Credit rationing and the adjustment of the loan rate: An empirical investigation

PETER KUGLER Universitiit Bern

Bern, Switzerhd

Credit Rationing and the Adjustment of the Loan Rate: An Empirical Investigation *

In this paper, models of credit rationing are analyzed using quarterly data on domestic bank loans of four countries. First, models of temporary (dynamic) credit rationing are considered. The price (interest rate) equation proposed by Bowden is estimated assuming equal and unequal adjustment speeds under excess demand and supply conditions. Second, the stability of the interest-rate equation is tested. We motivate this test by the fact that permanent (supply-side equilibrium) credit rationing implies instability of this regression relationship. Statistically sign&ant credit- rationing effects are found for the countries considered, with the exception of the U.S.

1. Introduction The theoretical and empirical analysis of credit rationing has

advanced significantly in recent years.’ The theoretical models of Keeton (1979), Stiglitz and Weiss (1981, 1983), and Devinney (1983) provide supply-side equilibrium explanations of credit rationing. In this framework, credit rationing stems from rational profit-maximiz- ing behavior of lenders under imperfect information: lenders are not willing to increase the interest rate enough to eliminate excess demand if the adverse selection and incentive effects of such an increase lead to decreasing expected profits. By contrast, the earlier

*Earlier versions of this paper were presented at the Second International Meet- ing on Monetary Economics and Banking, Nice (June 6-7, 1985), at a seminar at the University of Zurich, at a joint seminar of the University of Basle and the BIS, and at the Disentis meetings (October 7-8, 1984). The author benefitted from many useful comments of participants of these meetings. Special thanks go to Stephen King, who, in addition to many helpful suggestions, provided me with U.S. loan rate data. I also acknowledge the helpful comments of an anonymous referee. Of course, the responsibility for remaining errors is mine. Financial support by the Swiss National Science Foundation (Grant No. 4.600-0.82-09) is gratefully acknowl- edged.

‘For a survey of recent theoretical work the reader is referred to Baltensperger and Devinney (1985).

Journal of Macroeconomics, Fall 1987, Vol. 9, No. 4, pp. 505-525 Copyright 0 1988 by Louisiana State University Press 0164-0704/88/$1.50

505

Peter Kugler

theoretical models of Freimer and Gordon (1965), Jaffee and Mo- digliani (1969), and Jaifee (1972) created credit rationing by legal or social restrictions on interest rates.

In recent empirical research, disequilibrium econometric methods have been used in order to establish the empirical significance of credit rationing. This approach offers a new and promising way to handle the problem of inadequate information on the quantity supplied and demanded, which was circumvented in earlier studies by the use of proxy variables (Jaifee and Modigliani 1969; JafIee 1972) or survey data (Harris 1974). The findings of the studies applying disequilibrium econometrics differ somewhat according to models, methods, and data used. The results of Laffont and Garcia (1977) for Canada, Sealy (1979) for the U.S., Ito and Ueda (1981) for Japan, and Artus (1984) for France point to substantial disequilibrium situations in the market for bank loans. By contrast, Ito and Ueda’s two-country study suggests that credit rationing is not an empirically important phenomenon in the U.S. Moreover, the econometric disequilibrium models applied do not correspond to recent theoretical models of supply-side equilibrium credit rationing. The former models deal with temporary disequilibrium situations brought about by lagged price adjustment or with permanent disequilibrium situations involving no price adjustment. By contrast, the latter class of models implies that the market clears immediately if the corresponding interest rate is lower than a supply-determined bound. If this condition is not fulfilled, the interest rate does not adjust in order to eliminate excess demand.

The purpose of this study is to present some empirical evidence on the extent of credit rationing in three European countries (U.K., West Germany, Switzerland) and, for reasons of comparison, in the U.S. Our analysis is carried out mainly in the framework of the alternative formulation of the usual price-adjustment equation proposed by Bowden (1978). This approach offers two important ad- vantages: the equilibrium hypothesis is easily tested, and the de- gree of (temporary) disequilibrium can be compared directly across markets. First, the loan-rate equation is estimated with equal adjustment speeds under excess demand and supply conditions. Sec- ond, the model is extended to the case of unequal adjustment speeds, and the resulting reduced-form price equation is estimated with a two-stage probit approach. Third, the case of supply-side equilibrium credit rationing is considered. This model, which implies that the parameters of the price equation are not stable over time, is tested by using the cusum and cusum of squares test suggested by

506

Credit Rationing and the Adjustment of the Loan Rate

Brown, Durbin, and Evans (1975). Finally, the alternative test of equilibrium against disequilibrium, proposed by Hwang (1980), is carried out for reasons of comparison.

The paper is organized into three main sections. First, the models and the estimation and test procedures are briefly presented. The empirical results are outlined in the next section. The paper closes with some conclusions and a data appendix.

