Dymski 1988

Journal of Post Keynesian Economics
ISSN: 0160-3477 (Print) 1557-7821 (Online) Journal homepage: http://www.tandfonline.com/loi/mpke20
A Keynesian Theory of Bank Behavior
Gary A. Dymski
To cite this article: Gary A. Dymski (1988) A Keynesian Theory of Bank Behavior, Journal of Post
Keynesian Economics, 10:4, 499-526, DOI: 10.1080/01603477.1988.11489704
To link to this article: http://dx.doi.org/10.1080/01603477.1988.11489704
Published online: 04 Nov 2015.
Submit your article to this journal
View related articles
Citing articles: 9 View citing articles
Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=mpke20
Download by: [University of California Santa Barbara] Date: 04 April 2016, At: 08:13
GARY A. DYMSKI
A Keynesian theory of bank behavior

Downloaded by [University of California Santa Barbara] at 08:13 04 April 2016
Dr. Herbert Bab has suggested to me that one could regard the rate of
interest as being determined by the interplay of the terms on which the
public desires to become more or less liquid and those on which the
banking system is ready to become more or less un liquid. This is, I
think, an illuminating way of expressing the liquidity theory of the rate
of interest . .. (John Maynard Keynes, The Collected Works of John
Maynard Keynes, Vol. 14, p. 219)
I. Introduction
This paper develops a micro level model of the banking firm rooted in
Keynes's insights as a microfoundation for post Keynesian theory. This
model substantiates Keynes's claims that banks' functions of liquidity
supply and credit creation are interdependent and that the banking
system is a crucial determinant of the level of economic activity. The
essential feature of this model is the explicit incorporation of "real
time. " By contrast, in a standard bank model which ignores real time,
the bank's functions are independent, and bank behavior is not a deter-
minant of the level of economic activity.
A number of post Keynesian theorists have given considerable atten-
tion to the macroeconomic aspects of banking; however, analysis of
micro level bank behavior has been relegated to intuitive comments. A
case in point is Minsky's work on the financial fragility hypothesis
The author is Assistant Professor of Economics at the University of Southern

California.
He would like to acknowledge the insightful comments of John Elliott, Hyman
Minsky, Chris Niggle, John Veitch, Michael Woodford, and an anonymous referee
of this journal on preliminary drafts of this paper. Remaining errors are his own.
Journal of Post Keynesian Economics/Summer 1988, Vol. X, No.4 499

500 JOURNAL OF POST KEYNESIAN ECONOMICS
(FFH).l Davidson is the outstanding exception; he considers how fi-

nancial intermediation emerges in an environment of real time and
irreducible uncertainty. But Davidson's work is foundational, not
microfoundational: it develops a non-Walrasian treatment of time and
ignorance, but suggests no specific behavioral model. So post Keynes-
ian theory contains ideas about banking (Keynes and Minsky) without a
model of the bank, and a theory ofmoney (Davidson) without a model of

monetary institutions. 2
This paper first (section 2) reviews Keynes's ideas on banking.
Keynes wrote that the banking system is a crucial determinant of
cyclical fluctuations, but discussed bank behavior only in suggestive
terms. He described banking as a "dual" activity, involving credit
creation and liquidity provision, and showed that these functions can
become uncoordinated-that is, they are interdependent.
Sections 3 and 4 develop a model of banking in two stages. Section 3
begins by showing how deviations from the Arrow-Debreu (A-D)
assumption set generate a role for both banking functions. Banks'
liquidity supply function emerges when some actions have stochastic
outcomes; their credit creation function results from private informa-
tion. "Dual" banks perform both functions.
The behavior of dual banks then depends critically on the treatment
of time. Section 3 develops a model of dual banking firms in a
"timeless" environment, wherein all activities are of equal duration,
with identical starting points: so activities are synchronized. A stan-
dard Klein/Monti model is used to depict individual bank behavior; the
1A second example of the lack of an adequate microfoundation in post Keynesian

monetary theory is the literature on endogenous money. For example, Moore's pa-
per (1983) on bank lending provides no behavioral account of bank behavior, which
is simply a backdrop for his discussion of endogenous money supply. Rousseas
(1985) offers a markup theory of bank loan behavior as a microfoundation for mod-
els of endogenous money. However, his model ignores the fact that the bank is a
monetary institution; no real-time/uncertainty macrofoundations inform the behav-
ior of his model. Pal ley (1987-88) uses Moore and Rousseas as a microfoundation
for a financial markets model; while not fully developed, his bank model improves
on Moore and Rousseas in that it incorporates bank expectations about future
borrowing-market conditions. However, this feature is discarded when Palley sets
up his [mancial-market equations, in which interest rates are determined by a set of
one-period structural market equations.
2Here money is defined as any completely liquid asset, and banking as the provision
of liquidity and the creation of credit; banks are firms that carry out banking func-
tions. Davidson defines liquidity as "hav[ing] the means of settlement to meet all
one's contractual obligations when they come due" (1978b, p. 61). A completely
liquid asset, then, is one defined as the means of settlement.
KEYNESIAN THEORY OF BANK BEHAVIOR 501
aggregate behavior of the banking system is depicted via an interbank

market for excess reserves. The key feature of this model is that assets
and liabilities are of matched duration; as a result, banks achieve ex
ante equilibrium in their loan and borrowing markets in every decision
period. It is shown that dual banks' two functions cannot become
uncoordinated in a timeless environment and that a banking system of
Klein/Monti banks adjusts passively to changes in the level of economic
activity. Both results are contrary to Keynes's depiction of banking.

Section 4 then inserts "real time" considerations-such that eco-
nomic activities are of differential lengths and hence unsynchronized-
into the framework of section 3. Specifically, the Klein/Monti model is
modified by assuming that loans are multiperiod, while all other bank
instruments are one period in duration. With this modification, dual
banks are susceptible to sustained disequilibria, and the adjustment
behavior of dual banks will be among the causes of upturns and
downturns. This suggests, then, that the key elements of a Keynesian
model of bank behavior are dual functions and real time.
Section 5 reconsiders the role of banking in the FFH. In expo siting
the FFH, Minsky defines a unitary banking function (credit creation)
and assigns banks no distinct role. But incorporating the dual model
into the FFH resolves two problems unexplained by Minsky: (1) what
initiates downturns (moments of "financial crisis"), and (2) how
adverse shocks are disseminated. Specifically, dual banks propagate
the effects of credit crunches or of real-sector defaults throughout the
system, explaining why financial commitments are "fragile."
ll. Keynes's conception of banking

Both before and after penning The General Theory, Keynes character-
ized banking as a fundamental determinant of the level of economic
activity. In the Treatise on Money (1930), he wrote: "By the scale and
the terms on which it is prepared to grant loans, the banking system is
in a position ... to determine-broadly speaking-the rate of invest-
ment by the business world" (1930, Vol. 1, p. 138; and see pp. 163-
165). He then wrote in 1937: "In general, banks hold the key position
in the transition from a lower to a higher level of activity. If they refuse
to relax, the growing congestion of the short-term loan market or of the
new issue market, as the case may be, will inhibit the improvement"
(1973, p. 222). Thus, the "power of the banks" stems from "their
control over the supply ofmoney-Le., of liquidity" (1973, p. 211).
So bank credit allocation is at the heart of cyclical fluctuations (1930,

