Philip Lowe and Thomas Rohling - Research Discussion Paper - 9206 - June 1992

Economic Research Department, Reserve Bank of Australia

We are grateful to Adrian Blundell-Wignall, Malcolm Edey, Jerome Fahrer and John Hawkins for helpful comments. Any errors are ours alone. The views expressed herein are those of the authors and do not necessarily reflect the views of the Reserve Bank of Australia.

Financial deregulation in the 1980s saw the lifting of regulations on interest rates charged by banks. In general, lending rates now respond more quickly to changes in banks' cost of funds than they did in the regulated period. However, lending rates still do not always move one for one with changes in banks' marginal cost of raising funds. This paper canvasses four theoretical explanations, other than collusive behaviour, for loan rate stickiness. These theories are based on -
(i)      equilibrium credit rationing,
(ii)     switching costs,
(iii)    implicit risk sharing and
(iv)    consumer irrationality.

Using regression analysis, we also examine the degree of stickiness of Australian interest rates on secured and unsecured personal loans, credit cards, small and large business overdrafts, and housing loans. We find significant differences in the degree of interest rate stickiness among the different rates, even after allowing for lags in adjustment. The rate on credit cards is found to be the most sticky, followed by personal loan rates, the housing loan rate and the small business overdraft rate. The large business overdraft rate is found to adjust one for one with banks' marginal cost of funds. We briefly examine the behaviour of selected U.S., U.K. and Canadian interest rates. The general order and magnitude of interest rate stickiness is similar to that found for Australia. Although it is not possible to empirically discriminate between the different theories of loan rate stickiness, we interpret the results as providing strong evidence for the switching cost explanation. In addition, implicit risk sharing probably plays an important role in the stickiness of the housing loan rate.

TABLE OF CONTENTS

1. Introduction

2. Theories of Loan Rate Stickiness

2.1 Adverse Selection

2.2 Switching Costs

2.3 Risk Sharing

2.4 Consumer Irrationality

3. Tests of Loan Rate Stickiness

3.1 Data and Estimation Procedure

3.2 Results

3.3 Discussion

4. Summary and Conclusions

References

LOAN RATE STICKINESS: THEORY AND EVIDENCE

Philip Lowe and Thomas Rohling

1. INTRODUCTION

The last decade has seen the re-emergence of interest in issues dealing with the operation of the financial system. This interest has taken both theoretical and empirical work in two broad directions. The first is an exploration of the links between the financial system and aggregate economic activity'. The second direction focuses on more micro-economic issues. Questions such as why do banks exist, how do they set interest rates and what type of principal-agent problems arise in banking have received considerable attention. This paper has its roots in this second area of research. It examines possible reasons, other than collusive behaviour, for the stickiness of banks' loan rates and uses data on the various lending rates of Australian banks to examine the degree and causes of interest rate stickiness2.

Price stickiness has long played a central role in macroeconomics.

Paralleling the recent renewed interest in financial markets, there has been renewed interest in the causes of price stickiness. Many theories of slow or incomplete price adjustment in goods and labour markets have

been suggested. These include theories based on market structure and

lack of competition, on implicit risk-sharing contracts, on costs of

changing prices and on consumer switching costs3. While, in the

banking sector, price stickiness has often been attributed to a lack of

competition (see Hannan and Liang (1991)), many of the explanations

advanced to explain price stickiness in goods markets are also applicable

to financial markets. For example, Hannan and Berger (1991) use the

For a recent empirical evaluation of market structure in the Australian banking

industry, see Fahrer and Rohling (1992). Testing three types of market structure, they

find that the hypotheses of both perfect competition and perfect collusion can be

rejected, but that Cournot oligopoly cannot be rejected by the data.

See Blanchard and Fischer (1989) for a summary of various theories of price

rigidity, menu cost model of Rotemberg and Saloner (1987) to explain stickiness

in bank deposit rates, while Klemperer (1987) suggests that his model of

switching costs could also be used for the same purpose. Fried and

Howitt (1980) apply the Azariadis (1976) model of implicit insurance

contracts in labour markets to explain loan rate stickiness as a method of

assuring risk averse lenders of a relatively constant interest rate. A

number of explanations of stickiness in lending rates which take into

account the special nature of a loan contract have also been advanced.

Amongst these explanations, perhaps the most well known is the work

of Stiglitz and Weiss (1981) which shows that in an equilibrium

characterised by credit rationing, the loan rate may not move when other

interest rates move.

In this paper we pay particular attention to four explanations of the

stickiness of loan rates charged by banks. These explanations are based

The empirical work in this paper examines the behaviour of different

lending rates in response to changes in various measures of the banks'

cost of funds. In contrast to the bulk of studies on interest rate

stickiness, we examine the behaviour of a number of different Australian

lending rates. These include the rates on housing loans, secured and

unsecured personal loans, business loans and credit cards. For purposes

of comparison, we also examine the behaviour of a number of interest

rates in other countries.

The various rates that we consider apply to loans with different risk

characteristics and different switching costs. An examination of the

various lending rate responses to changes in the cost of funds thus

allows tentative inferences to be drawn as to the source of any observed

rigidity. Our results do not, however, discriminate sharply between

different hypotheses. Such discrimination is made difficult by the

inability to observe the information costs involved in bank lending.

The remainder of the paper is structured as follows. Section 2 presents

the four theories of loan rate stickiness mentioned above. Section 3 then

discusses the data, our empirical strategy and our results. We find a

considerable degree of price stickiness in all lending rates except for the

indicator rates for business loans. We interpret our results as providing

some support for the switching cost explanation although we cannot rule

out other explanations. Finally, Section 4 concludes and sumrnarises.

In the textbook world of perfect competition with complete information,

price equals marginal cost and the derivative of price with respect to

marginal cost equals one. When the industry moves away from perfect

competition this derivative typically becomes less than one. For example,

in the case of a monopolist facing a linear demand curve, the derivative

derivative is generally less than one when the perfect information

assumption and other implicit assumptions in the classical result are

of a bank loan may not respond one for one with the cost of providing

a bank loan. Specific attention is given to those explanations which

consider the peculiarities of the market for bank loans.

The focus of this paper is solely on marginal pricing decisions. These

decisions affect the profitability of the marginal loan. Overall profitability

is determined by a comparison of average lending rates and the average

cost of funds. The behaviour of the spread between these average rates

is examined in some detail in Reserve Bank of Australia (1992).

2.1 Adverse Selection

Perhaps the most well known model concerning agency costs in banking

is that developed by Stiglitz and Weiss (1981). The firm is assumed to

know the riskiness of its project while the bank cannot distinguish

between projects. This information asymmetry introduces problems of

moral hazard and adverse selection. An increase in the interest rate at

which investors borrow reduces the expected profit on all investment

projects. The safer the project, the greater is this reduction in expected

world in which risky projects already fail, do not reduce the firm's return

in those states. Consequently, when the bank increases its loan rate,

adversely (adverse selection). Alternatively, faced with higher interest

rates, firms may decide to undertake riskier projects (moral hazard).

