BANKING STOCK RETURNS AND

THEIR RELATIONSHIP TO INTEREST RATES AND EXCHANGE RATES:

AUSTRALIAN EVIDENCE

John Simpson*
University of Wollongong
P.O. Box 20183
Dubai, U.A.E.
Telephone : +9714 3954422
Facsimile : +9714 3955662

John Evans
University of Wollongong
P.O. Box 20183
Dubai, U.A.E.
Telephone : +9714 3954422
Facsimile : +9714 3955662

JEL Classification: EO, E44. Key Words: Banking, stock market returns, interest rates,
exchange rates, cointegration

*Corresponding author

This paper can be downloaded without charge from the

1
BANKING STOCK RETURNS AND THEIR RELATIONSHIP TO INTEREST
RATES AND EXCHANGE RATES: AUSTRALIAN EVIDENCE

By
John L Simpson
&
John P. Evans
Abstract

Banks in Australia, as agents of the Central Bank, are key players in the implementation

of exchange rate and monetary policies. The purpose of this study is to examine dynamic

interactions among, and the long-term equilibrium relationships between, bank stock

returns and the key macroeconomic and monetary policy variables in interest rates and

exchange rates. The concern is whether or not current economic indicators, as reflected in

interest rates and exchange rates, can explain banking stock market returns or vice versa.

The statistical models used include regression models, cointegration tests and Granger

causality tests from vector autoregressive models. The study finds no evidence that

Australia’s bank stock market returns form a cointegrating relationship with short- and

long-term interest rates and exchange rates over the period of study and thus conclusions

may not be drawn relating to long-term rational expectations in the Australian banking

market. Evidence is presented that causality runs from bank stock returns to interest rates

and exchange rates. This indicates that Australian monetary authorities, over the past

decade, appear to have placed strong reliance on the health and performance of the

banking and financial sector as they formulate monetary and exchange rate policy

settings.

2
 Introduction

The Australian Stock Exchange (ASX) and the Australian Securities and Investment

Commission (ASIC) have a comparatively high degree of regulatory authority over the

stock broking industry and these bodies require substantial disclosure and transparency

from stock broking firms. The Australian Prudential Regulatory Authority (APRA), in

terms of bank capital adequacy, regulates banks in Australia, and APRA acts in concert

with Australia’s central bank. The latter has the responsibility of implementing monetary

and exchange rate policy. Banks in Australia remain agents of the RBA and as such they

are involved in the implementation of these policies. Banks remain the most important of

Australian financial intermediaries in that on average 60% of a firm’s financial structure

is debt. In 1998 the Australian Bank’s share of all financial assets stood at 41.4% as part

of their principle lending function. This appears to have fallen since 1955 when the share

was 55%, however, it does not mean a decline in the importance of banks. The decline

“… is explained by an increase in funds under management held by the banks and an

increase in the securitisation of assets off the balance sheet” (Viney, 2000, Pages 43-45).

In light of this regulatory and competitive environment there is the possibility that

Australian bank stock returns, as a proxy for the financial health of the banking sector,

drive interest rates and exchange rates. The question of perhaps greater relevance is

whether or not the latter two variables are exogenous to the Australian banking system

and therefore are those variables the key drivers of banking stock returns?

3
That macroeconomic variables1 are significant drivers of stock prices is now widely

accepted. Studies in the United States (USA) are well documented (for example, Fama,

1970; 1990; 1991). Other studies had similar objectives (for example, Huang & Kracaw,

1984; Chen, 1991; Pearce and Roley, 1998; Wei & Wong, 1992). The most often used

framework to empirically examine these factors was initially provided by Ross (1976) in

the arbitrage-pricing model. Chen et al. (1986) used the arbitrage-pricing framework to

show that economic variables imparted a systematic effect to stock price returns. It is

contended that economic forces have influence over discount rates as well as the capacity

of companies to create cash and future dividends. Macroeconomic factors can thus

become risk factors in the stock markets. The arbitrage-pricing framework attempts to

discover whether or not risk premia that attach to these various risk factors are priced into

stock market returns. An interesting question arises as to how smaller stock markets like

that in Australia respond to changes in fundamental economic variables compared to the

larger and arguably more efficient US stock market. Cointegration2 has become another

framework for the analysis of macroeconomic variables and the stock market. Chen et al.

(1986) provided evidence to suggest that there is a long-term equilibrium relationship

between economic variables and the stock market and Granger (1986) sought to verify

this through cointegration analysis.

A number of additional questions arise. Is there evidence to suggest a long-term

cointegrating relationship between these variables? Is new information about interest

1
The variables studied usually include money supply, industrial production, short- and long-term interest rates and exchange rates.
This study focuses on interest rates (short-and long-term) and exchange rates.
2
Cointegration implies that two variables have similar stochastic trends and exhibit a long-term equilibrium relationship (Hill,
Griffiths & Judge, 2001, pp 346-347).

