randall
Pozdena, Assistant Vice President, and Jack Beebe,
Senior Vice President and Director of Research.
5
desirable. The optimal extent and structure of such
guarantees, however, are issues that are addressed,
but not resolved, in the paper.
The purpose of this paper is to provide a conceptual framework for both understanding the changes
occurring in the financial system and analyzing the
policy implications of those changes. The paper is
organized as follows: in Section I, the economic and
I.
regulatory forces that are driving the evolution of the
financial system are described. Then the major
emerging trends are outlined in Section II. Finally,
the policy implications of these changes are considered in Section III. The paper concludes in Section
IV that policymakers should focus on reforming
three key areas: the federal safety net, bank powers,
and the payments system.
Forces for Change
The changes that are likely to take place in the
U. S. financial system are the result of the interaction of ba'iic economic forces with regulatory and
legal constraints. In some cases, the economic
forces will overwhelm these constraints; in other
cases, regulatory and legal forces will dominate.
Legal and Regulatory Forces
The present financial system has in place extensive legal and regulatory structures to promote the
stability of the banking industry and the payments
mechanism and to facilitate the conduct of monetary
policy. To the extent that these structures conflict
with private economic incentives, the private market will attempt to take advantage of or avoid them.
As a result, our system of laws and regulations itself
is a force for change in financial markets.
The first, and perhaps most important, example is
the effect of the federal safety net on bank behavior.
Underpriced deposit protection gives banks and
thrifts incentives to reduce equity capital and to
expand the scope of insurance coverage. Similarly,
nondepository institutions may seek to own depository firms in order to raise insured funds.
A second force for change is reserve requirements. It has been argued that reserve requirements
are needed for monetary policy purposes. Yet the
requirement to hold noninterest-earning reserves
against transactions deposits, together with the prohibition on the payment of explicit interest on
demand deposits, provides incentives to create
functionally similar instruments, such as overnight
RPs, that arbitrarily increase activity in the payments system.
A third regulatory force for change is the system
of legal restrictions on the ownership and powers of
banking firms, which is particularly cumbersome at
a time when nonbanking firms are expanding
rapidly into the provision of many banking services.
Geographic restrictions facing banking firms also
are affecting industry structure by shifting activity
away from traditional banks.
Fourth, payments system policies affect not only
the level of risk in the payments system but also the
Economic Forces
The following primary economic forces appear to
be at work today. First, electronic information processing is reducing the cost of gathering, managing,
and transmitting the data required to produce financial services. This trend not only makes it possible
to provide certain traditional financial services at
reduced cost but makes new financial services feasible, thereby altering ways of raising and investing
funds. Moreover, the effect of technological change
is amplified by the rising demand for time-saving
innovations, such as certain integrated financial
services, and devices that improve access to retail
funds balances, such as ATMs and point-of-sale
terminals.
Second, the growth in wealth and the level of
commercial activity worldwide is increasing the
number, size, and complexity of transactions in
financial markets. The changing nature of transactions, in turn, is stimulating the demand for sophisticated financial services.
Third, greater volatility in interest rates,
exchange rates, and asset prices is expanding the
demand for ways to manage risk. Not only is it
increasing reliance on specific risk-management
products, as exemplified by the growth of options
and futures markets, but also the complexity of
traditional financial instruments. Moreover, widespread loan losses in recent years have heightened
the need for risk management and diversification
within institutions.
6
structure of the financial services industry. The
absence of real-time settlement and mechanisms to
price intraday credit extended by the Federal
Reserve arguably contribute to the level of payments
system risk. The exclusion of nondepository firms
from access to the payments system may be inducing commercial enterprises to buy depository
institutions in order to gain the access they desire.
Finally, regulatory, tax, and other policies vary
across countries (and in some instances, states) even
II.
for the same class of institution. These international
(and interstate) differences affect not only the locus
of financial activity, but also the ability of anyone
government or agency to pursue policies that ensure
prudent practices in the financial system. However,
the recently proposed agreement between U.S. and
British banking authorities is an important step in
the attempt to internationalize supervision and regulation.
Emerging Trends
Several key trends are emerging from the economic, technological, regulatory, and legal forces
affecting the U.S. financial system. These include
trends toward direct placement and securitization,
functional realignment in the provision of financial
services, expanded access to the payments system,
and geographic integration of financial services and
markets. They are described below.
rating services and, in one sense, even by banks
themselves through the issuance of standby letters
of credit. The effects of declining information costs
are especially evident in the shrinking share oflarge
and middle-market corporate borrowers that continue to rely on banks as their primary source of
funds. More of these borrowers now obtain
Moody's or Standard and Poors' ratings in conjunction with standby letter of credit backing to enable
them to raise funds directly in the commercial paper
and bond markets.
In addition, the rapidly growing depth and liquidity of futures and options markets enable investors
to manage interest-rate risk inexpensively and
directly, thereby reducing the relative cost advantage that banks traditionally have enjoyed. Similarly, secondary markets for a broadening array of
primary securities provide a growing source of
liquidity which, in the past, only banks were able to
provide economically. Finally, pension funds, taxsheltered savings plans, and money market and
bond mutual funds now offer numerous channels to
satisfy the diverse denomination requirements of
borrowers and investors. In fact, money market
mutual funds continue to hold over $250 billion in
assets even though banks no longer are subject to
interest-rate ceilings on most deposit products. As a
result of these developments, financial market participants can now purchase
in relatively small
denominations - their desired mix of liquidity,
credit, and interest-rate risks directly, without having to tum to the banking system in the traditional
sense.
In addition to the economic forces favoring direct
Direct Placement and Securitization
Increasingly, borrowers are placing debt
securities directly with investors and relying correspondingly less on traditional financial intermediaries - most notably, commercial banks
as a
source of funds. This rise in direct placement is the
outcome of a number of underlying economic
forces, as well as constraints, imposed by the current financial regulatory framework. Growth in the
sheer size of transactions, together with the declining costs of transmitting credit information and
effecting transactions, have made direct placement
cost-effective for an increasing number of transactions.
Historically, banks have enjoyed cost-advantages
in evaluating a given borrower's creditworthiness
and in taking on interest rate, liquidity, and credit
risks. As a result, financial intermediation through
the use of two instruments - a deposit liability and
a loan asset - with the bank as a party to each has
been more economical than direct dealings between
borrowers and investors. Today, however, banks'
cost-advantages in executing intermediation functions are diminishing. The declining cost of technology permits the sale of credit information by various
7
in the primary market. However, banks still will
function as traditional intermediaries, as most will
continue to hold loans in portfolio, particularly
loans to borrowers whose creditworthiness is relatively costly for the market to evaluate or whose
funding needs are not standard.
placement, regulatory constraints continue to make
intermediation by depositories less attractive than
otherwise. Reserve requirements, in particular, raise
the cost of attracting reservable funds. Moreover,
the decision of bank regulators in recent years to
require banks to maintain higher minimum (book)
capital-to-asset ratios probably is raising the effective cost of holding assets in portfolio. With underpriced deposit insurance, it generally has been more
profitable for banks to raise insured deposits than to
raise capital to fund loans.
As a result of these economic forces and regulatory constraints, banks increasingly have been shifting to so-called off-balance-sheet activities. In
recent years, for example, banks have offered
standby letters of credit (for a fee) to back debt
securities of prime issuers rather than funding loans
to those issuers. By assuming these contingent
liabilities, banks can increase their effective leverage without actually violating formal capital
requirements. In effect, the implicit government
protection of large institutions encourages their
entry into the growing market for financial guarantees. Approximately 15 percent of all commercial
paper and 29 percent of newly issued municipal
bonds now have some form of financial guarantee as
backing. l
Because banks retain some comparative advantages in evaluating the creditworthiness of smaller
businesses and households as well as in servicing
debt obligations, they continue to originate loans
but are selling and then servicing an increasing
number of them. Similarly, banks are "securitizing" loans (that is, pooling and using them as
security for marketable debt instruments that are
sold outright to investors). In addition to residential
mortgage loans, which were first securitized in
1968, commercial mortgages, automobile finance
loans, and credit card receivables are also being
securitized now. For many of these products, the
rate of growth of the derivative security far exceeds
the growth rate of the underlying asset. For example, residential mortgage backed securities grew by
155 percent over the past five years, while real estate
loans in bank and thrift portfolios grew by only 24
percent. 2
Clearly, the trend is for banks to take on the role of
broker, or even underwriter, to facilitate transactions
Functional Realignment
Economic forces such as the demand for greater
convenience in financial services, the declining cost
of effecting transactions, and the growth of securitization and direct placement are causing a breakdown in institutional specialization. Commercial
banks, thrift institutions, securities firms, insurance
companies, and other types of financial and nonfinancial companies increasingly are offering products that overlap their traditional markets. Although
these developments do not necessarily portend fullscale integration of financial service firms, they do
suggest that the old institutional boundaries 'governing firms' activities are breaking down and that a
realignment of the types of services each firm
chooses to provide is taking place.
For wholesale commercial banks and investment
banks serving the needs of the corporate sector, this
process of realignment is especially apparent. For
commercial banks, the push towards investment
banking is a logical extension of their expertise in
lending as their corporate borrowers rely more and
more on direct placement. By the same token,
investment banks believe that their ability to offer
certain commercial banking services would be
advantageous. For example, investment banks want
to be able to offer payments services and to settle
transactions directly because doing so themselves is
more efficient and profitable than obtaining the
same services from commercial banks. 3
So far, nonbank firms have been more successful
at circumventing barriers than have commercial
banks simply because they are regulated less extensively. Technically, regulatory and legal restrictions
on the ownership of commercial banks prevent
nonbank firms from offering banking services.
However, . through such innovations as checkable
money market mutual funds and cash managementtype sweep accounts, nonbank firms now offer
services that are functionally similar to commercial
8
a few years ago induced corporations and households to invest in more sophisticated cash management technology, which they continue to use even in
the current lower interest rate environment. The use
of such technology makes direct access to the payments system for the purposes of consolidating and
investing idle balances especially attractive. Such
forces are behind brokerages' cash management
accounts and the establishment of nonbank banks
by brokerage firms.
Combined with these economic forces are several
regulatory constraints that encourage the use of
alternatives to bank-provided payments balances.
Noninterest-eaming reserves and the prohibition on
the payment of explicit interest on demand deposits
raise the effective cost of using demand deposits to
settle transactions. As a result, corporations in particular employ cash management techniques that
minimize such balances. Their actions, especially
those related to the overnight RP market, undoubtedly are part of the explanation for the extraordinary
volume of transactions over Fedwire, the Federal
Reserve's electronic funds transfer network.
Nonbank firms traditionally have been denied
direct access to the payments system in general, and
to Fedwire in particular, because of concerns about
increased payments system risk. However, as noted
above, ownership of thrifts and nonbank banks
enables nonbank firms to circumvent restrictions on
access and may, in time, render the current legal
framework governing access to the payments system obsolete.
banks' payments and deposit services. Moreover,
with the expansion of thrift institutions' lending,
payments, and deposit-taking powers, the ownership of nonbank firms (which is not restricted in the
same way as commercial bank ownership) confers
many banking powers. Likewise, the innovation of
nonbank banks, which skirt the legal definition of a
commercial bank by offering only commercial
loans or demand deposit services but not both, will
enable nonbank firms such as Merrill Lynch to offer
banking services.
While nondepository firms are now able to offer
virtually the full range of banking services - albeit,
less efficiently than through outright ownership of
commercial banks - banks are trying to broaden
their nonbanking activities. To a certain extent,
regulators are accommodating these pressures.
Bank regulators have expanded permissible
activities to include credit-related insurance, discount brokerage, (limited) securities underwriting,
data processing, financial planning, and investment
advisory services. Moreover, regulators now sanction bank holding companies' purchases of failing
thrift institutions, thereby expanding the opportunities for those banking organizations.
For the most part, regulatory accommodation still
does not allow banks the degree of freedom that
nonbank firms have. In particular, commercial
banks are prevented from underwriting the vast
majority of municipal debt and all domestic corporate debt and equities. Where they are not as constrained by Glass-Steagall restrictions, large commercial banks underwrite a significant and growing
proportion of securities in international capital markets, including interest rate and currency swaps.
(There are, however, limitations on the amount of
corporate debt and equity underwriting they can do
even in foreign markets. 4)
Geographic Integration
The growth of international trade and commerce,
the integration of financial markets and payments
media, and the declining cost of information technology appear to be increasing the optimal geographic scope of firms in banking and finance. As a
result, there is a trend towards internationalization
of capital markets and interstate provision of
domestic financial services. Domestic firms of stature can now raise funds economically in the rapidly
growing euromarkets. A large California utility, for
example, has at times raised a significant proportion
of its longer-term funding in the euronote and bond
markets even though its operations are confined
largely to domestic markets.
Expanding Access to Payments System
Many of the forces that are encouraging realignment in the provision of financial services also are
motivating nonbanks' desire for access to the payments system. In particular, the increasing integration of payments and securities activities and the
trend towards direct placement are making direct
access to the payments system more valuable than in
the past. Also, the high and volatile interest rates of
9
national deposit-taking networks to broaden and
diversify their core deposit, financial services, and
lending bases and to provide customers engaged in
interstate transactions with improved access to the
payments system. Regulations already have accommodated these forces to a large extent, although
perhaps not in the most economical way.
Through holding company subsidiaries, banks
and thrifts now can perform virtually all banking
functions across state lines except deposit-gathering. (With the advent of brokered deposits and
nonbank banks, they are not fully constrained even
in this last area.) Moreover, individual states are
now accommodating interstate entry. Thirty-seven
states have passed legislation permitting entry by
banking firms located out of state. Eighteen of those
states permit, or will soon permit, entry by banks
headquartered anywhere in the country.
Commercial and investment banking firms are
expanding their international activities not only to
"follow their customers" but also to take advantage
of international regulatory discrepancies. They
have, for example, been attracted to the London and
Tokyo markets, which recently have loosened
restrictions on the activities of market participants.
At the same time, domestic restrictions also are
pushing U. S. financial institutions overseas.
Restrictions on commercial banks' securities underwriting activities as well as reserve requirements
and deposit-rate regulations, for example, have
induced U. S. banks to shift business to international
markets where they can avoid domestic regulations. 5
The trend towards interstate provision of domestic financial services is even more pronounced. 6
Banks are seeking to establish regional and even
III.
Implications and Issues
These developments raise a number of public
policy concerns. First, the present approach to regulation of the financial system encourages an inefficient use of resources. For example, resources are
devoted to discovering and exploiting loopholes in
the current legal and regulatory system. More
importantly, the result of this process is a structurally inefficient financial industry that is characterized by a proliferation of new instruments, transfer of traditional banking activity to nonbanks, and
payments volumes that are excessive in relation to
economic activity.
Second, without deposit insurance reform, the
expansion of financial activity of banks or the integration of financial and commercial activities may
lead to an undesirable propagation of the deposit
insurance subsidy. One concern, for example, is that
a stressed nonbank affiliate might draw financial
support from the bank, endanger the bank, and
indirectly be supported by the deposit insurance
fund.
Third, the growth of international financial centers and of unregulated firms' involvement in the
provision of financial services implies diminished
federal supervisory leverage over financial activity
that may be essential to financial stability. Diminished supervisory control is particularly trouble-
some in light of concern about the potential for
undesired or unintended de facto extension of the
federal safety net.
The current legal, regulatory, deposit insurance,
and payments frameworks are. inadequate for
addressing these policy concerns. Reform is needed
to preserve financial stability and to accommodate a
changing financial environment. However, such
reform must balance the benefits from enhancing
stability against the costs. For example, stability
could be enhanced (in the short run) if deposit
insurance were extended to every financial firm.
Absent deposit insurance reform, however, such an
approach would distort risk-taking decisions (or
require a vast expenditure of supervisory resources
to prevent the distortion).
Since it is not feasible or desirable to insure every
firm or activity, we must decide what truly needs to
be protected. Although the extent of insurance
coverage is a subject of intense debate, nearly all
agree that protection of transactions balances
(whether held at commercial banks or at nonbanks)
is essential. 7 Also, because many observers are
concerned that a serious contraction in the availability of intermediated credit from depository
institutions could have destabilizing consequences,
there also is a view that a significant part of all
10
nontransactions balances needs to be protected as
well. 8 At the same time, however, there is the
concern that this protection not extend too far.
Clearly, we must not protect the owners of creditgranting intermediaries from the consequences of
their decisions lest we run the risk of excessive risktaking on their part.
Although the question of what ought to be protected has no simple answer, most observers conclude that both the payment and credit intermediation functions of depositories need partial, if not
fairly extensive, protection. Many of the issues
discussed below regarding the structure of the
deposit insurance system, the boundaries of bank
powers, and the operation of the payments system
are predicated on this conclusion. The conclusion
itself, however, is open to debate.
premia unrelated to risk, possible coverage of all
deposit and nondeposit liabilities (at least at large
banks), and a willingness to let insolvent banks and
thrifts continue to operate has seriously undermined
the discipline on
that would ottlerWl~;e
be imposed by the market. 9
As a result, regulation and supervision bear the
main burden of limiting risk-taking. However,
should the
of
insurance
continue to expand, the
of containing
bank risks with supervision and regulation would
dim and leave the government to underwrite risks
for larger and larger segments of the economy.
Thus, reform of the deposit insurance system is
central to and a prerequisite for financial reform.
Indeed, it may be needed just to deal with the
current economic environment, as exemplified by
the problems of the savings and loan industry and its
insurance fund.
Deposit Insurance and the
Federal Safety Net
Our present deposit insurance system actually has
performed remarkably well over the last fifty years.
Although there have been runs on individual banks,
spillover effects have been limited and there have
not been any banking panics at federally insured
institutions. In addition, payouts from insurance
funds were very modest prior to the 1970s.
More recently, however, many observers have
begun to question the viability of the system in the
wake of a record number of bank failures, the large
foreign debt exposures of the money center banks,
and the well-publicized problems of the FSLIC.
One can argue that some of the recent problems
stem from an implicit (and at times, explicit) extension of the federal safety net well beyond the stated
coverage of deposit insurance. One might even say
that the safety net has been spread so thinly it may
soon tear. Moreover, because the current system
relies so heavily on supervision and regulation, it
has become increasingly unable to accommodate
the market forces and trends enumerated above.
Approaches to Reform
There have been many proposals for reforming
the deposit insurance system. Some involve restricting the explicit or implicit scope of deposit insurance coverage while others seek to "reprice" insurance to reduce the moral hazard problem. Below,
the pros and cons of various alternatives are discussed.
Reducing the Scope of Deposit Insurance
Perhaps one of the oldest reform proposals dates
back to Henry Simons' 1948 proposal for 100 percent reserve banking as modified by Milton Friedman in 1959 to include the payment of interest on
reserves. IO This idea, which in essence has been
revived by Robert Litan and John Kareken among
others, 11 would turn banks into institutions similar
to money market mutual funds
that
banks'
liabilities would be used to fund
safe assets,
such as short-term government securities, cash, and
reserve balances at the Federal Reserve.
If banks were required to back their liabilities
with only "perfectly safe" assets, they could not
fail. Moreover, no restrictions on the ownership of
such "eunuch" banks would be necessary since
there would be no opportunity for the bank to
The Status Quo
It is now widely recognized that the current
deposit insurance system introduces a moral hazard;
that is, it gives insured institutions an incentive to
take on excessive risk. The combination of flat-rate
II
"payouts." If this were done, deposits not fully
insured would be subject to an immediate markdown and might never be fully repaid. (Under the
experimental modified payout plan, uninsured
qepositors of some closed banks received only a
prorated portion of the estimated value of the failed
banks' assets immediately. Uninsured depositors of
such failed banks could receive additional payments
if, upon disposal of the assets, the realized value
exceeded the FDIC's estimate. The actual plan,
however, did not shift losses to depositors if a
"failed" bank were handled through a purchase and
assumption.) 12
Increasing depositor surveillance through these
avenues would reduce ex ante risk-taking because
uninsured depositors would require premium rates
for the riskier institutions and the threat of a run by
depositors would induce institutions to operate prudently. Over time, the level of supervision and
regulation could be scaled back, although there
would be a need for more public information about
the conditions of banks. Both of these developments
would accommodate the natural evolution of the
financial system. However, the proposal provides
little protection against runs by uninsured depositors. To the extent that one believes that runs by
uninsured depositors are potentially destabilizing to
the financial system, as apparently was the view of
regulators in the episode involving the failure of
Continental Illinois in 1984 13, proposals of this
nature do not offer sufficient protection.
In sum, limiting the scope of deposit insurance,
either by limiting the functions of insured institutions or by limiting insured deposit coverage, could
reduce or eliminate the moral hazard implicit in the
current deposit insurance system. However, these
proposals would increase the potential for credit
runs that could destabilize the financial system. An
alternative approach to reform is to maintain fairly
broad insurance coverage of the payments and
credit functions of financial intermediaries while
"repricing" that coverage to reduce the moral hazard.
support failing nonbank affiliates. (The bank's
deposit liabilities would be used to fund only safe
assets and not to fund any form of credit to affiliates,
including intraday payments credit.) Under these
conditions, which imply complete legal and economic separation of the bank from nonbank affiliates, the failure of a nonbank affiliate could not
impair the bank. Of course, if banks held assets that
were "fairly," but not perfectly, safe and were
allowed to extend credit to subsidiaries, or to lend
their" good name" to the subsidiaries in a way that
implied legal liability, then the problems of controlling the risks undertaken by a diversified conglomerate would arise.
Implicit in this "safe assets" approach is the
notion that deposit insurance should protect only the
payments system or payments-related balances. In
fact, under this proposal, meaningful credit intermediation would take place only in uninsured financial institutions or subsidiaries similar to currentday banks in most respects except for their inability
to offer insured pure transactions accounts.
Although uninsured intermediaries presumably
would take on fairly conservative risk postures, they
probably would still use short-term liabilities to
fund risky, longer-term loans to some degree and
thus could be subject to problems with depositor
runs. These runs have the potential for destabilizing
the credit system. Thus, although the safe assets
proposal might provide adequate protection for the
payments function of depositories, it would offer no
protection for credit intermediation.