2. Models and Estimating and Testing Procedures Equal Adjustment Speeds

Consider the following standard econometric disequilibrium model for bank loans:

Lt’ = Xi, (~1 + P,RLt + qlt; 0)

G = .%,a, + P,RLt + qzt ; (2)

L, = min (LB, LS) ; (3)

ARL,=h(L:‘-G); (4)

where Ld and L” are the demand for the supply of loans, respectively; L is the observed quantity of loans assumed to be the min- imum of demand and supply; Xi and XZ are the vectors of predetermined variables; RL is the interest rate on loans; cri, 02, pi, and B2 represent the parameters to be estimated; and qi’s (i = 1, 2) are normal, nonautocorrelated error terms with expectation zero and contemporaneous covariance matrix a. The level of the loan rate, RL,, adjusts partially to disequilibrium conditions. The parameter, A, reflects the extent of disequilibrium in the loan market. If this parameter is zero, there is no interest-rate adjustment and, therefore, no tendency towards equilibrium. By contrast, if the condition A = m is valid, we have perfectly flexible interest rates and, therefore, always have equilibrium. This formulation of the model has two drawbacks. First, there is no well-defined statistical test for the equilibrium hypothesis, A = m. Second, the values of X depend on the unit of measurement of L and, therefore, are not directly com- parable for different markets. These problems can be avoided by using the formulation of the model proposed by Bowden (1978a). Inserting (1) and (2) in (4) and solving for RL, supplies the following expression:

507

Peter Kugler

RL, = pRL,-1 + (1 - k) F&2 - -Gel + rlzt - q13

Pl - P2 ; (5)

1

p = 1 - A(@, - pz) *

Noting that the last term on the right-hand side of Equation (5) corresponds to the solution RLP of the equilibrium model [(l) and (2)], we arrive at the following complete formulation of the interest- rate equation:

RL, = pRL,-I + (1 - p)RLf ; (6)

RLf = -GP-2 - Xf,~l + qzt - q1t

PI - I32

The actual loan rate is a weighted sum of its lagged and equilibrium values (that is, RL, is a geometric distributed lag of RLF).

In the equilibrium case, A = 03, we have lo, = 0, whereas in the case of complete interest-rate rigidity, A = 0 is covered by in = 1. Thus, the equilibrium hypothesis p, = 0 is statistically well defined and can be tested. In addition, the parameter, CL, is unit free and can be directly compared across markets.

The structural model consisting of (l), (2), and the new price- adjustment equation, (6), can now be estimated with different sin- gle-equation and system methods. However, the disequilibrium parameter, p, can be estimated without estimating demand and supply functions by the application of ordinary least squares (OLS) to the price equation, (8), if we assume that Xi, and Xzt do not contain RL,-1:

RL, = PRL,-~ + 2;6(1 - /A) + (1 - t.~)e~. (8)

Thus, 2: is the vector of the predetermined variables Xi, and Xzt; 6 is the corresponding reduced-form coefficient vector; l t is defined as hzr - qltMP1 - P2h and its variance is denoted by a:.

It has to be noted that the disequilibrium model presented above differs from the usual partial-adjustment model for bank loans.

508


The partial-adjustment scheme for the volume of bank loans has to be considered as an equilibrium model in our context. In this framework, partial adjustment is desired and caused by quantity adjustment costs and not by interest-rate stickiness. Formally, we get this model by including Lt.-l in X,, and/or X2, and the assumption A = ~0 (p = 0). H owever, the reduced-form interest-rate equation of a generalized volume-adjustment model including I&-, can no longer be empirically distinguished from our disequilibrium interest-rate equation. In this case, B&-i has to be included in Z,, and l.r, is, therefore, no longer identifiable in Equation (8).

In addition, credit rationing is tested jointly with the simple geometric price-adjustment scheme. Therefore, misspecifications of the latter-caused, for example, by nongeometric adjustment pat- terns or the relevance of additional variables representing administered pricing schemes2-lead to invalid results. Therefore, it seems advisable to compare the reduced-form results with the findings obtained by an alternative procedure. This can be done by specifying and estimating the full structural model or by testing for disequilibrium using the method proposed by Hwang (1980). The latter approach is based on the fact that the regression of L,, BL,, and 2, is stable (coefficients and variance) under the equilibrium hypothesis, whereas disequilibrium regime switches lead to instability in this regression equation. Thus, testing for disequilibrium amounts to testing the stability of a regression equation using standard procedures like the cusum and cusum of squares method suggested by Durbin, Brown, and Evans (1975). This approach, which will be briefly outlined in the context of the equilibrium credit-rationing models, is used in this paper to check the results obtained in the reduced-form framework.