4
rates and exchange rates instantaneously and fully reflected in banking stock returns or

vice versa? Do interest rates and exchange rates significantly and systematically over or

under predict the future banking stock returns or vice versa? The answers to these

questions have implications towards rational expectations and market efficiency3 of the

Australian banking and finance sector.

Literature Review

It is clear that the relationship between stock prices and returns in particular countries and

economic variables has received great attention over recent years. For example,

Mukherjee and Naka (1995) in a study that investigated the Japanese stock market returns

found, using a better performing vector error correction4 model (VECM) compared to the

vector autoregressive model5 (VAR) model, that the Japanese stock market was

cointegrated with a group of six macroeconomic variables. Their findings were robust to

different combinations of macroeconomic variables in six-dimension systems.

Kwan and Shin (1999) utilised a cointegration test, a Granger causality test (Granger,

1988) and a VECM to find that the Korean stock price indices were cointegrated with a

set of macroeconomic variables, which included exchange rates and money supply, and

that the set of variables provided a direct long-run equilibrium relationship with each

3
Following on from Fama (1976), an efficient banking stocks market will be one in which banking stock prices always fully reflect
available information.
4
The VECM is a full information maximum likelihood estimation model, which provides for hypotheses testing for cointegration in
an entire system of equations in one step. Variables do not have to be normalised and it requires no a priori assumptions of
endogeneity or exogeneity of the variables (Johansen, 1991).
5
The autocorrelation function is the sequence of correlations between the current value of variables and past values at specific
intervals. The first order regression process or time series model means that the observable variables are each assumed to follow an
auto regressive process and the study allows for the possibility of a non-zero mean for the intercept parameter in each case (Hill,
Griffiths & Judge, 2001, pp 341-343).

5
stock price index. They also found that stock price indices were not a leading indicator

for the macroeconomic variables.

Maysami and Koh (2000) when investigating the long-term equilibrium relationships

between the Singapore stock index and selected macroeconomic variables and

Singaporean stock returns found with the assistance of a VECM that the Singapore stock

market is interest and exchange rate sensitive. They also found that the Singapore stock

market was significantly and positively cointegrated with the stock markets of Japan and

the USA.

Hondroyiannis and Papapetrou (2001) examined macroeconomic influences on the stock

market for Greece. Among the macroeconomic variables investigated were interest rates

and exchange rates. They too found that stock prices do not lead changes in real

economic activity but that the macroeconomic activity and foreign stock market changes

only partially explained Greek stock price movements. They found that oil price changes

did explain Greek stock price movements and had a negative impact on economic

activity.

Greenbaum and Thakor (1995; page 127) suggest that the process of financial

intermediation is of “central importance to the functioning of a modern economy”. There

are various reasons for this that includes the notion that the central function of banks and

the regulation of banks are interrelated. Banking needs to be perceived as a viable

business and to be perceived in that way it must be regulated. This regulation becomes an

6
important part of monetary and exchange rate policy where, for example, banks are

required to maintain RBA exchange settlement accounts in credit through repurchase

agreements. Redemptions of investments in these areas are agreed at a penalty to the

banks, which must maintain a “safety valve” proportion of their risk weighted assets in

prime liquid assets that include government paper with a short-term to maturity. In

addition, banks have the ability to resolve incentive problems. The large Australian banks

have the ability to diversify and can reduce incentive costs and agency problems due to

their efficiency in processing and re-using information. The banks can derive economic

benefits from their size, and their direct lending function.

James (1987) found empirical evidence in the USA market that there is a positive and

statistically significant stock price response for a company, which acquires a bank loan,

compared to a negative stock price response for debt placed privately with insurance

companies. According to Greenbaum and Thakor (1995) it is quite clear that given

problems of private information and moral hazard, credit markets cannot be any more

than semi-strong form efficient. Banks therefore have an important role to play in

resolving problems of information asymmetry. Banks enhance the informational

efficiency of credit markets, as they possess privileged financial information that is then

passed on to others.

It could be argued that banks are special financial intermediaries whose operations are

unique in financial markets and impact strongly on an economy. The simplistic notion

that the economic health of a small, developed country (such as Australia) is vitally

7
dependent on the financial health of its banking system is the principal motivation for this

study. A review of the literature reveals that there has been no well-known study of the

strength and direction of relationships between Australian banking stock returns and key

macroeconomic variables.

Data and Hypotheses

The primary source of data for the study is DataStream. Monthly data for each of the

stock returns, exchange rates and interest rates is gathered for the period from January

1994 to February 2002. The data are analysed using the EViews 4 statistical package.

The Australian Dollar (AUD) floated in December 1983. By selecting a study period

commencing well after that date, a post-flotation of currency adjustment period has been

accommodated where the AUD initially depreciated substantially against the United

States Dollar (USD) particularly in the first two post-flotation years. A period of post-

deregulation adjustment has also been allowed for when Australian banks were no longer

restricted by a statutory reserve deposit (SRD) ratio and a liquid government securities

(LGS) ratio6. The adjustment period saw the establishment of a less restrictive prime

assets ratio (PAR) and the liberating effects of the removal of RBA sectorial lending

directions and interest rate ceilings.

