Defined Terms and Documents       

University of Wollongong  -  Research Online  Faculty of Business - Economics Working Papers Faculty of Business  -  January 2012

How to capture the full extent of price stickiness in credit card interest rates?

Abbas Valadkhani  -  University of Wollongong, abbas@uow.edu.au

Sajid Anwar  -  University of the Sunshine Coast

Amir Arjomandi  -  University of Wollongong, amira@uow.edu.au

Abstract: We present a new approach to evaluate the full extent of Price Stickiness in Credit Card interest rates by modifying the existing asymmetric models so that they can be adopted for testing both the amount and adjustment asymmetries as well as the lagged dynamic inertia. Consistent with similar studies, banks behave asymmetrically in response to changes in the Reserve Bank of Australia’s (RBA) target interest rate. Rate rises are passed onto the consumer faster than rate cuts and the credit card interest rate showed a very significant degree of downward rigidity. Based on the magnitude of the pass-through parameters obtained from short-run dynamic models, rate rises had a full one-to-one and instantaneous impact on credit card interest rates. However, in absolute terms the short-run effects of rate cuts were not only less than half of the rate rises but also were delayed on average by three months.

‐2‐

Analysing credit card interest rates has been one of the growing strands of research in

the banking and finance literature. Credit card suppliers generally operate within a

competitive market with many financial institutions offering their credit cards with limited

barriers to entry. However, in such a market it is puzzling as to why credit card interest rates

(compared to other lending rates) are so high and so sticky downwards, while banks enjoy

high profits at the same time. There are a large number of recent studies that have analysed

the interest rate pass-through of the central bank’s policy rate to various types of deposit and

Payne, 2006, 2007; De Graeve, De Jonghe and Vennet, 2007; Liu, Margaritis and Tourani-

Rad, 2008; Payne and Waters, 2008; Chong, 2010; De Haan and Sterken, 2011). However,

banks’ asymmetric behaviour in setting the credit card interest rate has received considerably

less attention in the literature, especially in the context of the Australian banking industry.

Previous studies in other countries have found that credit card interest rates are relatively

higher than other lending rates and show a very significant degree of inertia and downward

rigidity. This has resulted in the rising spread between credit card rates and other lending

rates. All major explanations for the asymmetric behaviour of credit card interest rates in the

literature can be grouped into six categories: (1) search costs, switching costs, and adverse

selection (Ausubel, 1991; Lowe and Rohling,1992; Lowe, 1995; Calem and Mester, 1995;

Stango, 2002; Berlin and Mester, 2004; Calem, Gordy and Mester, 2006; Agarwal,

Chomsisengphet and Liu, 2010); (2) rational or irrational borrowers (Brito and Hartley, 1995;

Della Vigna and Malmendier, 2004); (3) fixed and variable interest rates (Stango, 2000); (4)

tacit collusion between banks (Knittel and Stango, 2003); (5) option value of credit lines

(Park, 2004); (6) credit card penalty fees and risk (Massoud, Saunders and Scholnick, 2007).

‐3‐

Providing a comprehensive review of the existing literature is beyond the scope of this paper.

For a concise account of these studies see Scholnick et al. (2008).

Ausubel (1991), Lowe and Rohling (1992), Lowe (1995) and Calem and Mester (1995)

can be considered as four pioneering studies posing a combination of three explanations for

this puzzle, namely: search costs, switching costs, and adverse selection. In their views, to

change credit card providers, consumers may incur two possible types of costs: search costs

(associated with finding the information on alternative providers) and switching costs, (the

expenses incurring in the process of changing from one provider to another). High risk

borrowers are usually more inclined to search for lower interest rate cards because they think

it is more likely that they may use that debt in the future (Ausubel, 1991). Due to this likely

phenomenon, financial institutions hesitate to reduce their credit card interest rates because

this action may in fact encourage high risk borrowers to apply for their products. According

to Calem, Gordy and Mester (2005), borrowers who possess higher credit card balances also

have higher search costs. Thus, if a bank unilaterally reduces its credit card interest rate, it

will be appealing to mainly less profitable borrowers who have lower balances. Both Ausubel

(1991) and Calem and Mester (1995) therefore conclude that banks have no incentive to

unilaterally lower their interest rate.

In a competitive equilibrium and based on completely rational individuals, Brito and

Hartley (1995) suggested an alternative explanation to that of Ausubel (1991), Lowe and

Rohling (1992) and Calem and Mester (1995). In their view, for two reasons borrowers

continue to use credit card debt even if its interest rate is substantially higher than other

alternative forms of borrowing: (a) transactions costs associated with the use of non-credit

card sources for short periods of time are usually high; (b) borrowers try to smooth their

income and consumption streams through time. Brito and Hartley (1995) also provide an

explanation for the rising spread between credit card rates and other lending rates. Unlike

‐4‐

other bank loans, credit cards enable the flexibility to borrow more even if consumption

exceeds income, especially when other loans are more expensive to set up or the loan periods

are either relatively short or unpredictable. However, if the spread between credit card and

other forms of finance becomes significantly wider, the credit card benefits will diminish and

eventually low-risk borrowers, who have access to other loans, will stop using credit card

debt.

Credit card providers usually offer two types of interest rate: fixed rates which do not

change for long periods and variable rates which move concurrently with market interest

rates. Stango (2000) believes that such a pricing structure influences the competitive structure

of the market. Adopting a game theory approach, Stango (2000) demonstrates that the size of

firms can also exert a strong influence on the pricing structure chosen (i.e. fixed or variable

rates). Knittel and Stango (2003) offer another alternative explanation for credit card

stickiness based on the possibility of tacit collusion between banks. Numerous States in the

US introduced the State level price ceilings based on Usury Laws in the 1980s while in other

States there were no binding ceilings. Knittel and Stango (2003, p.1471) argue that “credit

card providers resort to tacit collusion by using the interest rate level of the binding ceiling in

some states as the focal point for interest rates in those states that did not have binding

ceilings”. The main conclusion of a comprehensive review of the literature by Scholnick et al.