Another approach to limiting the scope of deposit
insurance focuses on explicitly restricting the
payouts made to depositors to encourage depositor
surveillance of depositories' risk-taking. Traditionally, the insurance system has restricted payouts
by fully insuring each deposit only up to some
maximum amount. But other approaches could be
taken, such as also explicitly insuring a given percentage of deposits above the maximum amount. Of
course, to be meaningful, maximum insurance
coverage would have to be enforced strictly and
uniformly for all failed banks and "assisted" mergers that expose the insuring agency to losses.
The FDIC's "modified payout" propoool would
work well in this context if it were applied uniformly
to "purchases and assumptions" as well as to
Repricing Deposit Insurance
The most obvious way to reprice deposit insurance is to charge an insurance premium that rises
12
or acquired commercial firms. Another difficulty is
the lack of legal authority for the insuring agency to
require chartering agencies promptly to close
institutions deemed insolvent on the basis of a
market vahle assessment of equity.
As a result, any practical implementation of this
approach would have to allow for errors in closure.
If depositors believed a bank might be closed too
late,
would run unless they could
be assured that losses would be covered by a third
party, such as subordinated debt holders and/or the
deposit insurance fund. (To be effective, subordinated debt should be perpetual and subordinated to
both bank deposits and the insurance fund.)
There are other ways of accommodating errors in
assessing market values. One would be to give
regulators the authority to err on the safe side either
by closing a bank that might still have a positive
market value of equity or by requiring the bank to
increase its equityJo reduce the risk of ex post
uninsured depositor or insurance fund losses.
Although politically impractical today, yet another
method would be to hold bank equityholders liable
for losses exceeding their original capital, as was the
case prior to the 1930s, when stockholders of
nationally chartered banks were liable for losses up
to twice the par value of the stock owned.
The implementation of prompt market-value closure would raise many political problems,
especially during a transitional period. For example,
the closure of institutions that are currently insolvent would raise major problems for the FSLIC, and
possibly even the FDIC. However, these are the very
institutions that now pose the gravest threat to the
insurance funds. Nevertheless, it would be possible
and desirable to move closer to market-value
accounting and closure rules. Moreover, once
insured institutions adjust to a truly unforgiving
closure policy, they would voluntarily hold more
capital in relation to the riskiness of their portfolios
to reduce or eliminate the risk of being declared
insolvent.
The current risk-based capital proposal, which
requires banks with more risky assets and off-balance sheet activities to hold more capital, can be
considered a step in the same direction. Such proposals, however, will succeed in eliminating or
reducing the moral hazard in deposit insurance only
with the ex ante risk of the insured institution's
portfolio. This is a sound concept because it would
penalize bank equityholders for excessive risk-taking and thus would internalize the costs of risktaking along with the benefits.
In practice, however, this proposal could prove
extremely difficult to implement because it would
require charging an insurance premium based on
examiners' assessments of the ex ante market values
and risks of a bank's portfolio of assets, many of
which are not traded or readily marketable, as well
as judging the risks and potential profitabilities of its
non-portfolio activities. Moreover, to have a significant impact on ex ante risk-taking, examiners' risk
assessments would have to look well to the future,
and premia might have to be adjusted fairly dramatically on the basis of subjective risk assessments.
(The FDIC's legislative proposal to double the
annual premium to one-sixth of a percent of deposits
for banks in the high-risk category would not be
sufficient to deter risk-taking.)
A second method of internalizing risk requires
that insured institutions be closed before the market
value of their equity could fall below zero. If this
could be accomplished without error, a closed
institution's assets necessarily would be sufficient to
discharge its liabilities at the time of liquidation. As
a result, failed (that is, closed) institutions would
not impose losses on the insurance fund. Instead,
bank equityholders would bear the full costs and
benefits of their decisions and would have no incentive to take excessive risks.
Moreover, as long as depositors were confident
that regulators would be successful in closing banks
before the market value of equity became negative,
and thus assure them of protection from losses, they
would not run on a "troubled" bank. In this manner,
it would be possible, in concept, to protect deposits
and prevent runs while simultaneously confining
risk to bank equityholders.
To be effective, however, this approach would
require increases in both the scope and frequency of
federal supervision of insured institutions to monitor the market values of their equity closely. One
major practical difficulty in increasing supervision
lies in assigning accurate market values to nontraded assets and liabilities. Such valuation might
be even more difficult ifbanks took on added powers
13
if they help to ensure that insured institutions maintain a positive market value of capital over a wide
range of possible ex post outcomes. Since riskier
assets have a higher probability of declining in
value, requiring additional capital for these assets ex
ante increases the probability of a positive market
value ex post.
Like a scheme of risk-based deposit insurance
a true risk-based capital approach would
require ex ante estimates of the value and riskiness
of each type of asset as well as its contribution to the
overall riskiness of the portfolio. However, an
approach requiring banks to hold additional capital
(even if based on a fairly crude assessment of risk)
probably would be easier to implement and less
likely to generate errors that cause major distortions
than a system of risk-based deposit insurance premia.
In sum, there are practical and political problems
with each of the approaches" to insurance reform
described. But if we wish to maintain deposit
insurance coverage that is as extensive as what we
have now, reform is necessary. The optimal
approach probably will involve a blend of reforms.
ing firms to shift activities to nonbank subsidiaries.
In fact, one bank consulting firm has advocated a
corporate restructuring dubbed "double de-banking" in which the bank holding company relinquishes its commercial bank charter in favor of a
nonbank bank charter (to retain payments system
access) while placing all of its other financing,
underwriting, and loan servicing activities in separate nonbank subsidiaries. 14
The basic conflict between economic forces and
regulation extends beyond domestic markets. As
activity continues to shift to less-regulated international centers, bank regulators will find themselves
regulating and supervising a shrinking share of total
financial activity. To the extent that the quality of
supervision deteriorates because of the difficulty of
supervising an international banking organization
in its entirety, the stability of the financial system
could be threatened. These challenges to supervision could be overcome, in part, by coordinating
supervision and regulation in the world's three most
important financial centers - New York, London,
and Tokyo. The U.S.-U.K. risk-based capital proposal is a first step. Nonetheless, because of restrictions on domestic banking powers, there remain
strong incentives to shift activity toward less regulated environments.
Resolving the bank powers issue requires careful
balancing of disparate concerns. On the one hand,
because federal oversight and protection of some
portion of financial activity is essential to stability,
regulation must not be so at odds with market forces
that important financial activities shift away from
federal control. On the other hand, because the
provision of a federal safety net creates incentives
for excessive risk-taking, some minimum level of
regulation, or at least supervision, is necessary.
Bank Powers
At the heart of the conflict between the natural
evolution of the financial system and the legal and
regulatory structure governing that system is the
issue of bank powers. The current restrictions on
bank ownership and powers, enumerated in the
Glass-Steagall and Bank Holding Company Acts,
stand in the way of the trend towards functional
realignment in the provision of financial services.
While market forces will foster the development of
alternatives to bank-provided payments and credit
services, these alternatives may not be the most
efficient from society's perspective.
Specifically, preservation of the current restrictions on bank powers will cause financial activity to
continue to shift away from banks to nonbank
banks, thrifts, and investment banks. This shift
implies both a relative decline in business transacted
by banking firms and a rearrangement of activity
within the corporate structure of bank holding companies. Failure to resolve the nonbank bank issue
will lead to a decline in the value of the traditional
commercial bank charter, and may even cause bank-
Separation of Powers
Before we consider the extent to which bank
powers ought to be expanded in response to market
pressures, it may be useful to reconsider the original
rationale for separating banking from other financial
services and from commerce. Of primary concern to
legislators in the 1930s were the problems associated with concentration of resources and the potential for self-dealing. Such problems have been
14
addressed, with varying degrees of success, in other
countries without completely separating banking
and securities markets. IS Moreover, since the
1930s, the problems may have been mitigated to
some extent in the U.S. by SEC regulations and
surveillance. Likewise, antitrust restrictions should
serve to prevent excessive concentration and anticompetitive behavior. Finally, if greater integration
of financial services were allowed, the concentration of total financial resources might increase,
whereas the concentration for particular services
actually might decrease because a wider variety of
firms would be providing them.
Unlike the 1930s, a key concern regarding bank
powers today is the possibility that banking organizations would shelter additional activities under the
federal safety net. For this reason, some have argued
against expanding the powers of banking organizations, while others have argued that new powers be
carried out only in separate subsidiaries. Most
observers agree, however, that the type of corporate
separability that we have today is not very likely to
insulate the bank from losses of a nonbank affiliate
in times of stress. 16 Truly effective corporate separateness might require completely separate identities for the bank and nonbank affiliates, separate
boards of directors, and severe limitations on interaffiliate transactions. Such an approach might
severely restrict or even eliminate any potential
synergies the consolidated organization otherwise
might enjoy. 17
There is yet another view on the bank powers
problem. Reform of the deposit insurance system to
reduce its risk-taking incentives would make it
easier to expand bank powers in response to market
forces. With fewer limitations on bank powers, there
would be less incentive for financial activity to shift
away from federally supervised institutions.
This approach would accommodate the trend
towards functional realignment in the provision of
financial services. It also would enhance the efficiency of the financial system by, among other
things, enabling banks both to originate underlying
assets and then to underwrite and sell derivative
securities.
This approach may require increased surveillance
of the activities of the consolidated enterprise since
it increases a bank's opportunities for risk-taking.
Such surveillance need not be a major stumbling
block, however. In other countries where greater
integration of financial services is allowed, regulators apparently have been able to supervise the
activities of financial conglomerates. 18 Of course,
such supervision may be easier to carry out in
countries where there is only a handful of large
banks.
Expand the Commercial Powers of Banks
A second general approach to the reform of bank
powers would be to expand both the financial and
commercial powers of banks. This approach would
enable banks to own and control commercial firms
and vice versa. Concerns regarding increased concentration of corporate control could be resolved
through ownership limits, as have been established
in West Germany, to prevent banks from exercising
too much influence over the economy.
Once again, however, expanding bank powers in
this way could complicate the assessment of the
risks borne by the deposit insurance system. For
example, the pressure to lend to troubled "house"
firms may increase affiliate banks' risk unless
federal supervisors can evaluate the soundness of all
inter-affiliate transactions. (Alternatively, bank regulators could ban all inter-affiliate transactions, but
if the ban were effective, it would severely reduce
the benefits of conglomeration.) Reform of the
deposit insurance system would, in theory, reduce
the problem of increased risk. In practice, however,
fully effective reform rests on the ability of regulators to monitor the market value of the consolidated
enterprise
a difficult task, at best.
Given these difficulties, it is debatable just how
far we should proceed in the direction of allowing
banks to affiliate with commercial firms. One
Expand the Financial Powers afBanks
Along with a program for meaningful insurance
reform, two broad reforms of bank powers might be
considered. First, we might consider expanding the
financial powers of banks. In other words, banks
might be allowed to underwrite and trade securities,
underwrite and sell insurance, manage mutual
funds, and offer other financially related services.
15
advantage of such affiliations would be the reduction of risk through the conglomeration of dissimilar
activities. However, the operating synergies
between banking and commerce do not appear to be
great. Instead, there is some evidence that commercial firms are seeking banking powers primarily
because they desire access to the payments system
and wish to take advantage of related marketing
synergies. If this were true, one way to resolve the
issue of integrating banking and commerce would
be to grant nondepository firms access only to the
payments system provided they collateralized their
transactions.
drafts. Although the Fed is charging an implicit
price on very large intraday credit activity as the
result of a policy of limiting daylight overdrafts, it
does not price most intraday Fed credit. Second, the
Federal Reserve does not charge for the default risk
itassumesby offering finality of payment on Fedwire. Thus, receivers of funds on Fedwire are not a
potential source of discipline in the payment-credit
decision. This distribution of risk differs from that
ofprivate networks where the provisional nature of
transactions makes receivers evaluate the creditworthiness of payors.
Finally, some argue that there are externalities
associated with payments activity that lead to the
underpricing of a credit associated with payments
even on wholly private networks. In particular, they
argue that individual payments are transacted in
ignorance of the burden that would be imposed on
others should that transaction fail. If this view were
correct, private charges for payments credit would
be lower than the social cost of that credit. This and
other causes of underpriced payments credit encourage the use of intraday credit that may be too large
from the viewpoint of economic efficiency.
The delayed settlement feature of present day
private payments systems adds to the concerns
raised by underpricing. Delayed settlement
increases the chances that an adverse event will
nullify transactions that have already taken place.
As payments activity grows and the interconnectedness of the payments system increases, some argue
that the likelihood of such disruptions will increase.
Combined with excessive use of payments system
credit because of underpricing, this additional concern raises the risk of coincident liquidity or solvency problems for participants that could, in tum,
precipitate a general loss of confidence in the payments system.
Although intervention by the central bank should
be able to protect the economy from such liquidity
problems, such intervention is not costless and, if
perfonned frequently, could create additional incentives for risk-taking, particularly if the intervention
extends beyond providing liquidity to ensuring solvency. Thus, the problem of excessive payments
system risk- like excessive risk-taking in other
facets of banking - is a serious concern of current
payments system policy.
The Payments System
The major trends enumerated here bear importantly on the functioning of the payments system.
Increased financial activity, securitization, and
internationalization of markets presage a growing
payments volume. There is legitimate concern that
these trends may increase both the possibility and
consequences of losses arising from a payments
system malfunction or from the failure of a major
participant in the system. 19
In a payments system that uses the creation and
extinction of credit to facilitate payments activity,
such failures can generate liquidity problems for
participants. With highly interconnected payments
flows that rely on credit, a single failure can cascade
into liquidity problems throughout the payments
network. One of the functions of a central bank, of
course, is to provide liquidity to sound institutions
in such circumstances. However, central bank payments system policy should not imply protection
against insolvency or even encourage frequent use
of the emergency liquidity facility.
The Status Quo
The consequences of maintaining current payments conventions in light of anticipated growth
trends in the volume of payments are worth considering. The payments system now entails underpriced intraday credit, delayed settlement, and
access that is limited to depository institutions.
Underpriced intraday credit arises in several
ways. First, the Federal Reserve encourages use of
intraday credit by not charging for daylight over-
16
The third feature of concern in the current payments system is access that is limited to depository
institutions. Nonbank institutions have been overcoming the limitation through thrift and nonbank
bank ownership: In addition, nondepository firms
are using sweep-type arrangements to provide payments services to their customers. These arrangements, however, may affect the size and timing of
payments activity· in an undesirable way from the
standpoint of payments system risk. For these reasons, payments system access is of increasing concern whether or not there is a change in explicit
access policy.
risk in the private market, funds receivers in the
private market do have an incentive to monitor and
control their risk exposure. Private bilateral payments decisions, however, may not automatically
take· into account the total "social" credit risk
involved. Reduction of this risk requires surveillance by the appropriate regulators and the principal participants in private payments networks.
Such surveillance may require minimum participant
capital (or liquid reserve) requirements, net debit
limits,or other risk-limiting devices such as those
currently employed by private intraday credit systems such as CHIPS and Euroclear.
Analogous to charging interest on intraday overdrafts, interest also should be paid on positive
balances. Symmetry in the treatment of borrowing
from and lending to the Federal Reserve System
would improve the functioning of the private intraday credit market. It also would decrease Fedwire
congestion associated with attempts to maintain
minimum required reserve balances at the end of the
day, and would enhance the attractiveness of holding corporate demand deposits at banks (which also
should be allowed to yield explicit interest).
Pricing Fed Credit
Further progress toward the pricing or rationing
of intraday Federal Reserve credit would remove a
major stimulus to the overuse of intraday credit,
both on Fedwire and on private wholesale networks.
These additional steps should be taken despite
issues raised by Federal Reserve payments queues
and computer malfunctions, although improvements in these areas should be made in conjunction
with pricing efforts.
Ideally, intraday credit pricing would embody not
only the time value offunds, but also the value of the
default risk implicitly assumed by the Federal
Reserve in granting finality of payment on Fedwire.
This approach would simulate the discipline exerted
by receivers of funds in the private intraday credit
market, and reduce the direct credit risk to which the
Federal Reserve System would be exposed.
With a positive price for intraday credit, overall
use of such credit would decline. In the short run,
this decline may retard the growth of activities that
have become reliant on underpriced intraday credit,
such as churning in the securities market and the
corporate cash management process generally. It is
not clear, however, that the current level of payments
activity (involving daily flows of $1 trillion or more)
is efficient, whereas it is clear that the current
system induces inadequate credit evaluation. The
latter increases the risk of payments failure on
private networks, or, alternatively, the risk to the Fed
on Fedwire.
Pricing Fedwire intraday credit presumably
would push more payments activity into the private
credit market. Although such a shift might increase
Real-Time Settlement
In addition to providing better management of
risk in a delayed-settlement environment, an
increased price for intraday credit will encourage a
transition toward "real-time settlement" whereby
both monitoring of positions and matching of payments flows will occur on a continuous basis. A
payments system should be a credit system only if it
is more efficient to bridge temporal gaps between
the payment and receipt of funds through borrowing
than through expending resources to make transactions synchronous. Under the current system, borrowing and asynchronous payments are favored.
With costly intraday credit, participants will seek
the means to synchronize transactions and settle
obligations in "real time." For example, repayment
of funds borrowed overnight will be more closely
matched in time with funds inflows that reflect
borrowing for the next night. Such operations, if
exactly matched in time, will reduce overdraft
exposure by substituting a relatively small net transfer (the difference between the two borrowings) for
two gross transfers mismatched in time.
17
and real-time payments technology increasingly
will be needed to manage risk economically.
Finally, by resolving the problems of underpriced
intraday credit and delayed settlement, there would
be •less .need to~ontinue. to limit· access to the
payments system. An orderly expansion of payments system access, in conjunction with these
other reforms, likely would not pose undue risk and
would resolve the problems created by non-depository firms exploiting various loopholes in the current policy.
Real-time settlement is becoming increasingly
feasible as communications and electronic accounting technologies advance. Since real-time settlement eliminates, by definition, temporal risk in the
the evolution toward real-time
settlement will contribute significantly to reducing
payments system risk. Many transactions may be
quite costly to settle in real time, of course, and the
will continue to involve credit
extension to some degree. However, as around-theclock and global securities trading progresses, the
importance of managing temporal risk will mount,
IV.
Summary and Conclusions
could be left underwriting the risks of an everincreasing share of the economy.
Similarly, the implicit government guarantee
behind the payments system may prove to be unsustainable in the face of rapid financial innovation.
Underpriced intraday credit in conjunction with
delayed settlement appears to be a major part of the
problem. Without reforms in these areas, expanded
payments system access poses further risks.
Finally, banks are experiencing economic pressures to expand into nontraditional activities. A
major reason for preventing them from doing so is to
protect the deposit insurance and payments guarantees. However, many observers question whether
the U. S. banking industry will be able to compete
effectively if it continues to be regulated more
stringently than domestic nonbank firms and banking firms in other countries.
Clearly, market forces for change are posing
serious challenges to the current financial regulatory framework and safety net. By reforming the
legal and regulatory framework to accommodate
these forces and to encourage more market discipline of risk-taking, we can move toward a more
efficient and stable financial system. Undoubtedly,
a blend of many of the approaches touched upon
here will be needed to reach these goals.
A financial revolution is underway. Already we
see glimpses of the new financial world in the forms
of increased securitization, a diminished role for
bank-provided intermediation, functional realignment in and geographic integration of financial
services, and expanded access by nonbank firms to
the payments system. These are trends driven both
by fundamental economic forces and attempts to
circumvent regulation and to exploit government
guarantees.
Many of these changes have not resulted from
explicit policy choices. While most would admit
that a thorough reform of financial regulatory and
legal policy is long overdue, the continuing debate
over just what changes are necessary apparently has
paralyzed the policymaking process.
Although there are no easy or simple solutions,
the time has come to move forward because failure
to make the needed changes may threaten financial
stability. Three areas are especially in need of thorough reform: the federal safety net, the payments
system, and bank powers.
The
problem with our current deposit
insurance system is that it provides an incentive for
excessive risk-taking, which could propagate
throughout the economy as distinctions between
banks and nonbank firms diminish. Without
changes in the insurance system, the government
18
FOOTNOTES
1. Summary of the Results of the August 1985 Senior Loan
Officer Opinion Survey on Bank Lending Practices conducted by the Federal Reserve System.
10. Milton Friedman, A Program for Monetary Stability,
New York: Fordham University Press, 1959; and Henry
Simons, Economic Policy For a Free Society, Chicago:
University of Chicago Press, 1948.
2. Source: Federal Reserve Bulletin, various years. Additional sources on securitization are: Christine Pavel,
"Securitization," Economic Perspectives, Federal Reserve
Bank of Chicago, JulylAugust 1986; and Randall Pozdena,
"Securitization and Banking," Weekly Letter, Federal
Reserve Bank of San Francisco, July 4, 1986.
11. Robert Utan, "Taking the Dangers Out of Bank Regulation," The Brookings Review, Fall 1986; and John
Kareken, "Ensuring Financial Stability," in The Search for
Financial Stability: The Past 50 Years, San Francisco: The
Federal Reserve Bank of San Francisco, 1985.
3. Michael Keeley, "Interest on Business Checking
Accounts?," Weekly Letter, Federal Reserve Biink of San
Francisco, May 2, 1986.
12. Frederick Furlong, "FDIC's Modified Payout Plan,"
Weekly Letter, Federal Reserve Bank of San Francisco,
May 18, 1984.
4. For additional information on functional realignment,
see: Bank for International Settlements, "Recent Innovations in International Banking," April 1986; Alan Daskin and
Jeffrey Marquardt, "The Separation of Banking and the
Securities Business: A View of the United Kingdom, West
Germany and Japan," The World of Banking, MaylJune
1984; Steven Felgran, "Bank Entry Into Securities Brokerage: Competitive and Legal Aspects," Bankers Desk
Reference: New Topics, 1985; Samuel Hayes, "Investment
Banking: Commercial Banks' Inroads," Economic Review,
Federal Reserve Bank of Atlanta, May 1984; and Christine
Pavel and Harvey Rosenblum, "Banks and Nonbanks: The
Horse Race Continues," Economic Perspectives, Federal
Reserve Bank of Chicago, May/June 1985.
13. Frederick Furlong, "Market Responses to Continental
Illinois," Weekly Letter, August 31, 1984.
14. "Continued Inaction by Congress May Spur Banks to
Drop Charters," American Banker, November 6, 1986.