6
Deregulation of the early 1980s replaced these monetary policy tools with the prime assets ratio.

8
Secondary markets in Australian government bonds have been active over the full period

of the study with the effects of the treasury bond and note tendering system. In addition it

must be concluded that following deregulation and the granting of additional banking

licenses in Australia in the mid-1980s, the Australian banking industry cannot be

described as oligopolistic, which was the case prior to deregulation. That is, the

competitive environment for all banks in Australia has been relatively open and stable

despite the existence of a small amount of takeover and merger activity and a movement

by other deposit taking financial institutions, for example, building societies, to obtain

banking licences.

The Australian banking stock returns index (coded BRI for statistical analysis) represents

the theoretical aggregate growth in the value of the constituents of the index in banking

stocks. The index constituents are deemed to return an aggregate monthly dividend,

which is included as an incremental amount to the monthly change in the price index.7

The Australian banking stock price index8 could also have been used in this study as the

dependent variable. The reasons for the use of the banking returns index as a proxy for

7
The calculation is as follows: BRI t =(BRI t −1 )(BPI t /BPI t −1 )(1+BDY/100n) where BRI t is the bank return index at time t;

BRI t −1 is the bank return index at time t-1 (months), BPI t is the bank price index at time t and BPI t −1 is the bank price index at
time t-1; BDY is the dividend yield on the price index and n is the number of days in the financial year. BRIo is the index value at
base date = 100.

∑ (P * N
1
t t
8
The Australian banking price index is BPI t = BPI t −1 *
n
where I t , I t −1 , Pt , Pt −1 are the index value at
∑P
1
t −1 * Nt * f
day t, the index value on the previous working day, unadjusted share price on day t, unadjusted share price on the previous working
day and Nt, f, and n are the number of shares issued on day t, the adjustment factor for a capital action occurring on day t and the
number of constituents in the index. BPIo is the index value at base date = 100.

9
the financial health of the banking sector is provided in the results section. First

difference data for the BRI (that is, ∆ BRI t ) is equal to Log (BRI t /BRI t −1 ).

Data Definitions

The exchange rate (coded EX) monthly nominal exchange rates AUD to the USD

represents a mid point determined by the Reserve Bank of Australia on the basis of

quotations in the Australian foreign exchange market at 4pm Eastern Standard Time on

the 15th of each month.

The first difference data or the change in the exchange rate ( ∆ EX t ) is equal to Log

(EX t / EX t −1 ).

Long-term interest rates (coded LB) are proxied by the monthly 10-year Australian

government bond mid point between bid and offered rates obtained by the RBA from

tenders of 10-year government bond sales. Short-term interest rates (coded CR) are

proxied by the monthly interbank call-lending rate quoted by the Australian banks

through the RBA.

The first difference data or return on government long-term bonds ( ∆ LB t ) is equal to

Log (GB t / GB t −1 ) and the return of the call money market ( ∆ CR t ) is equal to Log

(CR t / CR t −1 ).

10
Hypotheses

The USA market is a major market of Australian goods. When the AUD depreciates

against the USD, Australian products become cheaper in the USA and if the demand for

these goods is elastic the volume of Australian exports should increase causing higher

AUD denominated cash flows to Australian companies. More profitable companies

borrow from the banks for expansion in fixed assets, trade finance and working capital

and this creates greater profitability to banks which, in turn translates to an increase in

bank stock returns. The opposite should occur if there is an appreciation of the AUD. It

could also be that higher bank stock returns are a signal to policy makers to ease

exchange rates to in turn maintain momentum in an economy growing relatively strongly

in the absence of inflationary pressures. The ensuing hypotheses are stated in the null

format.

Ho 1a: There is a significant positive relationship between level series of bank stock

return index and exchange rates.

Ho 1b: There is a significant positive relationship between first differences (changes) in

bank stock returns and exchange rates.

Changes in both short- and long-term interest rates affect the discount (bank base lending

rate) rate in the same direction via their effect on the nominal risk free rate of interest.

Higher lending rates affects a bank’s ability to lend profitably and therefore translate to

lower bank returns. As both short-and long-term interest rates rise, bank stock prices fall.

Similarly higher bank stock returns may be a signal to policy makers to ease interest rates

11
at a time when there is minimal pressure for an increase in rates to maintain interest rate

differentials between Australia and her major trading partners (with substantial interest

rate falls in the USA particularly over the past two or three years). Again inflationary

pressures have not been apparent in recent years. Banking stock returns have increased

while interest rates have fallen.

Ho 2a: There is a significant positive relationship between level series bank stock returns

and short-and long-term interest rates.

Ho 2b: There is a significant positive relationship between first differences (changes) in

bank stock returns and short- and long-term interest rates.