(2008, p.1469) is “that while a large amount of research has already been undertaken on

credit cards, debit cards and ATMs, we believe that there are still a great many issues and

puzzles that remain to be resolved”.

In the contemporary literature the asymmetric behaviour of credit card interest rates

could also be referred to as “rockets-and-feathers hypothesis”. In a similar context, Bacon

(1991) argued that gasoline prices “shoot up like rockets” in response to a positive rise in oil

prices and “float down like feathers” in response to a fall. Hannan and Berger (1991) and

‐5‐

Neumark and Sharpe (1992) are among earlier studies that have tested the rockets and

feathers hypothesis in the context of the banking industry. Their thorough investigation

suggests that, as compared to negative shocks, consumer deposit interest rates respond much

slower to positive shocks. Mojon (2000) examined the pass-through parameters in a multi-country

setting (i.e. Belgium, France, Germany, Italy, Netherlands and Spain) by reporting

only short-run multipliers for a panel of 25 credit market rates and 17 deposit rates and for

the period 1979–1998. He found that the volatility of the money market rate lowers the pass-through

parameters for both credit and deposit rates, whereas increasing banking competition

can increase it.

This paper aims to develop a general modelling framework that allows one to examine

the full asymmetric effects of changes in the cost of funding on the variable-interest rate

charged by Australian banks and non-bank financial institutions for credit cards. We consider

two specific issues. First, in overall terms, does the credit card interest rate respond

asymmetrically to changes in the funding cost? If the funding cost changes, will Bacon’s

(1991) “rockets-and-feathers hypothesis” be applicable in the context of Australia’s retail

credit card rates? Second, when the cash rate, which is known as the federal funds rate in the

US, increases by one per cent, on average, by how much and how quickly do the standardvariable

rate for credit cards rise? One should note that significant asymmetric rate

adjustments can also adversely affect the efficacy of the RBA’s (Reserve Bank of Australia)

The cash rate is not the only factor that affects the lenders’ behaviour. The decision of

individual lenders to change their interest rates is also influenced to varying degrees by a

number of other factors such as the extent of securitisation and the individual bank’s

exposures to different types of external and internal sources of borrowing. However, the cash

‐6‐

rate has become increasingly politicised and the focus of much attention by media

commentators and the public alike. This is not hard to understand as for some families

interest payments constitute a substantial part of their income and interest rate changes have a

direct and appreciable effect on their consumer decisions. During periods of both increasing

and decreasing rates, the media not only focuses on which lenders are raising or lowering

their rates more quickly in response to the direction provided by the cash rate, but also the

extent to which the change in the cash rate is passed on to borrowers. In the context of

mortgage rates banks have responded in a variety of ways, with some justifying apparently

excessive rate rises outside the RBA cycle as being due to the increased cost of funding,

while others adopt a strategy of rate rise restraint to capture market share. Particular groups of

lenders (especially, credit unions and building societies, mortgage originators, and so on)

have also used their varying responses to the cash rate for publicity purposes aimed at

levering their ‘consumer-friendly’ credentials, especially against the major banks.

Most of the previous studies have focused on the asymmetric pass-through of funding

costs into mortgage-interest rates, an issue which is by no means Australia specific. Other

developed countries are also facing the same dilemma. For instance, Corvoisier and Gropp

(2002) and Bikker and Haaf (2002) have highlighted substantial differences across European

countries in terms of the pass-through of monetary policy interest rate changes into money

market rates. Bikker and Haaf (2002) examined the banking sector competition in 23

countries and discovered that competition is much weaker in local markets.

Payne and Waters (2008) thoroughly analysed the long-run interest rate pass through of

the federal funds rate to the prime rate over the period February 1987 to October 2005. They

state that the response of the prime rate to changes in the federal funds rate appears

asymmetric. Payne (2007, 2006) has also proposed several useful momentum threshold

autoregressive models in the literature on interest rate pass-through from the federal funds

‐7‐

rate to mortgage rates. He has found that the mortgage rates are co-integrated with the federal

funds rate in the long run but with incomplete interest rate pass through in the short run. A

recent study found that the pass-through of official interest rate changes into mortgage rates

in the Netherlands is only about half of the Euro-zone average (De Haan and Sterken, 2011).

Little empirical work has been conducted regarding the dynamic effects of positive and

negative changes in the central bank’s policy rate on credit card rates. However, there are

numerous studies in the literature that found evidence of asymmetric pricing for other lending

rates (e.g., Haney, 1988; Allen, Rutherford and Wiley, 1999; Hofmann and Mizen, 2004; and

De Haan and Sterken, 2011). Liu, Margaritis and Tourani-Rad (2008) conducted a

comprehensive study by using New Zealand’s monthly data on various mortgage, deposit and

lending rates for the period 1994–2004. They quantified the extent of pass-through and

adjustment speed of retail interest rates in response to changes in benchmark market rates by

using the Phillips and Loretan (1991) methodology. They found that not only the long-term

pass-through of retail rates do vary significantly across different rates but similar to Kim and

Nguyen (2008) they also argue that changes in the policy rate can exert more influence on

short-term interest rates. With some evidence of asymmetry in the adjustment of retail rates,

they conclude that since 1999 the Reserve Bank of New Zealand’s transparency in setting its

target policy rate has reduced the instrument volatility and enhanced the efficacy of monetary

policy.

Lowe and Rohling (1992) and Lowe (1995) investigated the degree of stickiness in

Australia’s various deposit and lending rates including credit cards by comparing them in the

card interest rate was the stickiest rate and there was no sign of observing less rigidity in the

the Reserve Bank of Australia. In April 1985, this rate was deregulated.