Unequal Adjustment Speeds Consider the following reformulation of the price-adjustment

equation:

ARLt = { MC-G), L;’ > L; A,(Lf - L;) ) L:‘< L; * I

(44

*The introduction of an error term in Equation (4) does not invalidate the reduced-form approach in the model with equal adjustment speeds. However, the consideration of unequal adjustment speeds outlined in Section 3 depends on the exact proportionality of excess demand and price changes.

509

Peter Kugler

This model allows for different downward (excess supply) and upward (excess demand) adjustment speeds of the loan rate. There- fore, only excess demand can lead to disequilibrium situations (A, # m), whereas excess supply is immediately eliminated by a decline of EL, (A, = m). Thus, this model comes closer to recent theoretical models of supply-side equilibrium credit rationing. By contrast, it is assumed that excess demand always creates a disequilibrium.

The combination of (l), (2), and (4a) and the definition of two price-adjustment parameters,

1

I4 = 1 - Ai@, - pJ ’ i=l,2,

gives the following reduced-form equation for RL,:

RL, = ~1 (RLt-1 + (1 - PI)=:: > PzRL,-1 + (1 - t-dRLt* >

(64

Inserting (7) in (6a) and arranging the inequalities yields the following formulation of the reduced-form price equation, which is suit- able for its estimation:

RL, = (84 PIRL,-1 + G’S0 - ~1) + (1 - IJ& >

The easiest way to estimate the model with unequal adjustment speeds consists of the separate application of OLS using excess demand and excess supply observations, respectively. How- ever, this approach has two serious drawbacks. First, it yields biased estimates as the regime selection is endogenous. Second, the re- striction that the coefficient vector, 6, is the same across regimes is not taken into account.

The application of standard results for switching regression (for example, Maddala 1983, 223-28) supplies the conditional expecta- tions of the disturbance given the regime:

510


E [

(1 - /.&,I ; (RL-1 - z; 6) h $ c 1 4% = -0 - I”&-& & ; t

E (1 - t.+,] ; (RL,-1 - Z:6) < ; I

= (1 - I+, & . (9) c I

$t and @‘t are the values of the standard normal density and dis-

tribution function evaluated at - ’ VGI - Z:6), respectively. Equa- WE

tion (8a) can now be estimated, taking into account (9), by the following two-stage probit approach. The first stage consists of the estimation of the probit model for the regime selection variable, I,,

1, t (RL,-l - Z; 6) 2 ; (ARL, 5 0) c

I, =

1

> 00)

0, b (RL,el - Z;S) < ; (ARL, > 0) E E

as a function of RL,-, and Z,. This procedure results in estimates of C& and at, which are denoted by 4, and &*, respectively.

The second stage can be carried out in two different ways. Bowden (197813, 208-10) proposes the separate estimation of the

4% two equations of (Sa) supplemented by the terms -(l - bz)oE c

t

+t and (1 - ~~)a, -

1 - @,’ respectively.

This approach has, however, the drawback that the cross regime coefficient restrictions are not taken into account. Thus, it seems preferable to proceed as follows. The unconditional expectation of RL, is given by

E(R&) = E(RL,(Z, = l)@, + E(RL,(Z, = O)(l - at,,) . (11)

Inserting (8a) and (9) in (11) yields

Peter Kugler

E(RL,) = [

pzRL,-1 + z;6(1 - tJ.2) - (1 - &$ (I$ t I

+ [ PlRL1 + z:ao - 111) + (1 - Plb.& 1 (1 - @,I ; t or after rearrangement,

+ (CL2 - Plbe+t~ 02)

Thus, we have a restricted nonlinear regression equation for RL,. After replacing +t and @‘t by their first stage estimates, l.~.i, I& - pi), 6, and u, can be estimated by a nonlinear regression approach.

Equilibrium Credit Rationing and the Loan Rate Equation In the sequel, the econometric problems arising with recent

theoretical equilibrium models of credit rationing are considered in the frame of a model proposed by Stiglitz and Weiss (1981). This model assumes that lenders maximize expected profits under imperfect information. On this condition, adverse selection (borrowers willing to pay higher interest are, on average, greater risks) and adverse incentive effects (the higher the interest rate, the riskier are the projects taken by borrowers) lead to a backward bending loan supply curve as illustrated in Figure 1. The expected profit rate of lenders increases with the loan rate on the upward-sloping segment of the curve. However, the positive effect of an interest- rate increase on the expected profit rate is outweighted by negative adverse selection and incentive effects on the downward-sloping segment of the supply curve. This implies that lenders are not willing to increase the interest rate above RL”. Therefore, the market clears at RL$ if the demand schedule is given by G,“, whereas we will have credit rationing at RL” in the case of demand curve Lf.