The problems induced by asymmetric information mean that an increase

in the loan rate charged by the bank will not necessarily result in a

proportionate increase in the expected receipts of the bank. If the

probability of default rises sufficiently, the bank's expected receipts may

actually fall when it increases its loan rate. Faced with this situation, the

bank will elect not to increase its lending rate even if its cost of funds

increase. In such an equilibrium, the bank will set the loan rate below

the market clearing rate and ration credit. The interest rate will exhibit upward stickiness.'

This price stickiness result does not, however, necessarily hold up in

equilibria in which credit is not being rationed. Consider a world in

~illiamson(1 987) demonstrates that a credit rationing equilibrium can exist in

the absence of adverse selection and moral hazard problems, although the

assumption of asymmetric information remains critical. He derives debt contracts as

an optimal arrangement between borrower and lender. Lenders are assumed to face

a monetary cost (for example, bankruptcy costs) of borrower default. At some point,

the probability of default on a given loan increases to such a point that the expected

additional bankruptcy costs outweigh the additional return. At this point, the bank

will find it optimal to no longer increase its lending rate even if the costs of funds increases.

which there are two broad classes of borrowers to which a bank can

lend. For the first class of borrower (such as governments), the

probability of default is zero, while for the second class of borrowers, the

probability of default is positive and increasing in the loan interest rate

(through adverse selection or moral hazard). Assume that the bank can

distinguish between the two classes of borrowers, but not between

customers within each class. Further, assume that the bank is risk

neutral and thus must earn the same expected return on both classes of

loans. Given perfect competition, that rate must be equal to the bank's

where R, is the rate charged on the riskless loan, P(.) is the probability

of default on the second class of loan and R2 is the rate charged on this

bank's costs of funds get transmitted one for one into changes in the

lending rate to the riskless borrower. However, provided the bank is

these loans the bank must increase the lending rate by an amount greater

than the increase in the cost of funds to compensate for the decrease in

the probability of repayment. At some interest rate the bank will not be

able to put the rate up enough to compensate for this risk and all lending

will be made to the first type of borrower. However, until this happens,

the interest rate should not be sticky on the risky loans. In fact, the

reverse is true; the rates on these loans should be very sensitive to

changes in the banks cost of funds.

This model can also be used to examine, more generally, the relationship

between business risk and the spread between the lending rate and the

marginal cost of funds. To do this, suppose that the probability of

default is a function of the state of the economy as well as the interest

rate. As the state of the economy worsens, the probability of loan default

increases. In this case, a deterioration in the economy is likely to lead to

a widening of the spread between the lending rate and the banks

marginal cost of funds5. This can be seen from equation (1). With the

marginal cost of funds (RJ held constant, the business lending rate (R,)

must increase if the probability of default P(.) increases. This issue is

explored in greater detail in Blundell-Wignal and Gizycki (1992).

2.2 Switching Costs

In typical markets, say the market for oranges, the seller does not care

who buys her product; one customer is the same as the next. Anyone

who wants to buy oranges at the listed price can do so. This is not

always the case in the bank loan market; banks are concerned with the

risk profile and potential behaviour of their customers. The bank needs

to find out some information about the characteristics of each and every

buyer. This is a costly activity for the bank. This cost of acquiring

information is often passed onto the buyer by way of a fixed up-front

fee. This fee makes it costly for a buyer to switch from one bank to another.

In addition, there are the regular search costs, or "shoe leather" type costs

of moving from one supply source to another. Such costs include the

costs of learning the different rates and conditions on the new loan.

There are also costs in filling out loan application forms, obtaining the

relevant documentation, and the time involved in attending interviews

with the lending agent. These "search and application" costs are often

more significant in banking than in most goods markets because of the

bank's need to discover the risk characteristics of its customers.

Klemperer (1987) shows that, in general, the existence of switching costs

leads to market segmentation, and reduces the elasticity of demand

facing each firm. Even with non-cooperative behaviour, the switching

costs lead to outcomes similar to the collusive solution, with the

derivative of price with respect to marginal cost being less than one.

Klemperer's model, applied to the banking industry, is set out below.

then it is possible that a deterioration in the state of the world could actually lead to

a fall in the business lending rate. This outcome is, however, extremely unlikely.

Consider two banks, A and B, producing functionally identical products,

with bank B6. Further assume that q consumers have reservation prices

r greater than or equal to f(qI7. Because of the need to obtain

information about a customer, banks charge a fee for new loan

applications. In addition consumers face search costs. Assume that these

search costs vary across individuals. Let T(w) be the cumulative density

function of consumers whose total cost of switching (that is search costs

plus establishment fees) to the other bank's loan product is less than or

loan, or the interest rate.

The demand for bank A's loans is given by

and the demand for bank B's loans by

Since pb is less than pa, some of A's borrowers (and potential borrowers)

will switch to B. Borrowers will, however, only switch if their

reservation prices are greater than or equal to pa and switching costs are

less than or equal to pa-pb. This loss in demand is given by the second

term in equation (2).

Demand for bank B's loans comes from three sources. First, it sells to its

own initial customers (the first term in equation (3)) and to those

customers that were initially borrowing from bank A and who find it

This association may come from having a deposit history with a particular bank.

optimal to switch to bank B (the second term). It will also lend to those

customers who were originally associated with bank A, but who did not

borrow from it, and who now find it optimal to switch to bank B (the

third term). Customers who have a reservation price (r) between pa and

pb, and a reservation price less switching costs greater than pb, fall into

this class.

Given these demand functions, it is possible to derive the non-cooperative

price setting equilibrium. Choosing bank B, the first order

condition for bank B's profit maximisation problem is given by:

Using equation (3)' equation (4) can be rewritten,

can be rewritten,

Suppose that all customers face some switching costs, and that switching

equation (6) can be solved for the equilibrium price,

collapses to p=m. That is, price equals marpal cost. This is same

outcome as that which is obtained under perfect competition.

-

oligopoly) outcome. In general, the higher the switching costs, the fewer

consumers are attracted to a price cut, and thus the less likely a bank is

the pricing solution for the bank lies between the perfectly competitive

outcome and the monopoly outcome.

The derivative of price with respect to marpal cost is given by:

changes in marginal cost are translated one for one into changes in price.

As increases, @/am falls and loan rates become more stickf.

The model above describes a world where people are initially associated

with a particular bank. One criticism of the model is that it assumes

For example, if a=0.30 (at an interest rate of 30%, the demand for funds is zero),

rn=0.13 (the marginal cost of funds equals 13.0%) and ;=0.10 (the maximum

some initial exogenous market share. However, the market shares may

be endogenous. That is, banks may compete more vigorously in the first

period in an attempt to gain market share, thereby increasing second

period monopoly power. In aggregate, however, banks cannot increase

market share, but the increased competition will dissipate any second

period rents.