Vol. 1, pp. 250-262).
Despite this reiterated emphasis on banking per se, Keynes wrote
little on the behavioral aspects of banking. He characterized banking as
a dual activity:
... the modern banker performs two distinct sets of services. He sup-
plies a substitute for State money by acting as a clearing house . . .
But he is also acting as a middleman in respect of a particular type of
lending, receiving deposits from the public which he employs in purchas-
ing securities, or in making loans to industry and trade .... (1930, Vol. 2,
p. 191)
Banks face a "dilemma" in combining these two functions, for "the

complete attainment of one of its duties is sometimes incompatible with
the complete attainment of the other" (1930, Vol. 2, p. 193). Keynes
elaborated further on the tension between banking functions in an
August 1931 essay, suggesting that it is central to banks' role as a
determinant of the level of economic activity:3
. . . consider what happens when the downward change in the money

value of assets within a brief period of time exceeds the amount of the
conventional "margin" over a large part of the assets against which
money has been borrowed ... [T]he banks, being aware that many of
their advances are in fact "frozen" and involve a larger latent risk than
they would voluntarily carry, become particularly anxious that the remain-
der of their assets should be as liquid and as free from risk as it is possible
to make them. This . . . means that the banks are less willing than they
would normally be to finance any project which may involve a lock-up of
their resources. (1963, pp. 171-173)
Banking figured only marginally in The General Theory. Keynes did

write that banks making loan commitments necessarily take on
"lender's risk": "This may be due either to moral hazard, that is,
voluritary default or other means of escape, possibly lawful, from the
fulfillment of the obligation, or to the possible insufficiency of the
3This passage refers specifically to the impact on banks of the "loss of money
value"-that is, of deflation. However, this passage can be generalized readily to
encompass a situation in which bank assets lose their value. Note that Keynes's
writing on banks is treated herein as an integrated whole.
margin of security, that is, involuntary default due to the disappoint-

ment of expectation" (1936, pp. 144-145). He went on to comment:
"During a boom the popular estimation of the magnitude of ... both
borrower's risk and lender's risk is apt to become unusually and impru-
dently low" (pp. 144-145).
m. A model of banking in a timeless environment

Our starting point in developing a Keynesian banking model is the A-D
equilibrium, wherein banking functions are redundant. It is shown that
less restrictive treatment of certainty and knowledge creates an analyt-
ical role for the two banking functions Keynes identified-respectively,
liquidity supply and credit creation. We develop a simple banking
model that performs both functions, but find that it does not embody
Keynes's ideas: the two functions cannot become uncoordinated, and
the behavior of this banking sector will not trigger turning points in the
level of economic activity. The reason for these disappointing results is
that the model of this section leaves undisturbed the A-D treatment of
economic activities as "timeless" or synchronous in the sense intro-
duced above-they are of equal length, with coordinated starting
points. 4
Banking functions in economic analysis
The assumptions underlying the benchmark model in contemporary

economic analysis, the A-D model, rule out the nonredundant existence
of both money and banking. The A-D model assumes, inter alia, com-
plete spot and forward markets, no uncertainty, no transactions costs,
and complete and costless information. Agents use full information to
costlessly choose any feasible activities over their planning horizon
without fear of disappointment. With costless transactions and pre-
coordination, money need not be held-it is merely a numeraire. No
events can disrupt agents' plans, so liquid reserves-and hence banks-
are unnecessary.
Money and banking can be introduced by relaxing assumptions made
in the A-D construct about knowledge and about the determinateness of
4The role of money is necessarily incorporated into this analysis of the role of bank-
ing in economic analysis. This follows naturally from the fact that banks are funda-
mentally monetary institutions.
action. Relaxing each assumption in turn generates a distinct banking

function. 5
Suppose first that all information is public, but that the A-D assump-
tion of determinate action is replaced by the assumption that some or all
actions are stochastic. Agents cannot know in advance whether their
plans will be realized. This deviation from A-D assumptions is suffi-
cient to generate a role for money and banking. With stochastic out-
comes, mismatches may arise between available resources and planned
expenditures at any point in time. This creates a precautionary demand
for liquid reserves-an analytical role for money. A role for banks
arises if it is assumed (as in Gurley and Shaw, 1960; Parkin, 1970; and
others) that banks arise because economies of scale are to be had in
liquidity supply. The law of large numbers suggests that these econo-
mies result when the bank attracts a sufficiently large number of de-
SIt is important to note that a nonredundant role for money and banking can also be
generated by substituting for the A-D assumption of costless transactions the idea
that the economy's transactions mechanism is either costly or flawed.
Benston and Smith (1976) introduced tlle costly-transactions approach to banking
in a partial eqUilibrium context. More recently, in an approach characterized as the
"New Monetary Economics" (Cowen and Kroszner, 1987), costly transactions-
alternatively, decentralized trading (Sargent, 1987)-have been introduced into an
A-D or EM (efficient-markets) construct (see note 6). This leads to a "unitary"
banking model: banks' role is to provide payments services (Fama, 1980).
Townsend explains the existence of banks (1983) by postulating that barter equilib-
ria are suboptimal when transaction costs increase monotonically with geographical
distance between agents. We ignore this approach here because its view of banking
is completely at odds with Keynes's. For example, Fama has written that "banks re-
main passive intermediaries, with no control over· any of the details of a general
equilibrium .... the real activity that takes place, the way it is financed, and the
prices of securities and goods are not controlled either by individual banks or by the
banking sector" (p. 48).
The idea of a flawed transaction mechanism underlies much recent work on
"overlapping generations" (OG). OG models generate a role for money and bank-
ing by assuming that multi-period-lived agents cannot achieve Pareto efficient equi-
libria autarchically because of incomplete markets. This is equivalent to assuming
that the economy's transactions mechanism is flawed. Given market incompleteness,
a trusted "social contrivance" is needed to effect optimal transfers of wealth. A fiat
money in the form of accounts at an eternal "bank" both provides this contrivance
and guarantees agents' trust; so the bank's existence assures a socially optimal
equilibrium.
We ignore the explicitly dynamic OG conception of banking because including it
here would force us to consider the relationship between explicitly dynamic and
"real-time" analysis. This goes well beyond the goals of this paper. Further, the is-
sues raised are not simple. For example, Michael Woodford has pointed out in con-
versation that a temporary equilibrium approach which is explicitly dynamic while
rejecting the possibility of steady-state equilibrium could simulate "real time." The
problem of explicitly dynamic monetary frameworks deserves separate treatment.
positors whose stochastic outcomes are uncorrelated; in effect, banks

are risk poolers.
Suppose, alternatively, that some agents have private information.
The immediate consequence of private information is that contracts
based on that information cannot be written at arms' length between
contracting parties, even if such contracts would be welfare-improv-
ing. In banking models, private information about credit-worthiness is
assumed to exist for a subset of the agents seeking borrowed funds.

Direct borrowing/lending contracts are inefficient because of either:
(1) adverse selection, wherein lenders do not know with certainty the
parameters of the probability distribution from which their borrowers
are drawn (Akerlof, 1970); or (2) moral hazard, wherein the sum
risked by the lender on a project with uncertain return exceeds that
risked by the borrower (Stiglitz and Weiss, 1981). Intermediaries then
arise as information specialists performing the costly monitoring such
contracts necessarily entail; this role is feasible either because of econo-
mies of scale in monitoring (Diamond, 1984), or because loan contracts
entail savings on the indirect cost of transmitting information through
market signals (Leland and Pyle, 1977). Money can be understood, in
models of this type, as the liquidity created by banks to transfer pur-
chasing power to bank-funded borrowers. 6
So the first deviation from A-D assumptions establishes a role for
liquid reserves; the second implicates bank credit creation in the level
of income, since bank decisions to extend credit cannot be replicated by
nonbanks making credit commitments at arms' length. Combining
these features suggests a dual-function bank (as in Klein, 1970) which
supplies liquidity and creates credit on the basis of stochastic outcomes
and private information, respectively. We now construct a simple mod-
el of a banking sector containing dual banks, based on the well-known
model of Klein (1971) and Monti (1972).
A dual bank in a timeless environment
Consider a banking firm whose asset position of small loans, L, securi-

ties, S, and reserves, H, is fmanced by three liability instruments-
deposits, D, borrowings, B, and equity, K. The bank must maintain an
H/D ratio at least equal to h, where 0 < h < < 1; since H pays no
6The bank models cited here, while representative, are drawn from a vast literature.
Baltensperger (1980) and Santomero (1984) provide extensive reviews of contempo-
rary models of the banking firm.
explicit interest, a minimal amount of it, hD, is held; excess reserves

are lent out in the borrowing market. Borrowings are defined as
interbank loans of excess reserves; they may be either positive or
negative. 7 All instruments are denominated in normalized units, so the
bank's balance sheet equals:
(1) L +S+ B = D(1 - h) + K.