This model can be written as follows:

L;i = f”(&t, RL,, rllt) ; (lb)

CW

512


L

\

LS I

f

\

\ Ld 1

RL*O RLU RL*l RL

Figure 1. Supply-Side Equilibrium Credit Rationing

FK, = min (RLf, EL:) ;

f dKt> RL:, -%t) = fS&> XP> rl2t) ;

W4

where f" is a normal, downward-sloping demand schedule and f” is a backward-bending supply curve, which has a unique maximum for % > 0. X1,, -Gt, qlt, and qzt have the same meaning as in Section 2.

513

Peter Kugler

After linearization of the two equations determining RL: and RL;, we arrive at the following three equations for RL,:

RL: = Xzt~ + St ; ,

(W

RL, = min (RL:, RLF) ; (84

where y is an additional unknown vector of regression coefficients, and St is a normal and nonautocorrelated error term with expectation zero.

lU,* and RL: are, of course, not observable. Thus, the three equations determining RL, have the same structure as the usual econometric disequilibrium model with an exogenous price variable. Therefore, the methods available to estimate and test this model can be applied to this model too.3

This framework is straightforwardly extended to cover the case of temporary or dynamic credit rationing. Consider the following augmented model:

RL: = PRL,-~ + (1 - p)RL: ; (8)

RL: = P~RL~-~ + (1 - pe)RL; ;

The first two equations of this system describe the adjustment of the loan rate to its market-clearing equilibrium value under circum- stances of temporary or dynamic credit rationing. Equations (8~) and (8e) display the adjustment of the loan rate to the equilibrium credit rationing solution. In this frame, RL, adjusts only to the supply- determined upper bound RLY and there is a permanent or equilibrium excess demand for loans.

3For a survey of these methods the reader is referred to Quandt (1982).

514


The equilibrium credit-rationing model consisting of 8, 8b, 8c, 8e, and 8f can be tested against the dynamic credit-rationing hypothesis,

RL, = FRL,-1 + (1 - /-l&z; + (1 - i.L)$,

with an approach suggested by Hwang (1980) for standard disequilibrium models. Let us define the sample separation variable:

Now (8) can be written as

RL, = Z,[pRL-l + (1 - p)Z;6 + (1 - &]

+ (1 - Z,)[P”RLI + (1 - $)-G.,Y

+ (1 - P”)edl ;

RL, = [Ztp + (1 - Z,)F.“]RL~-~ + X;,ZJr(l - p)

+ -Gm2(1 - I4 + 0 - ZtMl - 1171

+ ItO - I.45 + 0 - It)0 - CL%5 ;

RL, = pLt RL,-, + 2; b, + e, . (124

A comparison of (8) and (12a) shows that the dynamic credit-rationing hypothesis implies a stable regression relationship (coefficient and variance) of ZK, on RL,-, and Z,, whereas equilibrium credit rationing leads to instability of this very regression equation.4

In order to test the parameter stability of (6a) or (8), the methods suggested in the relevant literature can be used. In this paper, the cusum and cusum of squares test suggested by Brown, Durbin, and Evans (1975) is applied.5

Finally, a caveat remains to be mentioned. Instability of the

‘It is easily seen that equilibrium credit rationing implies instability of the regression of L, on RL, and Z,, as in the standard disequilibrium model.

‘The Monte Carlo results obtained by Hwang (1980) indicate that this test per- forms well in the context of disequilibrium models.

Peter Kugler

parameter of the loan-rate equation may arise from other reasons apart from equilibrium credit rationing. Changes in the targets and operating procedures of monetary policy may lead to parameter shifts of the loan demand and supply function. In addition, the adjustment process may be misspecified or relevant variables may be omitted from the loan demand and supply specifications used. These conditions will result in parameter instability of the loan-rate-adjustment equation, which is erroneously attributed to equilibrium credit rationing by the test proposed.

3. Empirical Results In this section, the results of the estimation of the disequilib-

rium interest-rate equation with quarterly data for the U.S., U.K., West Germany, and Switzerland are reported. The following specification of the equilibrium interest-rate equation is used:

RL: = a,, + S1 Z,-, + a2 RG, + 13~ RD,

+ &,Dp: + &,xt + 6,p, + 6,d, + Et. 03)

1 is the log of the nominal stock of bank loans, RD the interest rate on bank time deposits, RG is the yield of government securities, Dp: is the expected inflation rate, x is the log of real GDP or GNP, p is the log of the deflator of GDP or GNP, and d, is the log of the nominal stock of demand and time deposits. The coefficients &, &, s4, and a6 are expected to have a positive sign; 8, should be of negative sign; whereas the sign of a1 and 8s is indeterminate.