One response to this criticism is that since customers are aware that the

switching costs will make them captives of the bank, and will be under

possible monopoly power in the second period, they will be less tempted

to purchase from a bank who has initiated a price cut in the first period.

The price cut is a signal to the customer that the bank is attempting to

increase its market share with a view to increasing prices in the second

period. First period demand will then be less elastic than in an

otherwise identical market with no switching costs. This behaviour

results in price stickmess as described above. Further, in the banking

industry in particular, it is not unreasonable to assume that customers

w,ho wish to borrow are initially associated with some bank; either

through previous lending or the bank providing deposit facilities.

In the version of the Klemperer model presented above, the bank's need

for information causes part of the switching costs. Banerjee and

Summers (1987) present a model in which there are no information or

search costs of switching, but firms introduce artificial switching costs as

a loyalty inducing device. This enables firms to split the market and

thus charge a higher price, as in Klemperer's model. Any price cut by

a single firm must be greater than the switching cost before the firm

begins to attract consumers from other firms. With sizeable switching

costs, there is no incentive to cut prices marginally, (or chisel as in a

collusive market) because it would not gain the firm any customers.

Furthermore, it does not pay a firm to lower its price by enough to

capture the entire market. This would leave the other firm with no

customers, and in a position to lower its price below the first firm's price.

The Bertrand pricing solution would result, with price equal to margnal

cost.

The positive profits generated in this artificial switching cost model lead

to the question of entry. Normally, the threat of entry would force the

incumbents to price at marginal cost. However, if entry of new firms is

costly, it will not pay to enter, as Bertrand competition will result,

leaving no profits to cover the cost of entry. Given the high costs of

bank entry, especially at the retail level, banks may have some incentive

to introduce artificial switchmg costs.

An earlier model of markets with switching costs by Von Weizsacker

(1984) focuses on a firm's reputation, and is based partly on work by

Klein and Leffler (1981). Given a market where there are costs of

substituting between different products, customers are unwilling to enter

into a long term arrangement with a firm for fear of losing rents to the

firm at a future date. Finns are able to overcome consumers' reluctance

by reducing the uncertainty associated with price changes by holding

prices constant. In this way, firms may gain valuable reputations by

acting consistently. Observed price inertia may thus be an indicator of

competition, and not an indicator of collusion. However, this model

assumes prices will be fixed in all periods; an unlikely occurrence. If

consumers are risk neutral, an alternative outcome is that banks commit

to tie the interest rate to the observable cost of funds.

If borrowers are more risk averse than the shareholders of the bank,

there exists an implicit risk insurance argument for the sticluness of

interest rates. Fried and Howitt (1980), apply the implicit labour contract

model of Azariadis (1976) to model this effect. Given that the borrower

is risk averse, she prefers stable interest rate payments. As a result the

bank charges a less variable interest rate than its margnal cost of funds,

and the bank is compensated for the additional risk by receiving a higher

average rate than would be charged to risk neutral borrowers.

Customers treat this difference as an insurance premium. Fried and

Howitt argue that customers will not change banks when the lending

rate is higher than the marginal cost of funds because of the existence of

switching costs. Since both parties face these switching costs, it is

mutually advantageous to maintain a long-term relationship. The result

is interest rate stickiness.

2.4 Consumer Irrationality

Ausubel (1991) argues that search or switch costs, although present,

cannot provide a full explanation of credit card rate stickiness? He

argues that there is a class of borrowers who repeatedly believe that they

will pay the outstanding balance before the due date but fail to do so.

These consumers are insensitive to interest rate changes, and are the class

of borrowers that the banks prefer. High risk credit card borrowers, on

the other hand, are more likely to be interest rate sensitive because they

fully intend to borrow on their cards. A credit card rate reduction will

only attract customers who fully intend to borrow (i.e., the high risk

customers). This "reverse" adverse-selection problem makes banks less

likely to compete on credit card rates and thus rates are likely to be

sticky, especially in the downward direction.

3. TESTS OF LOAN RATE STICKINESS

The empirical literature on price stickiness in banking has typically

focused on a single deposit or lending rate. Yet, casual observation

suggests considerable variation in the degree of interest rate stickiness

across different products. The interest rate charged on credit cards

remains constant for long periods of time while the rate charged on

overdrafts changes regularly. In this section we formally examine

interest rates on a number of different types of bank loans and, by

examining differences in the degree of loan rate stickiness, draw some

tentative inferences concerning the cause of the stickiness.

3.1 Data and Estimation Procedure

Prior to the mid 1980s most bank lending rates were the subject of

regulation. For most types of lending these regulations were lifted in

Calem (1992) argues that switching costs are important in the US credit card

market. When a customer wishes to change credit cards, the new issuer may require

her to pay off the balance on the existing card. This may involve several months of

curtailed spending, and this constitutes a considerable switching cost. Such

conditions generally do not exist in Australia.

April 1985. In the case of overdrafts the maximum rate on all overdrafts

was set by the Reserve Bank prior to February 1972". At that time

interest rates on overdrafts drawn on limits over $50,000 became a matter

for negotiation between the banks and their customers while those drawn

under limits less than $50,000 remained regulated. In February 1976 the

threshold level was increased to $100,000 and in April 1985 all

regulations were lifted.

From 1966, when personal loans were introduced, the maximum rate that

banks could charge was set by the Reserve Bank. Once again, in April

1985, the controls were removed. At the same time, the maximum

interest rate that could be charged on credit cards was deregulated. Prior

to this time the maximum rate had been set at 18 per cent per annum.

The period of housing loan rate regulation extended beyond that for the

other lending rates. Until 1973, the maximum rate that could be charged

on housing loans was the same as that on overdrafts although banks

typically charged a lower rate. In October 1973 banks agreed to a

"consultative maximum" on housing loans which was below the

overdraft rate. This was formalised in December 1980 when the

maximum rate that could be charged on owner-occupied housing was set

one percent below the maximum overdraft rate. The ceiling on new

owner-occupied housing loans was finally removed in April 1986.

In the deregulated period, data on certain actual lending rates is readily

available. For example, the actual rate charged on credit card loans is

directly observable and is the same for all classes of borrowers. In

contrast, banks generally do not publish data on rates actually charged

on overdrafts. Instead, they typically quote some base or reference rate to

which a margin is added to obtain the actual loan rate. These margins

are determined on a case by case basis and are a function of the

perceived credit worthiness of the borrower.