where all instruments except B are constrained to be non-negative.
Making loans and servicing deposits is costly; so there is a cost
function, C = C(D, L), where CD > 0, CL > 0, CDD > 0,
CLL > 0, and CDL > 0.
The per-period rate of return on any instrument I is i. Securities and
borrowings are sold in perfectly competitive markets at s and b, respec-
tively. Bank managers must plan their position while some variables
remain stochastic. Both s and b are stochastic in the planning period
because conditions in the Sand B markets cannot be perfectly anticipat-
ed. Expected values, se and be, are generated by a process of expecta-
tion formation not theorized here (see note 17).
The bank faces a loan demand curve, L = LUC), where lC is the
°
contractual loan rate and where it is assumed that L I < and L I I > 0.
This curve is composed of borrowers unable to access credit markets
directly because of the existence of private information and assumed
for simplicity to be of uniform "type." The imperfectly competitive
nature of loan demand follows directly from the assumption of private
information (Stiglitz and Weiss, 1981; Diamond, 1984). And while the
bank makes contracts at lC, i-the realized return on loans-is
stochastic because of default risk, OB, where OB > 0. It is immediate
that:
(2)
The deposit market is imperfectly competitive due to geographic

restrictions on bank activity. The bank faces a deposit demand curve,
7Thls simplified treatment of the bank's short-term borrowing markets is severe, but
not without justification. Bank borrowing at the discount window is quantitatively
unimportant-in a typical week, discount window borrowings average less than 3
percent of interbank borrowings. Examination of interest rate data on two other
short-term borrowing markets frequently accessed by banks-the market for nego-
tiable certificates of deposit and for Eurodeposits-reveals that the rates in these
markets have tracked the interbank rate very closely since June 1970 (when the reg-
ulatory rate maximum on the former instrument was removed). See Dymski (1988).
D = D(d), whereD ' > OandD" < O. The 10cationofD(d) is known

with certainty by the ballie deposit holders know their D drawdown at
the same moment that other commitments of uncertain return are being
made. This assumption expresses most graphically the strength of the
assumption that all relevant activity is synchronous.
The bank's equity owners earn no explicit return on their invest-
ment; in effect, k is fixed at zero for planning purposes. Instead, they
claim the bank's residual income after obligations have been met.
The bank's expected profit flow in any period is then given by:
Equation (3) can be used to impute the expected return on K:
(4)
We can now consider the bank manager's decision problem. The

conventional assumption that the bank is a risk-neutral profit maximiz-
er (as in Klein, 1971, and Monti, 1972; and see Santomero, 1984) is
used here for simplicity. The bank maximizes equation (3) over r, d, 5,
and B, subject to equation (1). If the bank's equity owners have a
required rate of return, kthen equation (4) must also be satisfied (with k
substituted for kj. The bank's maximization problem can be simplified
by using equation (1) to eliminate 5 from equation (3). This gives:
(3 ') 'Ire = (Ie - oe - s")LW) + (be - se)B

- (d - se(1 - h))D(d) + seK - C(D, L).
The first-order conditions for ZC, d, and Bare, respectively:8
(5)
(6)
and
(7)
where €I = (oJ/oz)(i/l).
8Derivatives of functions with single arguments are indicated by primes ('). Deriva-
tives of functions with multiple arguments are indicated by using subscripts for
functional operators.
The systemic aspect of bank behavior is depicted via a simple repre-

sentation of the B market. Suppose there are N banks, indexed by
i = 1, ... , N. Each bank's demand for or supply of reserves equals
B i. Then the B market equilibrates when
N N
(8) E Bi = E (Di(1 - h) + Ki - Li - Si) = 0,
i=l i=l
given that (5)-(7) are satisfied for all banks. Also, we assume through-
out this analysis that net planned borrowing demand in the B market
does not exceed net planned S holdings. So the following consistency
condition holds:
N N
(9) E (Bi + Si) =E (Di(1 + h) + Ki - L i) ~ O.
i=l i=l
Equilibrium is achieved when there exists a vector (d*, IC*, B*, be *)
which satisfies condition (8) and which for all N banks satisfies equa-
tions (5)-(7) and generates expected residuals high enough that equa-
tion (4) does not bind. 9
An important feature of this equilibrium is the relationship of the B
and S markets. 10 By equation (7), be and se are equal in equilibrium.
This relation holds because B and S are structurally indistinguishable:
both are costless assets sold in competitive markets. Tnen (7) and
condition (8) jointly imply th.at the bank can effectively borrow and
lend at S8. This interbank market functions like the domestic capital
market in the open-economy Mundell-Fleming model. That is, sup-
pose that initially there is excess demand for interbank borrowings;
then b rises infinitesimally, causing ban.1(s with S positions to sell those
91n the case of the market for bank credit, equation (5), equilibrium does not imply
market clearing. Several authors (for example, Stiglitz and Weiss, 1981) have ob-
served that credit rationing, apparently a disequilibrium event, can be understood in
equilibrium terms. Note that second-order conditions are guaranteed by assumptions
made about the slopes of the L, D, and C functions.
IORelaxing condition (9) would open up a richer set of possibilities. For example, if
banks' aggregated initial plans resulted in the violation of (9), perhaps due to un-
usually strong credit demand, then the interbank rate would rise to accommodate the
borrowing pressure in the B market. This would drive a wedge between the securi-
ties and interbank rates, forcing all banks to cancel their S positions and dropping
equation (7) from the system. A model which explicitly incorporates this possibility
is examined in Dymski (1988). Condition (9) is not relaxed in this analysis to focus
attention on the role of factors related to "real-time" considerations. All N banks
are still, however, able to achieve ex ante equilibrium; only introduction of the as-
sumptions special to section 4 creates the possibility of ex ante bank diseqUilibrium.
off and shift into the B market. Given condition (9), the interbank
market always generates whatever volume of funds will satisfy (7) in
this way. So the thinness of the B market, due to restricted participa-
tion, is irrelevant; the B market functions as an extension of a deep,
broad S market.
A well-known result for this model is that the bank's asset-composi-
tion and portfolio-size decisions are independent. 11 This is readily
seen. The bank makes a decision on its (D) portfolio size using equation
(6): it sets d just equal to its expected return on S, net of the marginal
cost of D. The bank's L-market decision is based on equation (7): the
loan-contract rate is set where the net marginal return on L just equals
SB. SO independence here means that the size of the bank's loan portfo-
lio is not limited by its D volume: the bank can borrow in the B market,
at the expected S-market rate, whatever funds beyond its D base are
required for loan-market eqUilibrium.
This result leads to three further results contrary to Keynes's depic-
tion of banks and the banking sector. First, no tension arises between
banking functions in these timeless models. The bank discharges its
liquidity-supply function through its d decision, and its credit-creation
function through its Ie decision. Independence rules out tension, be-
cause the bank's functions are independently satisfied. Second, the
bank can achieve an equilibrium for each function based solely on
presently prevailing conditions. This is true no matter what previous ex
post experience has been; the essential aspect of timelessness is the
absence of lingering effects or commitments from one period to the
next. 12
Third, this banking sector is not a determinant of economic activ-
ity-it simply adjusts passively to structural conditions originating else-
where. Equations (5)-(7) clearly show this. Conditions in the financial
markets-that is, sB-determine the size of the banking sector; and
llSee Klein (1971) and Baltensperger (1980). Langohr (1982) shows that the inde-
pendence result is due to the assumption of imperfectly competitive L markets.
l20ne period's experience may, of course, affect bank behavior in the next. For ex-
ample, the bank may respond to bad draws in one period by increasing its expected
default rate, or even by changing its degree of risk aversion.
Also note that the bank's ability to achieve equilibrium in each planning period
is no guarantee of its ability to survive or discharge its obligations. Its ability to
meet condition (4)-with kB replaced by .t-and satisfy its equity owners, to pay its
deposit holders, and even to satisfy liquidity demand fully depends on the realiza-
tions of borrowers on funded projects. With sufficiently bad draws, the bank may
even fail to survive.
credit volume is determined by the position of the loan demand curve,

which is presumably given by "real" factors, and by se (the opportuni-
ty cost of making loans). Banks mechanically translate signals about
financial-market and real-sector conditions into a decision about how to
divide assets among securities, reserves, and 10ans. 13
In sum, Keynes's ideas about the banking system are not embodied
in this model. There is no interaction between the liquidity-supply and

credit-creation functions. In performing these functions, banks mirror
whatever conditions exist in nonbanking sectors; they have no separate
effects on the rate of income flow. 14 What is missing here is an adequate
treatment of time.
IV. A model of banking in a "real-time" environment