15. Daskin and Marquardt, op cit.
16. See Anthony Cornyn, Gerald Hanweck, Stephen
Rhoades and John Rose, "An Analysis of the Concept of
Corporate Separateness in BHC Regulation from an Economic Perspective," Appendix C. to the Statement by Paul
A. Volcker, June 1986. For an opposing view, see Samuel
Chase and Donn Waage, "Corporate Separateness as a
Tool of Bank Regulation," Samuel Chase and Company for
the American Bankers Association, October, 1983; and
Carter Golembe and John Mingo, "Can Supervision and
Regulation Ensure Financial Stability?," in The Search for
Financial Stability; The Past 50 Years, San Francisco: The
Federal Reserve Bank of San Francisco, 1985.
5. See also, Bank for International Settlements, "Recent
Innovations in International Banking."
6. See Donald Savage, "Interstate Banking Developments," Federal Reserve Bulletin, February 1987; Federal
Reserve Bank of Chicago, Toward Nationwide Banking: A
Guide to the Issues, 1986; Constance Dunham and
Richard Syron, "Interstate Banking: The Drive to Consolidate," New England Economic Review, Federal Reserve
Bank of Boston, May/June 1984; and Frank King, "Interstate Banking: Issues and Evidence," Economic Review,
Federal Reserve Bank of Atlanta, April 1984.
17. Paul Volcker, "Statement Before the Subcommittee on
Commerce, Consumer and Monetary Affairs of the U.S.
House of Representatives," June 11, 1986.
18. Daskin and Marquardt, op cit.
19. For further discussion of payments system risk, see:
David Humphrey, "The U.S. Payments System: Costs,
Pricing, Competition and Risk," Monograph Series in
Finance and Economics, no. 1984-1/2, New York: Salomon
Brothers Center for the Study of Financial Institutions,
Graduate School of Business Administration, New York'
University, 1984; David Mengle, "Daylight Overdrafts and
Payments System Risks," Economic Review, Federal
Reserve Bank of Richmond, May/June 1985; and E.J.
Stevens, "Risk in Large-Dollar Transfer Systems," Economic Review, Federal Reserve Bank of Cleveland. Fall
1984.
7. See, for example: E. Gerald Corrigan, "Are Banks Special? A Summary," Federal Reserve Bank of Minneapolis
Annual Report, 1982; and Michael Keeley and Frederick
Furlong, "Bank Regulation and the Public Interest," Economic Review, Federal Reserve Bank of San Francisco,
Spring, 1986.
8. Ben Bernanke and Mark Gertler, "Agency Costs, Collateral and Business Fluctuations," a paper presented at the
Federal Reserve Bank of San Francisco's Fall Academic
Conference, 1986.
9. For discussion of this issue, see: Frederick Furlong and
Michael Keeley, "Bank Runs," Weekly Letter, Federal
Reserve Bank of San Francisco, July 25, 1986; and Barbara Bennett, "Bank Regulation and Deposit Insurance:
Controlling the FDIC's Losses," Economic Review, Federal
Reserve Bank of San Francisco, Spring 1984.
19
Frederick T. Furlong and Michael C. Keeley*
This paper examines theoretically the effects ofmore stringent capital
regulation on a bank's incentive to increase asset risk and on the expected
liability ofthe deposit insurance system. Our analysis shows that regulatory increases in capital standards will not require greater regulatory
efforts to restrain asset risk because a bank's incentive to increase asset
risk declines as its capital increases. Thus, as long as regulatory efforts to
contain asset risk, such as bank examinations, are not reduced, more
stringent capital regulation will reduce the expected liability of the
deposit insurance system.
1970s, rose to about $2 billion per year in 1985 and
1986. Such increases in expenses for the deposit
insurance system have focused attention on increasing the stringency of bank capital regulation to limit
the FDIC's exposure to losses and to blunt the
incentives for "excessive" risk-taking by federally
insured banks.
The move to more stringent capital standards in
banking, however, has met with considerable controversy as well as some skepticism. Some argue
that higher capital requirements will cause banks
simply to invest in more risky assets, and thereby
offset, or even more than offset, the desired effects
of higher capital. This view often is echoed in the
financial press. In a New York Times article about a
Federal Reserve proposal to require banks to hold
capital in connection with agreements involving
interest rate and currency swaps, William
McDonough, vice chairman of First National Bank
of Chicago, is quoted as saying that " ... the
proposal could lead banks to take on riskier business
to compensate for the lower returns they would
almost assuredly get by having to maintain more
capital. "4
The effectiveness of capital regulation also has
come under question in the academic literature. A
study by Koehn and Santomero (1980), which
assumes that banks maximize utility in a meanvariance framework, 5 is representative of the literature on the theoretical relationship between capital
Over the past several years, bank regulators have
placed greater emphasis on the regulation of bank
capital. For example, the three federal bank regulatory agencies have raised capital requirements for
banks and bank holding companies and established
more uniform standards among themselves. I Most
recently, the federal bank regulatory agencies have
put forth proposals for risk-based capital requirements that would be coordinated with the Bank of
England. 2
These regulatory measures, in part, are reactions
to deteriorating capital positions, particularly
among the larger banking organizations. For example, among the twenty largest bank holding companies, the average ratio of the book value of
common equity to assets was over 6 percent in 1968
but only about 4 percent in 1980. 3 The increase in
the number of bank failures and the correspondingly
sharp rise in the Federal Deposit Insurance Corporation's (FDIC's) expenses in recent years also have
intensified interest in capital regulation. Total
expenses of the FDIC, which fluctuated between
about $50 million and $200 million per year in the
* Senior Economists, Federal Reserve Bank of San
Francisco. Comments from Jack Beebe, Christopher
James, David Pyle, Anthony Santomero and John
Scadding on earlier drafts are appreciated. Research
assistance from Alice Jacobson and John Nielsen is
gratefully acknowledged.
20
requirements and bank asset risk.. They conclude
that ". . . a case could be argued that the opposite
result can be expected to that which is desired when
higher capital requirements are imposed."6
In this paper, we evaluate the popular view of
ca,pjtal rygulatipn and the conclusions of earlier
theoretical studies on the effectiveness of capital
regulation. In contrast to the popular view and the
earlier academic work, we find that more stringent
capital standards alone would not give a bank more
of an incentive to increase the riskiness of its assets.
In fact, the incentive to increase asset risk falls as a
bank's capital increases. This implies that, as long
as regulatory and supervisory efforts to limit asset
risk in banking, such as bank examinations, are not
relaxed, increasing a bank's capital will lower that
bank's chance of failure and reduce the expected
liability of the deposit insurance system.
We also show in the Appendix that the conclusions reached by earlier theoretical studies using the
mean-variance framework were derived from internally inconsistent assumptions. In essence, these
studies implicitly (but unintentionally) assume that
bank failure is not possible by assuming that borrowing costs are unrelated to bank risk. Yet, they
seek to analyze the effects of capital regulation on
the probability of bank failure. Moreover, these
studies fail to incorporate the effect of the deposit
insurance guarantee on risk-taking. Although the
results of these studies regarding the effects of
capital regulation on the incentive to increase asset
risk are technically correct when bank failure is not
possible, such findings are of little policy relevance
since capital regulation and concern over risk-taking are relevant only when banks can fail.
The Appendix also contains an example to show
that the results of these earlier studies do not generany hold when subsidized deposit insurance and the
possibility of bankruptcy are taken into account.
Specifically, we show that when the asset return
distribution is binomial, the incentive to increase
asset riskdoes not increase as the stringency of
capital regulation increases.
The analytic framework used in the body of this
paper is the state-preference mode! rather than the
mean-variance model used in the older literature on
the topic. One reason for this choice is that the statepreference model, unlike the mean-variance model,
can easily accommodate the possibility of bankruptcy and an analysis of the effects of mispriced
deposit insurance on a bank's choice ofleverage and
asset risk. 7 Moreover, with the state-preference
model, the effects of changes in capital requirements on both banks' gains from increasing asset
risk and the expected liability of the deposit insurance system can be evaluated directly.
Another advantage of the state-preference framework is that it can be applied to the analysis of both
value-maximizing and utility-maximizing banks.
Utility maximization might be appropriate for certain smaller, closely held' banks where the owners'
risk preferences affect the riskiness of the banks'
portfolios, whereas value maximization is more
suitable for most other banks, particularly the large
publicly held banks whose stockholders can hold
diversified portfolios. Value-maximizing banks
would seek to maximize the current market value of
their equity, which is independent of the risk preferences of the owners. 8
In the next section, we start with a discussion of
the nature of bank capital and the issues that higher
capital requirements raise for bank regulators. In
Section II we introduce the state-preference model
and use it to analyze the effects of capital regulation
on asset risk and the liability of the federal deposit
insurance system, under the assumption that banks
choose to maximize the value of stockholders'
wealth. Section III contains a similar analysis,
applying the state-preference model to utilitymaximizing banks. The conclusions and policy
implications are presented in the final section.
21
I.
Issues in Capital Regulation
A bank's financial capital
that is, its equity
is the difference between the value of the institution's assets and liabilities. Banks use both capital
and liabilities to finance loans and investments. 9
The two sources of funding are distinguished in that
variations in earnings on assets are borne first by
capital holders. The larger the proportion of assets
funded by capital, the greater the range of returns on
assets that will be borne solely by equity holders and
the more likely the promised obligation to liability
holders will be met. Thus, if banks were not
insured, both equity and liability holders, including
depositors, would have an interest in the level of a
bank's capital. As with other firms, the stockholders
and liability holders (depositors) of unregulated
banks would be expected to "decide" on a satisfactory combination of capital financing and promised
return on bank debt.
The regulation of bank capital, then, must be
predicated on the assumption that a market determination of the level of capital would not be satisfactory from a public policy perspective. While
capital regulation predates federal deposit insurance, partly because of the externalities argued to be
associated with bank failures, the provision of the
federal deposit guarantee commonly is cited as the
main reason that the level of bank capital must be a
regulatory concern.
II.
The federal deposit insurance system, by guaranteeing deposits, in essence takes on the role of a
bankJiability holder and has an interest in bank
capital similar to that of private liability holders in
an uninsured firm. Indeed, some have argued that
the deposit insurance system has taken on the role of
covering virtually all bank liability holders in the
event of an insolvency. If so, the insurance system
would be the only liability holder with an interest in
bank capital.
From a regulatory perspective, a bank with more
capital relative to assets will be less likely to fail,
and, if it does fail, will impose smaller losses on the
insurance fund, all other things equal. However, the
probability of failure and the contingent liability of
the insurance system also depend on the variability
of the return on assets. 10 The higher the variability
of the return on assets for a given amount of capital,
the greater the chance of bank failure. 11
Consequently, a central issue in capital regulation
is whether banks would respond to higher regulatory capital requirements by choosing riskier assets
to offset or even more than offset the effects of
higher capital on the exposure of the deposit insurance system to bank risk. Below, we consider this
issue and examine under what conditions, if any,
regulation-induced increases in bank capital would
lower the expected losses of the deposit insurance
system.
Value Maximization
A value-maximizing bank chooses its portfolio
solely to maximize the current market value of
equity. Such a bank's portfolio decisions are independent of the risk preferences of its individual
owners because the owners are fully able to adjust
the composition of their personal portfolios to attain
any level of risk they desire. Therefore, even though
actual returns on the bank's portfolio are uncertain
(risky), a value-maximizing bank does not consider
the risk preferences of the owners.
Some of the implications of bank capital regulation for value-maximizing banks within the statepreference framework are discussed in Dothan and
Williams (1980), Sharpe (1978), and Kareken and
Wallace (1978).12 All of these studies provide
theoretical support for restricting leverage in banking when there is subsidized deposit insurance.
They do not, however, deal with the issue of how the
asset investment strategies of such insured banks
might be altered by capital regulation. Nor do they
consider how the behavior of a utility-maximizing
bank in the state-preference framework might differ
from that of a value-maximizing bank.
This portion of the paper addresses the first of
these two issues by extending the examination of
bank capital regulation within a state-preference
framework. For the reader who is not familiar with
this framework, a brief description of the statepreference model is presented in Box 1. Below, we
first show why leverage constraints are necessary
22
for insured, value-maximizing banks. Then, we
assess the likely effects of changes in capital
requirements on the asset risk of such banks and on
the liability of the deposit insurance system. The
discussion in Section III turns to the implications of
deposit insurance and capital regulation for utilitymaximizing banks within a state-preference model.
The calculations in Table 1 demonstrate what
happens to the total net worth (current value) of the
bank's equity as leverage increases. In line 2, the
bank increases leverage to 2 by issuing deposits with
a current value of $500 and purchasing an additional
100 shares of security A. In both states I and 2 the
bank promises to repay depositors $526.32
($500/.95). The net claims (future wealth) of the
bank in each state after paying off deposits are
shown in column 6. The current value of the bank to
the owners, column 8, is derived by multiplying the
net claims by the price of a dollar claim in the
appropriate state. The addition to the value of the
bank from the free deposit insurance, presented in
the last column, is derived by subtracting the initial
capital, $500, from the total value of net worth.
As Table 1 shows, initially the bank's value (net
worth) is not affected by issuing deposits. At the
lower levels of leverage, the bank would be indifferent to the amount of borrowing because its value
(column 8) would be unaffected. I5 Although the
risk of the bank increases with leverage, as reflected
in the growing disparity between the claims in the
two states (column 6), bankruptcy could not occur
and the deposit insurance fund would not be at risk
with leverage of 4 or less. Moreover, depositors
would be indifferent to the risk of the bank whether
or not their funds were insured as long as leverage
was less than or equal to 4.
It is easy to see why the insurance fund as well as
depositors are not at risk at low levels of leverage.
Up to a point, the bank is able to meet its promised
payments to depositors in both states 1 and 2. The
bank would not fail in either state since its liabilities
would not exceed its assets. Therefore, while risk
increases with leverage, as long as the bank's capital
is sufficient to ensure payment, the added risk is
borne only by the bank. 16
As leverage continues to increase, the bank eventually will be unable to meet its promised obligations to depositors in the first state. Without a third
party guarantee such as deposit insurance, depositors would not be willing to lend to a bank in return
for a promise to pay only $1 (per share) in each state.
With leverage equal to 5, for example, the bank
would issue a deposit with a current value of $.95
per share but would have to promise to pay about
$1.03 in each state instead of $1 (the actual pay-
Value-Maximizing Banks
Although the state-preference modcl can be
applied to an individual investor's decisions, it also
can be used to analyze the portfolio and leverage
decisions of an insured bank that maximizes its
current value (the market value of its equity). Since
the current value of such a bank is independent of
the risk preferences of the owners, we can put aside
any consideration of utility functions and focus
instead on how an insured bank's investment opportunity frontier itself is affected by leverage and
capital regulation.
The effects of leverage and the role of capital
regulation can be seen most easily through a numerical example with two states and two securities. In
this example, security A represents a promise to pay
$4 if state 1 occurs and $6 if state 2 occurs, and is
summarized as A(4, 6). The second security,
security B, is summarized as B( 1, 1). Security A is
a risky investment (a different payout in each state)
and security B is a riskless security (the same payout
in each state). For expositional purposes, we assume
that the current market price of a dollar payment in
state 1 is $.35 and the price of a dollar payment in
state 2 is $.60. The current price of a share of
security A then is $5. ($.35x4+$.6x6) and the
current price of security B IS $.95
($.35 xl + $.6 X 1).13
The bank is assumed to invest only in the risky
security. The bank's purchases of that security are
funded with a combination of capital and the proceeds from issuing shares of security B. Shares of
security B can be thought of as deposits that are
"insured" at a fixed-premium rate by the federal
government. In the example, the premium is set at
zero, but the analysis and conclusions would hold
even with a positive, fixed-rate premium. 14 Initial
capital is set at $500 by assumption. With no
deposits, the bank would have 100 shares of security
A, and leverage would be one.
23
24
25
ments would be about $1 .03 in state 2 and about. 95
in state 1). The current value of the depositor's
claims would be unaffected because the higher
payment in state 2 would compensate for the lower
payment in state 1. The deposit guarantee, however,
would allow a bank with leverage greater than 4 to
continue promising $1 to depositors in both states
because the deposit insurance fund would cover the
shortfall in state 1.
As seen in column 8 of Table I, once leverage is
extended to a point at which bank failure becomes
possible, the current net worth of the bank begins to
increase with leverage. The addition to net worth
represents the current value of the deposit insurance
guarantee (column 9). A bank gains from increasing
leverage and simultaneously investing additional
deposits in the risky security because the net claims
of the owners in state I can never be less than zero,
no matter how large the "promised" payments,
while the potential claims in state 2 are unlimited.
The state-preference model therefore predicts that a
value-maximizing bank with an insurance premium
less than the current value of the insurance payout
would limit its leverage only if forced to do so by
regulation.
26
Capital Requirements and Risk
Table 2 demonstrates how the value of a bank
with an initial leverage of 3, initial capital of $500,
and underpriced deposit insurance is affected by
shifting from holding only security A to holding
greater proportions of its assets in security D.
Parallel to the case of increased leverage with asset
risk held constant (Table 1), a bank with a given
level of leverage benefits from increasing its asset
risk with underpriced deposit insurance only when
bankruptcy is possible (that is when deposit claims
exceed the bank's gross claims in state 1). Once
bankruptcy is possible, the value of the bank
increases with asset risk (that is, with higher proportions of security D). Therefore, even if leverage were
limited through regulation, a value-maximizing
This brings us to the main question facing regulators: will regulatory efforts to force banks to hold
more capital be partially or even totally offset by
banksthaUhenacquire riskier assets?
To answer this question, anotherriskyassethas to
be introduced. In addition to security A(4, 6), we
assume that the bank also can hold the more risky
security, security D(O, 8.33), where the numbers in
Pllrentheses are the dollar claims per share of the
securities in the two possible states . The price of
security Dis $5 ($.35xO+$.6x8.33). A bank
with a given degree of leverage can alter its net
claims in future states by investing available funds
in different combinations of.these two risky assets.
27
bank with deposit insurance would want to hold the
risky security that maximized the value of the
deposit guarantee. (In the example, this would be a
portfolio that includes only security D.)
Figure 1 shows how the gains from increasing
asset risk are affected by leverage. Each of the lines
in the figure tracks the current value of the deposit
guarantee to the bank that invests greater proportions of funds in security D (and correspondingly
smaller proportions in security A), for a given
degree of leverage. The marginal value to a bank
from increasing asset risk (holding greater proportions of its asset in security D) is represented by the
slope of a line.
With low levels of leverage and asset risk, the
marginal value to increasing asset risk is zero (the
lines are horizontal). However, for higher levels of
leverage, the slopes of the lines increase as leverage
increases, indicating that the marginal value of
increasing asset risk increases with leverage. Put
another way, as the capital of an insured bank
increases, the marginal value to that bank of shifting
to a riskier composition of assets falls. This means
that more stringent capital requirements would not
give banks a greater incentive to invest in riskier
assets, and would reduce the liability of the deposit
insurance system. I?
With regulatory constraints on leverage, a bank
stilI would want to hold the risky asset, security D.
Figure 1
Effects of Leverage and Risk on the
Value of the Deposit Insurance Guarantee
Dollars
1,000
900
800
~ 700
~ 600
~ 500
400
'0 300
1l
~ 200
100
Leverage
~
6
i
o
- 100 L...--L..----''--.......--L_..&-........_.L-.-L.._L...-....
o 10 20 30 40 50 60 70 80 90 100
Riskiness of Asset Portfolio
Percent of shares held in security 0
('100% minus the percent of shares held in security A)
28
29
As a result, regulators also might put controls on
asset risk to reduce the liability of the deposit
insurance system. For example, regulation might
limit a bank to holding less than 30 percent of its
assets in security D. However, if regulatory limits on
the composition of bank assets and monitoring
(examinations) of banks were sufficient to constrain
the asset risk of a bank for a given level of leverage,
they would be sufficient for any lower level of
leverage because banks would have even less of an
incentive to evade them. Consequently, as long as
regulators did not react to lower leverage (higher
capital) by relaxing their efforts to limit asset risk, a
bank would not increase its asset risk, and the
liability of the insurance fund would decline. 18
III.
Summary
Not surprisingly, capital regulation is necessary
with subsidized deposit insurance to limit the liability of the insurance fund. However, more stringent capital standards for banks do not confound
regUlatory efforts to limit the riskiness of bank
assets because higher capital does not increase the
incentives of a value-maximizing bank to hold
riskier assets. In fact, the marginal value from
increasing asset risk for an insured bank declines as
leverage is lowered. This conclusion does not
depend solely on the state-preference framework.
As discussed in Box 2, a positive relation between
leverage and the gains from risk-taking also can be
derived from an options approach to evaluating the
gains from risk and leverage in banking.
Utility Maximization
In this section we incorporate utility maximization into the state-preference model. With utility
maximization, the state-preference model implies
that capital regulation is either irrelevant because
risk-averse owner-managers will hold sufficiently
conservative portfolios to make bankruptcy impossible or capital regulation will limit the liability of
the deposit insurance system in the same way that it
can for value-maximizing banks.
investor will allocate his capital between the available securities to maximize utility along the investment opportunity frontier.
The introduction of underpriced deposit insurance expands the opportunity frontier. The opportunity frontier with free deposit insurance can be
derived from the types of calculations presented in
Table 1. (Recall there is only one risky security
A(4,6), riskless borrowing by issuing security
B(l,l), free deposit insurance, and initial bank
capital of $500.) Figure 2 shows the various combinations of wealth (W I and W 2) that can be attained
by increasing leverage and investing in the risky
security (security A). Point A in Figure 2 indicates
the combination attainable with no leverage, and the
shaded segment includes points attainable by
increasing leverage.
The key difference between the choice set with
deposit insurance and the set without deposit insurance is that there is no limit to the amount of wealth
that would be attained in state 2 once bankruptcy
would occur in state 1. That is, various wealth
outcomes (WI' W 2 ), such as (0,895), (0,968),
(0,1042), can be attained only with free deposit
insurance. With free (or underpriced) deposit insurance and no capital regulation, there is no limit to
the amount of wealth that could be attained in state 2
by increasing leverage.
Whether the effects of deposit insurance on the
opportunity set will influence the owner-manager's
Utility-Maximizing Banks
Utility maximization has been rationalized as
being more applicable than value maximization to
smaller, owner-managed banks because the investment opportunity set for such banks and their
owners may be one and the same. The assumption
behind this rationalization is that the owner-manager cannot attract capital in addition to his own and
that most of his portfolio is invested in the bank.