Based on the foregoing literature, if interest rates and exchange rates are significantly and

consistently priced in bank stock returns they should be cointegrated and this relationship

will provide evidence that these key economic variables significantly explain expected

bank stock returns.

Ho 3a: There are zero cointegrating relationships between level series bank stock price

returns and the key economic variables in exchange rates and short- and long-term

interest rates.

12
Ho 3b: There are zero cointegrating relationships between changes (first differences) in

banking stock returns and changes (first differences) in short- and long-term interest rates

and exchange rates.

Granger (1981, 1988) and Granger and Weiss (1983) provided evidence that if a pair of

variable series are cointegrated the bivariate cointegrating system must possess a causal

ordering in at least one direction. If the evidence is such that bank stock returns

variability is linked to these key economic variables it can also be shown that the change

in the bank stock returns either lag or lead movements in these economic indicators.

Based on this theory and the foregoing literature the fourth hypothesis of the study is

developed.

Ho 4a: There is zero one-way or two-way causality between level series bank stocks

returns and two key economic variables in interest rates and short- and long-term interest

rates.

Ho 4b: There is a zero one-way or two-way causality between changes in banking stock

returns and changes in short- and long-term interest rates and exchange rates.

Methodology

The underlying objectives of this study are to ascertain the significance of the strength

and direction of the influence of interest rates and exchange rates on banking stock

13
returns or vice versa. Specifically the investigation seeks answers as to whether or not

there is a long-term cointegrating relationship between banking stock returns and the

most important macroeconomic variables that dominate monetary policy settings.

Causality is also a matter of concern in that evidence is sought as to whether interest rates

and exchange rates drive bank stock returns or vice versa. The level series are first tested

and an unrestricted VAR 9 model applied. If it is necessary after considering those results,

then the first differences series are to be similarly tested using either a VAR (1) model or

a VECM.

The use of ordinary least squares (OLS) regressions assumes that all level variables are

normally distributed based on skewness and kurtosis10. The Jarque-Bera test (Jarque &

Bera, 1987) is used to test the normality of the distributions for each level series. In each

case for the level series of banking stock returns, interest rates and exchange rates the test

null hypothesis that the residuals are not normally distributed is rejected at the 5% level

of significance. The test null hypothesis that the residual of the level series regression is

not normally distributed cannot be rejected at the 10% level of significance.

As previously mentioned it would have been equally feasible to use the banking stock

price index as the dependent variable. The level series for the banking stock price index

behaves in a similar way to the banking stock returns index. A regression of one series

against the other shows an adjusted R Square value of .9985 with a coefficient t statistic

of 180.6027 which is highly significant when p is equal to 0.0000. The F statistic (at

9
This is a VAR model of the level series.
10
Skewness is how symmetrical the residuals of the variables are around zero and kurtosis relates to the “peakedness” of the
distribution.

14
32617.35) is highly significant to a similar level. When a Granger causality test is applied

it is found that there is two-way causality between the two series which are both highly

significant at the 1% level (F statistics for causality of returns to prices is 13.3747 (p =

0.0042) and the F statistic prices to returns is 12.8599 (p= 0.0005).

As previous studies appear to have analysed share prices, this study selects the banking

stock returns index as the dependent variable. The stages of analysis for the level series

and, if necessary, the first difference series, are as follows:

1. Multivariate Regression (OLS)

The regression of bank stock returns against short- and long-term interest rates and

exchange rates is analysed.

The variables have been defined in hypotheses and data.

LnBRI t = α + β 1 LnEX t + β 2 LnCRt + β 3 LnLBt + et …………………………………….1)

Of concern are the adjusted R Square value, the significance of the F statistic, the

magnitude of and signs of the regression coefficients, the t-statistics and their

significance as well as the magnitude of the Durbin Watson (DW) statistic (Durbin &

Watson, 1971)11. If, for example, the F and t-values are significant and the adjusted R

11
This remains as one of the most important tests in testing for autocorrelation and for spurious regressions when the variables are
integrated stationary processes (Hill, Griffiths & Judge, 2001, pp.271-274).

15
Square value shows reasonable explanatory power but the DW statistic is lower than the

adjusted R Square value, the regression may be spurious. Such a regression may not

possess the implied explanatory power if the level series are integrated, nonstationary

processes. There may be a serious time dependence in the level series. Further testing of

the level series for integration is the next step. The data are also tested for autoregressive

conditional heteroskedasticity12.

2. Integration and heteroskedasticity

For each of the level series in bank stock returns index, long-term interest rates, short-

term interest rates and exchange rates, tests are undertaken for stationarity13 using the

Augmented Dickey-Fuller (ADF) and the Phillips-Perron (PP) unit root tests (Dickey &

Fuller, 1981; Phillips & Perron, 1988)14. For example, if in each case the value of the

ADF and the PP statistics were greater than the 1, 5 or 10 percent critical values15 this

would indicate that the level series are nonstationary. The test null hypothesis of a unit

root is accepted. It does not necessarily follow that, if the level series are nonstationary,

the error terms of the multivariate regression of the level series are nonstationary. Unit

root tests are applied to the error terms of the regression of the level series variables.