Simply introducing banking functions into the A-D assumption set is
then not sufficient to construct a Keynesian banking model. To obtain
Keynes's results, "real time" must replace the timelessness of the A-D
framework.
Numerous authors have used Keynes's rejection of calculable risk in
favor of fundamental uncertainty (in The General Theory and else-
where) as the entry point for a non-Walrasian conception of time. In
particular, Davidson has emphasized that uncertainty and irreversible
time explain the existence of money, "sticky" wages, markets, and
financial intermediaries. These institutions are not second-order con-
structs supplementary to a prior concept of equilibrium; rather, their
very existence indicates that such an equilibrium is irrelevant as a
model of an observed economy. Money is a "time machine" for
transmitting purchasing power; other durable goods or assets are
unsafe because of illiquidity risk. Financial institutions "make an
13This conclusion would be modified if the assumption of a constant level of bank
risk preference were weakened.
14Banks' purely passive or reflective role here is emphasized when the parallel be-
tween this construct and the model of efficient markets (EM) which underlies mod-
ern finance theory is considered. The EM model (see Fama, 1976) modifies the A-D
eqUilibrium by assuming that: (1) information is incomplete (stochastic), though it re-
mains costless and freely available (so that expectations are homogeneous); and (2)
agents can borrow and lend at a risk-free rate. In practice, the EM and A-D con-
structs are virtually identical, since equilibrium can be shown to exist in a stochastic
Walrasian framework essentially equivalent to the efficient-markets assumption set
(Radner, 1968). What is interesting for our purpose is that the risk-free borrowing/
lending market in the EM construct discharges both banking functions. The risk-
free asset is effectively money; agents' unrestricted access to the market for this as-
set obviates the need for a banking firm. The similarity of the model of this section
to the EM model should alert us that its conception of banking is shallow.
orderly spot market and operate as a residual buyer and seller" of

"placements" - ' 'titles to capital goods and debt contracts" (David-
son, 1978a, pp. 63,62). Banks, in turn, expand the pool of funds in the
financial system by selling liabilities which are perfect substitutes for
money to nonbank units. IS
We have already introduced uncertainty into the model of section 3.
We now augment that model by simulating irreversible or real time; in
particular, we posit that economic events of different durations coexist.

Complex actions, such as production and investment, take "time" to
complete; others, such as decisions to convert liquid assets (deposits)
into money, are effectively instantaneous. So banks create and supply
liquidity for agents carrying out both instantaneous and time-using
activities. "Real time" here means that agents, because they are in-
volved in nonsynchronous activities, cannot pre-coordinate their posi-
tions at each moment in time. Our focus is on banks, which are obliged
to supply liquidity on demand while making new loans and funding loan
contracts of lengthy duration.
The bank's roster of instruments remains as in section 3. Three
changes are made. First, loan contracts may extend over multiple
periods; thus, at any point in time the bank carries a fixed volume of
vintage loans. Second, we simulate the essential property of deposits-
their instantaneous convertibility into reserves, H-by assuming de-
posit volume is stochastic when loans are made. Third, a two-period
framework replaces the one-period construct of section 3. The bank
now operates in two periods-a loan period and an adjustment period. 16
lSDavidson goes on to depict banks' effect on economic activity in a manner that

echoes Keynes's insights. He writes that, while financial intermediation is expan-
sionary in making" 'bear' hoards" available for investment fmancing, it has a
cyclical aspect. Financial firms' willingness to make (or finance) placement markets
varies: "at times of general insecurity, these institutions may magnify the rush to li-
quidity and thereby accentuate slumps" (1978a, p. 317).
16The real-time model developed here is similar to the models in Deshmukh,
Greenbaum, and Kanatas (1983), Tobin (1982), and Sprenkle (1987). Deshmukh,
Greenbaum, and Kanatas (1983) relax the assumptions of both costless information
and uniform agent decision horizons. Borrowers in loan markets need funds for a
longer duration than other security-market participants, so banks making loan con-
tracts transform maturity and take on rate risk of the type modeled here. Tobin
(1982) and Sprenkle (1987) add to this assumption set the idea that bank liability
volume in the second period is stochastic because deposit rates are fixed below mar-
ket levels. The model developed in this section differs from these earlier efforts in
four ways: (1) it considers the impact of non-zero loan default probabilities on bank
behavior; (2) it explicitly takes account of the adding-up constraint on the B market;
(3) it explicitly considers the problem of disequilibrium adjustment and second-best
equilibria; and (4) it adds vintage loans to the model.
In the loan period, the bank decides on the volume of new loans it
will add to its stock of vintage loans. Deposit volume remains
stochastic until the adjustment period: so realized D, B, and S volumes
may diverge from those expected. Security and borrowing volumes can
be varied without penalty in both periods; for this reason, the differ-
ence (S - B) is termed (after Tobin, 1982) the bank's "net defensive
position" (NDP). Interest rates on S and B remain stochastic until the

adjustment period.
The bank enters the loan period with vintage loans made inj previous
periods, wherej is indexed by j = t - 1, ... ,t - J. For simplicity,
all loans made in period j, LV, are assumed to have: an identical
contract rate, the discounted present value of which equals Ij; and an
identical default probability given by oj. Thus, the expected return on
period j loans, r;
= Ij - oj, in any period depends on the current
valuation of oj; we ignore variations in the discount rate applied against
future earnings. The bank faces an imperfectly competitive loan-de-
mand curve for new loans in the loan period L = L(le), with the same
properties as above. The per loan rate of default on new loans remains
oe. Some vintage and current loans terminate in the adjustment period.
We assume that some of the loan volume contracted in the loan period
consists of informal understandings (perhaps based on traditional loan-
customer relationships).
In the loan period, bank managers bound their ignorance about
deposit volume by developing a subjective probability distribution and
a point estimate, 15, for D. Here 15 represents the bank's estimate of the
value of deposit demand in the adjustment period for any deposit rate d.
Then D varies directly with d and with the random variable x, which
represents deposit variability:
(10) D(d, x) = 15(d) (1 + x).

A probability density function fix) is defined over [ -1, 1], with x = 0
elsewhere, to exclude negative deposit outcomes; so j _~ f(x) dx = 1.
This distribution is symmetric around zero. The function F(x) is the
cumulative density function (c.d.f.) for X.17
17Sections 4 and 5 use the probability calculus to depict the way in which agents form
understandings of the likely values of stochastic events. We note, however, that some
scholars who have reflected on the problem of Keynesian "real time" have suggested
replacing the apparatus of probability theory with an alternative conception of poten-
tial surprise (see, for example, Shackle, 1967). While probability and potential sur-
prise are, in a sense, competing algebras for handling stochastic events, Katzner
(1986) has recently shown that they are fundamentally different in conception and can
lead to different results. Nonetheless, the probability calculus is used here because of
its familiarity.
In section 3, B was a choice variable. Now, the level of B is jointly

determined by the loan drawdown and the stochastic process governing
D. The bank's d and ZC decision determines not B, but the value of x
below which the bank will have to borrow in the adjustment period to
fund its loan portfolio. To see this, consider the bank's adjustment-
period balance sheet, which equals:
(11) LV + L + S + B = D(1 - h) + K,
where
t-J
LV = 1: Lj.
j=t-l
Then set NDP, S - B, equal to zero; substitute for D using equation

