Asymmetric adjustment of commercial bank interest rates: evidence from Malaysia and Singapore pdf

Trang 1

S0261-5606(96)00016-2 0261-5606/96 $15.00 + 0.00

Asymmetric adjustment of commercial bank interest rates: evidence from

Malaysia and Singapore

B A R R Y S C H O L N I C K *

Faculty of Business, University of Alberta, Edmonton T6G 2R6,

Canada

This paper examines the rigidity of commercial bank interest rates, using

evidence from Singapore and Malaysia An asymmetric error correction

technique is used to test whether mean adjustment lags are different

when retail rates are above or below their equilibrium levels It is

concluded, in both countries, that the mean adjustment lag is shorter

when the deposit rate is above its equilibrium than when it is below its

equilibrium Using the framework of Hannan and Berger (1991), this

finding implies that the hypothesis of collusion cannot be rejected (JEL

This paper examines the rigidity of commercial bank (retail) interest rates in Singapore and Malaysia While the rigidity of commercial bank interest rates has received some attention within the context of developed countries (e.g Hannan and Berger, 1991; Neumark and Sharpe, 1992) it has received much less attention within the context of developing economies This is unfortunate given that the relative rigidity of commercial bank rates has important implica- tions for programs of financial liberalization The standard policy proposition from the financial liberalization literature (Fry, 1988) is that administered lending and deposit rates result in the misallocation of credit and thus the distortion of investment decisions A point that it is often ignored in the financial liberalization literature, however, is that even after state controls on retail interest rates have been lifted, interest rate rigidity may still persist because of the behavior of the banks

Most of the existing literature on interest rate rigidity and bank monopoly power uses cross-sectional data to examine issues of market concentration in different geographical markets (Gilbert, 1984) This type of data, however, is typically unavailable for developing countries The aim of this paper is to

* Very helpful comments were received from an anonymous referee, as well as from Maxwell

J Fry, Kate Phylaktis, Paul Boothe, Kanhaya Gupta, Rolf Mirus and participants at the 1994 meeting of the Canadian Economics Association in Calgary

Trang 2

Asymmetric adjustment of commercial bank interest rates: B Scholnick

introduce a different approach, using only the time-series data that are available It examines interest rate rigidity using the cointegration and error- correction methodologies The main innovation in the paper is the use of an asymmetric error-correction methodology, which makes it possible to examine differences when interest rates are above or below their equilibrium levels As

in Neumark and Sharpe (1992) and Hannan and Berger (1991), this paper examines how retail (lending and deposit) rates respond to changes in the wholesale rate (e.g the interbank rate), which is treated as exogenous How- ever, it addresses the important question of how long it takes for retail rates to respond to changes in the wholesale rate under situations when they are either above or below their equilibrium levels

An examination of interest rate rigidity in Singapore and Malaysia is particularly important in light of the interest expressed in the liberalization experi- ences of Asia-Pacific economies This interest has culminated in the widely distributed World Bank study 'The East Asian Miracle' (1993) This study argues that policies concerning the liberalization and repression of interest rates have played a very important role in the economic growth of the Pacific Rim economies Given the importance ascribed by the World Bank to the problems and benefits of financial liberalization in the Asia-Pacific, an under- standing of retail rate rigidities after liberalization is clearly of interest

I Retail interest rates in Singapore and Malaysia

Both Singapore and Malaysia began liberalizing their banking systems in the late 1970s Before that, all retail interest rates were set by a banking cartel In the 1980s and 1990s Singapore developed a highly sophisticated and internatio- nally competitive banking system The last exchange control regulations were dropped in 1978 (World Bank, 1993) The aim of the authorities was to develop

an international financial centre to rival Hong Kong This led to a large increase in the number of foreign banks in the financial system Singaporeans have no restrictions placed on them in terms of using domestic or foreign banks, thus ensuring a highly competitive system (Claassen, 1992) with a low degree of industry concentration