12
“With time series data where we have data over time on one economic unit, it is possible that the error variance will change and this
would be true if there were an external shock or change in circumstances that that created more or less uncertainty about the variable”
(Hill, Griffiths & Judge, 2001, p. 238).
13
“A stochastic time series variable is stationary if its mean and variance are constant over time and the covariance from the two
values from the series depends only on the length of time separating the two values and not on the actual times at which the variables
are observed” (Hill, Griffiths & Judge, 2001, pp.335-336).
14
These tests, in addition to testing whether or not a series is a random walk, also test critical values of variables for a unit root (or a
random walk process) in the presence of a drift. Testing is thus undertaken for stationarity and therefore for integration in a series.
15
These tests both use MacKinnon critical values for rejection of the hypothesis of a unit root (MacKinnon, 1991).

16
In other words and in equation format, whether or not any two vectors16 are cointegrated

can be tested by analysing the error terms of the regressions of these level series vectors

for stationarity. The errors are tested using the ADF and/or PP test. The following

regression is in turn estimated:

∆eˆt = α 0 + γeˆt −1 + vˆt ……………………………………………………………………2)

Where,

∆eˆt = eˆt − eˆt −1 is the change in the value of the error term or residual at time t to time t-1.

α 0 is the intercept of this regression.

γ is the regression coefficient.

eˆt −1 is the error term at time t-1.

vt is the error term of this equation at time t.

The ADF and the PP statistics are then compared to the MacKinnon (MacKinnon, 1991)

critical values at 1%, 5% and 10%. If the test statistic is less than the critical values the

residual of the level series of those vectors is stationary.

16
The vectors in this study are BRI (banking stock returns index) LB (long-term interest rates), CR (short-term interest rates) and EX
(exchange rates). The examination is made after the level series and first difference series are logarithmically transformed.

17
Our interest also lies in testing for autoregressive conditional heteroskedasticity using the

ARCH ML17 test. If there are no integrated, nonstationary processes in the level series the

examination could proceed to a VECM specified on first differences of the level series

assuming that the first difference series were found to be integrated, nonstationary

processes.

3. Cointegration

Assuming that the level series are integrated, nonstationary processes, the study proceeds

with an analysis of the unrestricted VAR model and applies the Johansen (1988)

maximum likelihood (ML) procedure (Hansen & Johansen, 1993) to test for

cointegration. The optimal lag period is found using the maximum likelihood (ML)

tests18 over lags one to four and assuming an intercept and a linear deterministic trend in

the data.

4. Causality

Working on the level series in the VAR, Granger causality is examined in order to

describe the short-term dynamics of the model and two-way causality is investigated. It

would be concluded, for example, that if causality were stronger, running from interest

17
ARCH models were introduced by Engle (1982) and were specifically designed to model and forecast conditional variances. The
variance of the dependent variable is modelled as a function of past values of the dependent variable and independent or exogenous
variables.
18
These tests are used to test for cointegration between first differences of the specified variables. When results show that the
likelihood ratio is greater than the specified critical values the null hypothesis of zero cointegrating equations should be rejected.
When eigenvalues are less than one this means that the cointegrating relationship is stable based on an optimal lag interval (Johansen,
1993).

18
rates and exchange rates to bank stock returns, that interest rates and exchange rates

would be exogenous variables and thus determined outside the Australian banking

system. If it is clear that interest rates and exchange rates drive bank stock returns, the

concern lies in the sign of the coefficients (indicating positive or negative correlation

with bank stock returns in each case), the adjusted R Square value and the significance of

the F value (indicating the strength of the causality). In other words, the concern is

whether or not volatility of interest rates and exchange rates is transmitted to, and

reflected in the volatility of bank stock returns or vice versa and what is the strength and

direction of those factors?

If the results indicate that the level series regression is spurious the specification of a

VAR model is undertaken. For analytical and comparative reasons first differences of the

series are also analysed using either a VAR (1) model or a VECM.

Results

The results of the stages of the investigation are herewith reported. Significance levels of

10% or better are sought in view of the medium to long-term period of the study. The

analysis was divided firstly into an analysis of level series and then into first differences.

19
Level series

Multivariate Regression (OLS)

When OLS regression analysis is run on the level series, the adjusted R Square value is

found to be .8453, with an F value at 177.7254 (highly significant at p = 0.0000). With

regard to the coefficients (the significance level is in parenthesis), the intercept t statistic

is 31.6560 (p = 0.0000), exchange rates t statistic is –7.7747 (p = 0.0000), short-term

interest rates t statistic is 1.7648 (p = 0.0809) and long-term interest rates t statistic is

-9.0517 (p = 0.0000). The correlation matrix against bank stock returns (the correlation

measure is in parenthesis) shows exchange rates to be highly negatively correlated

(-0.8427), and long-term interest rates to be highly negatively correlated (-0.8675). It is

noted at this early stage that there is a significant positive relationship between exchange

rates and short-term interest rates (0.5305) and exchange rates and long-term interest

rates (0.7327) and between long- and short-term interest rates (0.6403). Clearly

multicollinearity between interest rates and exchange rates is evident in the level series

model.