‐8‐

post-deregulation era. Lowe and Rohling (1992) stated that switching and search costs can be

Lim (2001) examined the asymmetric adjustments between three Australian bank

interest rates: a bank bill rate, a loan rate and a deposit rate. She uses a multivariate

asymmetric error-correction model to capture the long- and short-run relationships between

the levels of the rates and short-run relationships between the changes in the rates. Her

empirical results indicate that “banks value their borrowing customers and tend to pass on

results are in sharp contrast with previous Australian studies (Lowe and Rohling, 1992;

Lowe, 1995; Kim and Nguyen, 2008) as well as international studies (Hannan and Berger,

1991; Neumark and Sharpe, 1992; Payne, 2006, 2007; Allen and McVanel, 2009; De Haan

and Sterken, 2011).

Kim and Nguyen (2008) examine the effects of the RBA and the U.S. Fed’s target

interest rate announcement news on the Australian financial markets over the period 1998–

2006, including the 90–day bank bills, 3–year and 10–year bond rates. They found evidence

of asymmetric news effects on interest rates where they respond more strongly to unexpected

rate rises than rate falls. However, “the news effect is stronger at the short-term ends of the

interest rates” (Kim and Nguyen, 2008, p.392).

In this paper we address an important policy issue not recently tackled in the Australian

context. The majority of previous studies have tested for asymmetric effects of changes in the

central bank’s policy rate by using a Wald test in which two coefficients corresponding to

positive and negative changes in the lagged error correction terms were assumed to be equal.

term (see Table 5, p.145).

‐9‐

In other words, as it is demonstrated in the next section, they have tested for the adjustment

asymmetry (e.g. see Sarno and Thornton, 2003; Chong, Liu and Shrestha, 2006; Liu,

Margaritis and Tourani-Rad, 2008; Chong, 2010).

If the policy rate was not co-integrated with a particular lending rate, then the proposed

asymmetric models in such studies would have become symmetric. However, in this paper

we argue that even if the lending rate is not co-integrated with the policy rate, it is quite

possible that short-run positive and negative changes in the central bank’s policy rate may

still exert different effects on the corresponding lending rate. In other words, the amount

asymmetry should also be tested but within a dynamic framework. It should be noted that

Lowe and Rohling (1992) proposed a simple asymmetric model which was used for testing

the amount asymmetry. However, in their model only positive and negative changes in the

Rohling, 1992, p.25), whereas in our proposed models the amount asymmetry is tested

allowing up to 12–month lag.

Our analysis is undertaken within a sample period during which the RBA has conducted

monetary policy by setting the desired interest rate on overnight loans in the money market.

This paper offers an analytical modelling framework to quantify the full extent of asymmetric

rate changes (if any), resulting in greater efficiency and transparency of the lending market.

The results and policy implications of this paper increases our understanding of the credit

card lending market in Australia and are beneficial to all borrowers and various government

regulators, which can play an important role in market efficiency and consumers’ protection.

The rest of this paper is structured as follows. In Section 2 various theoretical models are

postulated which capture two possible forms of asymmetric behaviour (i.e. the amount and

adjustment asymmetry) as well as the dynamic asymmetric responses of the lagged dependent

variable. Section 3 discusses the choice of our sample period, the descriptive statistics of the

‐10‐

data employed followed by the results. Section 4 presents the empirical results of the longand

short-run credit card interest rate models as well as the policy implications of the study.

Section 5 provides some concluding remarks and the final section presents the agenda for

future research.

In order to capture the long-run relationship between the central bank’s policy rate and

the credit card cash rate, following Rousseas (1985), we utilise equation (1) in which

financial institutions set their standard variable rate as a mark-up on the cash rate. That is:

where:

=the white noise error term.

Several other studies in the literature have also specified similar long-term relationships

Chong, Liu and Shrestha, 2006; Toolsema and Jacobs, 2007; Liu, Margaritis and Tourani-

Rad, 2008; De Haan and Sterken, 2011). Standard unit root tests, such as the Augmented

Dickey-Fuller (ADF) test, the DF test with GLS detrending (Elliott, Rothenberg, and Stock,

1996) and the Kwiatskowski-Phillips-Schmidt-Shin (KPSS, 1992) test as well as the Lee and

Strazicich (2003) test, which endogenously incorporate two structural breaks in the testing

procedure, are used in the next section to ensure that our empirical results are not biased

towards the erroneous non-rejection of the non-stationarity hypothesis. Since both variables

11‐

integrated with the cash rate. Following the Engle-Granger two-step procedure if the two

variables are co-integrated, then the stationary residuals resulting from equation (1) could

equilibrium.

Standard co-integration tests (Johansen, 1995) implicitly assume a symmetric adjustment

process but if the adjustment process is asymmetric or if the loan rates are sticky downwards,

these tests can lead to misleading results. In other words, the Engle-Granger type tests with a

linear adjustment procedure will be inappropriate when the dynamic adjustment of loan rates

in fact exhibits non-linear behaviour. One alternative is to use the threshold co-integration

test proposed by Granger and Lee (1989), Enders and Granger (1998) and Enders and Siklos

(2001).