Consequently, unlike a bank that is maximizing its
current market value, the owner's preference toward
risk would influence the bank's portfolio decisions. 19
It is assumed that the owner-managers are riskaverse. As pointed out in Box I, in a simple twostate world, risk aversion means that utility functions are convex with respect to the origin. That is,
future wealth has diminishing marginal utility in
each state of the world. As is shown in the figure in
Box 1, in a world without deposit insurance, an
30
Figure 2
Subsidizecl Deposit Insurance Expands
the Opportunity Set
bank, a utility-maximizing bank that is willing to
accept bankruptcy will maximize leverage.
Capital Requirements and Risk
w,
The predictions of the. state-preference model
regarding the effect of capital regulations on asset
risk for utility-maximizing banks (that will accept
bankruptcy) also are similar to those for valuemaximizing banks. As just stated, maximizing utility is equivalent to maximizing wealth in state 2 (the
nonbankruptcy state) for a bank that has underpriced deposit insurance and will accept a nonzero
risk of bankruptcy. Table 2 shows that if leverage
were restricted by regulation to 3, wealth in state 2
could be increased by increasing asset risk (holding
a larger proportion of assets in security D). Therefore, it could be necessary to regulate even utilitymaximizing banks' asset portfolios to prevent such
banks from increasing asset risk. 20
Table 2 also indicates that wealth in state 2 is
directly proportional to the current value of the bank
when bankruptcy is possible in state 1. (With bankruptcy possible in state I, wealth in state 2 is equal
to the current value of the bank divided by .6.) This
means that the marginal effect on wealth in state 2
for a given increase in asset risk declines as leverage
declines, just as does the marginal effect on the
current value of the bank (see Figure 1).21 Therefore,
as long as the regulatory efforts to prevent a bank
from increasing asset risk are not lessened, imposing a lower leverage position would not increase the
incentives for a utility-maximizing bank to increase
asset risk.
700
600
500
400
300
200
100
500
600
700
800,
900
1000
1100
1200
W2
Expanded OVpportunity Set
leverage decisions depends on the owner-manager's
preferences. To benefit from deposit insurance, the
owner-manager must be willing to risk bankruptcy
- that is, there must be some amount of wealth in
state 2 that will compensate for zero wealth in state
1. If there were no such level of wealth in state 2,
utility functions would approach the axes
asymptotically and interior solutions (some wealth
in both states) would be obtained (that is, a point on
the frontier to the left of point C). Without the
possibility of bankruptcy, the deposit insurance
fund would not be at risk, and no capital regulations
would be required. Moreover, even if capital regulations were imposed, portfolios that make bankruptcy possible would not be held.
Alternatively, if there were some level of wealth in
state 2 that could compensate the bank owner for
zero wealth in state 1, the utility functions would
intersect the axis. Such a set of preferences is
depicted in Figure 2 as indifference curves Uland
U2 . In Figure 2, raising leverage would increase
utility (U 2 > U I) because wealth in state 2 would
increase while wealth in state I would remain zero.
As shown in the previous section, this is the same
reason that the current value of the bank increases
with leverage. Consequently, maximizing the utility
of future wealth in state 2 for the type of preferences
depicted in Figure 2 (that is, indifference curves that
intersect the axis) is equivalent to maximizing the
current value of the bank. Like a value-maximizing
Summary
In sum, incorporating utility maximization into
the state-preference model does not affect our conclusion that more stringent leverage requirements
will reduce payouts from the deposit insurance fund
as long as the stringency of portfolio regulation
remains unchanged. Some owner-managers might
be so risk-averse that they would be unwilling to risk
bankruptcy even with deposit insurance. However,
the owner-manager that will risk bankruptcy in one
state for a sufficiently high claim in the other state
would seek to maximize wealth in the non-bankruptcy state. For such persons, utility maximization
and value maximization are comparable and all of
the results of the earlier section apply.
31
IV. Summary and Conclusions
This paper analyzes the theoretical relationships
among' capital regulation, bank asset risk, and the
liability of the federal deposit insurance system. We
demonstrated that a bank can benefit from underpriced deposit insurance by increasing leverage and/
or asset risk. As a result, some degree of capital
regulation is needed to limit the liability of the
deposit insurance fund.
More importantly, the analysis suggests that regulatory increases in capital standards will not
require greater efforts to restrain asset risk. On the
contrary, the marginal value of increasing asset risk
declines as leverage falls - that is, less leverage
(more capital) reduces the gain from risk-taking. In
other words,. banks with the least capital have the
most incentive to increase asset risk.
We have shown under the assumption of value
maximization that more stringent capital regulation
lowers the contingent liability of the deposit insurance system as long as the stringency of asset
portfolio restraints is not reduced. This result follows for value-maximizing banks in both the statepreference and options models. Moreover, incorporating utility maximization into the state-preference model does not change this conclusion.
The key policy implication that stems from our
analysis is that regulatory efforts to raise capital
standards in banking would not by themselves lead
to more risky asset portfolios.
FOOTNOTES
1. Actions taken in 1981, 1983 and 1985 raised capital
requirements for banks and bank holding companies, and
made the federal bank regulatory agencies' definitions of
capital more uniform.
10. From an economic standpoint, a bank (or any other
firm) fails when the value of its capital falls below zero.
Mathematically, the probability of failure is the probability
that the value of end-of-period assets is less than that of
end-of-period liabilities:
2. The proposal for risk-based capital standards was
made public in January 1987.
Prob [Failure]
3. The measure of common equity used in these ratios is
not the current regulatory definition of equity capital that
includes loan loss reserves and preferred stock.
where:
+
p)A < (1
+ r)lJ,
(1)
rate of return on assets (which is
assumed to be random),
A
4. See, "Fed Urgues Swap Plan for Banks," New York
Times, March 5, 1987.
= Prob [(1
= initial assets,
promised rate on liabilities,
5. Other studies that consider the effects of capital regulation on bank asset risk within the mean-variance framework
are Kahane (1977), Blair and Heggestad (1978), and
Hanweck (1984).
l
Prob [
= initial value of liabilities, and
] = denotes the probability of [
].
Without information on the type of probability distribution
governing pA, the probability of failure can be bounded by
using the Tchebichef inequality (see Koehn and San- '
tomero, 1980). However, by assuming that the normal
distribution approximates the distribution of pA (i.e.,
PA - N [E(pA), O'2 (pA))), we can solve forthe probability of
failure:
6. Koehn and Santomero (1980), p. 1244.
7. For purposes of this paper, failure and bankruptcy
occur when the market value of a bank's liabilities exceeds
that of its assets.
8. We recognize that the utility maximization model also
might be rationalized by appealing to the notion of the
separation of ownership and control so that the firm's
operating decisions depend on the manager's risk preferences.
Prob [Failure] = F[l - A
+ _rl - E(pA)]
O'(pA)
(2)
where: F = the standard (J-L = 0, (J'2 = 1) unit normal
cumulative distribution function.
9. Some articles appear to confuse financial capital with
physical capital. For example, Santomero and Watson
(1977) view financial capital as a physical investment that
could have been made in other sectors of the economy.
Financial capital, however, is not directly related to physical investment, and higher bank capital does not limit the
amount of physical investment in other sectors of the
economy. The amount of capital relative to liabilities is
simply a reflection of the way a bank finances its assets.
Bank capital as well as liabilities are available to be invested in nonbank physical investment through bank loans, for
example, as well as in bank facilities.
11. Equation 2 in footnote 10 indicates that the probability
of failure increases as the riskiness of the asset portfolio,
O'(pA), increases, and as leverage (as reflected in the
quantity of liabilities relative to assets) increases. To prove
this, the equation can be differentiated with respect to the
applicable parameter as follows:
ilProb [Failure]
SO'
-f (
SProb [Failure] = f (
ill
32
)[l - A
+
(J'2
) (1
0'2
+
r)
>0
rl - E(pA))
>0
(1)
'
(2)
where f( ) is the standard normal density function evaluated at the initial position. The first inequality holds
because the term in brackets is negative. Also note, in the
second inequality, assets are held constant and, thus, an
increase in liabilities reflects a corresponding decrease in
capital.
oHailure would decline with decreased leverage as the
number of states in which the bank was able to meet
promised payments increased.
18. In this paper, we have not dealt directly with bankruptcy costs. As shown in Dothan and Wil.liams (1980),
such costs can lead to a determinate degree of leverage
for a value-maximizing uninsured bank. However, bankruptcy costs are not sufficient to limit leverage if banks
have access to a subsidized deposit insurance. Although
bankruptcy costs are not incurred in all future states, they
nonetheless can be evaluated in terms of their effects on
the current value in the state-preference framework. It can
be shown that, with free deposit insurance, bankruptcy
costs that are fixed or proportional to assets generally will
be insufficient to limit leverage. This result holds whether
bankruptcy costs fall on the bank or on the depositors.
12. Other studies such as Merton (1977) and Pyle (1984)
provipElLJseful insights into the regulation of bank leverage
and ass~t risk by using options models to analyze the value
of the federal deposit guarantee. In Box 2, we analyze the
effects of capital standards on asset risk in banking using
an options model.
13.. This implies a risk-free real interest rate of 5.26 percent, (($1.001$.95) -1] x 100%.
14. The analysis would be essentially the same if the
premium rates were variable as long as the premiums paid
were less than the value of the deposit guarantee.
19. Despite legitimate questions as to whether this
assumption would apply to any real-world banks since
owners of small, privately held banks can diversify their
portfolios, we hold to it.
15. The indifference of a low leverage bank in our example
to the degree of leverage parallels Proposition I (the market
value of a firm is independent of its capital structure) in
Modigliani and Miller (1958). While Modigliani and Miller
do not use a state-preference model, Hirshleifer (1966)
uses the state-preference approach to show that Proposition I still holds in that framework. In fact, the state-preference model can be used to show that the Modigliani-Miller
theorem holdswith or without bankruptcy, when there is no
subsidized deposit insurance.
20. It is possible that, with leverage held sufficiently low,
the wealth attainable in state 2 from investing in the riskier
security would not be adequate to compensate a utility
maximizing bank owner for risking zero wealth in state 1,
even if the utility curves crossed the axis. In such a case,
the bank would choose a portfolio along the AC portion of
the opportunity frontier in Chart 2, and no other portfolio
restraints would be required. However, at some higher
level of leverage the same bank would begin to take
advantage of the opportunity to increase wealth in state 2
through investing in the riskier asset, security D.
16. One policy implication here is that the distortions of
deposit insurance could be eliminated if risk in banking
were borne only by the banks. Along this line, it has been
suggested that risk would not be shifted to the insurance
fund if there were timely closures of banks. With continuous
(and costless) monitoring of banks, this would correspond
to closing a bank before its market net worth reached zero
(Furlong and Keeley, 1985). With periodic examinations of
banks, the state-preference approach indicates that
losses to the insurance fund could be avoided only if banks
had enough capital to ensure their solvency in all possible
states.
Similar results hold in a model with more than two states.
With very low leverage, a bank may not be able to realize
sufficient compensation in the nonbankruptcy states to
justify risking bankruptcy in even one possible future state.
It would not be necessary to regulate the composition of
such a bank's assets. At higher levels of leverage, the bank
ultimately would increase asset risk and allow for bankruptcy in at least some states.
21. In a multi-state world it also is the case that the
marginal effect on wealth in each of the nonbankruptcy
states with positive payouts would increase with leverage.
17. In this two-state model, the probability of failure actually would decline only if leverage and risk were restricted
in such a way that the bank could meet obligations in each
state. In a model with more than two states, the probability
33
ApPENDIX
Capital Regulation and Asset Risk
in a Utility-Maximization, Mean-Variance Framework
Introduction
A number of studies have attempted to analyze
the effects of bank capital regulation on asset risk
and the probability of bankruptcy while assuming
that banks maximize utility in a mean-variance
framework. This literature is best typified by articles
by Kahane (1977), and Koehn and Santomero
(1980).
We show below that the conclusions reached by
these studies were derived using internally inconsistent assumptions. Both studies assume that a bank's
borrowing cost would be unrelated to bank risk.
That is, a bank's borrowing cost (per dollar of
liabilities) is assumed to be constant regardless of its
asset risk or leverage. Thus, these studies implicitly,
but unintentionally, assume that bank failure cannot
occur. Yet, they seek to analyze the effects of capital
regulation on the probability of bank failure. Moreover, these studies fail to take into account the effect
of underpriced deposit insurance on the incentive to
take on excessive risk.
A possible explanation for why these studies
overlook the effects of bankruptcy on the bank's
borrowing cost is that the basic mean-variance
framework used is adapted from the finance literature on investment, which does not allow for bankruptcy since borrowing and lending are assumed to
take place at the risk-free interest rate. While this
simplifying assumption may be appropriate for certain investment decisions, it is not appropriate for
the analysis of banking with underpriced deposit
insurance. The reason is that concern over the
exposure of the deposit insurance system to risk in
banking arises only when bankruptcy is possible.
In this Appendix, we first construct a prototype of
the utility-maximization, mean-variance model
used in past studies to analyze the effects of bank
capital regulation on asset risk. We show that when
bankruptcy is not possible, and, thus, when there is
no deposit insurance subsidy, the results from our
prototypical model are identical to those of the
previous studies. Specifically, the effect on asset
risk of moving from one binding capital constraint
to a more stringent one depends on the nature of the
preferences of the bank's owner-manager. Restricting such an analysis to situations where bankruptcy
is not possible, however, makes these conclusions
irrelevant for policy purposes since capital regulation is needed only when bank failures and deposit
insurance payouts are possible.
In the next section of the Appendix, we add the
possibility of bankruptcy and subsidized deposit
insurance to the model. Doing so changes markedly
the bank's investment opportunity set. In addition,
we present a specific numerical example to illustrate
that, when the asset return distribution is binomial,
the incentive to increase asset risk does not increase
as the stringency of capital regulation increases
regardless of the nature of the bank owner-manager's preferences.
Background
The utility-maximization framework assumes
that the bank owner-manager's preferences toward
risk can be characterized by the expected rate of
return on capital, E(z), and the standard deviation of
the rate of return (f(z). Moreover, assuming riskaversion, these preferences can be depicted by a set
of concave, upward sloping iso-utility functions
depicting the tradeoff between the expected rate-ofreturn and risk.
Such iso-utility functions, Uland U2 (U i =
U[E(z), (f(Z)]), are depicted in Figure AI. The
vertical axis represents the expected rate of return
and the horizontal axis represents risk as measured
by the standard deviation of the rate of return. Along
indifference curve U l' the investor is indifferent
among the various combinations of expected returJ;l
and risk. However, the investor prefers the points on
U2 to those on U 1 because, for any given level of risk
(standard deviation) the expected rate of return on
U2 is larger.
The ideas behind this characterization of preferences are that initial wealth is given and that the
34
capital, K, or
Figure Ai
Utility-maximization Framework
z= [Ap -
U2
Expected
Rateot
Lr]/K
(AI)
where
A= initial assets
L
initial liabilities
K
initial capital
P rate of return on assets, assumed to be
random
z rate of return on capital, which is random
r - promised (which equals actual) rate of
return paid on (and cost of) liabilities.
Return
E(Pol
"(Po)
"(2)
(As discussed more fully later, if bankruptcy were
possible, the cost of liabilities would no longer be
fixed since actual payments to depositors would
sometimes be less than promised payments. This
implies that the cost of deposits to the bank would be
a random variable, which depends on the rate of
return on assets realized and leverage chosen. Consequently, equation Al would not apply to realizations of p when bankruptcy occurred.)
Equation A I may be rewritten by noting that
L=A K to give
Standard Deviation ot
the Rate of Return
investor is concerned about the expected value of
future wealth along with its standard deviation.
Future wealth is equal to one plus the rate of return
times current wealth. Consequently, with current
wealth (capital) assumed fixed, the mean and variance of future wealth are mathematically equivalent
to the mean and variance of the rate of return on
capital, respectively. Thus, similar to the statepreference model, utility maximization in a meanvariance framework also focuses on the probability
distribution of future wealth.
z= [AlK] [p-r] +r.
(A2)
The expected rate of return on capital, E(z) may be
found by taking expected values of both sides of
equation A2. This gives:
E(z) = [A/K] [E(P) - r] + r,
(A3)
Utility-Maximization Without Bankruptcy
as long as r is fixed and not random, which it would
be as long as bankruptcy were not possible. AI Thus,
increasing leverage, as measured by the asset-tocapital ratio increases the owner's expected rate of
return on capital linearly as long as default is not
possible. A2 (We later show this result changes when
bankruptcy is possible).
Similarly, the standard deviation of the return on
capital, (T(z), may be derived from equation A2 by
noting that when bankruptcy is not possible, the
covariance of rand p is zero. In this case,
A bank must decide on how risky an asset portfolio to hold and by how much to leverage that
portfolio. Given the owner's preferences toward
risk, utility will be maximized subject to a constraint that relates the expected return on capital,
E(z), to the standard deviation of the rate of return
(T(z). To derive this constraint, note that if the bank
chooses sufficiently low leverage and asset risks to
make bankruptcy impossible (that is, promised obligations to depositors are always met regardless of
the asset return that is realized, the rate of return on
capital, Z, is given by the gross returns on assets,
Ap, minus promised (which equals the actual) paymentsto liability holders, Lr, divided by initial
(T(z) = [A/K](T(p).
35
(A4)
(This equation is not valid when bankruptcy is
possible since the covariance of p and the cost of
deposits is not zero.)
Equations A3 and A4 may be jointly solved to
eliminate the [A/K] term to give
E(z)
[a(z) / a(p)][E(p) - r] + r.
utility function with the risk-return frontier for capital (the straight line in Figure AI). However, the
assumption here is that the unconstrained bank
would choose a degree of leverage for which bankruptcy is not possible.
Capital Requirements and Risk
In Figure A I, we showed how a particular asset
portfolio may be leveraged (assuming no bankruptcy) to obtain the capital risk-return frontier. Of
course, any asset portfolio may be leveraged
although there would be no reason for a bank owner
to leverage any asset portfolio other than the most
efficient one in a world without capital or asset
portfolio regulation. When capital constraints are
imposed, however, the bank owner generally will be
able to increase utility by leveraging asset portfolios
other than the one characterized by the parameters
E(po), a(po)· For example, suppose that the maximum asset-to-capital ratio allowed were 5. Then
the standard deviation of return on capital would be
5 times the standard deviation of return on the asset
portfolio chosen, and the expected return on capital
also would be five times greater.
(A5)
In other words, expected return on capital varies
linearly with the standard deviation of return on
capital for a given expected asset return and standard deviation of the return.
This linear relationship, equation A5, is plotted as
the straight line intersecting the vertical axis at r in
Figure A I. It is assumed that the particular asset
portfolio with expected return E(po) and standard
deviation a(po) is being leveraged. With no leverage
(A = K), the expected rate of return and standard
deviation of return on capital are equal to the
expected rate of return and standard deviation of
return on that particular asset portfolio - E(po) and
a(po), respectively. As leverage increases, the
expected rate of return and standard deviation of
return both increase linearly.
In general, it is assumed that a bank faces a
variety of different asset risk-return combinations as
determined by the availability of investment alternatives in its market (known as the asset risk-return
frontier). As shown in Figure A I, asset portfolios
with more risk are assumed to yield larger expected
returns. Also, it is assumed that the banking sector
is small enough that the asset risk-return frontier is
unaffected by banks' behavior. Thus, that frontier is
taken as given by banks in their optimizing decisions.
In this framework, the most efficient asset portfolio is the one where a line from the constant
borrowing rate, r, is tangent to the asset risk-return
frontier. This is depicted as point E(po), a(po) in
Figure AI. By leveraging this asset portfolio, the
bank can obtain the highest expected return on its
capital for any degree of risk. Since this tangency
point does not depend on the bank owner's preferences, the asset portfolio (that is, the particular
combination of assets) chosen depends only on the
risk-free interest rate and the asset risk-return frontier. The degree of leverage chosen, however, is
determined by the tangency of the owner's iso-
Figure A2
Imposing a Binding Capital Constraint
Causes Asset Risk to Rise
Expected
Rateo!
Return
Standard Deviation of
the Rate of Return
36
bankruptcy could not occur. Such a bank could
attract deposits at the risk-free rate because it would
always make the payments promised regardless of
the return on assets realized. Consequently, the cost
of deposits to such a bank would be fixed at the riskfree rate r, and not be a random variable.
With free deposit insurance, a bank could issue
deposits at a fixed risk-free promised rate even if
bankruptcy were possible. However, the cost of
deposits to the bank would no longer necessarily
equal the risk-free rate. When bankruptcy occurs,
the bank effectively would pay less than the promised rate on deposits, r. Only when bankruptcy does
not occur does the cost of deposits to the bank equal
the risk-free rate. Put another way, the excess of
contractual debt obligations over assets when bankruptcy occurs corresponds to the option value of
deposit insurance (see Box 2). Thus, the effective
deposit costto the bank is a random variable related
to the rate of return on assets and leverage.
As a result, the expected cost of deposits to the
bank would decline with increasing leverage and
would be less than the risk-free rate. This means that
neither the expected rate-of-return equation, A3,
nor the standard-deviation of the rate-of-return
equation, A4, would hold. Instead, the rate of return
on capital, Z, to a bank with free deposit insurance is
given by:
Figure A3
Increasing the Stringency of Capital Regulation
Expected
Rate of
u,
/._ _ c,
Return
U2
Standard Deviation of
the Rate of Return
In geometric terms, the risk and return on capital
for a given asset portfolio and leverage can be found
by extending a ray from the constant borrowing rate
through the asset portfolio chosen up to the maximum leverage allowed. As Figure A2 shows, when
leverage is limited by regulation, the capital-risk
return frontier becomes convex once the leverage
constraint becomes binding. As a result, if a binding
capital constraint were imposed on a previously
unconstrained bank, the bank would choose a more
risky asset portfolio. This is shown as a move from
V o to V 1 in Figure A2.
In Figure A3, two binding capital constraints, CI
and C2, and a particular set of preferences are
depicted. However, as one moves from one binding
capital constraint, CI, to a more stringent one, C2,
the effects on asset risk depend on the owner's risk
preferences
asset risk can either increase,
decrease, or remain the same, which is the basic
conclusion reached by the traditional literature.
[Ap - Lr]/K if bankruptcy does
not occur, that is, if
A(l +p)~L(l +r).