The banking system in Malaysia is significantly more concentrated than that

in Singapore (Tseng and Corker, 1991) For example, in 1982 the top three banks held 44 percent of deposits (Cho and Khatkhate, 1989) although this figure has been declining slowly in the 1980s Also, during the 1980s the central bank has argued that Malaysia is 'overbanked', thus it has sought to discourage the establishment of new banks, both domestic and foreign (Marashdeh, 1994) Furthermore, while the central bank in Malaysia has liberalized the market for deposits, and allowed the banks to determine their own deposit rates, this has not been the case with the lending rate In 1983 the central bank introduced a Base Lending Rate (BLR) system, whereby they set the 'floor' for commercial bank lending rates Thus the lending rate in Malaysia is substan- tially different from the other retail rates being analyzed in this paper, in that it

is largely determined not by the commercial banks but rather by the central authorities The market for loans in Malaysia was subject to further regulation

Trang 3

regarding the inflow of foreign loans Claassen (1992) notes that restrictions were imposed on Malaysians wishing to bring loans acquired abroad into the country No such restrictions were placed on Malaysians depositing funds abroad This is an attempt by the authorities to set the floor for lending rates, without cheaper loans being imported from abroad

II Conceptual framework

This paper uses the basic framework presented by Hannan and Berger (1991), who differentiate between different types of deposit rate rigidities This framework is similar to that used by Neumark and Sharp (1992) This paper extends the Hannan and Berger (1991) paper, however, by examining both deposit rates

as well as lending rates The first Hannan and Berger hypothesis concerns symmetric relationships, while the second set concerns asymmetric relationships

The symmetric Hannan and Berger hypothesis concerns differences between banking firms in different market areas This is stated by them as follows: 'to the extent that firms in more concentrated markets exhibit higher price conjectures as a result of greater recognised interdependence, operation in a more concentrated market implies greater price rigidity' (1991, p 940) Their argument is that the greater the degree of interdependence between banks, because of the high market concentration, the greater the reluctance to adjust deposit rates after exogenous changes to the wholesale rate This is because of the perception that deposit rate changes may lead to adverse reactions from the other banks which may affect their own supply of deposits

On the other hand, banks in highly competitive markets will be less concerned about the perceived behavior of other banks, which implies lower interest rate rigidity in response to exogenous shocks

An important problem with the hypothesis as defined above, however, is that

it is not clear that the existence of market concentration should lead to

symmetrical rigidity when banks are raising or lowering their interest rates Clearly, banks that are able to engage in collusive relationships with their competitors will have a tendency to rapidly lower deposit rates, but will be relatively sluggish when raising them It is thus argued here that results from tests of symmetric relationships are significantly less powerful than tests of asymmetric relationships Symmetric tests, however, are able to provide some evidence on issues where asymmetries may be less important, such as a comparison between rigidity of rates controlled by the authorities (e.g the BLR in Malaysia) and market determined rates

Hannan and Berger (1991) develop a further set of hypotheses which deal with asymmetric rigidity They examine whether interest rate rigidity is different when rates are increasing or decreasing In their model of deposit rates they propose two competing hypotheses On the one hand, they argue that greater rigidity in deposit rate increases could arise from the collusive pricing behavior of banks Such arrangements may break down if prices are changed, thus banks consider the cost of breakdown before adjusting prices The expected costs are higher for deposit rate increases relative to decreases,

Trang 4

because of increasing payments to depositors Thus, deposit rates will be relatively more rigid when they are increasing On the other hand, Hannan and Berger also propose an alternative argument for greater rigidity in deposit rate

decreases If banks perceive that an important cost to them of changing the deposit rate comes from the negative reaction of customers, then they will be reluctant to decrease deposit rates hence greater downward rigidity These two competing arguments can be called the 'collusive' and the 'customer reaction' hypotheses