The quoting of reference rates for an increasing variety of lending

lo Between 1956 and 1962 a maximum for the average overdraft rate charged was

also set.

products makes it difficult to determine the degree of stickiness of certain

actual loan rates. Conclusions regarding the stickiness of the reference

rate do not necessarily translate into conclusions regarding the stickiness

of the actual rate charged. With this caveat in mind we examine three

overdraft rates. Two of these rates are the advertised overdraft reference

rates of one large Australian bank. The first of these two is the reference

rate for large corporate overdrafts while the second is for small business

overdrafts. The third rate used is the rate most commonly charged on

small business overdrafts by another Australian bank. This rate includes

the margin but it is only available for a subset of our sample period. We

refer to this rate as the "standard rate". In addition to the above

overdraft rates we examine the degree of stickiness in the housing loan

rate, the credit card rate and a variety of personal loan rates.

As detailed above, most lending rate ceilings were lifted in April 1985.

Where data permits, we begin our sample period in January 1986. This

allows a period of adjustment to the deregulated environment. For the

housing loan rate series, the sample period begins in July 1986, while the

four personal loan rate series are only available after September 1987.

The following table summarises details of the rates used for the

deregulated period.

I !---------------------+------------------------------------------+---------------------- I

4

I

I 1 I

I I I

I 1

1 I I

I 1 I

For the period prior to deregulation we examine the unsecured personal

loan series published by the OECD, the housing loan rate, the credit card

rate, the minimum rate charged on overdrafts greater than $100,000 (the

prime rate) and the rate most commonly charged on small business

overdrafts (the standard rate). Further details of all interest rates used

are presented in Appendix 1.

A comparison of the movements of selected rates can be seen in Figure

1 for the period 1979:l to 1985:3 and in Figure 2 for the period 1985:4 to

1991:B.

International comparisons of lending rate behaviour are made difficult

by the fact that lending practices differ across countries. In particular, for

a number of countries, the bulk of personal and housing lending is done

by way of fixed interest loans. The response of interest rates on new

loans of this type to changes in the cost of funds, will be different to that

of the response of interest rates on variables rate loans. Given that in

this study we use variable rate loans for Australia, the international rates

that we examine are restricted to those on variable rate loans. We

examine five such rates: the prime lending rates in the United States,

Canada and the United Kingdom, the mortgage lending rate in the

United Kingdom and the United States credit card rate. Further details

of these rates, together with details of the relevant marginal cost of

funds, are available in Appendix 1.

The standard price stickiness tests involve regressing the loan rate on a

measure of the banks' marginal cost of funds. If price equals marginal

cost, changes in the marginal cost of funds should be transmitted one for

one into changes in the lending rate. It is a difficult task to measure the

exact marginal cost of funds for a bank given the range of funding

sources and deposit products that are available. Nevertheless, there are

observable interest rates which provide satisfactory proxies. These are

the bank bill rate and the certificate of deposit rate (CD rate). We also

construct a third measure of the marginal cost of funds. This measure

is a weighted average of the interest rates paid on fixed deposits over

$50,000, the CD rate and the bank bill rate. The weights used are the

using existing liability shares may not capture the true marginal cost of

funds. Its advantage is that it uses a wider range of interest rates than

any single interest rate measures. Our three measures of the marginal

cost of funds all show very similar time profiles and our empirical tests

revealed similar results for all three measures".

For brevity, we only report the results using the CD rate. We also report

initial results using two other measures of the costs of funds. Both are

weighted average interest rates that banks pay on various classes of

deposits. The first rate, which is labelled AVERAGE, is the average

interest rate that banks pay over all deposits. The second rate, labelled

RETAIL, is a weighted average rate that banks pay on their retail

deposits. A more complete description of these two rates is given in

Appendix 1. Neither of these rates are likely to represent the banks'

marginal cost of funds but are included for completeness and as a basis

for comparison. The discussion focuses on the results obtained using the CD rate.

All estimation is carried out using ordinary least squares. All hypothesis

tests are conducted with a covariance matrix which is robust to

conditional heteroskedasticity and serial correlation. The covariance

matrices are calculated using the Newey-West procedure with 6 lags.

"errors in variables" problem with the parameter estimates being biased and

inconsistent. In simple one variable regression models it can be shown that (see

Johns ton (1972)):

a'x

explanatory variable (in this case the measured marginal cost of funds). It can be

seen that the coefficient is asymptotically biased towards zero and that the extent of

the bias is a function of the ratio of the measurement error variance to the variance

in the cost of funds. To the extent that any measurement error exists in our measure

of the marginal cost of funds, its variance is likely to be small relative to that of the

variance of the measured cost of funds. Any asymptotic bias due to measurement

error is thus likely to be small.

There is some debate over whether nominal interest rates are

characterised by a stationary or by an integrated process. Unfortunately,

the tests which discriminate between these two alternatives are of low

power and of questionable use over sample periods as short as those

used in this paper. We take the view that interest rates are stationary

and thus classical inference is valid. For completeness, however, we also

report in Appendix 2 selected results for regressions where the interest

rates have been first differenced.

We begin by examining the deregulated period. Table 1 presents the

results of regressing the lending rates on the contemporaneous cost of

funds variables. When the CD rate is used, the estimates of the

coefficients on the CD rate are in all cases less than one. In almost every

case the estimates are sigruficantly less than one. All the personal loan

rates, the housing loan rate, the credit card rate, the most commonly

charged small business rate and the base rate for small business loans all

show some degree of stickiness. For the large corporate base rate it is

possible to reject the hypothesis that the coefficient equals one at the 7

per cent level of significance.

When the retail cost of funds and the average cost of funds are used, we

find higher coefficient estimates overall; with the retail cost generally

yielding the highest estimates. For these two cost of funds variables, the

coefficient estimates are both above and below one.

In Table 2 we report marginal significance levels for tests of the

hypotheses that coefficients on the cost of funds variables are equal for

various pairs of lending rates. The personal loan rate, the housing loan

rate and the credit card rate each exhibit significantly more stickiness

than the base rates and the standard overdraft rate. We also find that

there are significantly different degrees of stickiness between the three

retail rates with the credit rate being the stickiest and the owner-occupied

housing rate the least sticky. Although not reported, we find no

statistically sigruhcant difference in the degree of price stickiness between

any of the four personal loan lending rates.

PERIOD

Secured (rnin.)

Secured (max.)

Unsecured (min.)

AVERAGE COST

Unsecured (rnax.)

Personal (OECD)

RETAIL COST

CD RATE

Base (Small)

Base (Large) 3.03 1.29 0.95 2.09 1.71 0.84 3.88 0.90 0.94

(0.68) (0.07) (1.07) (0.12) (0.82) (0.05)

Credit Card 22.87 0.01 -.01 23.65 -0.07 -.01 24.34 -0.10 0.01

(2.13) (0.20) (261) (0.31) (2.13) (0.15)

Housing Loans

- - - - - -- - - - -

Notes: Standard errors appear in parentheses below the coefficient estimates.