With some modifications the majority of previous studies (see inter alia Sarno and

Thornton, 2003; Chong, Liu and Shrestha, 2006; Liu, Margaritis and Tourani-Rad, 2008;

Chong, 2010) have used the following short-run dynamic model based on the assumption that

the corresponding two interest rates are co-integrated:

0 1 1

0 1

+

   

 

‐12‐

equilibrium is asymmetric and hence there is evidence of adjustment asymmetry. If

equilibrium path results in a relatively swifter error correction as compared to the case of a

lagged positive disequilibrium. This can easily be justified as financial institutions prefer to

hand, when the lagged-actual retail rate is above the equilibrium path, banks may have less

desire to lower their rate to its equilibrium path, causing the speed of adjustment to be more

rewritten as:

0 1

0 1

+

 

and one may use equation (6) instead:

0

0 1

+

 

Therefore, if two interest rates are not co-integrated or if there is no adjustment asymmetry,

one may end up using a symmetric model such as (6) and (5), respectively. However, it is

have asymmetric impacts on the dependent variable. Similarly positive and negative changes

‐13‐

alternative and flexible framework which can be used not only for testing both the adjustment

0 1 1

0 0 1 1

+

     

   

(7)

are the effects of positive

It should be noted that Allen and McVanel (2009) and De Haan and Sterken (2011) have

recently used a model which allows us to test both the amount and adjustment asymmetries in

a short-run dynamic model for mortgage rates. However, in their models it is assumed that

model as flexible as equation (7).

In equation (7) the extent of the amount asymmetry depends on the difference between

0

and

0

. In other words, the higher the difference, the more the extent of the amount

asymmetry. It is also hypothesised that rate rises have some immediate effects on the retail

within the first month to be statistically significant),

whereas in the case of rate cuts the lagged effects are exhausted within a longer period such

changes in the cash rate on the credit card rate are allowed to be different in magnitude as

well as through time.

‐14‐

Using equation (7), a Wald test can also be employed to test the amount asymmetry. If

amount asymmetry. This means that short-run changes in the cash rate can exert asymmetric

effects on the credit card rate. Accordingly, if

12 12

0 0

 

interest rate rises more than rate decreases. For a detailed discussion of the distinction

between amount and adjustment asymmetries in the literature see Chen, Finney and Lai

(2005), Bachmeier and Griffin (2003), Bettendorf, van der Geest and Kuper (2009), Allen

and McVanel (2009) and De Haan and Sterken (2011).

one may argue that positive and negative lagged values of the dependent variable have

then conclude that the retail interest rates are sticky downwards.

It should be noted that Australia’s approach to monetary policy has undergone

significant changes over time. From the mid-1970s until 1985, based on the assumption of a

strong and persistent relationship between inflation and the supply of money, monetary

policy was conducted by targeting the annual growth of M3. However, in 1985 this policy

was abandoned because deregulation of the financial system made M3 a misleading guide to

the stance of monetary policy (Grenville, 1990). From 1985 to 1988 a “checklist approach”

was adopted, whereby a multitude of indicators such as, monetary aggregates, the GDP

growth rate, the shape of the yield curve, exchange rates, and the unemployment rate were

considered prior to the implementation of monetary policy. The checklist approach was also

‐15‐

unsuccessful and finally discontinued in 1989 due to the impossibility of monitoring the

above indicators which could provide contradictory policy signals.

Since 1990, the RBA has conducted monetary policy by setting the desired (target)

interest rate on overnight loans in the money market. This year has been considered as the

starting point for our sample period. Through the monetary policy transmission mechanism,

changes in the cash rate are ultimately reflected in the rates on all lending instruments in line

with the desired policy intent. The cash rate in Australia is now considered the baseline for

the various interest rates paid by borrowers. Fig. 1 suggests that during our sample period,

which consists of monthly data from January 1990 to October 2011, there is a very close

relationship between the cash rate and credit card rate but at the same time the spread

our monthly data have been obtained from the RBA (2011, Statistical Tables F1 and F5).

[FIG. 1 ABOUT HERE]

Brännäs and Ohlsson (1999) suggest that the detection of asymmetry depends on the

sampling frequency of the series. They argue that the use of aggregated frequencies (i.e.,

annual) may obscure the nonlinearities or asymmetries that exist in a series. They found that

“asymmetric monthly series may become symmetric when aggregated to quarterly or annual

frequencies” (Brännäs and Ohlsson, 1999, p.341). Therefore, in this paper, we use monthly

data to detect any discernable asymmetric behaviour not easily observable in the raw and

aggregated data. As can be seen in our results presented in the next section, the effects of

positive and negative changes in the cash rate on the credit card rate appear to be exhausted

within the first three months. Therefore, this confirms Brännäs and Ohlsson’s point that the

14.95 per cent in October 2011.

‐16‐

use of annual time series in lieu of monthly could have masked the existing nonlinearities or

asymmetries in the series.

Table 1 presents the descriptive statistics and the unit root test results for all of the

variables employed in this paper. During the sample period the average credit card rate was

17.7 per cent, ranging from a minimum of 14.4 in May 1994 to a maximum of 23.5 in

January 1991. With a standard deviation of 2.44 per cent, the credit card rate showed a

slightly less variability than the cash rate (3.20 per cent), particularly after the adoption of the

RBA’s inflation targeting policy in 1993. Based on the reported Jarque-Bera statistics, the

null hypothesis of normality is rejected at any conventional level for all series as their

distributions are positively skewed and show a typical leptokurtic pattern with the kurtosis

statistic well exceeding 3.0.

An important step before undertaking our empirical investigation is to determine the

time series properties of the data. This is an important issue since the use of non-stationary

data in the absence of co-integration can result in spurious regression results. To this end,

three unit root tests, i.e. the ADF test, the DF test with GLS Detrending (Elliott, Rothenberg,

and Stock, 1996) and the Kwiatskowski-Phillips-Schmidt-Shin (KPSS) (Kwiatkowski et al.

1992) test, have been adopted to examine the stationarity, or otherwise, of the time series

data. In this paper the lowest value of the Schwarz criterion (SC) has been used as a guide to

determine the optimal lag length. Unlike both the ADF test and the DF test with GLS, the

KPSS test has the null of stationarity, and the alternative indicates the existence of a unit root.