It is possible to extend the work of Hannan and Berger (1991) by relating this discussion to lending rates In this case the collusive hypothesis would be supported by evidence of greater rigidity in lending rate decreases as banks would expect higher costs from the breakdown of collusive arrangements if lending rates were lowered Alternatively, evidence of rigidity in lending rate

increases would support the adverse customer reaction hypothesis

llI Testing methodology

As in the paper by Hannan and Berger (1991), this paper will examine the relative rigidity of lending and deposit rates in response to shocks to the wholesale interest rate, which is assumed to be exogenous The money market

or call interest rate can be thought of as the marginal cost of funds for banks

or as the wholesale interest rate This paper differs from the Hannan and Berger paper in that it adopts a different econometric methodology It analyzes time series data using the well known cointegration and error-correction mechanism approach Neumark and Sharpe (1992) note that the short-run variability of the wholesale-retail rate spread is a well known phenomenon, as

is the fact that retail rates will often be determined by averaged wholesale rates over some time horizon In other words, it could be expected that the short-run relationship between wholesale and retail rates is relatively volatile, while the long-run relationship is relatively stable This dichotomy between short-run variability and long-run stability has a very close analogy in the dichotomy between the short-run error correction model, and the existence of long-run cointegrating vectors

The cointegration methodology has been extensively used, thus a full de- scription will not be given here Once it has been determined that all the data series are integrated of order one I(1), the Johansen technique is used to test whether cointegrating vectors exist between the different retail rates and the wholesale rate These can be thought of as long-run relationships between the different variables This technique uses maximum-likelihood procedures to determine the number of cointegrating vectors among a vector of time series

If it is assumed that the vector of n I(1) time series, Yt, can be written as a VAR,

( 1 ) Yt = 1-IlYt-1 + IIzyt-2 + " ' I I k y t - k d- et,

then the vector can be reparameterized as

q

(2) A y t = I~ + IIyt_g + v~ FiAYt_ i + ut '

i=l

Trang 5

where q = k - 1, II = E{= 1Bi - 1, Bj is an (n x n) matrix from the lags of the VAR, and F = Ek=i+~Bj for i = 1 q

The rank r of the matrix II in (2) determines the number of cointegrating vectors in the VAR If r = 0 then there are no cointegrating vectors The Johansen procedure is designed to statistically determine the number of cointegrating vectors r in the VAR The complete testing procedure is reported in Johansen (1988) Johansen provides two different likelihood ratio tests to determine the value of r These are the trace test, with a test statistic

n

i = r + l and the maximum eigenvalue test

where A are eigenvalues from II As is usual in this literature, both statistics are reported here Tests are conducted both under the null that r = 0, and then that r - 1

Once the cointegrating vectors have been determined, the usual procedure is

to implement an Error Correction model to examine short-run dynamics In its simplest form this can be written

( 5 ) A i t = 8 0 + 81 mw t + 8 2 R t _ 1 + et,

where i is the endogenous retail rate, w is the exogenous wholesale rate and R are the residuals from the cointegrating vector between the retail and the wholesale rates It can be shown (Doornik and Hendry, 1994) that the mean lag for this simple model is

This model is thus used to determine the adjustment lag for the symmetric model The drawback with this specification is that it assumes that adjustment

is symmetric when the retail rate is above or below its equilibrium level

IV Asymmetric adjustment

This paper examines asymmetric adjustment processes, similar to those used by Neumark and Sharpe (1992) However, it differs in that the cointegration rather than the partial adjustment methodology is used The equilibrium relationship between the retail and wholesale rate is defined simply as the mean of the residual series from the appropriate cointegrating equation It is well known that the cointegrating residuals are by definition I(0), which implies that the series is mean reverting Thus if residuals are above their mean they will tend to move back towards the mean (or equilibrium relationship), and vice versa for residuals below their mean Because of the structure of the cointegrating relationships, when a residual is above its mean, this can be interpreted

as the retail rate being above its equilibrium level with respect to the wholesale rate This notion of an equilibrium relationship follows directly from the definition of cointegration

Trang 6

The procedure used to examine asymmetric adjustment is to divide the single series of residuals (R) from the cointegrating equations into two separate series R + and R - , where

R + - - R , if R > / z

and

and /z is the mean of R The asymmetric specification in (7) and (8) can be thought of as a dummy variable procedure to split R into two series It does not affect the estimation of either R or the cointegrating vector Recall that the only restriction on R in terms of the cointegration methodology is that it

be stationary or I(0)