(A6)
z=
-1
if bankruptcy does
occur, that is, if
A(l +p)< L(l +r).
The expected rate of return on capital, E(z), found
by taking the expected value of equation A6 is:
E(z) = -I Prob[Failure]
(A7)
+ E[(Ap -rL)/K I p>p*](l- Prob[Failure])
Introducing Bankruptcy and
Deposit Insurance
The analysis above was derived under the
assumption that an unconstrained bank would
always make asset and leverage choices such that
where
p*_
- -K(l+r)+
--- r
A
37
(A8)
depending on the asset rate-of-return distribution,
the probability of failure can increase up to a point as
leverage increases. However, in the limit as leverage
increases (and KIA goes to zero), the probability of
failure approaches the probability that the rate of
return on assets, p, is less than the promised rate on
deposits, r.
Consequently, by increasing leverage, the owner
can increase without limit the expected rate of return
on capital as long as at least some part of the asset
rate-of-return distribution exceeds the promised
deposit rate. Thus, even if the expected rate of return
on assets were less than the promised rate on
deposits, a bank with underpriced deposit insurance
would gain from leverage as long as this condition
held. This conclusion contrasts with the results
obtained when bankruptcy is not possible. In that
is the level of p above which bankruptcy does not
occur and
Prob[Failure]
== Prob[p < p*] ==
Prob[p <
(A9)
K(l+r) +r].
A
Equations A6 and A7 indicate that the simple linear
relationship between E(z) and a(z) presented in
equation A5, generally would not be valid. A3
Moreover, Equation A7 indicates that the bank
owner would never lose more than his or her initial
capital (that is, the minimum z would be - I, even
though z would be less than minus 1 if the promised
obligation to depositors were met in the event of a
bankruptcy). Also, equation A9 indicates that,
38
case, leverage can increase expected return only if
the expected return on assets exceeds the promised
rate.
deviation of return both increase linearly and there is
no bankruptcy. After this point (when leverage
exceeds 26), there is a 50 percent chance of realizing
the low asset return (denoted as outcome 2) and
going bankrupt. However, there is also a 50 percent
chance of drawing the high asset return and profiting
from leverage. As leverage increases, end-of-period
capital increases without limit as long as bankruptcy
does not occur. A4
Rationality implies that a person will prefer a
lottery that pays $100 with a 50 percent chance and
$0 with a 50 percent chance to one that pays $10
with a 50 percent chance and $0 with a 50 percent
chance. AS Thus, this example illustrates that even a
risk averse bank owner that is willing to risk bankruptcy (that is, one who is willing to participate in
the type of lottery just described) in return for a
sufficiently high payoff when the higher asset return
is realized would prefer unlimited leverage. Consequently, in this example, maximizing utility is
equivalent to maximizing value as long as the bank
owner is willing to risk bankruptcy.
In this simple model, a similar result would apply
to asset risk under leverage constraints. As long as
some non-zero probability of bankruptcy were
acceptable, a bank owner would maximize asset risk
since that would maximize end-of-period capital if
bankruptcy did not occur and would not affect endof-period capital if bankruptcy did occur. These
implications are in sharp contrast to those from the
cited mean-variance literature which claims that
risk aversion would limit leverage and a~set risk.
Moreover, as in the state preference model presented in this paper, it can be shown that the gain
from increasing asset risk is positively related to
leverage. Thus, in the case of binomially distributed
asset returns, more stringent capital regulation does
not increase the incentive to increase asset risk.
A Numerical Example
Below, a simple numerical example is presented
that allows for both subsidized deposit insurance
and the possibility of bankruptcy. This example
shows that a bank can gain from subsidized deposit
insurance only by assuming a nonzero risk of bankruptcy. The example also shows that when asset
returns are distributed binomially, more stringent
capital regulation will not increase the incentive to
increase asset risk. Thus, this example demonstrates
that the results in the cited mean-variance literature
do not hold generally.
The calculations presented in Table Al demonstrate what happens to the expected return and
standard deviation of return on an initial $100
investment of capital in a bank (with deposit insurance provided to it at no cost by the government) as
leverage increases. To simplifY the calculations, we
assume that the rate of return on the asset being
leveraged is drawn from a binomial probability
distribution with a .5 probability of a 10 percent rate
of return and a .5 probability of a 0 rate of return (for
an expected rate of return of 5 percent). It is assumed
that the bank is able to attract any quantity of
deposits it wants at a promised 4 percent rate of
return because deposit insurance (which is provided
free to the bank) covers any shortfalls when bankruptcy occurs.
The calculations in the Table demonstrate what
happens to expected end-of-period capital and its
standard deviation as leverage increases. (The rate
of return on capital, in percentage terms, is just endof-period capital minus 100). In line 2, the bank
increases leverage to 2 by issuing $100 of deposits
and promising to return $104 to depositors at the
end of the period. The end-of-period capital for each
of the two possible asset returns after paying the
deposit claims is shown in column 6. The expected
end-of-period capital and standard deviation of endof-period capital are shown in columns 8 and 9. The
payment from the deposit insurance fund is shown
in column 7.
As the Table indicates, initially, as leverage
increases up to 26, the expected return and standard
Summary
The results of previous studies using the meanvariance framework regarding the effect of capital
regulation on asset risk can be replicated assuming
that bankruptcy is not possible. However, when
bankruptcy is possible and underpriced deposit
insurance is provided to banks, the results of these
studies no longer hold generally.
39
APPENDIX ENDNOTES
REFERENCES
Arrow, Kenneth J. "The Role of Securities in the Optimal
Allocation of Risk-bearing," Review of Economic Studies, April 1964.
Blair, Roger D. and Arnold A. Heggestad. "Bank Portfolio
Regulation and the Probability of Bank Failure," Journal of Money Credit and Banking, Vol. 10, No.1,
February 1978.
Debreu, Gerard. Theory of Value: An Axiomatic Analysis of
Economic Equilibrium, New York: John Wiley & Sons,
1959.
Dothan, Uri and Joseph Williams. "Banks, Bankruptcy, and
Public Regulation," Journal of Banking and Finance,
NO.4,1980.
Furlong, Frederick T., and Michael C. Keeley. "Bank
Runs," Weekly Letter, Federal Reserve Bank of San
Francisco, July 5, 1986.
Hanweck, Gerald A. "A Theoretical Comparison of Bank
Capital Adequacy Requirements and Risk-Related
Deposit Insurance, Premia," Research Papers in
Banking and Finance, Board of Governors of the
Federal Reserve System, No. 72, May 1985.
Hirshleifer, Jack. "Investment Decisions Under Uncertainty: Applications of the State Preference
Approach," Quarterly Journal of Economics, Vol. 80,
May 1966.
Kahane, Yehuda. "Capital Adequacy and the Regulation
of Financial Intermediaries," Journal of Banking and
Finance, No.1, 1977.
Kareken, John H. and Neil Wallace. "Deposit Insurance
and Bank Regulation: A Partial Equilibrium Exposition," Journal of Business, Vol. 51, No.3, 1978.
Koehn, Michael and Anthony M. Santomero. "Regulation of
Bank Capital and Portfolio Risk," The Journal of
Finance, Vol. XXXV., No.5, December 1980.
Merton, Robert C. "An Analytic Derivation of the Cost of
Deposit Insurance and Loan Guarantees," Journal of
Banking and Finance, No.1, 1977.
Modigliani, Franco, and Merton H. Miller. "The Cost of
Capital, Corporation Finance and the Theory of Investment," The American Economic Review, Vol. 48, June
1958.
Pyle, David H. "Deregulation and Deposit Insurance
Reform," Economic Review, Federal Reserve Bank of
San Francisco, Spring 1984.
Santomero, Anthony M. and Ronald D. Watson. "Determining an Optimal Capital Standard for the Banking
Industry," The Journal of Finance, Vol. XXXII, No.4,
Sept. 1977.
Sharpe, William F. "Bank Capital Adequacy, Deposit Insurance and Security Values," Journal of Financial and
Quantitative Analysis, Nov. 1978.
Sharpe, William F. Portfolio Theory and Capital Markets.
New York: McGraw-Hili, 1970.
A1. An uninsured bank could attract deposits by paying a
fixed rate of interest, independent of its leverage or asset
risk as long as its asset risk were low enough relative to
capital that the probability of bankruptcy were zero. As
long as. bankruptcy were not possible, bank liability
holders would not be at risk of loss due to default and
would accept bank liabilities as riskless. At some point,
however, as leverage increased (for a given nonzero asset
risk), bankruptcy would become possible and the bank
would have to pay a higher deposit rate to compensate
depositors for the risk of default.
The utility-maximization literature cited assumes a constant borrowing rate environment,' but does not explicitly
acknowledge that this would be consistent only with a zero
probability of bankruptcy. Kahane does allow for a
stochastic deposit rate but assumes the promised rate
equals the rate the bank expects to pay. Moreover, he
assumes the expected cost of deposits and the promised
rate are independent of leverage and asset risk. These
assumptions would hold only if bankruptcy were not possible.
A2. Thus, it is crucial to distinguish the asset from the
capital risk-return frontier. Blair and Heggestad (1978) fail
to do so.
A3. However, we do not mean to imply that equations A6
and A7 necessarily can be used to derive the appropriate
risk-return constraint for utility maximization in a meanvariance framework. One reason is that variance no longer
adequately characterizes risk when bankruptcy is possible.
A4. After the point where bankruptcy becomes possible,
the relationship between the expected rate of return on
capital (column 8 minus 100) and its standard deviation
changes (the expected rate of return rises more rapidly
and the standard deviation rises less rapidly with leverage.)
A5. This is true even though the standard deviation of the
first alternative is larger.
40
Randall Johnston Pozdena*
The semiconductor industry has played a key role in international trade
disputes. Using data on Dynamic Random Access Memory devices and
semiconductor chip fabrication facilities, this analysis examines the
behavior of the industry for evidence of the influence of time-related
technological change, economies of scale, learning curve behavior, and
international differences in strategic pricing behavior. The analysis finds
only weak evidence of anti-competitive behavior.
The pace of innovation in the field of electronics
has been extremely rapid in the last thirty years, and
high technology electronics has been a major source
of strength for the American economy. The development of solid state devices - and integrated circuits
in particular - has been the major contributor to the
startling evolution of this field and the entry of high
technology electronics into so many aspects of daily
life. In addition, many place their hopes for continued growth of the national and regional economies on intensified innovation in and application of
high technology electronics. Along with biotechnology, high technology electronics is seen as a
kingpin of the future of the American economy.
The purpose of this paper is to evaluate popular
claims that the semiconductor industry is susceptible to anticompetitive behavior, particularly on the
part of foreign competitors. Specifically, we will
examine the market for a particular integrated circuit (Ie) device for evidence of imperfectly competitive performance. Learning and scale economies
are found to be significant in this industry, and
market structure
while not showing excessive
exhibited rigidity.
concentration of market share
Combined, these observations are consistent with
what one would find where inefficient forms of
strategic pricing behavior are practiced.
Production functions associated with integrated
circuit fabrication facilities located in the United
States and Japan are estimated to provide an insight
into the origins of alleged international differences
in pricing strategies. Only weak evidence is found to
support the notion that Japanese integrated circuit
(IC) fabrication costs are below those of their U.S.
counterparts.
In Section I of this paper, a brief description of the
semiconductor industry and its products is presented. Section II contains a description of the IC
production process and an economic characterization of this process. Section III discusses potential
implications of the production environment on
industry structure and performance. In Section IV,
several simple empirical investigations are performed to assess the importance of learning and
scale economies in the IC industry and to investigate the origins of differences in U.S. and foreign
firm pricing strategies. The paper concludes with a
summary of findings and their implications for the
future of the U.S. semiconductor industry.
* Assistant Vice President, Federal Reserve Bank of
San Francisco. This paper benefitted from the valuable assistance of numerous people. I wish particularly to thank Dan Hutcheson of VLSI, Inc.,
lone Ishii of the Semiconductor Industry Association and Lane Mason of Dataquest, Inc. for their
generosity in making available much of the data
used in this study. The paper also benefitted from the
assistance provided by Bill Arnold of the Electronic
Materials Report, Inc., Don Larsen of Tektronics,
Inc., and Carolyn Rogers of Hambrecht and Quist,
Inc. Finally, I wish to thank William M. Robertson
for providing excellent and diligent research assistance.
41
I.
The Industry and Us Products
The history of the semiconductor industry, its
technology and products are discussed in a number
of published sources l . It is useful, however, to
review the basic features of the industry and its
technology both to support the logic of subsequent
discussion and to delimit the economic issues we
will address.
The semiconductor industry is so named because
it produces devices that exploit the special electrical
characteristics of a class of natural elements and
compounds known as "semiconductors" (such as
silicon, germanium and gallium arsenide). The
materials have the property that they can be made to
behave alternately as conductors or barriers to the
flow of electrical current. In the late 1940s, discovery of a means of managing the behavior of semi.conductor crystals led to the development of the
transistor - a device that uses small currents to
control the conduction behavior of the material.
Thus, the transistor can form the basis of an amplifier or electronic switch.
Through the 1950s and 1960s, the transistor
rapidly replaced the vacuum tube because of its
superior ruggedness, smaller size, lower power consumption, and ability to execute tasks more rapidly.
Before 1958, functional electronic devices were
built by connecting a number of transistors and
other electronic components in a discrete manner.
Then, two scientists developed the "integrated circuit" or IC, which is a single device combining the
II.
functions of a number of transistors. By so doing,
ICs opened the possibility of constructing more
efficient and compact electronic devices.
The first ICs were produced in commercial volume in the mid-1960s. They are produced by a
complex process of etching, "doping" the crystalline material with other elements, and heat-treating the surface of a semiconductor crystal wafer.
Today, a 5-inch diameter wafer of silicon can yield
one hundred or more "chips", each of which may
contain as many as 1 million transistors. Although
"discrete" devices are still produced, the IC is now
the dominant semiconductor product and has revolutionized industrial and consumer electronic products. In 1985, approximately $16.5 billion in IC
shipments were made worldwide, against about $5
billion in shipments of discrete devices 2 .
Despite wide variation in the types of functions
that ICs can perform, the same basic production
process is used in their manufacture. Microprocessors (the "brain" of computational devices),
memory devices (for storing information), and a
wide variety of standard circuits used in consumer
electronics, telecommunications devices, and military hardware all involve similar production procedures 3 . By focusing our attention on ICs in general, and memory devices in particular later in the
paper, we hope to make useful generalizations about
the semiconductor industry.
The Economics of Ie Production
The focus of this paper is on factors influencing
the structure and future international competitiveness of the American semiconductor industry. We
begin with a brief description of the IC production
process. Certain aspects of this process are unusual
and, when considered in light of U.S. patent law
and the alleged industrial policies of foreign competitors, may be important determinants of the
structure, performance, and international competitiveness of the American semiconductor industry.
ment takes the form of circuit layout development,
development of "maskworks" or templates to
imbed the circuitry in the surface of the semiconductor material, and development and testing of
prototypes. Because the prototypes often do not
behave as modelled during the layout development
process, many cycles of the prototype development
process may be required before a useful design
evolves. This basic circuit design process interacts
with the design of the fabrication process and, in
some cases, with the design of other chips or
"firmware" (programming incorporated into ICs).
In total, this preproduction investment may cost as
much as $100 million in the case of a new micro-
Major Features of the Production Process
The production of integrated circuits involves
very large pre-production investment. Such invest-
42
processor chip4.
Actual fabrication of the integrated circuits takes
place in a fabrication line ("fab line") facility.
Wafers of the semiconductor crystal (predominantly
silicon) enter one end of the fab line and the fabricated IC exits the production process after various
stages of chemical and heat treatment, "dicing" of
the wafer into constituent chips, and electrical and
physical attachment of the chip to its plug-like base.
It is conventional to describe the capacity and
activity levels on a fab line in terms of "wafer starts"
per week. The relationship between wafer starts and
actual production flow ofICs, however, will depend
upon the design of the device being fabricated, the
size of the wafer stock, and the efficiency of the
fabrication process, which generally is higher on
lines with newer vintage fabrication equipment and
higher quality labor.
Labor and capital are substitutable to some
degree in most of the steps of the fabrication process. Once a fabrication process has been configured, however, significant changes in the process
can be costly and time-consuming. Similarly,
although a single fab line can, within limits, be used
to produce a variety of devices, different types of
devices involve different processing steps and
sequences, new computer programs to guide those
steps, and can involve changes in the degree of
cleanliness of the fab line environment. Crossovers
to radically different devices, therefore, also are
costly5.
fundamental electronic building block of the IC,
namely the transistor.
In the short run, fab line capital and the capital
representing the design of the IC (the maskworks)
are fixed. Output is varied by the firm by manipulating labor and materials inputs. It seems clear that
average total short-run costs decline sharply with
increased output because of large, fixed maskwork
and fab line capital costs. At production levels
above the design capacity of a firm's fab line facilities, however, problems of congestion likely arise.
Each of the 50 to 100 processing steps takes a finite
amount of time and few opportunities exist in the
short run to accelerate the processing or to improve
the yield of useful output from wafer starts~. Thus,
in the short run, rising average variable costs likely
cause average total costs to rise at high output levels.
In the long run, both fab line capital and maskwork capital are variable, and there are several
potential sources of increasing returns to scale. One
is that larger fab line facilities offer lower unit
fabrication costs than smaller ones. The industry's
practice, however, has been to manipulate the number rather than the size of fab lines to alter fab line
capacity, suggesting that individual fab line scale is
not a major source of economies of scale generally.
Of the 1,500 or so fab lines in existence in 1986,
two-thirds had design capacities between 1,500 and
4,300 wafer starts per week? If fabrication were
an important source of scale economies, its effects,
therefore, must be derived from firm-level synergies
from operating multiple lines. (The issue of fabrication scale economies is explored further below.)
Increases in the firm's stock of "maskwork capital" also could result in lower long-run average
costs. Conceptually, we might view improvements
in maskworks and manipulation of processing steps
(that is, alteration in the design of the Ie) as either
an increase in the employment of maskwork "capital" or a change in technology. Technological
change is usually assumed to be exogenous to the
firm's labor and capital allocation decisions (that is,
technical change depends only upon the passage of
time) whereas investments in what we are calling
"maskwork capital" have been an important component of IC firms' cost-minimization strategy.
Indeed, the commitment of resources to chip (and
fabrication process) design is probably responsible
Short-Run and long-Run Costs of
Production
The characterization of IC products and the production process made above can be re-stated in
conventional economic terms as follows. First, the
product in the IC industry is probably best thought
of not as the IC itself, but rather, the units of memory
storage, switching, or logical processing functions
it provides. Although there are qualitative differences across IC devices providing these various
functions (such as access speed in memory devices
or the compactness of the IC device that contains
them), it is helpful to think of the market as demanding memory storage or other functions rather than
ICs per se. Then, within gross functional categories
at least, the elemental unit of output relates to the
43
for most of the widely touted, sharp declines in IC
product costs that have been observed over time. IC
engineers have succeeded in increasing the number
of elemental components ("transistors") that can be
accommodated by a single semiconductor chip of
given physical dimensions and, hence, reducing the
unit cost of fabricating IC products. (Empirical
evidence will be presented in Section IV below on
the relative contribution of scale economies and the
passage of time to the decline in the cost of IC
memory products.)
"cloning" of maskwork capital by foreign or
domestic competitors, although the more general
practice of reverse engineering remains legal 1I.
A second often-cited feature of the IC industry is
the relevance of "learning curve" phenomena to IC
production. The notion is simply that the cost of
production may be related not only to the rate of
output of a firm (economies of scale) and changes in
technology over time but also to the independent
effect of accumulated production experience. Such
a phenomenon is considered to be relevant to complex manufacturing technologies: as output experience increases, the firm better understands the technology involved and technical efficiency
increases 12.
Since integrated circuit manufacturing is an
extremely complicated technical process, it seems
likely a priori that learning-related cost adjustments
may occur. The implications ofleaming phenomena
and a test for their existence are presented below.
To summarize, the IC production process is a
highly technical one involving large investments in
maskwork capital that are difficult to recover if an
enterprise fails and difficult to protect from exploitation by other firms. Potentially significant economies of scale are likely, and probably flow mainly
from economies at the level of the firm rather than
the plant (that is, the fab line). Costs also may
decline over time because of technological progress
and with accumulated output experience because of
"leaming curve" phenomena.
Technological Diffusion and learning
Two other aspects of the IC production environment are relevant to understanding the current and
likely future performance of the IC industry. The
first is that property rights in "maskwork capital"
historically have been poorly defined, making it
difficult for one firm to prevent access to the fruits of
its investment by other firms. It is relatively easy to
reconstruct the design and manufacturing steps
involved in an IC product through a process known
as "reverse engineering"9. By a sequence of photographic analysis, disassembly by etching, and
materials analysis, a rival firm can reconstruct the
architecture of a functional chip and the maskworks
and processing steps necessary to reproduce it. Such
"reverse engineering" can cost as little as one-one
thousandth of the original firm's investment lO and
permit "pirate" firms to enjoy lower total costs. The
passage of the Semiconductor Chip Protection Act
of 1984 foreclosed the possibility of precise
III.
Implications for Industry Structure and Performance
Scale Economies and Contestability
The economic characteristics of the IC industry
make it likely that the production of ICs is characterized by economies of scale. Significant scale
economies, in tum, would mean that markets for IC
products will tend to be concentrated in the long
run, and thus have the potential for inefficiency.
Baumol and Baileyl4, however, have argued that
high levels of concentration (or even monopoly) in
production need not have serious effects on market
efficiency if the market were "contestable". For an
The preceding discussion of the economic
characteristics of the IC industry may help explain
the likely structure and behavior of the industry,
particularly whether the industry exhibits characteristics that may make it vulnerable to anticompetitive behavior. It has been widely alleged, for exampIe, that Japanese producers have pursued predatory
pricing strategies in certain IC products 13. In this
section, the implications of the postulated economic
characteristics for structure and performance are
discussed as a prelude to attempts at empirical
verification.
44
industry to be considered contestable (in the sense
that Bailey and Baumol use the term), it must be
possible for new firms to enter a market displaying
abnormal profits and earn normal profits or
if
extant firms cut prices to thwart the new entry leave the· industry without losing the investment
associated with entry.