21

Notes:

1. The entries in each cell are the marginal significance levels for the tests of the

same across the two relevant lending rates. The three entries in each cell relate to the

three cost of funds variables. In order these three variables are:

1. Average cost

2. Retail cost

3. CD rate.

2. The estimations involving the standard rate are from 1986:l to 1990:4.

Comparing the base rates for small and large business loans we are

unable to reject the hypothesis that the two rates exhibit the same degree

of stickiness. We are, however, able to reject the hypothesis that the

standard overdraft rate is characterised by the same degree of stickiness

as the base rate for large corporate loans.

In summary, the results in Tables 1 and 2 suggest the following ranking

in terms of the degree of price stickiness. The credit card rate is the

stickiest followed by the personal loan rate, the housing loan rate, the

standard overdraft rate and finally the base rates.

The above regressions assume that adjustment of the lending rate occurs

in the same period as changes in the cost of funds. Such speedy

adjustment may not always take place. The transmission of the change

in the cost of funds may be spread out over a number of months.

Accordingly, we included a number of lags of the cost of funds in the

estimated equations. Table 3 presents estimates of the sum of the

coefficients on the contemporaneous and lagged variables for different

lag lengths. It also reports the marginal significance levels for tests of the

hypotheses that the sum of the coefficients on the lags equal one. The

same number of observations have been used for all lag lengths so that

the sum of the coefficients is directly comparable across different

numbers of lags.

In all cases adding lags increases the sum of the coefficients, suggesting

some delay in adjustment of lending rates. However, in general, the

basic conclusions drawn from using only the contemporaneous rate

remain unchanged. For the housing, credit cards and personal loan rates,

the ranking in terms of the degree of stickiness is maintained. Even after

nine lags are included the sum of the coefficients on all three of these

rates remain significantly less than one. The same is true for the

standard overdraft rate. In the case of the small business base rate it is

possible to reject the hypothesis that the sum of the coefficients on the

contemporaneous and lagged cost of funds equal one when only one lag

is included, but it is not possible to do so when three or more lags are

included. For the large loan base rate the sum of the coefficients

Notes:

1. The estimation period for these regressions is from 1986:lO to 1991:8 for all rates

except the housing rate, which is estimated from 1987:4 to 1991:8, and the standard

rate, which is estimated from 1986:lO to 1990:4.

2. The marginal significance levels are in parenthesis below the coefficient estimates.

LENDING

RATE

Personal

Loans

Standard

Overdraft

Base

(small)

Base

(large)

Credit Card

Housing

Loans

Sum of Coeffiaen ts

(marginal significance level for hypothesis test that CP=l)

Number of Lags

0

0.26

(0.00)

0.75

(0.00)

0.85

(0.01)

0.92

(0.16)

0.003

(0.00)

0.45

(0.00)

6

0.38

(0.00)

0.85

(0.04)

0.94

(0.26)

0.98

(0.69)

0.04

(0.00)

0.57

(0.00)

9

0.43

(0.00)

0.83

(0.04)

0.94

(0.35)

0.97

(0.66)

0.04

(0.00)

0.63

(0.00)

1

0.28

(0.00)

0.76

(0.00)

0.87

(0.01)

0.93

(0.20)

-0.01

(0.00)

0.48

(0.00)

3

0.31

(0.00)

0.81

(0.01)

0.91

(0.06)

0.96

(0.46)

0.005

(0.00)

0.52

(0.00)

generally increases with the addition of lags, however, it is not possible

to reject the hypothesis that the sum of the coefficients equals one even

when no lags are included.

If interest rates are sticky, declines in the banks' costs of funds will not

be passed completely into lending rates. This incomplete pass-through

during the interest rate reduction phase has sometimes led to the claim

that banks are exploiting their customers by increasing their lending rates

when the cost of funds increase, but not reducing their rates when the

cost of funds declines1'. One way to test such a claim is to estimate the

stickiness equations with separate parameter estimates for the cases when

the cost of funds decrease and increase. This is done by defining two

dummy variables, one of which takes a value equal to one when the cost

of funds declines, and the other takes a value equal to one when the cost

of funds remains the same or increases. The cost of funds is then

multiplied by each of the dummy variables to obtain two new variables

which replace the original cost of funds variable in the estimated

equation.

The results are reported in Table 4. For the majority of the interest rates

examined it is possible to reject the hypothesis that interest rates respond

symmetrically to cost of funds increases and decreases. However, in all

cases the coefficient on the cost of funds is hgher when the cost is

decreasing. There is no evidence that banks are consistently slower to

bring down their lending rates than they are to increase them. If

anything, the reverse is true. For the personal lending rates, the

reference rates and the home loan rate, the coefficient on the cost of

funds variable is significantly greater when the cost of funds is falling

than when it is increasing. However, the differences in the speed with

which rates are adjusted up and down are quite small. For the credit

card and the standard overdraft rate, there is no signiiicant difference

between the two coefficients.

12 For example, when monetary policy was eased in November 1991, some

consumers complained banks were not passing on the interest rate cuts fully to

mortgage rates. See Gittins (1991) for an account of the debate.

Unsecured (min.)

Unsecured (max.)

Personal (OECD)

Notes:

1. Standard errors appear in parentheses below the coefficient estimates.

2. UP is a dummy variable (1 when the CD rate increases) multiplied by the CD rate.

DOWN is a dummy variable (1 when the CD rate decreases) multiplied by the CD rate.

In Table 5 we report the results of the tests of interest rate stickiness for

the pre-deregulation period commencing in January 1979 and ending in

March 1985. Although there were ceilings on all lending rates (with the

exception of the prime rate) before 1985, these ceilings moved to some

extent with the cost of funds. Figure 2 shows the movement of the

various lending rates.

As expected, the results in Table 5 show that the prime lending rate is

again the most flexible interest rate. However, it is less flexible than in

the deregulated period. Even when nine lags are included, the full effect

of changes in the marginal cost of funds is not translated into the prime

rate. The personal loan rate, the housing rate and the standard overdraft

rate all show the same degree of stickiness. Again, when lags are

included, the sum of the coefficients increases, but for all three interest

rates the sum is always less than 0.5. The credit card rate is constant at

18 per cent through the entire sample. Comparing the pre- and postderegulation

periods, we find no change in the stickiness of the personal

loan rate. We do, however, find that the housing rate and the standard

overdraft rate are more flexible in the deregulated period. For the

standard overdraft rate this difference is particularly pronounced. In

summary, a comparison of the pre- and post-deregulation results

suggests that deregulation has meant that rates on housing and business

loans now move more closely with the cost of funds. In contrast, the

rates on personal loans and credit cards do not appear to be more

flexible in the deregulated period.