RD and RG are interest rates on alternative assets and liabil- ities affecting demand and supply considerations, and their presence in (13) is, therefore, immediately clear. 1,-i stems from stock- adjustment schemes for demand and supply. The sign of &, therefore, depends on the difference of the adjustment speed of demand and supply. Real GDP (GNP), the corresponding deflator, and expected inflation enter (13) separately because a real loan demand function is appropriate, whereas supply is to be formulated in nominal terms (Melitz and Pardue 1973). The positive sign of I& is brought about, therefore, by the partial effect of the expected inflation rate on real loan demand.6

‘The formulation of real loan demand and supply (2 - p) as function of real interest rates (RL - Dp”, RG - Dp”, RD - Dp”), real GDP (r), and real deposit

516


The indeterminate sign of 8s reflects the presumption that a rise in GDP (GNP) does not only increase the demand for loans, but may be perceived as a signal for lower loan risks and increase the supply of loans. The stock of bank deposits, D, is perceived as a pure supply-side variable, which has a positive influence on loan supply. Finally, it has to be mentioned that the estimation of an extended model brought no significant results for the change of X and P as additional variables.

Details on the data are given in the Appendix. Briefly, the loan rate, RL, is the average quarterly Federal Reserve Bank survey rate (U.S.), the base rate (U.K.), the rate charged for large bank loans (West Germany), and the interest rate on unsecured credits (Switzerland).7 The stock of commercial bank (domestic) loans, L,, contains commercial and industrial loans (U.S.), advances to industry (U.K.), short-term business loans (West Germany), and loans to private nonbanks-those without mortgages (Switzerland). RG is the treasury bill rate (U.S., U.K.) or the yield on long-term government bonds and RD is given by the interest rate on three-month, Euro-Swiss-Franc deposits for Switzerland, certificates of deposits (U.S., U.K.), and interbank deposits (West Germany). Dp” is obtained by fitting a fourth-order autoregression to the rate of change of the GDP (GNP) deflator. The data are seasonally adjusted, with the exception of the interest rates. The estimation period is 1970:i to 1983:iu (West Germany) and 1972:iii to 1983:iu (Switzerland). In the case of U.K., data from 1972:i to 1983:iu are used. This is motivated by the introduction of “Competition and Credit Control” in September 1971, which abandoned direct controls of bank lending. The availability of the loan series we use restricts our U.S. sample to 1973:ii-1983:iu.

The estimates received for a simple interest-rate equation with equal adjustment speeds (8), taking into account a first-order au- toregressive error term are presented in Table 1.’ The U.S. and U.K. estimates of the interest-rate adjustment parameter, TV, are very small and not significantly different from zero. Thus, the equilibrium hypothesis cannot be rejected in these cases. By contrast, the German and Swiss data result in p-estimates which are rather

stock (d - p) results in an interest-rate equation which is a restricted version of (13). The constraints are given by 6, = 1 - a2 - 8s and a6 = -8, - 6,.

‘In the case of Switzerland, no quarterly loan rate series is published by public authorities. Thus, a series kindly made available by Swiss Bank Corporation is used.

‘All calculations were done with the MIT-NBER software package TROLL.

517

TABL

E 1.

Es

timat

e of

th

e Di

sequ

ilibriu

m

inte

rest

Ra

te

Equa

tion

with

Eq

ual

Adju

stm

ent

Spee

d-s

RL.,

= PR

,L-~

+

(1 -

CL

) (6,

+ 61

Z,-1

+

&RD,

+

&RG,

+

6,Dp

: +

&xt

+ as

p,

+ 6,

d,)

U.S

. U.

K.

0.02

3 (0

.051

) W

est

Germ

any

0.60

(0

.039

) Sw

itzer

land

0.64

(0

.079

)

8.06

(4

.37)

0.

62

(0.9

2)

-3.6

4 (2

.28)

1.

21

co.9

4

1.93

(0

.31)

0.

48

(0.0

74)

0.46

(0

.16)

0.

25

(0.0

96)

-0.7

6 (0

.14)

0.

45

(0.0

67)

0.79

(0

.083

) 0.

019

(0.0

28)

-0.0

18

(0.0

13)

0.01

8 (0

.019

) -0

.012

(0

.12)

0.

024

(0.0

22)

13.8

(1

0.4)

12

.6

(2.1

9)

2.92

(8

.96)

0.

25

(2.3

3)

0.43

(3

.37)

1.

41

(0.4

8)

12.4

(1

1.8)

-0

.40

(1.8

3)

-13.

1 (6.4

8)

-2.8

5 (0

.98)

-3

.81

(8.4

8)

-0.6

7 (0

.56)

0.87

0.

88

0.42

0.

13

(0.1

8)

0.36

0.

99

-0.2

4 0.

32

(0.1

6)

0.30

0.

99

-0.3

1 0.

36

(0.1

5)

0.17

0.

90

0.50

0.

89

(0.2

1)

‘The

fig

ures

in

pare

nthe

ses

are

the

estim

ated

sta

ndar

d er

rors

. ‘E

stim

ated

re

sidua

l sta

ndar

d er

ror.

3Est

imat

ed

first

-ord

er

auto

regr

essiv

e co

effic

ient

of

the

re

sidua

ls.