In Table 1 we have also reported the results of the Lee and Strazicich (2003) test, which

endogenously incorporates two structural breaks in the testing procedure. The resulting two

break dates for individual series are also reported in Table 1. Irrespective of which test is

considered, it appears that our two interest rate series are I(1).

[TABLE 1 AND FIG. 2 ABOUT HERE]

‐17‐

In this section, we first use monthly data for the period January 1990–October 2011 to

OLS method. The estimation results are presented in Table 2. The average mark-up

personal loans, the estimated high mark-up is quite justifiable. The long-run pass-through

parameter is estimated at 0.67 per cent, which appears to be of consistent sign and order of

magnitude and highly significant. This means that on average the long run credit card rate is

7.1 per cent (13.6–6.51) more than the cash rate. We also tested the null hypothesis that the

was rejected at the 1 per cent level of significance (see Table 2) and as a result the spread

between the credit card rate and the cash rate has been widening (see Fig. 1).

We have also provided the recursive estimates of both the mark-up and the pass-through

parameter in Fig. 2. It seems that both the mark-up and pass-through parameters enjoy

relative stability around the estimated mean value since 1993, when the RBA adopted the

inflation targeting policy. According to the results of four unit-root tests presented in Table 2,

it seems that the resulting residuals obtained from the OLS estimation of equation (1) are I(1)

despite considering two endogenous breaks in the Lee and Strazicich (2003) test. Therefore,

one cannot argue that the credit card rate is co-integrated with the cash rate. The widening

spread between the two series presented in Fig. 1 could also partly explain the absence of cointegration.

Similar results were obtained by Lowe (1995, Table A1) 17 years ago in his

comprehensive study of the linkage between the cash rate and Australia’s lending rates

correction terms) are not statistically significant in the estimated Model I (see Table 3 in the next section).

‐18‐

As discussed in the previous section, we first estimate equation (7) and present the

that we are using monthly data, an upper band of 12 lags was allowed. Following the generalto-

specific methodology, insignificant variables in equation (7) were omitted on the basis of a

battery of maximum likelihood tests. We imposed joint–zero restrictions on explanatory

variables in the unrestricted (general) model to obtain the most parsimonious and robust

equation. We found that the cash rate is weakly exogenous with respect to the dependent

variable. This justifies the use of the OLS method to estimate our short-run dynamics models

based on an asymmetric version of the Engle-Granger two-step procedure. This is not

counter-intuitive as the cash rate is directly controlled by the RBA as a policy variable. The

weak exogeneity test results are available from the authors upon request.

the error correction series is I(1) (see Table 2) and at the same time it is not statistically

It should be noted that the other estimated coefficients in both Models I and II remain very

we only focus on the interpretation of Model II below.

[TABLE 2 AND FIG. 3 ABOUT HERE]

According to the results of Model II presented in Table 3, all of the estimated

coefficients are statistically significant at the five per cent level or better and have the

expected theoretical signs. Model II also performs well in terms of goodness-of-fit statistics.

level of significance. Furthermore, it passes a battery of diagnostic tests and shows no sign of

‐19‐

serial correlation (see the Breusch-Godfrey LM tests), misspecification (see the Ramsey

RESET test), heteroskedasticity (see the ARCH, White and Breusch-Pagan-Godfrey tests)

and instability (see the Chow forecast tests in three different forecasting periods). The only

diagnostic test that Model II could not pass was the Jarque-Bera normality test of the

residuals. This was unavoidable given the use of monthly data.

One problem associated with the analysis of the relationship between the cash rate and

the credit card rate is non-constancy or instability of the estimated coefficients which can

create economic and econometric complications in deriving any inference from the empirical

model. Therefore, parameter constancy is pivotal in modelling changes in the credit card rate.

We have evaluated our preferred estimated short-run model (i.e. Model II) by a number of

recursive stability tests which are displayed in Fig. 3 in the following order:

a b c d

e f g h

i j k l

 

 

 

 

where panel (a) displays the recursive residuals; (b) depicts the CUSUM test; (c) shows the

one-step forecast test; (d) is the n-step forecast test; and panels (e) to (l) present the

recursively estimated eight coefficients in exactly the same order that these coefficients

appear in Model II in Table 3 (from top to bottom). These evaluative tests are useful in

assessing stability of a model, as recursive algorithms avoid arbitrary splitting of the sample.

Overall, the graphical tests for stability reported in Fig. 3 support the in-sample constancy of

particular month, this would have immediately led to a rise of 1.032 per cent in the credit

card rate. On the other hand, a similar one per cent rate cut would have resulted in only 0.163

associated with the RBA’s rate cut would then be only 0.458 per cent within the first three

‐20‐

months (0.163+0.295), whereas the corresponding effect for a rate rise would be an

immediate 1.032 per cent increase. Thus, in absolute terms the sum of short-run effects of

rate cuts were less than half of the rate rises.

are statistically

adjustment asymmetry is not applicable. However, it appears that the credit card rate is very

sticky downwards because

1 1

ˆ ˆ

 

1990 era, the estimated coefficients of the lagged dependent variable are also statistically

the downward stickiness hypothesis. This proffers support for the short-run applicability of

the rockets-and-feathers hypothesis in the context of the credit card market in Australia. The

of the lagged dependent variable coefficients

1

0.292 0.196 129 0.617.