As one final exercise, we examine the behaviour of a number of lending

rates in other countries. The results are reported in Table 6. The degree

of loan rate stickiness in the other countries examined is similar to that

for Australia. In all three cases, the coefficient on the prime rate exceeds

0.9. For the U.S. and the U.K., it is not possible to reject the hypothesis

that the coefficient on the cost of funds is different from one. While the

hypothesis can be statistically rejected for the Canadian prime rate

regression, the coefficient is economically close to one. As is the case in

Australia, the housing rate in the U.K. exhibits considerably more

stickiness than the prime rate; the hypothesis that the coefficient equals

one can be easily rejected. Similarly, the credit card rate in the U.S.

Notes:

LENDING

Personal

Loan (OECD)

Standard

Overdraft

Overdrafts

Credit Card

Housing Loan

Sum of Coefficients

Number of Lags

No Lags

13.27

(1.62)

7.63

(1.90)

5.17

(0.87)

18.0

(0.00)

7.01

(1.03)

0

0.21

(0.00)

0.23

(0.00)

0.62

(0.00)

0.00

(0.00)

0.28

(0.00)

0.31

(0.10)

0.36

(0.12)

0.66

(0.06)

0.00

(0.00)

0.34

(0.07)

3

0.29

(0.00)

0.29

(0.00)

0.71

(0.00)

0.00

(0.00)

0.36

(0.00)

6

0.39

(0.00)

0.39

(0.00)

0.76

(0.00)

0.00

(0.00)

0.45

(0.00)

9

0.44

(0.00)

0.43

(0.00)

0.78

(0.00)

0.00

(0.00)

0.48

(0.00)

TABLE 6: TESTS OF LOAN RATE STICKINESS: OTHER COUNTRIES

Notes:

1. Standard errors appear in parentheses below the coefficient estimates.

2. The U.S. credit card rate equation is estimated using quarterly data.

exhibits extreme stickiness, bearing virtually no relationship to the

marginal cost of funds; again a similar result to the Australian case.

3.3 Discussion

Clearly, there are different degrees of flexibility in the loans rates of

various lending products. While our tests do not allow us to distinguish

accurately between the different theories of loan rate stickiness discussed

in Section 2, the results do suggest a number of conclusions.

First, the fact that the minimum and maximum secured and unsecured

personal loan rates all exhibit the same degree of stickiness, suggests that

the credit rationing argument is not solely responsible for the stickiness

of personal loan rates. If the credit rationing argument were correct, then

one would expect that the maximum rate on unsecured loans would

exhibit greater stickmess than the minimum rate on secured loans. The

maximum unsecured rate is charged to those customers whose "type" the

bank is most unsure of. In contrast, the bank is more likely to be able to

observe the type of customers who provide the bank with collateral and

are charged the minimum rate. Thus, if any class of borrowers were to

be credit rationed, it should be those customers paylng the highest rate.

There is no evidence that the interest rate charged to these customers

responds any differently than that charged to other personal loan

customers. While the possibility exists that all personal loan customers

are credit rationed and thus all rates behave in the same manner, we

view this as unlikely. Instead, we regard switching costs, especially

search costs, as the likely explanation of the stickiness in personal loan rates.

We suggest a similar explanation for the stickiness in housing loan rates.

Housing lenlng is considered to be amongst the safest forms of bank

lending. Banks are able to inspect and value the collateral for these

loans, and the actions of the borrower are unllkely to prejudice the value

of the collateral or the "outcome of the project". Neither are there

significant problems in working out the type of the borrower, as most

projects (i.e., houses) are of similar risk. Thus, we view the credit

rationing explanation as being an unlikely source of the stickiness in the

owner-occupied housing rate.

As discussed above, the switching costs of moving from one housing

loan to another are high, especially when mortgage stamp duty is taken

into account. Mortgages typically have had loan establishment fees

which vary according to the size of the mortgage. For a $100,000

mortgage loan establishment fees have typically exceeded $1,000

although in recent times there appears to have been some reduction in

these fees. In addition, to move a mortgage of $100,000 from one bank

to another costs approximately $340 in stamp duty13. Furthermore,

some banks charge an early repayment fee amounting to one months

interest payment if the loan is paid out early. Finally, there are the

standard search and information costs of applying for a new housing

loan. All these various costs make it a costly exercise for the borrower

to switch banks. The theory presented in Section 2 suggests that these

costs cause stickiness in the housing rate.

The risk sharing argument may also play some role in explaining the

stickiness of the housing rate. The household sector is, in all likelihood,

more risk averse than shareholders of the banks. As discussed in Section

2, the stickiness of the loan rate on owner-occupied housing may be, in

part, due to an implicit insurance contract between the bank and its

customers. Conditional upon the CD rate being the appropriate marginal

cost of funds for housing lending, the implicit ex post risk premium does

not seem to have been very large. Over the period January 1987 to

August 1991 the housing rate was, on average, 1.13 percentage points

above the CD rate, while the prime rate was 2.03% above the CD rate".

l3 The establishment fee on mortgages is based on the amount of the loan with

the fee increasing with the loan size but at a decreasing rate. The stamp duty is

calculated as $7.50 on the first $16,000 and $4.00 on each $1,000 thereafter. Some

banks also charge a regstration fee, a discharge fee, and a title search fee. These fees

amount to approximately an additional $100.00.

14 The National Australia Bank's submission to the "Martin Inquiry" (1991) into

the banking industry, stated "we see housing lending as involving a long term

relationship with customers and the household sector preferring a degree of stability

in the interest rates which they face."

In recent times, competition in the housing and business lending sector

has seen reductions in switching costs. At least one bank has waived the

usual establishment fee for business customers of another bank. It has

also promised to pay the government stamp duty and financial

institutions duty on opening a new account. Such developments, if they

become more widespread, should ensure that interest rates follow the

cost of funds more closely.

However, there is the possibility that these policies may only be an

attempt to increase market share, with the special discounts being

removed in the second period when the bank has "captured" its new

customers. Second period prices can then be increased without fear of

losing customers. If consumers anticipate this behaviour, then these first

period cost reducing policies will be of limited success in gaining market share.

The credit card industry is another market where both search and

switching costs are likely to be present. Consumers face some

information costs when determining the lowest interest rate offered by

banks on their credit cards. They also face the costs of time and effort

of applying for a card, and the cost of the time lag between applying for

a card and receiving one. These costs are, however, small compared to

those incurred in switching a mortgage from one bank to another.

Ausubel (1991), using U.S. data, finds that switching costs are not large

enough to be the sole explanation for the credit card rate stickiness.