At least one attribute of the IC industry suggests
that it may not be ideally contestable: a major cost
of entry preproduction research and development
- is difficult to recover if the firm is unsuccessful in
competing against extant producers and must exit
the market. In this respect, the IC industry contrasts
with most manufacturing, transportation and service industries for which acquisition of re-sellable,
fixed assets is the dominant cost of entering a
market. Thus, "contesting" for markets may not be
an effective means of imposing competitive discipline within the IC industry. This makes it
especially important to investigate that industry's
scale economies.
produce more) in the leamingstage of production
than dictated by short run profit maximization considerations alone.
Not all environments are conducive to such learning curve pricing behavior, however. Whereas firms
in an industry composed of just a few firms are able
to exploit a learning-curve pricing strategy, Spence
argues that strategy is less effective in unconcentrated production environments where there is assertion of competitive price discipline with successful
entry.
In addition, whether learning is important at all to
either pricing behavior or market structure depends
upon how "rapidly" learning takes place. If learning were very rapid (that is, the effects of accumulated production experience are small relative to the
effects of current production on costs), then there
would be few strategic advantages to deviating from
short run profit maximization. A related point concerns how rapidly learning reaches other firms. If
such diffusion is very rapid, accumulated output
might help explain cost and price trends for the
industry as a whole, but current prices would be
determined by current costs.
The implications of Spence's view for the IC
industry might be summarized as follows. !flearning (cumulative output) were important to firms in
the IC industry, then early entrants able to survive
initial periods of low prices might gain an (at least
temporary) advantage over later entrants. Market
structure would be less fluid than otherwise, and
such "first movers" could enjoy higher profits than
subsequent rivals. Firms, in tum, would have a
strategic incentive to pursue an early-entry strategy.
learning and IC Market Efficiency
In Section I, we also postulated that IC production occurs in the presence of a learning curve. The
existence of learning effects on unit production
costs may have a bearing on both industry structure
and pricing behavior and, thereby, on the efficiency
of the IC industry. The logic of these effects in a
model of dynamic entry and pricing behavior has
been demonstrated rigorously by Spence 15. The
implications of his model will only be summarized
briefly here.
First, if learning (production experience) reduces
costs, Spence has demonstrated that, under certain
theoretical conditions, learning can confer some
protection from competitive entry to the first firm
into a market, in effect, simulating an entry barrier.
If "first movers" do enjoy such advantages in the IC
industry, then market structure might be expected to
be rigid over time
that is, show little change in
the rank and share of firms in the market.
The second aspect of learning curve theory of
interest here are the effects of the learning curve on
pricing behavior. In essence, because production
experience confers subsequent cost advantages on
the firm, a firm maximizing long lun profits in the
presence of a learning curve will charge less (and
International Competition in the IC Industry
For American IC producers, concerns over preproduction costs and scale economies, contestability, and learning phenomena seem to be at
the root of current debates over the marketing strategies of their Japanese competitors. Japanese producers have gained a growing share of the world
semiconductor market; their share of combined
U.S.-Japanese production has risen from 33 percent
in 1971 to over 50 percent in 1982. Since 1982, the
U.S. share of world IC sales has fallen from about
60 percent to 50 percent, while the Japanese share
has risen from 30 to about 40 percent 1Sa .
45
addition, the property rights traditionally extended
by the Japanese to foreign creators of intellectual
property have been criticized as being weak by
international standards. In the debate over software
copyright reform from 1983 to 1985, for example,
the Japanese proposed standards of protection were
weaker than both international copyright standards
and the standards applied to domestic (Japanese)
copyrights 20 .
As Dasgupta and Stiglitz21 have pointed out, illdefined intellectual property rights can reduce innovation below the socially optimal level. However,
for low barriers to cloning to have a permanent
effect favoring Japanese over U.S. production, the
ability to "reverse engineer" a competitor's product
must be asymmetric internationally which it is not,
and some other factor must operate to "cement"
market dominance once dominance is achieved by
this means.
Finally, it is possible that growth in the Japanese
IC market share flows from legitimate differences in
fabrication cost. These differences could arise from
lower costs for labor of a given quality or superior
Japanese management of fabrication facilities. (It is
unlikely that differences in materials or equipment
costs would be as important since most of the wafer
stock and fab line equipment has been manufactured
by one country: the U.S.) The theory of international factor price equalization22 argues against the
lower labor cost argument, but foreign producers
could still have the comparative advantage if they
have a greater endowment of relevant production
factors 23 .
It is frequently alleged that the Japanese have
obtained their growing share onc sales by pursuing
"predatory" pricing strategies. In particular, Japanese IC manufacturers have been accused of selling
Ie products in world markets at prices below their
cost of production. In 1986, for example, there were
three major International Trade Commission complaints alleging such behavior filed by the American
industry and the U.S. government 16.
If true, one explanation for such pricing behavior
would be the existence of exploitable learning curve
advantages, although any IC firm - not only the
Japanese ones - could exploit those advantages.
Nevertheless, it is argued frequently that Japanese
IC producers are better able to survive the early
periods of low profitability necessary to secure
market dominance in a "learning" dominated production environment. They are alleged to benefit
from their affiliation with conglomerate manufacturing organizations, which underwrite early periods of low profitability, and the availability of subsidies from the Ministry of International Trade and
Industry (MITI) and the banking industry 17 .
Whether such subsidization occurs (or differs dramatically from support the U.S. IC industry has
received from the military) has been debated extensively18.
A second alleged reason for the growth in Japanese IC market share is that low technological and
legal barriers to copying U. S. designs and processes
have unfairly teduced the total costs faced by Japanese producers. Particularly egregious cases of
apparent cloning indeed can be documented l9 . In
IV.
Empirical Examination of the
Several empirical investigations may inform our
understanding of the structure and performance of
the IC industry. First, a study of market share
rigidity may shed light on the structure of the IC
market. Market shares that appear to be rigid over
time might indicate that the market is not easily
contestable, or that learning phenomena operate to
retard entry.
Second, we could test directly for the existence of
learning phenomena by examining the response of
costs to cumulative firm output - costs should
decline with accumulated output experience. In
addition, if a high degree of market concentration
Ie Industry
were associated with lower IC prices (an association
not normally expected except in the presence of a
learning curve or other strategic pricing considerations), learning in the IC industry would be more
likely to be firm-specific (that is, it would not
diffuse so rapidly that it did not influence firm
behavior).
Third, the relationship between the scale of production and cost also would be of interest. If scale
economies were not extremely great, the industry
would be less likely to be concentrated, and the
potential distortions caused by lack of contestability
or other constraints on the fluidity of the industry
46
would be less important.
Finally, to address concerns about the behavior of
foreign competitors, it would be helpful to compare
fabrication costs in Japanese and U.S. facilities. If
Japanese IC prices were below their American
counterparts’ but fabrication costs were the same,
we might have evidence that Japanese producers
were pricing below the full cost of production,
including the cost of maskwork capital.
Unfortunately, the data available on the IC indus
try are not ideal for examining all of these relation
ships. No cost data are available and they would be
suspect in any case, since the cost of producing a
specific product is difficult to extract in a conglome
rate enterprise. This is a particularly serious short
coming in studying Japanese IC production costs.
Fairly good price and output data are available by
firm and device, however, as are data on the labor
and capital employed on individual fabrication
lines. In what follows, these data are exploited to
provide rough information on the relationships of
interest.
The Belhawior ©f the DRAS1 industry
We explore here the issues of market structure
rigidity, learning effects, and scale economies in the
context of a particular type of IC device — the
Dynamic Random Access Memory (DRAM) IC.
This device stores binary bits of information in a
randomly accessible manner. (The term “dynamic”
simply refers to the requirement that DRAMs be
powered continuously to retain implanted memory.
The term distinguishes them from a related device
— the Static RAM — that does not need continuous
electrical power.) The memory capacity of DRAMs
is measured in kilobits; each kilobit is 1,024 indi
vidual bits of memory capacity and is abbreviated
by a “K”. To date, DRAM devices have been
Chart 1A
Number of DRAM Devices Solid by Quarter
Chart 1B
Time Trend in DRAM Prices
Cents per 1000 Bits
S o u rce : D ataquest
47
manufactured in cdmmercial volume in 4K, 16K,
64K and 256K capacities.
We focus on DRAMs for a number of reasons
even though they represented only about 10 percent
of total IC sales worldwide in 1985. First, the
DRAM device is as close as the semiconductor
industry gets to a “commodity”-type of device.
Most other ICs have qualitative attributes that make
them difficult to study over time or across firms.
Second, unlike microprocessor ICs for example,
DRAMs have been produced in significant volumes
by non-U.S. firms, allowing some exploration of
the influence of foreign entry on industry behavior.
Indeed, DRAMs were involved in recent allegations
of “dumping” by the Japanese24. Finally, as a
practical matter, to expand sample sizes, it is neces
sary to combine data across devices. Such a com
bination is feasible with memory devices because
they are unambiguously “generic” in their essential
unit of service (the “bit”), and bits are substitutable
across devices.
Substitutability across types of DRAMs is illus
trated in Chart 1, which shows the actual quantity
shipped in Panel A and actual prices per bit for four
DRAM devices in Panel B. The sales of the 4K
DRAM, for example, peak and decline sharply
(note that all quantities are in log terms) when the
price per bit of the successor device (16K DRAM)
falls below the 4K price per bit. A similar pattern
holds for subsequent generations of devices. The
chart also illustrates vividly the observation made
earlier that increasing the bit density on the chip has
contributed importantly to the observed declines in
price per bit of DRAM memory.
Chart 2
3-Firm Concentration Ratios
for Each Type of DRAM
P ercent
Source: Dataquest, FRBSF
level. These findings probably are consistent with
the existence of important scale economies and,
therefore, limited “room” in the market for addi
tional firms. Indeed, as the market for 4K devices
matured and declined, the concentration ratio for
4K DRAMS gradually increased as firms exited the
market.
Although firm entry into the market for DRAMs
appears capable of reducing levels of industry con
centration for at least short periods of time, new
entrants have difficulty dislodging “first movers”.
To illustrate this observation, the orderings of firms
ranked by market share from one year to the next
were compared and a statistic measuring the correla
tion of these ranks — the Spearman Rank Correla
tion Coefficient — was computed for each pair of
adjacent years and for each device. (Unchanged
year-to-year rank ordering produces a Spearman
rank correlation coefficient of 1.0)25. The results are
presented in Chart 3.
The high correlation of firm market share rank
ings from one year to the next suggests that DRAM
market structure is not highly fluid after the initial
period of entry, and has with each successive
DRAM device reached this condition of structural
stability more rapidly. Moreover, data not presented
in the chart suggest that the first producers of a
device not only retain pre-eminence in the market
for that device but often are “first movers” into
production of the next generation device. Turnover
of producers is greatest among those firms that are
not first entrants. These observations may suggest
the existence of high fixed costs (either pre-produc-
Market Rigidity
In examining the DRAM market for evidence of
structural rigidity, it is instructive to trace the evolu
tion of market structure in DRAM manufacture. As
Chart 2 reveals, a new DRAM device typically is
introduced by one or two firms with entry occurring
gradually until concentration (as measured by the
share of the market held by the largest 3 firms)
declines to a relatively modest level.
In the cases of the early devices (such as the 4K
DRAM), entry occurred more gradually than with
subsequent generations of devices, and concentra
tion levels did not decline below the 50 percent
48
Chart 3
Time Trend in Market Structure
Rigidity as Measured by the
Spearman Correlation of Market Share
where
priced,t) = price per bit, in U.S. dollars for
device x at time t
timed,t) = t = the date of the price observation
on device x
output(x,t) = average industrywide output per
firm (in number of devices), for device x at
time t
cumulative output(x,t) = average device output
per firm summed over time by device, for
device x at time t
market structured,t) = 3-firm concentration
ratio in market shares of device x at time t
Japanese market shared,t) = fraction of total
production by Japanese firms, for device x at
time t
size(x,t) = the memory capacity in bits for
device x at time t
and
x = l , 2, 3, 4
t= 1 to 44 measured from 1974, quarter one
Source: Dataquest, FRBSF
tion or production) or learning phenomena that
benefit incumbent firms.
Scale Economies and Learning Effects
Without specific data on IC production costs, it is
not possible to test directly for the existence and
importance of learning in DRAM manufacture, or to
explore directly the magnitude of scale economies.
However, an examination of the behavior of DRAM
prices in addition to the inferences drawn above
from the behavior of market structure may shed
some light on cost behavior. In particular, except in
instances of coordinated or monopoly pricing,
prices and average costs are likely to move together
over time. This relationship suggests that some
inferences about costs can be derived from price
data if the circumstances that might lead to noncom
petitive pricing can be controlled.
To explore these relationships more formally (and
to control for the effects of the passage of time,
accumulated production experience, and market
structure), we studied a simple econometric rela
tionship using quarterly data on DRAM devices
over the 1976 to 1985 period. We studied the
variable the price of a bit of DRAM, and we pooled
time series data on four devices (the 4K, 16K, 64K
and 256K DRAM) to expand the sample size. The
general form of the relationship studied was:
Price(x,t) = h[time(x,t), output(x,t),
cumulated output(x,t),
market structured,t),
Japanese market share(x,t) and
size(x,t)]
Table 1 presents a regression analysis of a specific
configuration of Equation 1. In particular, ordinary
least squares were used to estimate a relationship
between the log of DRAM prices and the variables
identified in Equation 1. Most of the coefficients are
made a function of device size by interacting each
variable with the device size measure. All of the
coefficients in Table 1 are measured with consider
able statistical precision, permitting a number of
interesting observations.
First, the positive coefficient on the Size variable
suggests that (if prices follow costs) increases in
device size increase the cost of a bit of DRAM. That
is, with a given technology, it is more costly (per bit)
to produce large rather than small devices. The
negative coefficients on Time and Time times Size,
however, indicate that technological progress
decreases bit price and that technological progress
has been most important for large devices. The
positive coefficients on Time Squared and Time
Squared times Size indicate that the influence of
technological change in reducing the bit price are
diminishing with time and with increased device
size over time. This is consistent with the idea of
diminishing returns to size-related innovations and
innovations generally.
(1)
49
maximum of 8.53 million devices per quarter
observed in the sample. It thus offers some comfort
that the use of prices to study cost behavior may not
be unreasonable.
Third, the negative coefficient on Cumulative
Output indicates that, at least for the industry as a
whole, cumulative output has an effect independent
of that of current output. This finding supports the
hypothesis that learning contributes to IC production behavior. Although it is not evident that
Second, the negative coefficient on Output and
the positive coefficient on Output Squared implies
that the price per bit declines and then rises with
increased output. This is consistent with the notion
that there are economies of scale associated with IC
production over some range. The rate of firm output
at which prices begin to rise can be derived from the
estimated coefficients and is 5.4 million devices per
quarter for a 64 kilobit device. This rate is over four
times the observed average output and close to the
50
would be useful. The available data, however, does
not allow us to resolve all of these potential sources
of bias.
individual firms are able to exploit learning curve
phenomena in devising pricing strategies, it does
indicate that the theoretical potential to do so exists.
Fourth, the positive coefficient on the Concentration Ratio and the negative coefficient on Size times
the Concentration Ratio suggests that the influence
of market share concentration on DRAM prices is
positive for small devices, but negative for larger
effect becomes negative for a "size"
greater than about 40 kilobits.) As was observed
earlier in this paper, a negative effect of market
share concentration on price is suggestive of firmspecific strategic pricing in the presence of a learning curve. The negative coefficient for 64K and
256K devices is interesting because these are the
device sizes that came to be dominated by the
Japanese, suggesting, perhaps, that the Japanese
introduced a different pricing strategy into the
DRAM Ie market.
The independent effect of Japanese market presence on DRAM prices is given by the negative
coefficient on Japanese Market Share and the positive coefficient on Size times Japan Market Share. It
appears that, controlling for other market and production influences, the effect of the Japanese presence was to reduce prices for smaller devices
(empirically, smaller than 57 kilobits), but to elevate
prices for larger devices. Once again, the markets
for larger devices are ones that the Japanese are said
to dominate. Since we have controlled for industrywide learning by Cumulative Output, and early
strategic "underpricing" may be accommodated by
the concentration ratio variables, this finding could
be seen as evidence that prices can be elevated by
successfully dominating the market.
There are a number of important qualifications to
these findings. An obvious difficulty lies with the
study of average
and average firm output to
make inferences about what inherently is relevant
to individual firm costs and output. Many
relationships that hold at an individual firm level are
not appropriately aggregated or averaged. A similar
criticism attaches to the use of average cumulative
output instead of individual firm data to detect the
presence of learning curve phenomena. More complex functional forms, separate specification of the
estimated relationship for each device, and recognition of qualitative differences across devices also
A Study of Fabrication Facilities
Our second empirical investigation focused on
the fundamental unit of fabrication: the fab line.
This investigation is of interest both to verify the
casual observation made in· Section II about the
likely lack of scale economies in fabrication and to
study differences in fabrication activity on fab lines
operated by Japanese and American firms. If fabrication costs were the same in both countries, then
such costs would have to be eliminated as a source
of differences in pricing strategies.
Once again, quite severe data limitations restrict
the type and quality of analysis that can be performed. Fab line cost data are not available; only
data on installed capital equipment (in 1986 U.S.
dollars), the number of employees engaged on the
line, and wafer start activity are available. Data on
the specific type of IC device produced on the line
also are not available 26 . Nevertheless, the data do
permit two simple empirical tests
a comparison
of capital-labor ratios and a comparison of fab line
production functions.
First, using data on 386 fab lines from June 1986,
we computed the capital-labor ratios separately for
American and Japanese fab lines 27 , and found the
ratio in Japanese fab lines to be approximately 2
percent less than the ratio on American lines 28 . This
finding is likely an understatement of the actual
difference because the Japanese work week is one
day longer than the 5-day U.S. standard. Thus, for
Japanese lines, the number of employees on the fab
line is a downward-biased measure of labor input
flows (in man-days).
The finding of a 2 percent difference implies that,
assuming the same fabrication technology and
quality of labor, the unit cost of labor relative to
capital is lower for Japanese than American firms.
Since the capital equipment costs likely are very
similar (since much of the equipment is American in
origin), this, in turn, could be consistent with the
existence of absolutely lower labor factor costs in
Japan 29.
The second use of the data was to estimate a fab
line production function directly for a combined
51
sample of Japanese and American fab lines. The
estimated functional fonn was the Cobb-Douglas
representation of the production function:
Q = aLbKc
of these coefficients is a measure of economies of
scale30 . (Specifically, if b plus c in equation 2 is
greater than one, the production function exhibits
economies of scale; if they sum to less than one,
there are diseconomies to large scale production. ) In
addition, with simple assumptions about factor
prices and the profit-maximizing behavior of the
finn, cost functions can be derived 3l .
The estimates of the coefficients of equation 2
were obtained by taking the logarithm of both sides
and using ordinary least squares regression techniques. The results are presented in Table 2, with a
dummy variable introduced to identify possible
coefficient differences between American and Japanese fab lines. The coefficients on fab line employment and fab line capital are, respectively, .54 and
.35 for the American fab lines. It thus appears that
there are no scale economies from fabrication per
se. Indeed, the point estimate of the coefficients
imply very slight diseconomies of scale (.89 versus
1.00 for constant returns to scale) although the
estimate cannot be distinguished from 1.0 statis-
(2)
where
Q
= fab line output (measured in square inches
L
K
=
of wafer starts per week)
the number of fab line employees
the dollar value of fab line capital (in
millions of U.S. dollars).
The Cobb-Douglas representation has a number
of well-known limitations, the most important of
which is that the rate of substitution between factors
is constrained to be equal to one. It has the advantage, however, that the exponents of labor and capital provide estimates of the marginal products of
these respective factors. It also can be shown that
the relative sizes of these two coefficients are related
to the relative contribution of each factor to total
product under certain assumptions, and that the sum
52
ticaUy. The importance of the labor component of
fabrication inputs also is illustrated by the estimates, which show that labor's contribution to total
product is approximately fifty percent greater than
that of capitaPZ;
Important qualifications on this finding must be
offered, however. First, the output measure used in
estimating equation 2 is not the number of usable
ICs· completed but simply square inches of wafer
processed. To the extent that Japanese and Americanfirms differ in their ability to recover usable ICs
from this process, effective cost per IC would differ34 •
Second, Japanese and American producers may
emphasize different products not accommodated by
the simple production function estimated here.
Some data are available on coarse product categories associated with each fab line. However, the use
of separate dummy variables for these categories
did not significantly influence the estimated parameters, perhaps because the sample sizes for some of
those product variations were small.
Finally, if the differences in capital-labor ratios
on fab lines observed earlier in the two countries is
indicative of absolute differences in labor and capital costs in the two countries, production costs
would differ accordingly.
American vs Japanese Production
Functions
There also appear to be no differences in fab line
production functions between American and Japanese facilities. All three of the coefficients on the
variables designed to capture these differences (that
is, the Japanese dummy variable and its interaction
with fabrication employment and fabrication capital) are not statistically different from zero. From
this evidence alone, there is little to suggest that
fabrication economies can explain differences in
final product prices between the two countries. In
other words, if manufacturers in each country were
profit-maximizing and faced the same labor and
capital costs, there is no statistical evidence that
total fabrication cost relationships would differ33 .
V.
Conclusions
Although the available data do not permit definitive assessment of the factors that may affect firm
behavior within the IC industry, some light has been
shed on two major aspects of the performance of this
industry. The first is whether the industry is prone to
high levels of market share concentration or other
features that may result in inefficient performance.
Pricing behavior was consistent with sizeable, but
not extreme, overall scale economies, which
include pre-production costs. Market structure in
commodity-type DRAM devices appears to be concentrated only at low industry output levels.
Pricing behavior in the DRAM market is,
however, consistent with the existence of a firmspecific learning effect. In addition, large, "sunk"
pre-production investments are required to enter
new device markets. Both phenomena would tend to
give strategic advantages to incumbent firms and
hence to firms (of any nationality) that might be
supported through periods of negative earnings
while they acquire the advantages of production
experience and incumbency.
The second major issue confronting the IC industry is the conduct of foreign competitors compared
to U.S. manufacturers. The fact that estimated fab
line production functions did not uncover significant intercountry differences casts some doubt on
differences in fabrication costs as a source of competitive advantage for Japanese producers.
Weighing against this view is our finding (in the
context of fabrication lines) that firms in Japan
behave as if their labor is less costly (or of lower
quality) relative to capital than firms elsewhere.