Therefore,

one may conclude that Australian credit card rates face downward rigidity when the RBA

reduces its cash rate. We can use equations (8) and (9) to approximate the total effects of a

one per cent rise or fall in the cash rate on the credit card rate, respectively:

0

1

1.032

1.46

1 1 0.295













 

 



(8)

0

1

0.163

0.43

1 1 (0.292 0.196 0.129)









   





(9)

‐21‐

It should be noted that a one per cent rate cut (compared to a rate rise), not only has a

smaller total effect (0.43 vs. 1.46) but also it takes longer to eventuate. We have also formally

tested the absence of the amount asymmetry by using a Wald test. Using the estimated Model

positive and negative changes in the cash rate obtained in equations (8) and (9) are so

different. One can thus conclude that in the context of credit cards there is strong evidence of

the amount asymmetry. As to the stickiness hypothesis, we have tested another null

per cent level of significance. Hence, based on these results, one can argue that there is

convincing evidence for the existence of both the amount asymmetry and stickiness

hypothesis in the context of Australia’s credit card rates.

Although our results are consistent with previous studies, our proposed model is capable

of capturing all possible types of asymmetric movements in the data. An overwhelming

majority of previous studies suggest that there is a great deal of asymmetry in the short-run

changes in the credit card loan rates. Researchers from other countries which could be

exposed to banks’ asymmetric behaviour in setting their various interest rates may also find

our proposed model and results useful. This paper can assist borrowers and relevant

government regulators to quantify the extent of the asymmetric behaviour exhibited by the

banking industry as a whole. Borrowers are entitled to know why rate rises have been passed

onto them faster than rate cuts and vice versa. This paper benefits all borrowers through a

better understanding of credit card rates, the central bank through specific knowledge of the

effects of its monetary policy on money market interest rates, and the regulators in enhancing

‐22‐

transparency and competition in consumer lending. The results presented in this paper reveal

that the incomplete credit card rate pass-through is relatively large and persistent, requiring in

turn a closer government monitoring and scrutiny.

Banks’ asymmetric behaviour in setting various loan rates has been a major cause of

concern in the banking and finance literature. However, little empirical work has recently

been conducted regarding the dynamic effects of changes in the cash rate on Australia’s

credit card interest rate. This paper is among a few studies that have attempted to model these

two rates and trace out the lenders’ dynamic asymmetric responses to changes in the funding

cost over time. Specifically, this paper attempts to rigorously test the robustness of the

rockets-and-feathers hypothesis in the Australian banking context.

This paper uses all available monthly time series data (January 1990 to October 2011) to

model the intricacies associated with the dynamic interplay between the cash rate and the

retail-variable interest rate for credit cards. Our aim is to modify the existing asymmetric

models in the literature to capture two forms of asymmetries, namely, the amount and

adjustment asymmetries. Since the credit card rate is not co-integrated with the cash rate, one

cannot test for any evidence of the adjustment asymmetry. However, it is found that the credit

card interest rate rises more quickly than it falls in response to changes in the funding cost. In

other words, at an aggregate level, rate rises are passed onto the consumer faster than rate

cuts. In terms of the magnitude of the pass-through parameter in our preferred short-run

dynamic model (i.e., Model II), rate rises had a full one-to-one and instantaneous impact on

the credit card rate. However, the effects of rate cuts were delayed up to three months. Based

on our results, we found a very significant degree of downward rigidity and path-dependency

in terms of changes in the credit card rate. Similarly to previous studies we found that when

‐23‐

the RBA reduces the cash rate the Australian banks are reluctant to lower their retail credit

card rates and do their best to delay such an action for as long as possible. Our proposed

models can also be easily adapted to examine the full extent of price stickiness in the context

of other types of retail and wholesale interest rates, including those for personal loans,

mortgages and business loans.

In this paper we have established that the rockets-and-feathers hypothesis is applicable at

an aggregate level in Australia’s credit card market. On our agenda for future research, our

aim will be to purchase consistently-defined weekly (credit card) interest rate data for over

100 bank and non-bank financial institutions from CANSTAR (www.canstar.com.au). Such a

disaggregated and weekly study can reveal in which bank or non-bank institutions interest

rate pass-through is more incomplete and the extent of the stickiness and asymmetric

adjustment is greater. We can then describe the main characteristics of such lending

institutions. For example, are the leaders of rate rises mainly big banks, and if so, which one?

Have such banks maintained a similar position in the market through the sample period? One

can identify the bank-specific opportunistic behaviour in the credit card market and reveal

excessive profiteering by the parties involved due to market inefficiency or tacit collusion.

The results of such a large-scale project will enable us to identify the leaders and followers of

retail rate changes (if any). This type of information could provide consumers with more

specific forward information about the expected dynamics, the extent of interest rate passthrough

and the “best bank to choose”.

‐24‐

Agarwal, S., Chomsisengphet, S., Liu, C., 2010. The importance of adverse selection in the

credit card market: Evidence from randomized trials of credit card solicitations. Journal

of Money, Credit and Banking 42, 743–754.

Allen, J., McVanel, D., 2009. Price movements in the Canadian residential mortgage market.

Bank of Canada, Working Paper 2009-13.

Allen, M.T., Rutherford, R.C., Wiley, M.K., 1999. The relationships between mortgage rates

and capital market rates under alternative market conditions. Journal of Real Estate

Finance Economics 19, 211–221.

Ausubel, L.M., 1991. The failure of competition in the credit card market. The American

Economic Review 81, 50–81.

Bachmeier, L.J., Griffin, J.M., 2003. New evidence on asymmetric gasoline price responses.

The Review of Economics and Statistics 85(3), 772–776.

Bacon, R.W., 1991. Rockets and feathers: the asymmetric speed of adjustment of U.K. retail

gasoline prices to cost changes. Energy Economics 13, 211–218.

Berlin, M., Mester, L.J., 2004. Credit card rates and consumer search. Review of Financial

Economics 13, 179–198.

Bettendorf, L, Van Der Geest, S.A., Kuper, G.H., 2009. Do daily retail gasoline prices adjust

asymmetrically?, Journal of Applied Statistics, 36(4), 385–397.

Bikker, J.A., Haaf, K., 2002. Competition, concentration and their relationship: an empirical

analysis of the banking industry. Journal of Banking and Finance 26, 2191–2214.