(9); and solve for x. This gives that level of x, X, at which the bank's
exogenous liabilities will exactly cover its vintage and loan-period loan
volumes:
(12) x= (LV + L - K)/[D(1 - h)] - l.
If L < D(l - h) + K, so that the realized x exceeds X, the bank will

have excess reserves in the adjustment period, which can be used to
take a position in S or B. But if an x below x is drawn, the bank will have
to borrow in the B market. For simplicity, borrowings are used only to
fund L, not S. We can readily hypothesize that banks vary dramatically
in their approach to L and x: money-center banks that manage their
liabilities characteristically choose values of x > 0, whereas country
banks choose values of x < O.
Bank behavior in the loan period
We now consider the bank's problem. We retain the assumption of risk

neutrality. The bank has decision points in both the loan and adjustment
periods. In the former, the bank maximizes expected profits by setting
r based on [) and expected rates of return; in the latter, it maximizes
actual profits by choosing Sand B based on realized D. The bank's
loan-period LaGrangian function is of the form:
(13)
{2 = 7re + Al(L + se + K - D(l - h) - K) + A2(k - (7re /K»
where the form of 7I"e depends on the bank's choice of i and on which x
is realized.
To see this, we define two distinct profit equations, 71"1 and 71"2, as
follows:
t-J
(l4a) 71"1 = 1: W - oi)D
j=t-1 J J 'J
+ CZC - oe)LCZC) + beB + seS
- dD(d) - C(D, L)
if x > i
t-J
(l4b) 71"2 = 1: (/'j - oj)Lj + CZC - oe)LCZC) + beB
j=t-1
- dD(d) - C(D, L)
if x < i.
Note that B > 0 in (14a) and B < 0 in (14b). We then take expecta-
tions of (14a) and (l4b); it follows immediately that 7I"e = 7I"f + 7I"i.
The Appendix shows that the bank's loan-period maximand can be
written more succinctly as follows:
(15) 7I"e = ~ [I'j - oj - se - F(i)(b e - se)]Lj

J
+ W- oe - se - F(i)(b e - se)]L(ZC)
- [d - se(1 - h) - F(}()(b e - S)(1 - h)]D(d)
+ [se + F(i)(b e + se)]K - C(D, L)
where F(i) is the probability that the bank has to borrow in the adjust-
ment period to fund its fixed position. The bank maximizes equation
(15) over the decision variables ZC and d. The first-order conditions for
Ie and d are, respectively:
(17) d(l + Er/) = se(1 - h) + CD

+ (be - se){F(i) - [if(i)(LV +L - K)]/D}.
A loan-period equilibrium is established for a vector (ZC*, d*, if,

i = 1,. . . ,N) which satisfies equations (16) and (17) and which holds
for N equations defining an i for each bank via equation (12). Ex ante
consistency also requires that an expectational form of equation (9) be
satisfied by the equilibrium vector:
(9')
E (B~ + S~) = E (Di(1 - h) + Ki - Li - L i) ~ o.
i i
Bank behavior in the adjustment period
Two events of consequence occur for each bank in the adjustment

period: (1) a specific deposit volume, D i , is realized; and (2) updated

information on default rates for current and vintage loans is obtained.
The behavior of banks and of the B market depends on the aggregate
effect of deposit realizations. To see this, note that replacing 15 with D
gives each bank a determinate (plus or minus) posture vis-a-vis short-
term securities. The aggregate security-market position of the banking
sector is given by:
(18) E (Bi + Si) = E (D i(1 - h) + Ki - Li - Li).

i i
Equation (18) is identical to the consistency condition (9), with one

notable difference: there is no longer a requirement that the sum of
security demands be positive across all banks. If (18) remains positive,
then the B market will function as in section 3, and we can expect that
s "" b. We cannot rule out, however, the possibility that (18) is nega-
tive-that there are not sufficient excess reserves available to the bank-
ing system to allow banks with negative net defensive position to fund
their loan positions. 18
If the latter event occurs, the excess demand for borrowings will put
upward pressure on be. The upward movement of be, in turn, decreases
the opportunity cost of unwinding loan-period L contracts, especially
those made informally and/or those assigned a high default probability.
The cost of unwinding any loan contract equals c, a "bad faith" cost
figured as a proportion of the loan's value, plus ZC - 5e ; then a precon-
dition for loan contract unwinding is that Ie + c < b + 5e. The B
market equilibrates (at some b > s) when enough banks unwind
current-period loan contracts so that equation (8) is satisfied.
In considering loan-contract unwinding as an adjustment-period be-
havior, one must keep in mind that banks have obligations not only to L-
18The assumption of a lack of sufficient reserves to fund banks' illiquid positions is

generated, to some degree, by the artificial assumption that banks are dependent for
liquidity on excess reserves within the banking system. But even if banks can buy
liquid funds from nonbank agents, the latter may be unwilling to risk their funds
through short-term loans to banks except at onerous rates like those this model fore-
sees in the B market (see note 7).
market customers, but to equity owners as well. Loan-contract unwind-

ing may be rational behavior ifit increases expected profits in the adjust-
ment period and enables a bank to satisfy the condition that ke ~ k.
Discussion
This model has very different properties from the timeless model of
section 3. First, independence of asset-portfolio and portfolio-size

decisions is lost. Equations (16) and (17) clearly show that bank behav-
ior in the Land D markets takes into account both deposit and loan
volumes. The independence result obtains only if be = se; but, as
noted, the equivalence of band s is unlikely ifbanks get a systematically
bad deposit draw.
Not only are the bank's functions interdependent: they may conflict.
The more credit banks create to satisfy loan demand, the fewer funds
are available for redistribution to meet depositor demands for liquidity.
Conversely, reducing D and H volume to supply liquidity to depositors
undercuts the banking system's capacity to satisfy all loan contracts.
Banks may face a choice in the adjustment period between meeting all
explicit and implicit loan commitments and preserving flexibility by
maximizing asset liquidity in an adverse environment.
A second distinctive feature of this model is that optimal equilibria
may not be achieved. In the loan period, the bank's equilibrium may be
suboptimal because of the presence of vintage loans. The bank chooses
optimal values of ZC and d in the L and D markets which take into
account the stock of LV: note that the terms x and F(x) vary directly with
LV. If the term [F(x) + x/ex)] > 0 in equation (16), then Ie would be
lower in the absence of vintage loans; the same holds for d if the term
[F(x) - x/(x)(C + L - K)/D] > 0 in (17). Thus, loan and deposit
markets equilibrate at lower volumes and higher rates than they other-
wise would. With a sufficiently large LV, it is even conceivable that the
bank will make no loans in the loan period. Vintage loans may also be a
drag on bank profits. 19
19This follows, of course, when the expected return on vintage loans is less than that
on current-period loans. This immediately suggests that banks would be better off
securitizing such vintage loans. It must be kept in mind, however, that the bank can
only sell off multiperiod loans with rates below market by realizing capital losses.
This may result in unacceptable reductions in bank earnings, given the net-return re-
quirements of bank equity owners. So at any point in time a bank may be carrying a
volume of vintage loans which is effectively illiquid because selling it off would
force the realization of unacceptable capital losses. Since the need to finance this vin-
tage stock subtracts from the bank's ability to respond to current-period loan-demand,
this is another example of how functional independence is violated in this model.
Further, the suboptimal equilibrium of the loan period is likely to

become a disequilibrium state in the adjustment period. Indeed, banks
are normally in disequilibrium in the adjustment period; it is a virtual
certainty, barring perfect foresight, that neither s, b, nor x will equal
their expected values; hence the bank's ZC and d rates do not remain
equilibrium rates in the adjustment period. The degree of bank disequi-
librium will deepen if, as noted, the banking system has to make
adjustments spurred by liquidity shortages, triggered by the fact that
relation (18) is less than zero.
A third feature of this bank model is that its behavior is an indepen-
dent determinant of the level of economic activity. This follows, first,
because bank loan rate and volume decisions in the loan period are
based partly on assessments of the availability of banking-system li-
quidity in the adjustment period; in effect, expected liquidity consider-
ations influence current credit-creation activities. Second, as noted,
banks may be unable to meet all loan-contract obligations if the bank-
ing system is faced with a shortage of liquidity in the adjustment
period. Of central importance in banks' decisions about whether to
maintain asset positions or to maximize liquidity in adverse borrowing
environments is the amount of equity owner pressure on bank manag-
ers.
v. Bank behavior and fmancial fragility