An asymmetric short-run dynamic equation can thus be written as

where the asymmetric mean lags are defined as

and

Thus, in the same way that it is possible to derive an estimate of the mean adjustment lag from the coefficient of the symmetric residual series R, it is also possible to estimate two mean adjustment lags, one when the series is below its mean, and one when it is above it Once an equation such as (9) is estimated,

it is possible to use a Wald test with a X2(1) distribution to test the restriction that (10) is equal to (11), i.e that there is no asymmetry

V Results

All data are monthly, and are taken from the International Financial Statistics

CD-ROM, published by the IMF Tests are run from 1983:1 to 1992:11 for Malaysia, and from 1983:1 to 1994:4 for Singapore, the end dates being determined by data availability Lending and deposit rates are short-term rates

taken from lines 60p and 601 in IFS Wholesale rates are short-term money market or call money rates taken from line 60b in IFS

Initial test of the stationarity of the series are presented in Table 1 In all cases it can be concluded that the interest series are I(1)

Once it has been determined that the series are not stationary, it is possible

to test whether the different retail rates cointegrate with the wholesale rates in each country The Johansen cointegration tests, (3) and (4), are presented in Table 2 The hypothesis of at most one cointegrating vector ( H 0 : r < = 1) cannot be rejected in any of the tests The hypothesis of no cointegrating vectors ( H 0 : r = 0) can be rejected in all cases This finding is supported by both the eigenvalue as well as the trace tests Given the graphs describing the data in both countries, these results are not particularly surprising In both

Trang 7

1 8 -

16

14

12

10

8

6

4

2

0

1978

1980 1982 1984 1986 1988 1990 1992

M o n e y m a r k e t r a t e T i m e d e p o s i t r a t e - - B a s e l e n d i n g r a t e

FIGURE 1 Malaysia: interest rates

1 6 - -

14

12

10

8

6

M o n e y m a r k e t r a t e T i m e d e p o s i t r a t e - - L e n d i n g r a t e

FIGURE 2 Singapore: interest rates

Trang 8

TABLE 1 ADF unit root tests Singapore Malaysia Level Difference Level Difference Lending rate - 1.71 -3.15 - 1.52 -3.60

Deposit rate - 1.63 - 3.10 - 2.07 - 3.29

Wholesale rate - 1.45 - 4.21 - 1.70 - 4.55

Note: The ADF statistic is based on 12 lags and is calculated without a

trend The significance value is -2.87

S i n g a p o r e a n d Malaysia the retail rates clearly s e e m to have a long-run relationship with the wholesale rates O v e r time the rates t e n d to m o v e together V e r y similar results have also b e e n f o u n d by D i e b o l d and S h a r p e (1990) w h o c o n c l u d e that wholesale and retail interest rates in the U S A f o r m cointegrating vectors T h e intuitive i n t e r p r e t a t i o n o f these results is that o v e r time, in the long run, c o m m e r c i a l b a n k retail rates are largely d e t e r m i n e d by the marginal cost o f funds that banks face, i.e the wholesale rate

T h e o n e n o t e w o r t h y finding f r o m these c o i n t e g r a t i o n tests c o n c e r n s the cointegrating v e c t o r that is f o u n d b e t w e e n the wholesale rate and the lending rate in Malaysia Recall that this lending r a t e is not d e t e r m i n e d by c o m m e r c i a l banks, but r a t h e r is the base lending rate set by the m o n e t a r y authorities This finding is very useful w h e n deciding w h e t h e r the B L R policy was actively used

to repress the financial system Financial repression is usually t h o u g h t o f as actions t a k e n by the m o n e t a r y authorities to k e e p retail interest rates away

f r o m t h e i r m a r k e t clearing level T h e conclusions f r o m these results are that in the long r u n at least, the base lending rate did not diverge f r o m the m a r k e t