Instead, he argues consumer irrationality may exist, leading to the

reverse adverse selection problem described in Section 2. Ausubel

presents some evidence hom a respected consumers survey report

indicating that the majority of consumers say that they hlly pay the

outstanding balance on their creht card. However, bank data on credit

card usage indicates that the number of accounts incurring credit card

rate charges is in excess of 75 per cent., Consumers, in effect, say one

thing but do another. Some prima facia evidence for reverse adverse

selection can be found by looking at Figure 2. Over the period of study,

credit card interest rates have monotonically increased, even though the

cost of funds has both increased and decreased over the same period.

If reverse adverse selection characterises the credit card market, then in

certain circumstances, there may exist a role for government to encourage

lower rates of interest on cards. If a single bank attempts to lower its

credit card rate unilaterally, it will primarily attract the interest sensitive

customers. These customers represent high risk borrowers. Thus, when

a single bank reduces its credit card rate, it worsens its pool of

customers. In this case, declines in the marginal cost of funds may not

be translated in changes in credit card rates as no bank wishes to move

first. In such an environment, a co-ordinated reduction in interest rates

may be desirable. This co-ordination could come through government

initiatives.

The evidence on the base rates suggest that the small business rate may

move slightly more slowly than the large rate, although the differences

are statistically insignificant. They both appear to move in line with the

CD rate. On the other hand, the standard rate is considerably more

sticky. This apparent contradiction can be partially resolved by recalling

that the base rate is the rate offered to the bank's best small business

borrowers. The standard rate is the rate applicable to the bank's average

small business borrowers. The fact that the average rate is sticky

suggests that some of the rates across the spectrum of a banks small

business customer base are sticky. One would expect the banks to be

lending at or near the base rate to their best business customers. This

implies the remaining, (higher risk) borrowers face stickier interest rates.

This result is not inconsistent with credit rationing. Because of

information and monitoring problems, the riskier small business

borrowers are more likely to be credit rationed, causing their interest

rates to be sticky.

The move to quote all business lending rates as a margn over a base rate

may have caused lending rates to all business groups to move more

closely in line with the marginal cost of funds; that is, there may have

been a change in the pricing policy of banks. Given the lack of

appropriate data we are unable to test this hypothesis.

This paper examines the degree of price stickiness in the market for bank

loans. In the classical world of perfect competition, changes in marginal

costs are translated into similar changes in the price of the product. We

find that complete pass-through of changes in banks' marginal cost of

funds only occurs with the base or reference overdraft rates to large and

small business borrowers. For credit cards, personal loans, owner-occupied

housing loans and the standard overdraft rate, changes in the

banks' marginal cost of funds have not been translated one for one into

the contemporaneous lending rates.

We discuss four explanations for the stickiness of most of the lending

rates. The first explanation relies on the existence of equilibrium credit

rationing. In such an equilibrium, banks will be unwilling to increase the

lending rate, even when the cost of funds increase, for fear of reducing

their expected return. The second explanation relies on the fact that the

nature of a bank loan requires the bank to obtain information about each

and every customer. The incidence of these information costs falls on the

borrower in terms of upfront fees and search costs. These costs reduce

the elasticity of demand, giving the bank some market power. Third, the

stickiness in the loan rate may be the result of an implicit risk sharing

contract between the bank and its customers. Finally, we discuss a form

of consumer irrationality in the credit card market.

The results presented in this paper do not allow us to distinguish sharply

between these different hypo theses. To do this would require extensive

data on the cost of information collection by both banks and customers.

Evidence on the notoriously difficult to measure degrees of risk aversion

and consumer rationality would also be required. Nevertheless, the

results point in particular directions.

We find little support for credit rationing being the explanation of loan

rate stickiness. For the housing loan rate, switching costs and risk

sharing appear to be important causes of the interest rate stickiness. For

the personal loan rate, switching costs are again likely to play a role;

however, the failure of the behaviour of the personal loan rate to adjust

after deregulation may reflect a lack of competition in this market.

Evidence from the standard or most commonly charged rate to small

business borrowers suggests some interest rate stickiness.

In summary, there are solid reasons for bank lending rates not moving

one-for-one with the banks' marginal cost of funds. Incomplete

adjustment of lending rates does not necessarily imply collusive

behaviour amongst the banks. The peculiar nature of a banking contract,

in which the seller (the bank) acquires information about the buyer (the

borrower) but is not able to control either the buyer's actions, or

determine her true type, can help explain incomplete pass-through.

Whle little can be done about "switch costs" which arise directly from

the costs of information gathering, reducing artificial switch costs is likely

to reduce any market power that banks enjoy. This could be done by

eliminating mortgage stamp duty and by banks providing more extensive

and accessible information about the terms and conditions of various

loans.

INTEREST RATES

Personal lending rates were obtained from two sources. The first source

is from the OECD Financial Statistics, (Part 1, Section 2: Domestic

charged by major banks, as at the end of the month. It is an effective or

reducing rate, not a flat rate. The second source is the maximum and

minimum variable personal loan rates, secured and unsecured, obtained

from a major Australian bank.

(i) Standard Rate

This rate, obtained from internal RBA sources, is reported by one of the

major Australian banks. It is the most commonly charged rate to

borrowers for overdrafts of less than $100,000, typically small business

borrowers. In May 1990, a retail index rate was introduced. Loans to

small businesses are expressed as the retail index rate plus a margin.

(ii) The Rate on Overdrafts of $100,000 and Over

The minimum of a range of indicator rates reported by major banks.

This rate is used for the pre-deregulation period. RBA Bulletin, Table

F.3.

(iii) Reference Rate for Large Borrowers

The National Australia Bank Benchmark Rate applies to the Corporate

Clients accounts. End month or near end-month figures obtained from

the Monday edition of the Australian Financial Review.

(iv) Reference Rate for Small Borrowers

The National Australia Base Rate applies to the retail and commercial

accounts. End month or near end month figures are obtained from the

Monday edition of the Australian Financial Review.

This rate is from internal RBA sources, and is an average of rates

reported by the major banks on a bankcard with 55 day free credit

facility.

The housing loan rate to individuals for owner occupation is from the

RBA Bulletin, Table F.3. This is a predominant rate on variable interest

rate loans.

BANKS' COST OF FUNDS

We have constructed three different measures of costs of funds; retail,

wholesale and total. The "retail rate" is a weighted average interest rate

of current deposits, fixed deposits less than $50,000, passbook, statement

and investment accounts and "Other" (Cash management accounts). The

"wholesale rate" is a weighted average interest rate of fixed deposits

greater than $50,000, certificates of deposits and foreign currency

deposits. The broadest measure is the total weighted average deposit

rate. This rate takes into account all the major deposit sources available

to a bank.

The deposit categories and corresponding interest rates, along with their

source are listed below.