In addition, the DW statistic at 0.1994, is less than the adjusted R Square value,

indicating that, if the bank stock returns are integrated, the regression may be spurious.

Moreover, the DW statistic is sufficiently close to zero and substantially less than two.

This evidence supports rejection of the test null hypothesis of zero autocorrelation and a

conclusion that there may be serious time dependence in the series.

20
There is weak evidence in the high-adjusted R Square value and the significance of the

coefficients to support the rejection of Null Hypotheses 1a and 2a (that there is a positive

relationship between banking stock returns and the key economic variables), but this

evidence needs to be treated with caution in light of the spurious nature of the regression.

Integration

Each of the level series was tested for a unit root using the ADF and PP tests. The results

indicate that the level series in each case for bank stock returns, exchange rates, short-

term interest rates and long-term interest rates are non-stationary processes as, in each

case using both the ADF and the PP tests, the tests statistics were greater than the 1%, 5%

and 10% critical values leading to acceptance of the test-null hypotheses of a unit root.

In the case of bank stock returns the ADF statistic was -.0844, against the 1%, 5% and

10% critical values of –3.5000, -2.8918 and –2.5827 respectively. The PP statistic was

- 0.0103 against critical values of –3.4986, -2.8912 and –2.5824. In the case of exchange

rates the ADF statistic was 0.2757. The PP statistic was 0.0836. In the case of short-term

interest rates the ADF statistic was –1.2523 and the PP statistic was –1.0165 and in the

case of long-term interest rates the ADF statistic was –1.1744 and the PP statistic was

–1.1835. All test statistics are greater than the critical values.

21
When the residual of the bank stock returns regression was tested for a unit root it was

found that both the ADF and the PP test statistics were less than the1%, 5% and 10%

critical values at -4.2860 and –9.4382. This provides evidence that the residuals are

stationary and in-turn; evidence is provided that there are integrated, non-stationary

processes in the level series.

Testing was undertaken for autoregressive conditional heteroskedasticity (ARCH) after

perusal of the scatter plots of the OLS residuals of bank stock returns regressed on each

of the exchange rate and interest rate variables. These plots revealed the existence of

patterns in the residuals associated with exchange rates and those associated with long-

term interest rates.

Heteroskedasticity

An ARCH model was applied to the regression of banking stock returns against exchange

rates and interest rates. The ML-ARCH model (Engle, 1982) was applied to the data of

98 observations with convergence achieved after 206 iterations. The variance equation

coefficients for ARCH 1 and GARCH 1 respectively were 1.0395 and –0.1234. The sum

of the coefficients is close to unity revealing that volatility shocks are persistent in the

data. It is quite clear that a drawback in the level series data is that there is strong

evidence of autoregressive conditional heteroskedasticity. However, the OLS regression

was re-specified as an ARCH-ML model. The results of this analysis, shows that the

adjusted R Square value falls to .7681 with an F statistic of 54.5437, which is highly

22
significant (where p = 0.0000). The t statistics for long-term interest rates, short-term

interest rates and exchange rates at –175674, -2.139 and –11.3026 are each highly

significant (where p = 0.0000, p = 0.0324 and p = 0.0000 respectively). Nevertheless, the

DW statistic at 0.1362 remains significantly near zero and less than two to conclude that

the regression results remain spurious.

Again, with the use of the ARCH model there is weak evidence to support is the rejection

of the Null Hypotheses 1a and 2a that there is a positive relationship between banking

stock returns and interest rates.

Cointegration

Despite the foregoing, there is strong evidence to suggest that the level series are

integrated. The level series may be cointegrated. An unrestricted VAR model was applied

to the level series data. A VAR stability condition test19 was applied. The results of this

test show that the unrestricted VAR satisfies the stability condition as all polynomial

roots have a value less than one and lie within the unit circle. In addition the optimal lag

period was found to be one month using the Sims (1988) LR tests as expanded in the

following section on causality.

19
The lag structure/AR Roots Test (EViews 4) reports the roots of the characteristic autoregressive polynomial. The VAR is stable or
stationary if all roots have a modulus less than one and lie inside the unit circle.

23
However, the t statistics on one to four month lags in the VAR were not significant and

when the Johansen cointegration test was applied in each case it was found that the test

null hypotheses of zero cointegrating equations could not be rejected. In each case the

trace and maximum eigenvalue statistics were less than the 5% and 1% critical values of

3.76 and 6.65.