Brännäs, K., Ohlsson, H., 1999. Asymmetric time series and temporal aggregation. Review of

Economics and Statistics 81, 341–344.

Brito, D.L., Hartley, P.R., 1995. Consumer rationality and credit cards. Journal of Political

Economy 103, 400–433.

Calem, P.S., Gordy, M.B., Mester, L.J., 2005. Switching costs and adverse selection in the

market for credit cards: New evidence. Research Department, Federal Reserve Bank of

Philadelphia, Working Paper 05-16.

Calem, P.S., Gordy, M.B., Mester, L.J., 2006. Switching costs and adverse selection in the

market for credit cards: New evidence. Journal of Banking and Finance 30, 1653–1685.

Calem, P.S., Mester, L.J., 1995. Consumer behavior and the stickiness of credit-card interest

rates. The American Economic Review 85, 1327–1336.

Chen, L.H, Finney, M., Lai, K.S., 2005.A threshold cointegration analysis of asymmetric

price transmission from crude oil to gasoline prices. Economics Letters 89, 233–239.

Chong, B.S., Liu, M.H., Shrestha, K., 2006. Monetary transmission via the administered

interest rates channel. Journal of Banking and Finance 34, 1299–1307.

‐25‐

Chong, B.S., 2010. Interest rate deregulation: Monetary policy efficacy and rate rigidity.

Journal of Banking and Finance 30, 1467–1484.

Corvoisier, S., Gropp, R., 2002. Bank concentration and retail interest rates. Journal of

Banking and Finance 26, 2155–2189.

De Graeve, F., De Jonghe, O., Vennet, R.V., 2007. Competition, transmission and bank

pricing policies: Evidence from Belgian loan and deposit markets. Journal of Banking

and Finance 31, 259−278.

De Haan, L., Sterken, E., 2011. Bank-specific daily interest rate adjustment in the Dutch

mortgage market. Journal of Financial Services Research 39, 145–159.

Della Vigna, S., Malmendier, U., 2004. Contract design and self-control: Theory and

evidence. The Quarterly Journal of Economics CXIX (2), 353–402.

Elliott, G., Rothenberg, T., Stock, J.H., 1996. Efficient tests for an autoregressive unit root.

Econometrica 64(4), 813-836.

Enders, W., Granger, C.W. J., 1998. Unit-root tests and asymmetric adjustment with an

example using the term structure of interest rates. Journal of Business and Economic

Statistics 16, 304–311.

Enders, W., Siklos, P.L., 2001. Cointegration and threshold adjustment. Journal of Business

and Economic Statistics 19, 166–176.

Gans, J.S., King, S.P., 2003. A theoretical analysis of credit card reform in Australia, The

Economic Record 79(247), 462– 472.

Granger, C.W. J., Lee, T.H., 1989. Investigation of production, sales and inventory

relationships using multicointegration and non-symmetric error correction models.

Journal of Applied Econometrics 4, S145–S159.

Grenville, S., 1990. The operation of monetary policy. Australian Economic Review 2, 6–16.

Haney, R.L., 1988. Sticky mortgage rates: some empirical evidence. Journal of Real Estate

Research 3, 61–73.

Hannan, T.H., Berger, A.N., 1991. The rigidity of prices: evidence from the banking industry.

American Economic Review 84, 938–945.

Heffernan, S.A., 1997. Modelling British interest rate adjustment: an error correction

approach. Economica 65, 211–231.

Hofmann, B., Mizen, P., 2004. Interest rate pass-through and monetary transmission:

evidence from individual financial institutions’ retail rates. Economica 71, 99–123.

Johansen, S., 1995. Likelihood-based inference in cointegrated vector autoregressive models.

Oxford University Press, Oxford.

Kim, S.J., Nguyen, D.Q.T., 2008. The reaction of the Australian financial markets to the

interest rate news from the reserve bank of Australia and the U.S. Fed. Research in

International Business and Finance 22(3), 378-395.

‐26‐

Knittel, C.R., Stango, V., 2003. Price ceilings as focal points for tacit collusion: Evidence

from credit cards. The American Economic Review 93, 1703–1729.

Kwiatkowski, D., Phillips, P.C.B., Schmidt, P., Shin, Y., 1992. Testing the null hypothesis of

stationarity against the alternative of a unit root: how sure are we that economic time

series have a unit root?. Journal of Econometrics 54, 159–178.

Lee, J., Strazicich, M.C., 2003. Minimum lagrange multiplier unit root test with two

structural breaks. Review of Economics and Statistics 85, 1082–1089.

Lim, G.C., 2001. Bank interest rate adjustments: are they asymmetric?. Economic Record 77,

135–147.

Liu, M.H., Margaritis, D., Tourani-Rad, A. 2008. Monetary policy transparency and passthrough

of retail interest rates. Journal of Banking and Finance 32, 501–511.

Lowe, P., 1995. The link between the cash rate and market interest rates. Reserve Bank of

Australia, Research Discussion Paper.

Lowe, P., Rohling, T., 1992. Loan rate stickiness: Theory and evidence”. Reserve Bank of

Australia, Research Discussion Paper.

Massoud, N., Saunders, A., Scholnick, B., 2007. The cost of being late: The case of credit

card penalty fees. University of Alberta, Working Paper.

Mojon, B., 2000. Financial structure and the interest rate channel of ECB monetary policy.

European Central Bank, Working Paper No. 40.

Neumark, D., Sharpe, S.A., 1992. Market structure and the nature of price rigidity: evidence

from the market for consumer deposits. Quarterly Journal of Economics 107, 657–680.

Park, S., 2004. Consumer rationality and credit card pricing: Anexplanation based on the

option value of credit lines. Managerial and Decision Economics 25, 243–254.