Current: Not Table D.l plus gov't deposits,

bearing

interest

Trading Bank

Fixed: less

than $50,000

The breakdown of trading

government, into large and

internal RBA sources, prior to

January 1989. From this date,

All bank fixed divided using

last available proportions

Trading Bank

Fixed: over

$50,000

As above

Savings Bank Prior to January 1989, from

Includes gov't deposits. Now

in All Bank Liab. Table B.2

INTEREST

0 rate applied

Current account

rate

Supplied by a

major Australian

bank

maturity rate,

Table F.3

This is the most

common maturity

period. Data is

unavailable for

further breakdown

Fixed deposit,

weighted avg.

Table J.3

(discontinued

Dec 88), and

Table F.3

Weighted average

rate available to

Dec. 1988. From

month to maturity

is most common

3 month to

Maturity, less

than $50,000,

Table F.3

less than $50,000

Investment As above, Table B.2 Investment

accounts, Table

F. 3

Average of min

and max rates

Statement

Passbook

"Other"

As above, Table B.2 Statement

accounts, Table

F. 3

As above, Table B.2 Passbook

accounts, Table

F. 3

Internal RBA sources indicate

Management Accounts. Table

B .2

CMA rate

Average of min

and max rates

-

Internal RBA

sources

of Deposits

Foreign

Currency

Table B.2. Prior Jan. 1989, CD rate Table

Weighted average

Foreign currency liabilities,

Table B. 1

90 day bank

bill rate

No rate is

available. Assume

INTEREST RATES: OTHER COUNTRIES

UNITED STATES

The prime rate is from the Federal Reserve Bulletin, Table 1.33, "Prime

rate charged by banks on short-term business loans". The interest rates

are recorded on the date when they change, allowing an end-month

series to be constructed.

The credit card rate is from the ~ederalR eserve Bulletin, Table 1.56,

"Terms of Consumer Installment Credit", item 4, Credit card. The series

is available on a quarterly basis only.

The cost of funds measure is from the Federal Reserve Bulletin, Table

1.35, "Interest Rates Money Market and Capital Markets", item 12,

Certificates of Deposits, secondary market, 3-month.

UNITED KINGDOM

The prime or base rate is from Central Statistical Office, Finance

Statistics, published by the Government Statistical Senrice, England, Table

13.10 "Selected Retail Banks: Base Rates". The prime rate is recorded on

the date when they are changed, allowing an end-month series to be

constructed.

The variable mortgage rate on housing lending is from the OECD

Financial Statistics, Monthly, Part 1 Section 2. Table R.2/17. "Lending

nominal rate.

The cost of funds measure is the Sterling certificates of deposits 3

months, from Central Statistical Office, Finance Statistics, published by

the Government Statistical Service, England, Table 13.8 "Short Term

Money Rates: Last Friday of the Period". The minimum of the range of

rates is used.

CANADA

Prime Rate

The prime rate is from the Bank of Canada Review, Table F1 "Financial

Business. Figures are the last Wednesday of the month.

Banker's Acceptances

The measure of the cost of funds is from the Bank of Canada Review,

Table F.1 "Financial Market Statistics". As no Certificate of Deposit rate

is available, we use the Banker's Acceptances rate, 1 month. Figures are

the last Wednesday of the month.

Notes:

2. The parameters in the first two columns are estimated from 1986:l to 1991:8 for

except for the standard rate, which is estimated from 1986:l to 1990:4 and the housing

rate, which is estimated from 1986:7 to 1991:8. For all estimation in which lags are

included, estimation is 1986:lO to 1991:8, except for the housing rate, which begins

in 1987:4, and the standard rate, which is estimated from 1986:lO to 1990:4.

41

REFERENCES

Ausubel, Lawrence M., "The Failure of Competition in the Credit Card

Market", American Economic Review, March 1991, 81, 50-81.

Azariadis, Costas, "On the Incidence of Unemployment", Review of

Economic Studies, February 1976, 43, 115-125.

Banerjee, Abhijit and Lawrence Summers, "On Frequent Flyer

Programs and Other Loyalty-Inducing Economic Arrangements",

Discussion Paper, Harvard Institute of Economic Research, 1987.

Berger, Allen H. and Gregory Udell, "Some Evidence on the Empirical

Significance of Credit Rationing", Finance and Economic Discussion Series,

Federal Reserve Board, January 1990, #105.

Blanchard, Olivier and Stanley Fischer, "Lectures on Macroeconomics",

1989, MIT Press, Cambridge, Massachusetts.

Blundell-Wignall and Marianne Gizycki, "Credit Supply and Demand

and the Australian Economy", paper presented at the Australian

Conference of Economists, July 1992.

Calem, Paul, "The Strange Behavior of the Credit Card Market", Business

Review, Federal Reserve Bank of Philadelphia, January/February 1992.

Fahrer, Jerome and Thomas Rohling, "Some Tests of Competition in the

Australian Housing Loan Market", Research Discussion Paper 9202,

Reserve Bank of Australia, 1992.

Fried, Joel and Peter Howitt, "Credit Rationing and Implicit Contract

Theory", Journal of Money, Credit and Banking, August 1980, 12, 471-487.

Gertler, Mark, "Financial Structure and Aggregate Economic Activity: An

Gittins, Ross, "Bank Rates May Not Be in Your Interest", Sydney Morning

Herald, Nov. 13, 1991.

Hannan, Timothy H. and Allen N. Berger, "The Rigidity of Prices:

Evidence from the Banking Indusfxy", American Economic Review,

September 1991, 81,938-945.

Hannan, Timothy H. and Nellie J. Liang, "Inferring Market Power from

Time-Series Data: The Case of the Banking Firm", Finance and Economic

Discussion Series, Federal Reserve Board, January 1991, No. 147.

Johnston, J, Econometric Methods, 1972, McGraw Hill, New York.

Klein, Benjamin and Keith Leffler, "The Role of Market Forces in

Assuring Contractual Performance", Journal of Political Economy, 1981:4,

615-641.

Klemperer, Paul, "Markets with Consumer Switching Costs", Quarterly

Journal of Economics, May 1987, 102, 375-94.

Parliament of the Commonwealth of Australia, "A Pocket Full of

Change", Stephen Martin, M.P. Chairman, The House of Representatives

Standing Committee of Finance and Public Administration, AGPS Canberra,

Nov. 1991.

Reserve Bank of Australia, "Bank Interest Rate Margins", Reserve Bank

of Australia Bulletin, May 1992, 1-6.

Rotemberg, Julio J. and Garth Saloner, "The Relative Rigidity of

Monopoly Pricing", American Economic Review, December 1987,77,917-26.

Stiglitz, Joseph E. and Andrew Weiss, "Credit Rationing in Markets with

Imperfect Information", American Economic Review, June 1981,71,393-410.

von Weizsacker, Christian C., "The Costs of Substitution", Econometrics,

September 1984, 1085-1116.

Williamson, Stephen D., "Costly Monitoring, Loan Contracts, and

Equilibrium Credit Rationing", Quarterly Journal of Economics, February

1987, 135-146.