This section shows the relevance of the real-time bank model to the
Keynesian research agenda by reconsidering the role of banking in
Minsky's FFH. Minsky ignores banks in the FFH because he sees them
as performing no unique role. But the real-time model can illuminate
two unresolved questions in the FFH.
Minsky's FFH locates the cause of cyclical fluctuation in financial
relations: "Once [mancial considerations are integrated into the invest-
ment decision, it is evident that capitalism as we know it is
endogenously unstable" (1982, p. 81). In an expansion, units experi-
ence "boom euphoria" and convert sound financial positions into
increasingly fragile positions requiring ever higher proportions of cash
flow to meet contingent obligations. Finally, the financing web col-
lapses and recession sets in.
While this vision has been very influential, understanding of the
nature of FF remains incomplete. Two neglected questions, inter alia,
in the conception of financial fragility are: (1) what causes the cyclical
downturn? and (2) how is an adverse shock disseminated from its point
of impact with an economic system?
Minsky on the role of banking in the FFH
Like Davidson, Minsky envisions inherent tension between economic

functions: "A decision to invest is a decision to emit liabilities or
decrease liquidity" (1975, p. 89). Unlike Davidson, he views banks as
analytically indistinguishable from nonbanks. Borrowers' and lenders'

risk encompasses all units involved in financing processes. All units
finance long-term assets with short-term liabilities; hence, as Minsky
writes, "Corporations and households can be considered banks, in that
that they have cash flows to meet" (1975, p. 86).
Banks' role is to expand activity in the boom by more densely
layering asset positions rooted in increasingly optimistic assessments
of expected income flows:
The addition of layered financial intermediation to the financing process
adds further speculative elements. The speculative aspect of banking is
inherent in the very process of lending long and borrowing short. Howev-
er, in a boom the ingenuity of bankers is directed at turning every possible
source of temporarily idle cash into a source of financing for either real
operations or financial position making. (Minsky, 1975, p. 143)
Minsky, then, suggests that all economic units are bank-like. Bank-
ing is a "pervasive phenomenon" (1982, p. 77); banks are "generical-
ly defined as institutions specializing in finance" (1975, p. 57). In
effect, Minsky uses as his model for both banks and nonbanks the
timeless model of section 3: all units take liability positions to support
assets with uncertain return, but no units (for example, banks) are
burdened with the role of liquidity supplier of last resort. 20
Real-time banks and financial fragility
In contrast to Minsky, we have emphasized as banks' unique role their

performance of dual functions. Minsky ignores banks' liquidity-supply
function, which drives the real-time bank model. But it is precisely this
feature that makes the bank optimization problem unique: not only
2°Minsky's parallel treatment of banks and nonbanks extends even to his discussion
of banking runs: "Whenever a run on a bank or other organization takes place, ...
the unit losing liabilities either becomes a forced seller of assets, becomes a bor-
rower from other units at penalty terms, or fails because it is unable to meet its
commitments" (1975, p. 86; emphasis added).
must the bank take an asset position with a stochastic return; it must
finance that position with a liability base of uncertain composition.
The behavior of the real-time bank model is, however, consistent
with the broad outlines of the theory of fmancial fragility: (1) its asset
composition depends on its willingness to absorb risks associated with
uncertain future events; (2) its position thus depends critically on the
state of expectations; (3) when expectations are disappointed, the bank

attempts to reduce its degree of maturity transformation and maximize
its liquidity. While this provides some assurance about the "realism"
of some FFH propositions about aggregate behavior, explicit
incorporation of the real-time bank into the FFH framework can shed
light on some of its unresolved questions, such as those mentioned
above.
Banking firms and the moment of crisis
For Minsky, a financial crisis occurs when "rapid changes in desired

portfolios may be confronted with short-period inelastic supplies of
primary assets (real capital and government liabilities)" (Minsky,
1982, p. 131). It results from three causes: (1) the endogenous genera-
tion of a fragile fmancial structure based on "speculative finance" (p.
66); (2) "some triggering event that induces a reconsideration of de-
sired balance sheets" (p. 140), such that "management begins to view
[its] liability structures as too daring" (1975, p. 116); and (3) an
inelastic supply of the secure assets which can accommodate a "run to
safety. "
Minsky has written little about' 'triggering events"; in a characteris-
tic passage, he writes, "A sharp change occurs when position making
by refinancing breaks down; the change leads to . . . a sharp rise in
borrowers' and lenders' risk" (1975, p. 143). These two forms of risk
pertain, respectively, to intra-firm and external assessments of project
default risk; thus, both derive from real-sector disappointments. So
expectational shifts are singled out to the exclusion of other factors in
explaining banks' role in trigger events: "The state of credit reflects
bankers' views toward borrowers" (1975, p. 119).
But this neglects a second category of trigger event-the initiation of
a crisis by a contraction of monetary policy (a "credit crunch"). A
"credit crunch" has no impact on conditions in the real sector, but
affects banks' major source of liquidity, the interbank market (for
excess reserves). This market is neither "wide" nor "deep" because

of restricted participation. So the credit crunch immediately cuts into
the available supply in the B market. Further, banks cannot instanta-
neously adjust credit commitments in response to choked-off reserve
growth, and so the volume of funds demanded in this liquidity market
rises just as available funds are cut. This structural imbalance causes a
sharp rise (a "spike' ') in the interbank rate. 21 The effect of an increased
B-market rate on the real-time bank, in turn, is shown by this
comparative-static result for equation (15): a7r e /ab e = (ff+ S')F(x).
This expression is negative for a bank with negative NDP-that is, a
bank which depends on B-market borrowings to fund its asset position.
So it is precisely banks that do not match maturities and thus have taken
the risk that "required cash may be available only at penalty rates or
terms when refinancing is necessary" (Minsky, 1975, p. 88) that will,
then, be forced to contract credit. This credit contraction will then
bring about distress through the entire financial/real system. Note that
for the timeless model, by contrast with this result, an autonomous
upward shift in be is precluded by conditions (7) and (9).
Minsky neglects this event; his emphasis on the "subjective" risks
universally entailed in all maturity transformation leads him to ignore
the uniquely fragile structure of banks' liquidity supply.
The role of banks in the

dissemination of adverse shocks
The real-time bank model can also extend Minsky's conception of how
shocks spread from their point of impact to the financial system as a
whole, in a cyclical downturn. Two contributions of banks to the
dissemination process are developed here.
Minsky has argued that any shock is spread through revaluations of
21The evidence from the credit crunches of 1966-67, 1969-70, 1973-74, and 1980-
82 reveals that "spikes" in the interbank rate are a recurring feature of periods of
monetary contraction. Dymski (1987) shows that the interbank rate "spikes" above
other interest rates in crunch periods because of segmented participation in the
interbank market. Note that one effect of financial innovations since 1966 has been
an increase in the number of short-term liquidity markets to which banks have ac-
cess. Interestingly, however, the interest rates in these additional markets have
closely tracked the interbank rate (and, thus, have also exhibited "spiking" behav-
ior) since their inception, suggesting that participants in these markets key their be-
havior to developments in the interbank market proper (in effect demanding a pre-
mium when banks are vulnerable). Wolfson (1986) discusses the empirical aspects
of episodes of "credit crunch" in the postwar period.
expected default levels: "management begins to view liability struc-

tures as too daring" (1975, p. 116). Here again collapsing expectations
of future prospects alter perceptions about liability positions. But this
neglects the possibility that prices and perceptions can change faster on
banks' liability sides than on their asset sides. This possibility arises
because bank solvency is intertwined with the real-capital commit-
ments of all firms with which it holds credit contracts.

To see this, consider this comparative static result for equation (15),
the expected profit equation of the real-time bank: [hr e /
aBe = - L < O. This result implies that the expected profits and equi-
librium loan volume of a banking firm will decline as the bank's
perception of the likelihood of default among its loan commitments
increases. So any single firm's default is potentially harmful to the
status of bank loan contracts with multiple firms: through credit ration-
ing or through higher-than-anticipated loan rates, numerous firms'
positions are jeopardized. Banks are crucial "carriers" of this dissemi-
nation process because of the unique multifirm nature of their asset
commitments.
The second aspect of bank behavior in the midst of a cyclical
downturn concerns the downturn's differential cumulative impact
on banks and nonbanks. Thus far, we have investigated two char-
acteristic features of downturns (especially those accompanied by
a credit crunch)-increased default and bank borrowing rates. Two
additional features are increases in the securities rate and disin-
termediation-systematically lower deposit outcomes, caused by the
increased opportunity cost associated with deposit holding (and per-
haps by concern about bank solvency). The comparative static impact
of these two events on the real-time bank's expected profits-equation
(15)-are, respectively: a7r e /as e = (5" + 11")(1 - F(x» and a7r e /
a( -15) = d - se(1 - h) - F(x)(b e - se)(1 - h) < O. Disinter-
mediation clearly reduces expected profit and eqUilibrium L volume.
An increase in SB has the same effect for a bank with negative NDP.
We can now assess the cumulative impact of cyclical downturns on
banks. For a bank with negative NDP, all the behavioral aspects of
recessionary periods discussed in this section (see comparative-static
results above) push expected profits and eqUilibrium loan volume
downward. It is important to add that rising market interest rates
increase the bank's stock of illiquid vintage loans (see note 18). Thus
market forces in a cyclical downturn simultaneously drive eqUilibrium
loan volumes downward and the stock of illiquid loans upward. Fur-
ther, this stock of illiquid loans is characterized by below-market con-