Since the conventional wisdom is that Japanese
labor is not of lower quality, the behavior of Japanese firms argues in favor of a cost-advantage to
Japanese production of ICs. Perpetuation of this
disparity runs counter to the notion of international
factor price equilibration predicted by trade theory,
but without additional information, the argument
that cost differences are the basis for the growing
presence of the Japanese in the IC market must
stand.
53
of the U.S. Semiconductor
national semiconductor pricing policy. The agreement was reached by negotiation between the U. S.
and Japanese governments to resolve complaints
about Japanese IC pricing policy brought before the
U.S. International Trade Commission and inpetitions filed under Section 301 of the Trade Act of
197435 .
The ITC complaints and petitions were dropped
in return for agreements from the Japanese.to cease
the alleged practices of (1) retarding U. S. entry into
Japanese markets and (2) "dumping" of Japanese
products below cost in U.S. markets. In particular,
the agreement provides firms in both countries protection against "subsidized" sale of semiconductors (priced below "company specific cost of production plus 8 percent"). Also, as part of the
agreement, the Japanese government is charged
with monitoring the relationship between firm costs
and selling prices abroad.
The agreement potentially could proVIde a forum
for resolving the debate about whether the Japanese
producers are, in fact, subsidizing IC production. To
the extent that the concept "company specific cost
of production" is meant to refer to short-run average
costs, the agreement also could retard learning
curve pricing strategies and thereby improve prospects for American firms. At this writing, however,
the agreement had broken down because of the
alleged failure of the Japanese government to
enforce its terms.
Several of our findings suggest that semiconductor markets lost to foreign competition may be
difficult to recover. For one, assuming that produccost-advantages on· a firm
and
growing scale of "sunk" costs is associated with de novo entry, it follows that new nonsubincreasingly difficult to
dis;lo(ige incumbent firms, whether that incumthrough cost-advantages or
subsidized operation. Second, although the DRAM
Ulan..\-·., at least, is not especially concentrated at
markets in DRAMs have tended to
become rigidly structured over time.
On a more positive note, two recent policy
changes have important implications for American
firms. First, the passage of the Semiconductor Chip
Protection Act, by giving property rights to
designers of semiconductor chip maskworks,
should reduce significantly the more egregious
practic;es. The Act will reduce the likelihood
that different firms will face different effective costs
of maskwork capital. If, as is popularly alleged,
foreign firms previously acquired maskwork capital
through reverse engineering at the expense of American firms, the improvement in property rights in IC
maskworks should benefit American IC producers.
The second important policy initiative is the 1986
agreement reached between the U. S. and Japanese
governments regarding, among other things, inter-
54
FOOTNOTES
1. See, for example, "The Solid State Era," Electronics,
April 1980, M. S. Kiver, Transistor and Integrated Electronics, McGraw-Hili, 1972, and W. C. Hittinger, "Metal
Oxide Semiconductor Technology," Scientific American,
August 1973, pp. 48-57.
15. A. Michael Spence, "The Learning Curve and Competition," The Bell Journal of Economics, Spring 1981, pp.
49-70. For an application of the learning curve evaluation
process to another industry, see, for example, M. 8.
Lieberman, "The Learning Curve and Pricing in the Chemical Industries," The Rand Journal of Economics, Summer
1984,213-239.
2. World Semiconductor Trade Statistics Committee,
Semiconductor Industry Association, "Semiconductor
Forecast Summary," September 1986.
15a. Statistics on world semiconductor sales shares are
not known with precision. The statistics on US and Japanese semiconductor market shares are from "International Competitiveness in Electronics," US Office of Technology Assessment, Washington, DC, 1983. The data on
world IC shares are from "For Chipmakers, the Game has a
New Set of Rules," Business Week, January 13, 1986, p.
90. The trade balance in ICs also reflects Japanese competitiveness. In 1979, the US/Japan trade balance in ICs
was a surplus of $90 million. By 1984, this had deteriorated
to a deficit of $884 million. (Source: Industry Week, November 25, 1985.)
3. Excellent summaries of the technical relationships
among IC and other semiconductor devices are available
in the annual reports of VLSI, Inc., a San Jose, California,
research firm.
4. See, for example, the F. Thomas Dunlap, Prepared
Statement, "The Semiconductor Chip Protection Act,"
USGPO, J-98-39, pp. 152-168, and "Intel's Development
of 386 Chip Took 4 Years and $100 Million," Wall Street
Journal, August 29, 1986, p. 4.
5. See footnote 3.
6. Wafer processing into ICs involves such steps as
application and baking of special coating materials, deposition or doping selected portions of the surface, etching
and numerous measurement and testing steps. The speed
of each of these steps is not easily accelerated. In addition,
the wafer is carried from one step to the next mechanically
and production takes place, for many of the critical steps,
in hoods or rooms with highly processed atmospheres. It is
difficult to add equipment, work stations, or speed up
processing at will in such a confined and sequence-driven
environment.
16 These involved 64K DRAMS (ITC Docket #731TA-270), 256K DRAMs (ITC Docket #731-TA-300) and
EPROMs (ITC Docket #731-TA-288) DRAMs are Dynamic
Random Access Memory ICs and EPROMs are Erasable
Programmable Read Only Memory ICs.
17. See, for example, "The Effect of Government Targeting on World Semiconductor Competition: A Case History
of Japanese Industrial Strategy and Its Costs for America,"
Semiconductor Industry Association, 1983.
18. This debate is presented in Okimoto, Sugano and
Weinstein, eds, The Competitive Edge: The Semiconductor Industry in the US and Japan, Stanford University Press,
1984.
7. These data refer to MOS technology fab lines. The
source of the data is VLSI, Inc.
8. Specifically, the original contact photolithographic
techniques for transferring circuit designs to the surface of
the IC have been improved first through optical projection
techniques and today, electron beam lithography techniques.
19. In one seemingly egregious instance, an American
chip design was reproduced by a Japanese producer so
precisely that even a microscopic error in the chip architecture was replicated. See the document cited in footnote
4 above.
20 See J. Chesser, ibid
9. The prepared statement of the Semiconductor Industry
Association, ibid, pp. 122-128, discusses the reverse engineering process.
21. P. Dasgupta and J. Sitglitz, "Uncertainty, Industrial
Structure, and the Speed of R&D," The Bell Journal of
Economics, Spring 1984. See also J E. Tilton, "International Diffusion of Technology: The Case of Semiconductors," Studies in the Regulation of Economic Activity, The
Brookings Institution (Washington, DC), pp. 24-38 Gart
and Konakayama study the diffusion of the production of
an innovation using a simple model and several industries,
including the transistor
See Gort and
Konakayama, "A Model of Diffusion in
Production of an
Innovation," American Economic Review, December
1982, pp. 1111-1119.
10. From the testimony of F. Thomas Dunlap. See footnote
4.
11. Karen Ammer, "The Semiconductor Chip Protection
Act of 1984," Law and Policy in International Business, Vol
17, 1985, pp. 395-420. See also, J. Chesser, "Semiconductor Chip Protection: Changing Roles for Copyright and
Competition," Virginia Law Review, Vol 71, 1985, pp.
249-285.
12. This notion was studied in an early Federal Trade
Commission report and studies performed by the Boston
ConSUlting Group. See "Staff Report on the Semiconductor
Industry," Bureau of Economics, Federal Trade Commission, January 1977 and "Perspectives on Experience,"
Boston Consulting Group, 1972.
22. Miltiades Chacholiades, International Trade Theory
and Policy, McGraw-Hili, 1978, Chapter 10
23 M. Chacholiades, Chapter 11.
24 Specifically in the case of 64K and 256K DRAMs. See
footnote 16 above.
13. See, for example, the complaint before the International Trade Commission regarding 256K DRAMs (ITC
Docket number 731-TA-300)
25. The Spearman Rank Correlation Coefficient measures
the degree of correspondence between two series of
numbers by examining rank differences. This staUstic is
useful for studies of market rigidity because in a fluid
14. See E. Bailey and W. Baumol, "Deregulation and the
Theory of Contestable Markets," Yale Journal on Regulation, Vo11,1984, pp.111-137.
55
ignored and that estimation of production frontiersgenerally' raises problems associated with nonnormally dis'
tributed errors.
market, market shares held by individual firms would
change over time as.new entrants disturbed market shares
and place-switching occurred among extant firms. The
fact that the Spearman Correlations observed in the DRAM
market are. high is illustrated by a few simple examples.
Assume that an industry is structl.lred so that the .Iargest
firm has 30 percent market share, followed by firms with 25,
20, 15, 10, 5, and 2.5 percent respectively. If the first and
fourth firms switch places, the correlation between the old
and new structure would be only 67 percent. If, instead, the
smallest firm exits the market (with the next smallest firm
taking up its market share), the correlation would be 92
perGent. Both of these correlations are less than the yearto-year correlation observed in a mature DRAM market.
31.SeeH Varian, Chapter 1.
32. That is, the marginal product of labor is greater than
the marginal prOduct of capital by about· 50 percent. If
constant returns to scale were exhibited, it can be shown
that under competitive market conditions, the coefficients
on labor and capital could be interpreted as their respective factor shares.
33. The derivation of cost functions from assumptions of
profit maximization and Cobb-Douglas production technology are presented at length in H. Varian, ibid, Chapter
1.
26, The source of this data was VLSl,lnc., San Jose,
California. The data are not available from this author, as
per agreement with VLSI, Inc.
34.. Because of the extremely small physical size of the
constituent elements of an IC, small impurities or imperfections in the processed surface of the wafer can cause an IC
to perform inadequately, thereby reducing the effective
yield of ICs per wafer start. There may be international
differences in the ability of producers to improve effective
yields. In addition to eliminating or reducing the factors that
caused the imperfections in the first place, it is also possible to build in circuit redundancy and other means of
"salvaging" processed, but imperfect, chips. Even under
the best of circumstances, however, the ratio of the actual
to potential number of ICs per wafer may be as low as 50
percent for very large scale integrated devices sUGh as the
256K DRAM (Business Week, August 18,1986, p. 66).
27. The sample was confined to metal oxide semiconductor technology lines for comparability.
28. The difference is of only marginal statistical significance. Specifically, the difference is different from zero at
approximately the 80 percent confidence level.
29. The capital-labor ratio also could be affected by differences in the price of capital. Indeed, it is argued frequently that Japanese producers have access to subsidized financial capital. If true, the reduction in the capitallabor ratio stimulated by lower labor costs could be offset
by lower user costs of fab line capital.
35. See, "Chip Fight is Settled," Los Angeles Times,
August 1,1986, p. 11, and Semiconductor Industry Association, "US, Japanese Governments Reach Agreement
on Market Access, Prevention of Dumping," SIA Circuit,
Autumn, 1986, p. 1.
30. See, for example, H. R. Varian, Microeconomic Analysis, Norton Books, 1976, Chapter 4 for a discussion of the
econometric problems encountered in direct estimation of
production relationships. See also G. S. Maddala, Econometrics, McGraw-Hili, 1977, Chapter 13. Suffice it here to
say that the endogeneity of the right hand side variables is
56
Carolyn Sherwood-CaW
During the past ten years, voters in many states have passed measures
that limit the taxing or spending powers of local governments and thus
their average level of services and ability to differentiate themselves from
one another. This study of the effects of Proposition 13. a property tax
limitation initiative passed in California in 1978, concludes that the
initiative has reduced the overall size of local government, but that its
effects on fiscal differentiation vary considerably depending on the extent
of local governments' additional constraints.
During the past ten years, voters in many states
have passed measures that limit the authority of
governments in important ways. Most commonly,
these initiatives have taken the form of limits on
property tax rates or assessment practices, but
spending constraints also have been passed.
These statewide initiatives potentially could
interfere with the federal system of government.
Under the federal system, different levels of government are responsible for providing different types of
public or quasi-public goods and services. One can
think of all goods as lying on a continuum between
purely public goods and purely private goods. On
this continuum, the national government should
provide the purely public goods because the benefits
of such goods are dispersed and the costs of producing purely public goods do not increase as the
number of beneficiaries increases.
The degree of publicness of the goods provided
should decrease as the level of government progresses from national to state to local. Thus, most
services provided by local govemments are nearer
to the private good end of the continuum. Fire
protection, for example, can be provided privately
because it can be priced, its benefits can be limited
to those who pay for it, and the cost of service
provision increases as more households receive fire
protection. At the same time, it is public in the sense
that it benefits many who do not consume it directly.
That is, if one house on a block has adequate fire
protection, nearby houses are protected from the
spread of a fire that starts in that house.
Quasi-public goods such as fire protection are
particularly well-suited to local government provision because a rudimentary "market" allows individuals, by choosing a jurisdiction of residence, to
select a combination of taxes and public services
that suits their tastes and needs (Tiebout 1956).
Thus, the abilities of local governments to provide
different levels and mixes of services allows these
quasi-public goods to be provided more efficiently,
and the federal system to function more smoothly.
Statewide initiatives that limit local governments
could interfere with this federal system of providing
goods and services. Measures that tightly circumscribe local governments may limit the extent to
which jurisdictions can differentiate themselves
from one another. Residents of some communities
may not be able to consume the level of government
services that they would have chosen and for which
* Economist, Federal Reserve Bank of San Francisco. Martha Sellers provided invaluable research
assistance. The paper also benefited from conversations with Larry McCarthy and Rebecca Taylor,
both of the California Taxpayers' Association, and
from comments by the editorial committee.
57
they would have pa'id in the absence of such restrictions.
In this paper, I examine the effects of Proposition
13, a property tax limitation initiative passed by
California voters in 1978, both on the average level
of services provided by local governments in California and on the extent to which different localities
could continue to provide a variety of service levels.
The effects differ substantially among the various
types of local jurisdictions depending on their
reliance on the property tax, alternative financing
sources, and the extent of other constraints.
I.
To establish the context for the passage of Proposition 13, Section I describes California's local
government institutions and trends during the early
and middle 1970s. Then, Section II describes and
interprets Proposition 13 and other initiatives that
comprised California's "tax revolt." In Section
California data are used to examine the hypothesis
that the statewide limits on local government
reduced the extent to which local governments
could carry out their functions within the federal
system. Section IV summarizes and draws conclusions.
California local Government Before 1978
During the early and middle 1970s, California
was a high-tax state by almost any measure. For
example, in Fiscal Year 1970-71 (hereafter "FY
1971 "), California's total state and local tax burden
of $137 for each $1,000 in personal income was
high enough to rank eighth nationally. Of that total,
49 percent was generated by property taxes, 15
percent by sales taxes, and 10 percent by the state's
personal income tax I. The income and sales tax
burdens were not excessive relative to those in other
states, as California ranked 18th in income taxes per
$1,000 of personal income and 28th in sales taxes.
Instead, the burden of property taxes appears to be
largely responsible for the overall high tax burden
borne by state residents. California ranked second
in the nation in property taxes paid per $1,000 of
personal income. Thus, Californians easily could
have perceived their property tax burdens as excessive even before the years immediately preceding
their tax revolt.
Chart 1 illustrates changes that occurred in California's combined state and local tax sources prior
to the implementation of Propositon 13. By FY
1978, total tax collections for state and local governments had risen somewhat, to $158 per $1,000
personal income, from $137 in FY 1971. Meanwhile, property tax collections per $1,000 personal
income actually fell slightly, from $67 in FY 1971 to
$64 in FY 1978. As a result, the property tax
became relatively less important. Its share in total
tax collections fell from 49 percent in FY 1971 to 40
percent in FY 1978. At the same time, income and
sales taxes became relatively more important revenue sources, providing 17 and 18 percent, respec-
tively, of total state and local tax revenues in FY
1978.
The most striking observation regarding this
period is that, despite the rapid increase in property
values during the period and the widespread popular
impression of mushrooming property tax burdens,
property taxes as a proportion of personal income
were in fact shrinking. Property tax rates themselves
fell slightly, but much more surprising is the drop in
assessed value per $1,000 of personal income from
$2,487 in 1971 to $2,380 in 1978. Market values of
property were increasing rapidly during this period
but growth in assessed values did not exceed the
rapid growth in personal income which also characterized the period. One explanation for the slower
Chart 1
State and Local Tax Revenues
Per $1000 Personal Income
California excluding the City and County of San Francisco
Dollars
160
150
140
130
120
•
Total Stale and Local Tax Revenue
110
100
90
80
Tax
70
60
50
40
30
20
10 L..--'-_...L---I._...l----I_-L.=;.;.;:;,..;.;.;;;,.
1971
1972
1973
1974
1975
1976
1977
Tax
1978
Fiscal Year
"Gas, cigarette, alcohol and corporate income taxes
accounted for 60 percent of miscellaneous revenues in 1977.
58
Starting in FY 1974, the school funding process
changed considerably as a result of a 1971 ruling of
the California Supreme Court (Serrano v. Priest,S
Cal. 3rd 584). The Court ruled that it was unconstitutional for some school districts to provide
inferior education because low wealth in their areas
limited their property tax revenues. To implement
the Serrano decision, the state legislature placed a
cap on the amount of property tax revenues per pupil
that school districts could raise. Each school district's allotment was increased annually, but to
reduce the gap between high-wealth and low-wealth
districts, those districts that had raised less money
per pupil during FY 1973 received larger increases.
By capping the principal revenue source available to
school districts, Serrano effectively limited perpupil spending as well.
At the same time, Serrano imposed a minimum
per-pupil spending level on all school districts that
raised the share of state aid in school funding. For
those districts with low property tax revenues, state
funds filled the gap between the 1973 property tax
revenue base and the minimum per-pupil spending
level. As a result, the share of property tax revenues
in total school district funds fell to 46 percent in FY
1974 while the share of state aid rose to 39 percent.
During subsequent years, the property tax share
recovered somewhat, reaching 50 percent by FY
1978.
rise in assessed property values is that, at least in
some areas, assessed valuations were kept below
market value precisely to avoid the political consequences that would accompany sudden and sharp
increases in property tax liabilities.
Calculations by Oakland (1981) suggest that
assessed values of single-family homes were growing much more quickly during the immediate preProposition 13 period than were property values
more generally2. Therefore, individual homeowners
are likely to have experienced substantial increases
in their tax bills during this period.
Property Tax Structure
Prior to Proposition 13, property tax rates were
not coordinated among the various fiscally independent local jurisdictions such as cities, counties, and
school districts. As a result, voters who wished to
reduce their property tax burdens had no clear local
target to confront.
The governing body of each jurisdiction established its property tax rate annually to conform to its
budget requirements, and forwarded the rate to the
county assessor's office. The county then calculated
tax bills for individual parcels by adding up the rates
for all jurisdictions that serviced the area and multiplying by the assessor's office estimate of property
value. Thus, an individual property's tax bill would
likely include payments to a multitude of different
jurisdictions, including the county, city, and school
district as well as any number of special districts for
services such as water, lighting, fire protection,
streets, parks, flood control, cemeteries, or pest
control.
In practice, a little over half of all California
property tax revenues funded school districts. In FY
1972, school districts received 52 cents of each
property tax dollar, while counties received 32
cents, cities 10 cents, and special districts 6 cents.
Counties
Counties also rely primarily on property taxes
and intergovernmental grants for revenue. In California, counties provide a rather limited range of
services, the most important of which are public
assistance programs, judicial services, and health
services. Most of these functions are mandated by
the state and, as a result, the state traditionally has
viewed counties more as administrative arms of
state government than as autonomous local governments. A "rule of thumb" used in state government
is that 75 to 85 percent of county expenditures are
mandated by the state, leaving only 15 to 25 percent
of county spending under local control. As a result,
the level of differentiation among counties is limited 3 .
Consistent with these characteristics, counties
School Districts
Conversely, the property tax was a particularly
important source of revenue for school districts. In
FY 1972, for example, 54 percent of school district
funds in California were derived from the property
tax. Of the remainder, a majority (31 percent) came
from state aid.
59
Moreover, Cities can impose a relatively wide
range of taxes, including hotel, utility, and payroll
taxes. In addition, the state sales tax law allows
cities to receive sales tax revenue equal to 1.00
percent of sales6 . In 1978, property taxes provided
14 percent of total city revenues, while nonproperty
taxes provided 19 percent, current service charges
31 percent, state and federal grants 20 percent, and
miscellaneous other sources the remainder.
In summary, in the years immediately before
California voters passed Proposition 13, total state
and local tax revenues were growing modestly relative to personal income. At the same time, income
and sales taxes were becoming more important both
absolutely and relative to property taxes. More
surprisingly, property tax revenues actually were
falling as a proportion of personal income.
have little independent revenue-raising authority.
They can impose property taxes and a few other
taxes such as sales 4 , real property transfer, and
timber yield taxes, and they can charge fees for
services they provides. In 1978, before Proposition
13 was implemented, intergovernmental grants
provided 50 percent of county revenues, while property taxes provided 33 percent, and user fees 9
percent. Nonproperty taxes and miscellaneous revenue sources accounted for the remainder.
Cities
Cities have considerably greater leeway both in
terms of the services they provide and in terms of the
financing instruments available to them. With few
mandates from higher levels of goverment, cities are
relatively free to spend their money as they see fit.
Typical city services include fire and police protection, streets, parks, libraries, and museums.
II.
Changes in Local Finance
most jurisdictions9 . More recently, however, it has
become binding for many localities 10, and the number of jurisdictions constrained by the Gann spending limit is likely to mushroom over the next several
years ifrevenues grow as rapidly as expected.
There are several possible reasons for the passage
of these initiatives. Some argue that voters objected
to increases in their property tax bills that were not
the result of an explicit policy decision to increase
property tax revenues, but rather resulted simply
because property values rose. In fact, Oakland's
calculations (1981) suggest that many owners of
single-family homes did experience exorbitant
increases in the assessed value of their homes,
although the data presented in Section I suggest that
such increases did not occur among property assessments more generally. Moreover, those for whom
assessments did not rise dramatically still faced
large potential reassessments and, in many areas,
actual reassessments were not carried out uniformly
or equitably.
Others argue that the tax revolt was a response to
voters' perceptions that, despite a widespread
demand for smaller governments, government was
growing too large. This interpretation implies that
voters thought that governments had become unresponsive to them, and that government was growing
In 1978, California voters passed Proposition 13,
which placed a one percent ceiling on property tax
rates and stipulated that a higher rate could not be
imposed without a two-thirds majority of voters. 7
Since property tax rates in California had averaged
2.67 percent of assessed valuation in 1978 and
were in some cases over 3 percent, Proposition
13 immediately cut property tax rates substantially.