TABLE 2 Johansen cointegration tests

Cointegrating vector

Eigenvalue test Trace test

H 0 : r = 0 H 0 : r < = l H 0 : r = 0 H 0 : r < = l Panel A: Malaysia

deposit wholesale 18.98 2.37 21.35 2.37 lending wholesale 25.42 3.71 29.13 3.71 Panel B: Singapore

deposit wholesale 20.77 2.69 23.46 2.69 lending wholesale 27.22 2.72 29.94 2.72

The 5% critical values for the eigenvalue test are 14.09 (for Null r = 0) and 3.76 (for Null

r < 1)

The 5% critical values for the trace test are 15.41 (for Null r = 0) and 3.76 (for Null

r < = 1)

Trang 9

Asymmetric adjustment of commercial bank interest rates: B Scholnick

determined wholesale and deposit rates (a cointegrating vector can also be found between the lending and deposit rates) Thus the hypothesis of the BLR policy causing long-run or sustained financial repression can be rejected While the cointegration results can provide evidence on long-run behavior, it

is necessary to examine dynamic error correction equations to test hypotheses

of short-run behavior Equations of the form of (5) and (9) are reported in Tables 3 and 4 These results are used to derive the mean adjustment lags (6), (10) and (11) which are reported in Table 5

While an analysis of the symmetric models does not provide the richness of detail to be found in the asymmetric model, they do provide some interesting insights The most important point is that the lending rate in Malaysia has a mean lag that is considerably slower (11.4 months) than any of the other rates

in the study in both countries (between 6.2 and 7.5 months) This finding clearly indicates the effect of the base lending rate policy followed by the monetary authorities in Malaysia Thus while the existence of cointegration indicates that authorities may have considered the relationship between BLR and the market determined wholesale rate in the long run, there are clearly substantial short-run lags before the BLR is adjusted

The first stage in the examination of the asymmetric models is a Wald test to test the restriction that the two asymmetric mean lag estimates are equal, i.e that there is no asymmetry In all cases this restriction is rejected, except the case of the lending rate in Singapore where the X 2 statistic of 0.6 is below the critical value The rejection of the restriction implies that asymmetries exist between adjustment when the retail rates are above their equilibrium, when

Independent variables

Dependent variables

Constant

above mean

below mean

R 2

D W

S C o r r e l a t i o n F(12, 103)

0.23 1.75 17.6 0.86

- 0.30 ( - 3 9 )

- 0 1 2 ( - 4.6) 0.26 1.79 13.7 0.90

0.25 2.6 19.9 2.33

0.36 (3.5) 0.11 (5.6)

- 0.09 ( - 4 4 )

- 0 1 4 (3.0) 0.27 2.6 14.2 2.56

Note: t-statistic in parentheses

Trang 10

TAaLE 4 Singapore: Short-run dynamic equations

Independent Variables

Dependent variables

A deposit r A deposit r A lending r A lending r Constant

A wholesale rate

Symmetric ECM

Asymmetric ECM

above mean

below mean

- 0 1 0 -0.11 0.30 ( - 6 4 ) ( - 6 5 ) (6.8)

- 0 1 8 ( - 5 2 ) -0.11 ( - 7 6 )

0.25 (4.1) 0.11 (4.0)

- 0.10 ( - 5.7)

- 0.08 ( - 3.0)

F-statistic (F(2, 133) 44.8 30.7 39.5 26.8

S Correlation F(12, 121) 1.55 1.14 1.71 1.48

Note: t-statistics in parentheses

TABLE 5 Mean adjustment lags

Symmetric model months

Asymmetric model Wald test ML + M L - (no as't ~'mmet~z) menths months Panel A: Malaysia

Deposit rate

Lending rate

Panel B: Singapore

Deposit rate

Lending rate

Notes: t-statistics in parentheses

Wald test of the restriction of no asymmetry 5% critical value for X2(1) distribution = 3.84

they will tend to decline, compared to when they are below their equilibrium, when they will tend to rise

The conclusions from the market for deposits in the two economies are very similar Both cases result in significant asymmetries, and in both cases the

Định dạng
Số trang	12
Dung lượng	689,84 KB