There is no evidence of cointegration in the level series and thus no evidence in support

of the rejection of Null Hypothesis 3a, that there are zero cointegrating relationships

between banking stock returns and the key economic variables.

It was decided to advance the analysis to examine first differences and test again for

integration, cointegration and causality. Prior to these tests it is necessary to investigate

causality in the level series.

Causality

Pairwise Granger causality tests on the level series indicate as follows: For one lag it is

found that causality runs from bank stock returns to exchange rates, short-term interest

rates and long-term interest rates. The F statistics are 9.0045, 8.518 and 7.1393 with

significance levels of 0.0035, 0.0044 and 0.0089 respectively. For two lags, causality

runs from bank stock returns to exchange rates, short-term interest rates and long-term

interest rates with F statistics at 3.5867, 4.3019 and 5.1879 and with significance levels of

0.0317, 0.0164 and 0.0074 respectively. On two lags it is also found that causality runs

24
from short-term interest rates to bank stock returns with a F statistic of 3.3497 and a

significance level of 0.0395. For three lags causality also runs from bank stock returns to

exchange rates, and short- and long-term interest rates with less significance (at the 10%

level). It is confirmed that the minimum lag period in terms of causality is one lag based

on the significance of the F statistics (all F statistics are significant at the 1% level).

However, in addition, a Wald lag exclusion test (from EViews 4) was carried out for each

lag in the VAR20. The significance level of the joint Chi Square value when the lag is one

is higher at 0.0000 than lag two, lag three and lag four at 0.1812, 0.9627 and 0.7222. In
21
addition lag length criteria were also applied to the data. Based on the maximum or

highest value of LR (Sims, 1988), as well as the lowest values of Akaike, Schwartz and

Hannan-Quinn22 information criteria, the minimum lag is shown to be one.

The study needs to examine to three and four lags before any reverse causality is apparent

and that is shown from long-term interest rates to bank stock returns at the 5% and the

1% levels of significance respectively with three lags. The results show no significant

causality running from exchange rates to bank stock returns or from short-term interest

rates to bank stock returns over four lags.

It is apparent that causality is one-way, running from bank stock returns to exchange

rates, short-term interest rates and long-term interest rates with an optimal one-month lag.

20
Lag exclusion tests show for each lag the Chi Square (Wald) statistic for the joint significance of all endogenous variables at that lag
for each equation separately and jointly.
21
Lag-length criteria computes various criteria to select the lag order of an unrestricted VAR. The maximum lag can be tested. The
sequential modified likelihood ratio (LR) test is carried out where a comparison is made between the modified LR statistic and the 5%
critical values starting from the maximum lag and decreasing one lag at a time until there is a rejection.
22
These criteria provide a measure of information that strike a balance between measures of goodness of fit and parsimonious
specification of the model. The selection of the model is based on the smallest information criterion.

25
Support is therefore provided for the rejection of the Null Hypothesis 4a that there is zero

directional causality between the level series variables in the banking stock returns

model.

First Differences

Multivariate Regression

When OLS regression analysis is run on the first difference data it is found that the

adjusted R Square value is .1360 with an F statistic significant at the 1% level (with

p = 0.0008). Long-term interest rate changes are negatively related to changes in bank

stock returns index with a highly significant t statistic of –4.1181 (where p = 0.0001).

There is no significant relationship between short-term interest rate changes and

exchange rate changes. The DW statistic at 1.1928 is greater than the adjusted R Square

value and is sufficiently higher than zero and closer to two to conclude that the regression

may be relied upon (that is, it is unlikely to be spurious). Nevertheless, it is immediately

apparent that substantial information has been lost in the first differencing process.

It is apparent that long-term interest rate changes lead banking stock return index changes

in the short-term and that there is a significant negative relationship. There is weak

support for the rejection of Null Hypothesis 2b (that there is a positive relationship

between banking stock returns and one of the key economic variables in long-term

interest rates). Again, as in the case of the level series, this result needs to be treated with

26
caution due to the low explanatory power of the model. However, Null Hypothesis 1b

(that there is a positive relationship between the first difference series in exchange rates

and banking stock returns index) cannot be rejected.

Integration and Cointegration

After application of ADF and PP tests for stationarity, it is found that the first difference

series and the error terms of the regression of the first difference variables are stationary.

The first difference banking stocks index series ADF and PP statistics at –6.7990 and

–9.6809 are both less than the 1%, 5% and 10% critical values at –3.5000, -2.8918,

-2.5827 and –3.4993, -2.8915, -2.5826 respectively. For the highly significant first

difference series in long-term interest rates, the ADF and PP test statistics at –4.3672 and

–9.5383 were also less than the critical values in each case. The error term of the first

difference regression of banking stock returns showed ADF and PP test statistics of

–6.6427 and –9.4998, which were also less than the critical values in each case.