4Dur

bin

h-st

atist

ic.


large and highly significant. The current quarterly adjustment of the loan rate to changes of its equilibrium rate is only 40% (West Ger- many) and 36% (Switzerland). The estimates of the reduced-form coefficients, 6i, of the equilibrium model have, in most cases, the expected sign. However, in the case of West Germany and Switz- erland many bi-estimates are not different from zero at conventional significance levels.

The two-stage probit estimates of the model with unequal adjustment speeds are given in Table 2. To correct for autocorrelation of the error term, a generalized difference transformation (1 - pL) is applied to data series using the p-estimate given in Table 1. For all countries, the estimated difference of excess supply and demand price-adjustment parameter (i.~~ - pi) is very small, and the hypothesis p2 = pi can be accepted at all reasonable significance levels. Thus, these results are in favor of the simple interest-rate equation, assuming equal adjustment speeds.

Now let us turn to the results of the cusum and cusum of squares stability test of the interest-rate equation with equal adjustment speeds.’ As outlined above, this procedure amounts to testing temporary (dynamic) credit rationing (parameter stability) against permanent supply-side equilibrium credit rationing (parameter instability). The instability result is obtained for Switzerland (cusum of squares) and West Germany (cusum and cusum of squares); however, in the cases of the U.S. and U.K., no instability of the regression equation is indicated. Indeed, the equilibrium hypothesis can even be accepted for these countries, as the p-estimate is not significantly different from zero. By contrast, our results indicate that permanent credit-rationing effects are important in the market of domestic bank loans in West Germany and Switzerland.

Finally, the stability of the regression of 1, on RL,, RL,-1, and all other predetermined variables is tested by the cusum and cusum of squares procedure. As mentioned in Section 2, this amounts to an alternative test of credit rationing, which can be used to check the empirical results obtained. A generalized difference transformation (1 - pL) is applied to the data series using the p-estimate given in Table 1. The results of this additional check can be summarized as follows: for the U.S., no instability is indicated; whereas in the U.K., Swiss, and German cases, the stability hypothesis can be rejected. Thus, the findings reported above are confirmed by

‘The generalized difference transformation with the p-value equal to the estimate given in Table 1 was applied to the U.K., West Germany, and Swiss data series.

519

TABL

E 2.

Es

timat

es

of

the

Dise

quilib

rium

In

tere

st-R

ate

Equa

tion

with

Un

equa

l Ad

just

men

t Sp

eeds

U.S.

0.

17

-0.3

6 8.

04

1.99

(0

.12)

’ (0

.23)

(4

.19)

(0

.31)

U.

K.

0.02

9 -0

.076

0.

69

0.48

(0

.081

) (0

.14)

(0

.W

(0.0

75)

Wes

t Ge

rman

y 0.

61

-0.0

27

-3.2

7 0.

46

(0.0

43)

(0.0

77)

(2.3

5)

(0.1

5)

Switz

erlan

d 0.

70

-0.2

2 3.

12

0.68

(0

.091

) (0

.28)

(2

.25)

(0

.18)

-0.7

7 (0

.14)

0.

45

(0.0

65)

0.78

(0

.083

)

-0.0

11

(0.0

12)

0.01

6 (0

.019

) -0

.007

5 (0

.11)

0.

057

(0.0

51)

12.4

(1

0.6)

12

.5

(2.1

9)

0.85

(1

0.0)

-0

.20

(5.3

3)

0.50

(3

.81)

1.

45

(0.4

9)

11.1

(1

1.8)

-0

.85

(4.1

1)

-13.

2 0.

86

0.87

0.

42

0.01

05

.54)

-2

.97

0.36

0.

99

-0.2

4 0.

17

(0.9

6)

-2.6

9 0.

30

0.99

-0

.31

0.44

@

.W

-1.7

4 0.

17

0.87

0.

50

0.63

(1

.28)

‘The

fig

ures

in

pare

nthe

ses

are

the

estim

ated

sta

ndar

d er

rors

. *E

stim

ated

re

sidua

l sta

ndar

d er

ror.

3Firs

t-ord

er

auto

regr

essiv

e co

effic

ient

of

the

res

idua

ls.

‘Dur

bin

h-st

atist

ic.


this additional test for the U.S., West Germany, and Switzerland. By contrast, for the U.K. data, conflicting results are reported for the two approaches: the price equation estimates indicate no (temporary and equilib rium) credit rationing, while the cusum of squares test for the corresponding volume equation points to at least temporary credit rationing. It suggests that a nongeometric adjustment pattern, administered pricing effects, and/or the presence of RL,-, in the loan demand and/or supply function substantially biase the p-estimates in the U.K. case. Therefore, we have to dismiss our conclusion of no temporary credit rationing in the U.K. Neverthe- less, the stability of the loan-rate equation points against substantial equilibrium credit-rationing effects during the sample period.