tract rates in an environment of rising interest rates. 22 So, for a bank
with negative NDP, market forces are a deviation-amplifying mecha-
nism, driving banks steadily deeper into disequilibrium and inducing
the exhaustion of their credit-creation capacity just when the demand
for credit is greatest. The cumulative impact of the downturn on a bank
with positive NDP is ambiguous, since expected-profit losses from

disintermediation and heightened loan default will be offset by gains
on net holdings in short-term market instruments.
This brings us to the differential effect of downturns on banks and
nonbanks. Both types of units experience higher default rates and
disappointed expectations in downturns. Nonbanks can respond by
writing off assets or-more likely-by drawing on liquid reserves or
increasing short-term borrowing. Banks have the same options, but an
additional problem: the steps nonbanks take to strengthen their balance
sheets-drawing down liquidity or tapping lines of credit-increase the
balance-sheet disequilibria of banking firms. Banks alone are faced
with not just higher-cost liabilities, but an uncontrollably changing
liability structure, as disintermediation and liquidity drawdowns eat
into banks' deposit volumes.
The special vulnerability of banks in the downturn suggests why
financial fragility matters. In downturns, banks lose low-cost liabilities
and seek out additional liquidity in borrowing markets just when those
markets are tightest; and nonbank reliance upon bank borrowing to
make asset position is greatest when banks have least scope for lending.
Banks' capacity to satisfy credit demand is finite or shrinking; at some
point, it is exhausted. When the flow of credit through the banking
system is cut off, the financially fragile positions of both non banks and
banks cannot be further extended, and the chain of financial commit-
ments unravels. Banks' inability to respond to shifts in the volume of
credit demands explains why financial commitments become fragile in
the cycle.
Banking and fragility
Minsky's conception of banking emphasizes the interlocking and hence

fragile nature of credit commitments: "in a layered financial structure,
22These contradictory tendencies are heightened to the extent that nonbanks tap into
lines of credit in downturns; indeed loan-commitment drawdowns may be increas-
ing just as the bank's low-cost deposits are being fled away in a high-interest-rate
environment.
the unit acquiring a liability may have liabilities of its own, and its
ability to fulfill its obligations depends upon the cash flow it receives
from its assets, that is, other units' liabilities" (1975, p. 87). We have
suggested, however, that dual-function banks in real-time environ-
ments have a second source of fragility, discounted by Minsky: their
liquidity supply function. Contractionary monetary policy affects the
financial system through banks because banks' principal borrowing

market trades in excess reserves. And banks' distinguishing balance-
sheet feature, the inability to control liability composition, becomes a
critical asymmetry in a downturn. Nonbanks' need to borrow more
centers on banks whose capacity to meet credit demands is being
systematically eroded. Banks are the weak link because their liquidity-
supply function undercuts their credit-creation function.
VI. Conclusion
This paper has attempted to achieve goals at the analytical and method-
ological levels. Analytically, a Keynesian model of the banking firm
has been developed as a microfoundation for the macroeconomic
framework of Keynes and post Keynesian theorists. We have contrasted
a "real-time" model of a dual-function banking firm with the
"timeless" Klein/Monti model. This model is then used to reevaluate
the role of banking in the FFH. While Minsky implicitly uses the
timeless model in exposition of the FFH, the real-time model is consis-
tent with the broad outlines of the FFH. Further, the real-time model
suggests two refmements in the theory of financial fragility. First, with
real-time banks one can explain why monetary contractions (and not
just real-sector events) trigger downtowns. Second, it is precisely the
asymmetric impact of forces unleashed in downturns on real-time
banks' differentia specijica, the liquidity supply function, that explains
why a financially fragile set of fmancial relations can come unraveled.
Methodologically, this paper develops a microeconomic analysis on
the basis of some macrofoundational principles. This raises the ques-
tion of whether one or the other level is more fundamental. This paper
posits the primacy of the macroeconomic dimension. The limits on and
possibilities open to microeconomic units cannot otherwise be well
defined. Even the Walrasian general equilibrium model defines its
macrofoundational context by default.
This does not mean that macrofoundational analysis alone is suffi-
cient. For consistent modeling, correspondence between macroeco-
nomic aggregates and self-interested behavior by economic units must

exist. In effect, the microeconomic level acts as a reality check on a
macroeconomic framework. Further, macrofoundational principles do
not predetermine how a given economic system will respond to a given
set of conditions. Institutional details are critical in defining a model's
parametric context. Study of the behaviors and stresses on microeco-
nomic units may give insight about those units' reactions to changed
circumstances and about the momentum of the system as a whole.

This paper suggests four directions for further research. First, it
indicates that microeconomic analysis based on macrofoundational
constraints can delineate more precisely linkages that have only been
suggestively developed in macro-oriented post Keynesian theory. Sec-
ond, it suggests the importance of further defining and differentiating
categories of economic agents. Third, it demonstrates the need for a
consistent, careful analysis of the core propositions of the theory of
financial fragility. Fourth, empirical work that verifies, modifies, or
rejects the analytical assertions made here is essential. Elsewhere
(Dymski, 1988) I have successfully estimated equation (16) for aggre-
gated large-bank data; that effort is just a first step in this direction.
APPENDIX
Derivation of section 4 equations
To derive equation (15), the bank's maximand, we first take the expectation of
the budget constraint, equation (11), and substitute the resulting equation into
equation (14). This requires a simplifying assumption on equation (14a): the
bank's managers are assumed to expect the interest rates on Sand B to be equal
if x > x. So let se = be in equation (14a). Then we substitute for se
in (14a)
the budget constraint C + L + se
= De(l - h) + K; and we substitute for
Bin (14b) the constraint C + L = D e(1 - h) - Be + K, where we recall
that Be < O. Substitute jj for De (take the expectation of equation (9)). These
manipulations yield:
t-J
(AI) 1I"i = E (lJ - 5j - se)Lj + CZC - 5e - se)L(lc)
j=t-l
for x > x
t-J
(A2) 11"2 = E (lJ - 5j - bjLj + (lc - 5e - be)L(lC)
j=t-l
for x < x.
Recall that, for any constant C, I~ f(u)du = F(A) - F(B), where F(u) is
the c.d.f. for u; and note that, in this case, F(l) = 1 and F( -1) = O. Then
integrate equations (AI) and (A2) over the range of x, and add; this yields
equation (15) in the text. Note that we have made sufficient assumptions about
the curvature of the C, D, and L functions to satisfy second-order conditions.
r
In using (15) to derive first-order conditions for and d, one must recall
that F(x) is a function of both L(f) and D(d) through equation (12). Dividing
r
the first-order condition for through by L and rearranging gives:
I
(A3) [C(l + 1:2) = SB + f/ + CL
+ W- SB] [F(X) + f(x) (_U_+_L_-_K_-_D~(1_-----<.h)) ].