In addition, the initiative rolled back all assessed
property values to their 1975-76 levels. Annual
increases in assessed values could not exceed 2
percent or the inflation rate, whichever was lower.
When a property was sold, however, it automatically
would be reassessed at its market valueS.
In 1979, only a year after Proposition 13 was
passed, California voters approved another government limitation initiative. The Gann initiative
(Proposition 4) prohibited any state or local jurisdiction from spending more than it had spent during
FY 1979, adjusted for increases in prices and local
population. The Gann limit, like Proposition 13,
included a voter override provision. By a simple
majority, voters could approve spending in excess of
the limit for a four-year period.
The Gann spending limit was more or less forgotten for several years because revenues were growing
slowly enough that the limit was not binding for
60
ments unrelated to an increase in their desire to pay
for and receive local services. Alternatively, voters
could have sought to reduce the absolute size of
local (and state) governments by wresting control of
local governments from politicians and bureaucrats.
Under either interpretation, Proposition 13 is likely
to have resulted in smaller local governments and in
reduced property tax burdens.
If local governments have, in fact, become
smaller, we would expect to see lower revenues and
also less variation in the level of spending among
jurisdictions. The imposition of a new upper reve
nue limit would restrain jurisdictions from spending
larger than normal amounts of money and thereby
reduce the degree of variation among them. Such a
limit would compromise local governments’ roles in
the federal system by allowing voter-residents fewer
choices of tax-spending combinations than they
otherwise would have enjoyed.
“too fast.” The data suggest that state and local
governments in California were imposing an
increasing tax burden on state residents as tax reve
nues per $1,000 personal income grew from $137 in
1971 to $158 in 1978. Although the increase was not
related to the property tax11, the property tax
provided a convenient target because it was the
largest single source of state and local revenues, and
its administration was so disjointed. Since the Gann
initiative was unrelated to property taxation and
applied to the state government as well as to local
jurisdictions, its passage reinforces the idea that
taxpayers had grown dissatisfied with the size and
unresponsiveness of local governments.
Thus, two related explanations for the tax revolt
are consistent with both the public finance climate
during the middle 1970s and the passage of Proposi
tions 13 and 4. Homeowners could have been rebell
ing against an increase in their property tax pay
IIS. The impact of the Fiscal Changes
1979. By 1984, the burden had fallen still further, to
$115. California generally ranked among the top
five states in terms of total state and local taxes per
$1,000 in personal income during the pre-Proposition 13 years, and was consistently in the top ten.
The state’s ranking fell to 25 immediately after
Proposition 13 was imposed and averaged 20 during
Changes in the fiscal environment for state and
local governments were dramatic during the late
1970s and early 1980s, as Chart 2 illustrates. In FY
1978, just before Proposition 13 took effect, Cal
ifornia’s property tax revenue per $ 1,000 of personal
income was $64, and property taxes provided 40
percent of all state and local tax revenue in Califor
nia. In the following year, the property tax burden
had fallen by more than half, to $30, providing only
25 percent of California’s state and local tax reve
nues.
Sales and income tax burdens on California tax
payers remained relatively stable during the years
after Proposition 13 took effect. Sales tax revenue
per $1,000 personal income fell from $29 in FY
1978 to $26 in FY 1984, while individual income
tax revenue per $1,000 personal income rose only
from $27 to $28 during the same period. These two
taxes taken together comprised 35 percent of total
California state and local tax revenue in FY 1978,
and that proportion rose to 43 percent in FY 1979. It
rose slowly thereafter, reaching 47 percent in FY
1984.
As a group, these changes brought California’s
total state and local tax burden per $1,000 in per
sonal income down from $158 in 1978 to $121 in
Chart 2
State and Local Tax Revenues Per
$1000 Personal Income
California excluding the City and County of San Francisco
Dollars
*Gas, cigarette, alcohol and corporate income taxes accounted
fo r 60 percent of miscellaneous revenues in 1977.
61
the five subsequent years. In FY 1984, California’s
total state and local tax revenue per $1,000 personal
income was 99 percent of the national average, a
substantial decrease from 121 percent in FY 1977.
Chart 3
So urces of School District Revenues
1982 D o lla rs per P upil
School Districts
The discussion of the differences among jurisdic
tion types presented in Section I suggests that
school districts should have borne the brunt of the
impact from Proposition 13, since they previously
had relied most heavily on property taxation and
they had few alternative funding sources. However,
the impact of Proposition 13 on school districts was
greatly complicated by the Serrano decision.
Under the Serrano mechanism first implemented
in 1974, state aid filled the gap between locally
generated property tax revenues and minimum
funding levels for those school districts that needed
additional revenues. When Proposition 13 dras
tically reduced the amount of property tax money
available, the contributions from state aid increased
almost commensurately. Indeed, as Chart 3 shows,
the time paths of state aid and property tax revenues
(in real per-pupil terms) were almost perfect mirror
images of each other. In FY 1985, the state provided
63 percent of funding for schools, and the property
tax accounted for only 22 percent of school funds.
Total state funding for schools per $1,000 of
personal income has fallen since Proposition 13 was
passed, but school attendance has been falling
rapidly enough that inflation-adjusted per-pupil rev
enues actually have increased. In 1982 dollars, perpupil revenues fell from $2,841 in 1978 to $2,805 in
1979. Since then, however, they have increased
every year except one12, so that 1985 revenue per
pupil was $3,176 (in 1982 dollars)13.
Moreover, the evidence suggests that the degree
of differentiation among school districts has
declined during the period in which Serrano has
been effective. For example, when unified school
districts are ranked in order of their revenue limits,
the dollar difference among districts in the middle
50 percent was $256 in FY 1974. By FY 1983 the
difference had fallen to $129, despite a 92 percent
increase in the price level during the period14.
Because Serrano was aimed explicitly at limiting
spending differences among school districts, it
All California
Fiscal Year
likely bears more direct responsibility for this
change than does Proposition 13.
At the same time, Chart 3 suggests that, by
changing the major funding source from property
taxes to state aid15, Proposition 13 may have rein
forced the tendency toward greater uniformity
among school districts. As a result, the Tiebout
mechanism of “voting with the feet” probably is
less important in education provision than it was
before the Serrano and Proposition 13 changes took
place.
Counties
Because counties also derive a substantial propor
tion of their revenues from the property tax, one
would expect Proposition 13 to have made a sub
stantial impact on county finances. Indeed, county
revenues per $1,000 of personal income fell by 16.5
percent between FYs 1978 and 198016. However, as
discussed in Section I, counties traditionally have
had relatively little authority to differentiate them
selves from one another because an average of 75 to
85 percent of their spending is mandated by the
state.
As a result, the impact of Proposition 13 on
counties was different from that on school districts.
Overall, county funding fell whereas the same did
not occur for schools. At the same time, counties’
limited discretion before Proposition 13 suggests
62
personal income19 after Proposition 13 was imple
mented in 1979. They reached their nadir in FY
1981, but climbed thereafter; by 1985, total reve
nues stood significantly above their 1979 level.
Chart 5 breaks the change in total city revenues
into its components. Service charges are extremely
important sources of revenue. Because of the major
infrastructure requirements of services such as
water, utilities, and sewer, many cities must charge
high fees to recoup the costs of providing these
services. In addition, like counties, many cities
generate “profits” from city services.
The total revenue picture for California cities
changes, but not dramatically, when fee revenues
are excluded from the total. Chart 6 gives an indica
tion of the direction of changes, although a redefini
tion of data between 1981 and 1982 makes an
accurate appraisal difficult20. The revenue data
minus fees suggest, as do the total revenue data, that
revenue growth resumed after a sharp decline in the
late 1970s.
Table 1 sheds some light on how cities compen
sated for the declines in revenue from both property
taxes and federal grants. It lists total revenues per
$1,000 of personal income in 1978 and 1985 by
source.
Nonproperty tax instruments provide the most
obvious alternative revenue source. In addition to
Chart 4
Total City Revenues Per $1000 Personal Income
California excluding the City and County of San Francisco
Dollars
Fiscal Year
that any impact on the degree of differentiation
among counties would have been less dramatic than
the combined effects of Serrano and Proposition 13
on school district differentiation.
Cities
The discussion in Section I suggests that Proposi
tion 13 would have affected cities less than school
districts or counties because cities relied less on the
property tax for revenue. Nevertheless, cities in a
sense had the most to lose from Proposition 13
because they had been most able to differentiate
themselves from one another. Compared to other
types of jurisdictions, cities had had many revenue
sources and were not subject to as many statemandated programs.
In addition to the property tax revenues they lost
due to Proposition 13, cities lost a significant
amount of federal grant money as the federal gov
ernment rolled back its grants-in-aid programs.
Between FYs 1978 and 1985, inflation-adjusted
federal grants to cities17 fell by 50 percent18.
Whereas federal grants had provided 9.5 percent of
total city revenues in 1978, they provided only 5.9
percent in 1984.
Chart 4 illustrates the combined effects of Propo
sition 13 and reduced federal funds on city finances.
Total revenues fell substantially relative to estimated
Chart 5
City Revenues Per $1000 Personal Income
California excluding the City and County of San Francisco
'Includes Sales, Utility User, Business License and Other Taxes
"Includes Permits, Fines, Use of Money and Property
and Other Revenue
63
taxes accounted for 27 percent of total revenue for
those cities with fewer than 25,000 residents, far
more than the 10 percent that property taxes
provided during that year.
Sales taxes are a particularly visible source of
revenue for California taxpayers, but they were not
responsible for the increase in nonproperty tax revenues. Sales tax revenues accruing to cities 22 , as a
proportion of total city personal income, actually
fell 5 percent between 1978 and 1985, and their
share of total revenues remained constant during
that period.
Other revenue sources also grew substantially
relative to personal income during this period.
Although miscellaneous revenue sources provided
less than 10 percent of total revenues in 1978,
miscellaneous revenue grew by 62 percent between
1978 and 1985, and by 1985 actually provided 63
percent more money than did the property tax.
Miscellaneous revenue sources include such items
as investment and rental income, which cities began
to tap much more aggressively when their more
traditional revenue sources became scarcer.
User charges have provided another important
source of alternative revenue for California cities.
Service charge revenue as a proportion of personal
income rose 13 percent between 1978 and 1985.
Chart 6
General City Revenues Per $1000 Personal Income
California excluding the City and County of San Francisco
Dollars
Dollars
23
36
22
34
21
32
20
19
30
Passage of ..
Proposition 13
28
26
1973
1975
1977
1979
1981
Fiscal Year
1983
16
1985
sales taxes, many California cities impose taxes on
utility use, real property transfers, transient lodging
(hotels), and business licenses. These sources were
tapped liberally with the result that revenues from
nonproperty taxes rose 19.1 percent between 1978
and 1985 21 . While larger cities are in a better
position to take advantage of such alternatives,
nonproperty taxes can be important revenue sources
for smaller cities as well. In FY 1985, nonproperty
64
Some of the increase may have been associated with
increased spending on the services for which cities
levy charges, which is why this type of funding
should not in principle be included as general revenues. When. service charge funds are subtracted
from total revenues, adjusted revenues in 1985 were
13 percent lower. than their 1978 level.
Effect on Differentiation
One can get an idea as to whether the fiscal
constraints on cities affected the extent to which
they could differentiate themselves from each other
by examining variances of revenue across cities.
Calculating a consistent data set for all California
cities is a prohibitively time-consuming task, so
variances were calculated for two years using data
from two counties only. These two counties,
Alameda and Contra Costa, comprise the Oakland
Metropolitan Statistical Area (MSA), and in 1985
included 31 cities23 that varied substantially in
terms of their sizes and the incomes and other
characteristics of their residents.
Table 2 presents nonservice charge revenues per
capita (in 1984 dollars) for these cities 24 , with
means and standard deviations. The table shows
that total revenues on average fell about 10 percent
between 1978 and 1985, although there are significant differences among the cities. Real per capita
revenues rose for several cities, including Alameda,
Oakland, Walnut Creek and Pleasant Hill. However,
they fell substantially for several others, including
Berkeley, Newark, Union City, Antioch, and Pittsburg.
While changes in per capita revenues went in both
directions, the standard deviation fell slightly, from
$240 to $235. The ratio of standard deviation to the
mean rose slightly, from 0.617 to 0.671. These
small changes suggest that cities' abilities to differentiate themselves from each other did not
change significantly, and that Proposition 13 did not
interfere with the variety of tax-spending combinations available to consumers when they "vote with
their feet."
65
IV.
Summary and Conclusions
This paper has attempted to determine whether
the initiatives passed during California's "tax
revolt" had a significant impact on the level of
spending by local jurisdictions or on the variation in
spending levels among jurisdictions. The analysis
presented suggests that, on the whole, Proposition
13 appears to have reduced the size of local governments since the tax burden, as measured by taxes as
a proportion of personal income, has fallen since
Proposition 13 was passed.
Other effects of Proposition 13 have differed
considerably among the different types of jurisdictions, but for most localities, the changes brought
about by Proposition 13's limitations on property
tax revenue have been tempered by increased
reliance on other revenue instruments. For example,
school districts and counties now rely more heavily
on state aid. For school districts, the increase in
state aid was partly the result of the Serrano decision
that called for increased homogeneity among school
districts, and corresponded with an increase in state
authority over education at the expense of local
authority. For counties, increased state aid simply
reflected the previously ignored reality that California counties exist largely to carry out state mandates. The extent of heterogeneity among counties
was limited at the low end by these mandates before
Proposition 13 was implemented, and now is Iimitedat the high end as well, by revenue constraints.
Cities, in contrast, remain relatively autonomous
local governments, and they continue to derive their
revenues primarily from local sources. Proposition
13 did affect cities substantially, but its impact was
mitigated by cities' initial limited reliance on property taxes and by their ability to increase revenues
from alternative sources such as service charges and
nonproperty taxes.
FOOTNOTES
1. Most of the remaining tax revenues came from "selective" sales taxes on such items as gasoline. alcohol, and
tobacco, which accounted for about 10 percent, and
corporate income taxes, which accounted for about 5
percent. Motor vehicle licenses and other miscellaneous
taxes accounted for the remainder.
.
d h
h
h
f' I f'l
2 . 0 akl an d estimate t at t e s are 0 sing e- ami y
assessments in the total rose from 31.6 percent to 41.0
percent between FYs 1974 and 1978. Using his estimates
together with state total assessed valuations reveals that
single-family home values rose at an average annual rate
of 20.4 percent over the five-year period, while assessed
values of nonresidential property increased at an average
'annual rate of 12.7 percent during the same period.
3. While counties could charge higher taxes and spend
more money than state mandates called for, the existence
of an effective lower limit on spending meant that the
overall variation among counties would have been smaller
than for jurisdiction types with no spending floor.
4. All California counties receive revenues equal to 0.25
percent of taxable sales, but the funds must be spent for
transportation. In addition, residents of some counties
have voted additional sales taxes for transportation purposes, such as the 0.5 percent tax in the three counties
served by the Bay Area Rapid Transit District.
5. California state law prohibits local governments from
charging "excessive" fees for services rendered. In principie, this means that they can recoup the costs of service
provision but they cannot generate "profits" for use in other
areas of government. In practice, however, it is widely
acknowledged that many localities generate general revenues by selling such items as water and etectricity at prices
above true costs.
6. Counties receive all or part of the 1.00 percent that cities
do not opt for, and in addition receive the full 1.00 percent
in unincorporated areas.
7. Later the law was clarified to allow rates to exceed one
percent if the additional proceeds were used to retire
existing debt.
..
8. Thus, by the 1980s, ProposIlion 13 had created a large
gap in property tax bills between long-time owners and
new owners
.
9. Some argue that in addition local jurisdictions inflated
their spending during FY 1979 to make the Gann limit less
restrictive.
10. According to a recent study by the California Tax
Foundation, 119 jurisdictions have sought to override the
Gann spending limit at least once. Of the 60 elections held
through 1986 in which increased spending authority did
not require higher tax rates, only two resulted in a return of
funds to taxpayers rather than an increase in spending
authority. In contrast, tax-linked elections had passed 46
times and failed 33 times by early 1987.
11. Of the $21 increase, $12.50 was due to an increased
individual income tax burden and $8.54 was due to an
increased sales tax burden.
12. Per-pupil spending could increase from its 1979 level
despite the Gann limit because the school districts' own
limits include only locally generated funds. Since most of
the increase in school funding has come from the state
level, the funds are subject to the state's Gann limit, which
has not yet become binding, rather than to the school
districts'. Property tax revenues per pupil actually fell from
$704 to $694 (in 1982 dollars) between FYs 1979 and
1985.
13. The media have focussed considerable attention on
California's declining commitment to funding education.
The state's ranking in terms of per-pupil spending fell from
13 in 1974 to 26 in 1985. Correspondingly, California used
66
to spend more than the national average and now is
spending less. It is worth noting, however, that throughout
this period California has spent within five percent of the
national average, which suggests that any decline in education spending relative to that in other states has not been
particularly dramatic.
voters passed Proposition 62, which stipulates that any tax
increase is subject to approval of residents by a majority
vote. However, Proposition 62 applies only to general law
cities, and not to the 82 charter cities where most Californians reside.
22. In California, state law allows cities to receive revenues of up to 1.00 percent of taxable sales. This is part of
the statewide 6 percent sales tax. Counties receive revenues not claimed by cities and revenues generated in
unincorporated parts of the county, in addition to their own
0.25 percent transportation tax allocation.
14. These figures are cited in Osman and Gemello, 1984.
15. There still is some room for school districts to exceed
their Serrano limits. Since the Serrano case dealt only with
the inequitable distribution of real property, it placed an
upper limit on funding from property taxes only. School
districts can raise additional funds using alternative taxes,
and several school districts have passed parcel taxes that
tax all homeowners an equal dollar amount. Piedmont has
instituted a somewhat different parcel tax in which the tax
liability is based on lot size. Barring further litigation, school
districts may raise funds through taxes on anything except
the value of real property, and may spend the proceeds as
they wish.
23. In 1978 there were only 28 cities. Dublin in Alameda
County and Danville and San Ramon in Contra Costa
County were incorporated between 1978 and 1985.
24. Hercules (Contra Costa County) was excluded from
the calculations. Hercules is unusual in that its per capita
revenues in 1978 were over twice those of Emeryville, the
second highest per capita spender, and were almost nine
times the average of the other 27 cities. This was due to
sales tax revenues that accounted for over half of total
revenues. The excessive sales taxes ended in FY 1980.
Since they had nothing to do with Proposition 13 and since
including Hercules affected overall results substantially
the city was omitted from calculations to make the sample
more representative.
16. Tolal revenues fell 16.5 percent. General revenues,
which include only those that come with "no strings
attached", felt 24.4 percent. The difference between the
two is enterprise revenues, which consist of user fees. As
property tax revenues fell, counties had an incentive to
increase these fees (for water, sewer, utilities, etc.) in order
to make up for lost property tax revenues, so general
revenues fell more than did total revenues.
REFERENCES
Beebe, Jack. "Spirit of 13," Weekly Letter, Federal Reserve
Bank of San Francisco, September 28, 1979.
California Department of Finance, California Statistical
Abstract, Sacramento, 1986.
California Office of the Controller, Annual Report of Financial Transactions Concerning Cities of California, Sacramento, Fiscal Years 1973-74 through 1984-85.
California Office of the Controller, Annual Report of Financial Transactions Concerning Counties of California,
Sacramento, Fiscal Years 1973-74 through 1984-85.
California Office of the Controller, Annual Report of Financial Transactions Concerning School Districts of California, Sacramento, Fiscal Years 1973-74 through
1984-85.
California Senate Local Government Committee, "LongTerm Local Government and School Financing," in
Implementation of Proposition 13, volume II, Sacramento 1979.
California State Board of Equalization, Annual Report,
Sacramento, Fiscal Years 1977-78 and 1984-85.
California Tax Foundation, "Up to the Limit: Article XIIIB
Seven Years Later." Sacramento, CA, March 1987.
Oakland, William H. "Proposition 13: Genesis and Consequences," in George G. Kaufman and Kenneth T.
Rosen, eds., The Property Tax Revolt: The Case of
Proposition 13. Cambridge, MA: McGraw-Hili, 1981.
Osman, Jack W. and John M. Gemello. "California School
Finance: Policy Perspectives," in John J. Kirlin and
Donald R. Winkler, eds., California Policy Choices.
Sacramento Public Affairs Center and USC School of
Public Administration, 1984.
Tiebout, Charles M. "A Pure Theory of Local Expenditures," Journal of Political Economy, October 1956.
United States Department of Commerce, Bureau of the
Census, Census of Governments, Volume 4 Number 5
(Compendium of Government Finances), Washington
DC, Government Printing Office, 1967, 1972, 1977,
1982.
17. These figures exclude the City and County of San
Francisco. The joint city-county nature of Ihe government
makes comparisons with other California cities misleading.
18. Counties and school districts did not suffer as much as
cities did from reduced federal grants. For example,
federal grants comprised 27 percent of county general
funds in FY 1978 and 22 percent in FY 1985. The share of
federal grants in school funding actualty grew from 7.0 to
7.2 percent.
19. Personal income for California cities was estimated by
multiplying California personal income by Ihe proportion of
state population residing in cities for each year. This
accuralely represents city personal income if incomes do
not differ between incorporated and unincorporated
areas, and accurately represents changes in personal
incomes if income growth rates are equal in incorporated
and unincorporated areas. Since neither of these conditions is likely to hold exactly, the measure can only approximate city personal incomes. During the period in question,
between 70 and 75 percent of Californians resided in
incorporated cities.
20. Specifically, during the fiscal years 1981 and earlier,
cities reported revenues from certain functions inconsistently. For example, some reported "water" revenues
under general revenues, while others reported a special
"enterprise fund" for water services. The "general revenues" described in Chart 5 are those reported by cities so
they include some service charge revenue. Beginning in
1982, reporting conventions were standardized and all
revenues were reported either as "general" or as "functional" Under the new definitions, no service charge revenue is included as general. The data in the chart reflect
total revenues (general plus functional) minus service
charge revenues.
21. Proposition 13 did not limit local governments'
authority to increase nonproperty tax rates (Farrell v. San
Francisco, 32 Cal. 3rd 47,1982). Later, in 1986, California
67
@FedFRASER