Payne, J.E., 2006. The response of the conventional mortgage rate to the federal funds rate:

symmetric or asymmetric adjustment?. Applied Financial Economics Letters 2, 279–284.

Payne, J.E., 2007. Interest rate pass through and asymmetries in adjustable rate mortgages.

Applied Financial Economics 17, 1369–1376.

Payne, J.E., Waters, G.A., 2008. Interest rate pass through and asymmetric adjustment:

evidence from the federal funds rate operating target period. Applied Economics 40,

1355–1362.

Phillips, P.C.B., Loretan, M., 1991. Estimating long-run equilibria. The Review of Economic

Studies 58, 407–436.

Reserve Bank of Australia, (Statistical Tables F1 and F5), accessed 20 October 2011,

http://www.rba.gov.au/Statistics .

Rousseas, S., 1985. A mark up theory of bank loan rates, Journal of Post Keynesian

Economics 8(1), 135-144.

‐27‐

Sarno, L. Thornton, D.L., 2003. The dynamic relationship between the federal funds rate and

the Treasury bill rate: An empirical investigation. Journal of Banking and Finance 27,

1079–1110.

Scholnick, B., Massoud, N., Saunders, A., Carbo-Valverde, S., Rodríguez-Fernández, F.,

2008. The economics of credit cards, debit cards and ATMs: A survey and some new

evidence. Journal of Banking and Finance, 32, 1468–1483.

Stango, V., 2000. Competition and pricing in the credit card market. The Review of

Economics and Statistics 82, 499–508.

Stango, V., 2002. Pricing with consumer switching costs: Evidence from the credit card

market. The Journal of Industrial Economics L(4), 475–492.

Toolsema, L.A., Jacobs, J.P.A.M. 2007. Why do prices rise faster than they fall? with an

application to mortgage rates. Managerial and Decision Economics 28, 701–712.

‐28‐

0

4

8

12

16

20

24

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

Cash rate (r)

Credit card rate (i)

%

‐29‐

8

12

16

20

24

28

92 94 96 98 00 02 04 06 08 10

Recursive estimates of mark-up parameter (intercept)

± 2 S.E.

-0.4

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

92 94 96 98 00 02 04 06 08 10

Recursive estimates pass through parameters (slope)

± 2 S.E.

‐30‐

-1.00

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

1.00

96 98 00 02 04 06 08 10

Recu rsive Resid uals ± 2 S.E.

-60

-40

-20

0

20

40

60

96 98 00 02 04 06 08 10

CUSUM 5% Sig nifican ce

.000

.025

.050

.075

.100

.125

.150

-1.0

-0.5

0.0

0.5

1.0

96 98 00 02 04 06 08 10

One-Step Pro bability

Recursive Residuals

0

-1.0

-0.5

0.0

0.5

1.0

96 98 00 02 04 06 08 10

N-Step Pro b ab ility

Recu rsive Resid uals

-.15

-.10

-.05

.00

.05

.10

.15

98 00 02 04 06 08 10

Recursiv e C(1) Estimates

± 2 S.E.

0.0

0.4

0.8

1.2

1.6

2.0

2.4

98 00 02 04 06 08 10

Recursiv e C(2) Estimates

± 2 S.E.

-.5

-.4

-.3

-.2

-.1

.0

.1

.2

.3

.4

98 00 02 04 06 08 10

Recu rsiv e C(3) Estimates

± 2 S.E.

.1

.2

.3

.4

.5

.6

.7

.8

.9

98 00 02 04 06 08 10

Recu rsiv e C(4) Estimates

± 2 S.E.

-0.2

0.0

0.2

0.4

0.6

0.8

1.0

98 00 02 04 06 08 10

Recursiv e C(5) Estimates

± 2 S.E.

.0

.1

.2

.3

.4

.5

.6

98 00 02 04 06 08 10

Recursiv e C(6) Estimates

± 2 S.E.

-.1

.0

.1

.2

.3

.4

.5

98 00 02 04 06 08 10

Recu rsiv e C(7) Estimates

± 2 S.E.

-.2

-.1

.0

.1

.2

.3

.4

.5

.6

98 00 02 04 06 08 10

Recu rsiv e C(8) Estimates

± 2 S.E.

‐31‐

Descriptive statistics and unit root test results (January 1990–October 2011).

Mean 6.511 -0.043 17.691 -0.011

Maximum 18.180 0.720 23.500 1.050

Minimum 3.000 -1.260 14.400 -2.200

Std. Dev. 3.199 0.255 2.435 0.247

Skewness 2.330 -1.691 1.049 -2.957

Kurtosis 7.960 8.297 3.216 29.535

ADT test

Optimal lag 2 1 1 0

DF-GLS test

Optimal lag 2 1 1 11

KPSS test

Optimal Bandwidth 12 10 12 8

Lee and Strazicich test

Optimal lag 6 1 5 0

Break dates 1995:01 1992:03 1993:10 1994:02

2009:03 1995:01 2009:02 1997:09

and Strazicich (2003) test was based on the statistical significance of the

rejected at the 1 per cent level of significance.

‐32‐

Variable

Dependent variable:

0 1

1 1

: 1

: 1









Residual unit root tests:

ADF test

Optimal lag 1

DF-GLS test

Optimal lag 2

KPSS test

Bandwidth 12

Lee and Strazicich test

Optimal lag 8

Break dates 1993:09

1997:01

card rate.

break model in the Lee and Strazicich (2003) test was based

on the statistical significance of the corresponding parameters.

rejected at 1 per cent level of significance.

‐33‐

Estimated short-run dynamic asymmetric models.

Akaike info criterion -0.401 -0.415

Schwarz criterion -0.259 -0.302

Serial correlation LM

Test:

Heteroskedasticity test:

Chow forecast test:

per cent level of significance, respectively.