D(l - h)
Using equation (12) to simplify the term
[(LV +L - K - D(l - h))/(D(l - h))]
yields equation (16). An exactly analogous procedure for d gives equation

(17).
REFERENCES
Akerlof, G. "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism."
Quarterly Journal of Economics, 1970, 84, 488-500.
Baltensperger, E. "Alternative Approaches to the Theory of the Banking Firm." Journal of
Monetary Economics, 1980, 6, 1-37.
Benston, G., and Smith, C. "A Transactions Cost Approach to the Theory of Financial
Intermediation." Journal of Finance, 1976,31,215-232.
Cowen, T., and Kroszner, R. "The Development of the New Monetary Economics." Journal
of Political Economy, 1987,95, 567-590.
Davidson, P. Money and the Real World. 2nd ed. London: Macmillan, 1978a.
- - - . "Why Money Matters: Lessons from a Half-Century of Monetary Theory." Jour-
nal of Post Keynesian Economics, Fall 1978b, 1 (1), 46-70.
Deshmukh, S.; Greenbaum, S.; and Kanatas, G. "Interest Rate Uncertainty and the Financial
Intermediary's Choice of Exposure." Journal of Finance, 1983,38,141-147.
Diamond, D. "Financial Intermediation and Delegated Monitoring." Review of Economic
Studies, 1984, 51, 393-414.
Dymski, G. On the Role of Banks and the Interbank Market in Short-Run Financial Instability.
Mimeo, University of Southern California, 1987.
- - - . Banking, Bank Innovation, and Monetary Repression. Mimeo, University of South-
ern California, 1988.
Fama, E. Foundations of Finance. New York: Basic Books, 1976.
- - - . "Banking in the Theory of Finance." Journal of Monetary Economics, 1980, 6, 39-
57.
Gurley, J., and Shaw, E. Money in a Theory of Finance. Washington, D.C.: The Brookings In-
stitution, 1960.
Katzner, D. "Potential Surprise, Potential Confirmation, and Probability." Journal of Post
Keynesian Economics, Fall 1986, 9 (1), 58-78.
Keynes, J. M. A Treatise on Money. Vol. 1-2. London: Macmillan, 1930.
- - - . The General Theory of Employment, Interest and Money. London: Macmillan, 1936.
- - - . "The Consequences to the Banks of the Collapse of Money Values." In Essays in
Persuasion. New York: Norton, 1963.
- - - . The Collected Works of John Maynard Keynes. Vol. 14. London: Macmillan, 1973.
Klein, M. "Imperfect Asset Elasticity and Portfolio Theory." American Economic Review,
May 1970.
- - - . "A Theory of the Banking Firm." Journal of Money, Credit and Banking, 1971,3,
205-218.
Langohr, H. "Alternative Approaches to the Theory of the Banking Firm: A Note." Journal
of Banking and Finance, 1982, 6, 297-304.
Leland, H., and Pyle, D. "Information Asymmetries, Financial Structure, and Financial
Intermediation." Journal of Finance, 1977,32, 371-387.
Minsky, H. John Maynard Keynes. New York: Columbia University Press, 1975.
- - - . Can "It" Happen Again? Armonk, N.Y.: M. E. Sharpe, Inc., 1982.
Monti, M. "Deposit, Credit and Interest Rate Determination under Alternative Bank Objec-
tive Functions." In Mathematical Methods in Investment and Finance. Ed. by. G. Szego and
K. Shell. Amsterdam: North-Holland, 1972, pp. 430-454.
Moore, B. "Unpacking the Post Keynesian Black Box: Bank Lending and the Money Supply."
Journal of Post Keynesian Economics, 1983,5 (4), 537-556.
Pal ley, T. "Bank Lending, Discount Window Borrowing, and the Endogenous Money Supply:
A Theoretical Framework." Journal of Post Keynesian Economics, Winter 1987-88, 10 (2),
282-303.
Parkin, M. "Discount House Portfolio and Debt Selection." Review of Economic Studies,
1970,37,469-497.
Radner, R. "Competitive Equilibrium Under Uncertainty." Econometrica, 1968,36,31-58.
Rousseas, Stephen. "A Markup Theory of Bank Loan Rates." Journal of Post Keynesian Eco-
nomics, Fal11985, 8 (1), 135-144.
Santomero, A. "Modeling the Banking Firm: A Survey." Journal of Money, Credit, and
Banking, 1984, 16, Part 2, 576-602.
Sargent, T. Dynamic Macroeconomic Theory. Cambridge: Harvard University Press, 1987.
Shackle, G. L. S. The Years of High Theory. Cambridge: Cambridge University Press, 1967.
Sprenkle, C. "Liability and Asset Uncertainty for Banks." Journal of Banking and Finance,
1987, 11, 147-159.
Stiglitz, J., and Weiss, A. "Credit Rationing in Markets with Imperfect Information." Ameri-
can Economic Review, 1981, 71,393-410.
Tobin, J. "The Commercial Banking Firm-A Simple Model." Scandanavian Journal of Eco-
nomics, 1982,495-530.
Townsend, R. M. "Theories of Intermediated Structures." Carnegie-Rochester Conference
Series on Public Policy: Money, Monetary Policies and Financial Institutions, Spring 1983,
18, 221-272.
Wells, P. "A Post Keynesian View of Liquidity Preference and the Demand for Money."
Journal of Post Keynesian Economics, Summer 1983, 5 (4), 523-536.
Wolfson, M. Financial Crises. Armonk, N.Y.: M. E. Sharpe, Inc., 1986.

Dymski 1988

Uploaded by

Copyright:

Available Formats

Dymski 1988

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Dymski 1988

Uploaded by

Copyright:

Available Formats

Journal of Post Keynesian Economics

ISSN: 0160-3477 (Print) 1557-7821 (Online) Journal homepage: http://www.tandfonline.com/loi/mpke20

A Keynesian Theory of Bank Behavior

To link to this article: http://dx.doi.org/10.1080/01603477.1988.11489704

Published online: 04 Nov 2015.

Submit your article to this journal

View related articles

Citing articles: 9 View citing articles

Full Terms & Conditions of access and use can be found at

A Keynesian theory of bank behavior

The author is Assistant Professor of Economics at the University of Southern

Journal of Post Keynesian Economics/Summer 1988, Vol. X, No.4 499

(FFH).l Davidson is the outstanding exception; he considers how fi-

model of the bank, and a theory ofmoney (Davidson) without a model of

1A second example of the lack of an adequate microfoundation in post Keynesian

aggregate behavior of the banking system is depicted via an interbank

activity. Both results are contrary to Keynes's depiction of banking.

ll. Keynes's conception of banking

So bank credit allocation is at the heart of cyclical fluctuations (1930,

Banks face a "dilemma" in combining these two functions, for "the

. . . consider what happens when the downward change in the money

Banking figured only marginally in The General Theory. Keynes did

margin of security, that is, involuntary default due to the disappoint-

m. A model of banking in a timeless environment

Banking functions in economic analysis

The assumptions underlying the benchmark model in contemporary

action. Relaxing each assumption in turn generates a distinct banking

positors whose stochastic outcomes are uncorrelated; in effect, banks

assumed to exist for a subset of the agents seeking borrowed funds.

A dual bank in a timeless environment

Consider a banking firm whose asset position of small loans, L, securi-

explicit interest, a minimal amount of it, hD, is held; excess reserves

(1) L +S+ B = D(1 - h) + K.

The deposit market is imperfectly competitive due to geographic

D = D(d), whereD ' > OandD" < O. The 10cationofD(d) is known

Equation (3) can be used to impute the expected return on K:

We can now consider the bank manager's decision problem. The

(3 ') 'Ire = (Ie - oe - s")LW) + (be - se)B

The systemic aspect of bank behavior is depicted via a simple repre-

credit volume is determined by the position of the loan demand curve,

in this model. There is no interaction between the liquidity-supply and

IV. A model of banking in a "real-time" environment

orderly spot market and operate as a residual buyer and seller" of

particular, we posit that economic events of different durations coexist.

lSDavidson goes on to depict banks' effect on economic activity in a manner that

position" (NDP). Interest rates on S and B remain stochastic until the

(10) D(d, x) = 15(d) (1 + x).

In section 3, B was a choice variable. Now, the level of B is jointly

Then set NDP, S - B, equal to zero; substitute for D using equation

(12) x= (LV + L - K)/[D(1 - h)] - l.

If L < D(l - h) + K, so that the realized x exceeds X, the bank will

Bank behavior in the loan period

We now consider the bank's problem. We retain the assumption of risk

(15) 7I"e = ~ [I'j - oj - se - F(i)(b e - se)]Lj

+ [se + F(i)(b e + se)]K - C(D, L)

(17) d(l + Er/) = se(1 - h) + CD

A loan-period equilibrium is established for a vector (ZC*, d*, if,

Bank behavior in the adjustment period

Two events of consequence occur for each bank in the adjustment

A loan-period equilibrium is established for a vector (ZC, d, if,