4. Summary and Conclusions In this paper, the empirical significance of credit rationing was

analyzed using quarterly data on domestic bank loans of four countries. First, models of temporary (dynamic) credit rationing were considered. The price (interest rate) equation proposed by Bowden was estimated assuming equal and unequal adjustment speeds under excess demand and supply conditions. Second, the stability of the interest-rate equation was tested by the cusum and cusum of squares procedure. We motivated this test by the fact that permanent (supply-side equilibrium) credit rationing implies instability of this regression relationship. Third, the alternative test for disequilibrium proposed by Hwang (1980) is carried out as an additional check.

The main results can be summarized as follows. The estimates of the Bowden interest-rate equation point to no temporary credit rationing in the U.S. and the U.K. By contrast, substantial credit- rationing effects were found for West Germany and Switzerland. The hypothesis of equal interest-rate adjustment speeds, however, could be accepted for all four countries. The application of the cusum and cusum of squares tests indicated that the German and Swiss interest-rate equations are not stable over time, whereas the stability hypothesis could be accepted in the cases of the U.S. and the U.K. The Hwang test confirmed the interest-rate equation result for all countries except the U.K. In this case, this test points to (temporary) disequilibrium indicating that the interest-rate equation results are biased.

Our results suggest the following, conclusions. First, the de- gree of temporary credit rationing seems to be different on the mar-

521

Peter Kugler

ket for bank loans across the countries considered. The U.S. loan rate appears to adjust rather quickly to changing market conditions; whereas for the U.K., Swiss, and German loan rates, this is not true. In other words, disequilibria on the market for business loans seem to persist in West Germany, the U.K., and Switzerland; whereas competition among banks leads to quick equilibrium adjustment in the U.S. loan market. Second, the instability of the loan-rate equation for West Germany and Switzerland points to the occurrence of permanent (supply-side equilibrium) credit rationing during the sample period in these countries. The stability of the U.S. and U.K. loan rate equation indicates no disequilibrium of this kind during the sample time span. Of course, instability of the Ger- man and Swiss loan-rate equation may also be caused by parameter shifts or other misspecifications of the equilibrium model and the loan-rate adjustment process. All four countries considered experi- enced strong changes in the conduct of monetary policy during the sample period. Thus, these common shifts offer no convincing ex- planation for our results differing across the countries. However, the different institutional characteristics of the banking systems in the countries considered suggest a caveat which deserves further investigation. By contrast to the highly specialized U.K. and U.S. banks, German and Swiss banks have, as a rule, a wide range of activities. The loan supply of such “universal” banks depends on the volumes and interest rates of a wide range of assets and lia- bilities. Therefore, misspecifications of loan supply and, as a con- sequence, the loan rate equation are more likely for Switzerland and Germany than for the U.K. and the U.S.

Finally, our results with respect to temporary credit rationing in the U.S. confirm the findings obtained by Ito and Ueda (1981) with other data, models, and methods. For the other countries and with respect to supply-side equilibrium credit rationing, no com- parable results are published. Thus, it would be interesting to see whether the results obtained in this paper are robust with respect to the use of different models, data, and methods.

Receioed: August 1986 Final version: July 1987

References Artus, P. “Le fonctionnement du march6 du credit: diverses anal-

yses dans un cadre de desequilibre.” Revue bconomique 4 (July 1984): 591-621.

522


Baltensperger, E., and T.M. Devinney. “Credit Rationing Theory: A Survey and Synthesis.” Zeitschrift fur die gesamten Staatswis- senschafien 141 (1985): 475-502.

Bowden, R. J. “Specification, Estimation and Inference for Models of Market in Disequilibrium.” lnternational Economic Review 19 (October 1978a): 711-26.

-. The Econometrics of Disequilibrium. Amsterdam: North- Holland, 1978b.

Brown, R.L., J. Durbin, and J. M. Evans. “Techniques for Testing the Constancy of Regressions Over Time.” Journal of the Royal Statistical Society, Series B 37 (1975): 149-92.

Devinney, T.M. “A General Equilibrium Analysis of the Borrower- Lender Relationship: An Examination of the Credit Rationing Hypothesis.” Ph.D. diss., University of Chicago, 1983.

Freimer, M., and M. J. Gordon. “Why Bankers Ration Credit.” Quarterly Journal of Economics 79 (August 1965): 379-416.

Harris, D. “Credit Rationing at Commercial Banks: Some Empirical Evidence.” Journal of Money, Credit and Banking 6 (May 1974): 227-40.

Harvey, A.C. The Econometric Analysis of Time Series. London: Philip Allan, 1981.

Hwang, H. S. “A Test of a Disequilibrium Model.” Journal of Econometrics 12 (1980): 319-33.

Ito, T., and K. Ueda. “Tests of the Equilibrium Hypothesis in Dis- equilibrium Econometrics: An International Comparison of Credit Rationing.” International Economic Review 22 (1981): 691-708.

Jaffee, D.M. Credit Rationing and the Commercial Loan Market. New York: John Wiley, 1972.

Jaffee, D.M., and F. Modigliani. “A Theory of Test of Credit Ra- tioning.” American Econometric Review 59 (December 1969): 850- 72.

Keeton, W. Equilibrium Credit Rationing. New York: Garland, 1979. Kugler, P. “Ungleichgewichtsoekonometrie f&r den schweizerischen

Hypothekarzinssatz.” Schweizerische Zeitschrifi fiir Statistik und Volkswirtschaft 121 (1985).

Laffont, J. J., and R. Garcia. “Disequilibrium Econometrics for Business Loans. ” Econometrica 45 (July 1977): 1187-204.

Maddala, G. S. Limited-Dependent and Qualitative Variables in Econometrics. Cambridge, MA: Cambridge University Press, 1983.

Melitz, J., and M. Pardue. “The Demand and Supply of Commer- cial-Bank Loans.” Journal of Money, Credit, and Banking 5 (1973), 669-92.

523

Peter Kugler

Quandt, R.E. “Econometric Disequilibrium Models.” Econometric Reviews 1 (1982): l-63.

Sealy, G.W., Jr. “Credit Rationing in the Commercial Loan Mar- ket: Estimates of a Structural Model Under Conditions of Dis- equilibrium.” Journal of Finance 34 (June 1979): 689-702.

Stiglitz, J.E., and A. Weiss. “Credit Rationing in Markets with Im- perfect Information.” American Economic Review 71 (1981): 393- 410.

-. “Incentive Effects of Termination.” American Economic Re- view 73 (December 1983): 912-27.

Appendix. Data Interest rates, loan and deposits stocks are end-of-quarter fig-

ures. All data series are seasonally adjusted except interest-rate series.

U.S. RL: Quarterly loan rate survey average. (Source: Federal Reserve

Bulletin) RD: Certificates of deposits rate. (Source: Survey of Current Busi-

ness) RG: Three-month treasury bill rate. (Source: IMF-IFS) L: Commercial and industrial loans in bn. $. (Source: Federal Re-

serve Bank of St. Louis) D: Demand, time, and savings deposits in bn. $. (Source: IMF-

IFS) x: GNP in bn. 1972 $. (Source: Survey of Current Business) P: GNP deflator. (Source: Survey of Current Business)

U.K. RL: Base rate (London clearing banks). (Source: Economic Trends) RD: Certificates of deposits rate (three months). (Source: Economic

Trends) RG: Treasury bill rate (three months). (Source: IMF-IFS) L: Advances to industry in bn. 2. (Source: OECD-MEI) D: Demand, time, and savings deposits in bn. 2. (Source: IMF-

IFS) X: GDP in 1980 2. (Source: Macroeconomic Data Bank of the

CSO) P: GDP deflator. (Source: Macroeconomic Data Bank of the CSO)

524


Germany RL: Bate on business loans of 1 to 5 bn. DM. (Source: Monats-

berichte der Deutschen Bundesbank) RD: Three-month money market rate. (Source: Monatsberichte der

Deutschen Bundesbank) RG: Government bond yield. (Source: IMF-IFS) L: Short-term business loan in bn. DM. (Source: Monatsberichte

der Deutschen Bundesbank) D: Demand, time, and savings deposits in bn. DM. (Source: IMF-

IFS) X: GNP in bn. 1970 DM. (Source: Vierteljahrliche Volkswirt-

schaftliche Gesamtrechnung des Deutschen Institutes fiir Wirtschaftsforschung (DZW))

P: GNP deflator. (Source: Vierteljahrliche Volkswirtschaftliche Gesamtrechnung des Deutschen Institutes fur Wirtschaftsfor- schung (DZW))

Switzerland RL: Bate on unsecured credits including commissions. (Source: Swiss

Bank Corporation, unpublished) RD: Three-month Euro-Swiss-Franc rate. (Source: Monatsberichte

der Schweizerischen Nationalbank) RG: Government bond yield. (Source: IMF-IFS) L: Domestic (private sector) bank credits (without commercial bills

and mortgages). (Source: Monatsberichte der Schweizerischen Nationalbank)

D: Demand, time, and savings deposits in bn. 3%. (Source: IMF- IFS)

x: GDP in bn. 1970 Sfr. (Source: Vierteljahrliche Nationale Buchhaltung der Basler Arbeitsgruppe fiir Konjunkturfor- schung (BAK))

P: GDP deflator. (Source: Vierteljahrliche Nationale Buchhaltung der Basler Arbeitsgruppe fiir Konjunkturforschung (BAK))

525

Documents

Credit rationing and the adjustment of the loan rate: An empirical investigation