Thus evidence is provided that the first difference data are not integrated, non-stationary

processes and further integration and cointegration analysis thus ceases at this point,

except that a VAR (1) model is applied to arrive at a Granger causality test. Null

Hypothesis 3b (that there are zero cointegrating relationships between the first difference

series) cannot be rejected.

27
Causality

However, it remains appropriate to press on for evidence of causality in the first

difference data. When lag order selection criteria are applied to the first difference data it

is found that Akaike, Schwarz and Hannan-Quinn information criteria are at their

minimum at 0 lags with values of -20.6980, -20.5890 and -20.6540 respectively.

However, the maximum likelihood ratio exists at two lags at a value of 27.6853.

Using the lag exclusion Wald test, all lags except lag two are rejected (where lag two, Chi

Square value is at 24.7051). This value is significant at the 10% level (with p = 0.0752).

When the Granger causality/block exogeneity test is applied to the dependent variable

being banking stock returns index movements, it is found that long-term interest rates

have a significant Chi Square value of 10.1611 with four degrees of freedom and a

probability (significance level) of 0.0337.

The pairwise Granger causality test indicates that long-term interest rate changes Granger

cause banking stock index changes at two lags (F statistic 4.2597 and p = 0.0171), at

three lags (F statistic is 5.0995 with p = 0.0027) and at four lags (F statistic is 4.1935 and

p = 0.0038). However, according to the Wald test when all explanatory variables are

taken into account (recall the Chi Square value of the Wald test was 24.7051 with a 10%

level of significance where the joint probability taking all variables into account was

p = 0.0752) it is concluded that the optimal lag is two for long-term interest rates to

Granger cause banking stock returns index changes. There is no other significant causal

28
relationship in the first difference data for the model for bank stock returns index

changes.

It is apparent in the first difference analysis, that long-term interest rates changes are

exogenously determined with an optimal two-month lag. Evidence is therefore provided

for the partial rejection of Null Hypothesis 4b (that there are zero directional causality

between banking stock return movements and changes in the key economic variables,

specifically in long-term interest rates).

Conclusion

The underlying objective of the paper is to investigate the relationships between banking

stock returns and key economic variables of monetary policy in short-and long-term

interest rates and exchange rates. The motivation for the paper is that it is widely

recognised that banks are of crucial importance to financial sectors and to economies.

Evidence is sought within the study to support the simplistic statement that the economic

health of a country is dependent on the financial health of its banking system.

Evidence is not provided for cointegration of the subject level variables, and, thus, it is

not apparent that longer-term rational expectations play an important part in banking

markets in Australia. Overall, the lack of evidence of cointegration in the level data does

not support the evidence provided for Japan, by Mukherjee and Naka (1995), for Korea,

29
by Kwan and Shin (1999), for Singapore, by Maysami and Koh (2000) and for Greece,

by Hondroyiannis and Papapetrou (2001).

The level series regression results (which indicate weak support for short-term negative

relationships between the key economic variables and banking stock returns) should be

treated with caution in light of the spurious nature of the OLS and ARCH regressions.

When first differences are examined evidence is provided that long-term interest rate

changes appear to be significant and negatively related to banking stock return changes.

Again these results need to be treated with caution in light of the lower explanatory

power of the first differences OLS regression.

In USA studies, (Fama, 1970; 1990; and 1991; Huang & Kracaw, 1984; Chen, 1991;

Pearce & Roley, 1998; Wei & Wong, 1992), economic variables appeared to drive stock

prices and returns. Support is provided indirectly for the above research when first

differences causality is examined. On two month optimal lags it is found that causality

runs from long-term interest rate changes to banking stock returns changes and that the

relationship is negative. Again caution needs to be heeded as the first differencing of

variables loses substantial information making it difficult to draw conclusions about

economic relationships of level series.

The most significant and clear results for this study are that level series causality runs

from banking stock returns to key macroeconomic variables. It is apparent that

Australian monetary and exchange rate policy makers do take into account the financial

health and performance of the Australian banking system (as proxied by the banking

30
stock returns index) when they formulate exchange rate and interest rate policy settings,

or permit the market to adjust these endogenous indicators.

References

Chen, N. -F., 1991, “Financial Investment Opportunities and the Macro economy”,

Journal of Finance, 46, Pages 529-554.

Chen, N. -F., Roll, R., and Ross, S., 1986, “ Economic Forces and the Stock Market”,

Journal of Business, 59, Pages 383-403.

Dickey, D. A., and Fuller, W. A., 1981, “Likelihood Ratio Statistics for Autoregressive

Time Series Within a Unit Root”, Econometrica, Volume 49, Pages 1022-1057.

Durbin, J., and Watson, G. S., 1971, “Testing for Serial Correlation in Least Squares

Regression-111”, Biometrica, Volume 58, Pages 1-42.

Engle, R. F., 1982, “Autoregressive Conditional Heteroskedasticity with Estimates of the

Variance of United Kingdom Inflation”, Econometrica, Volume 50, pp. 1987-1006.

Fama, E. F., 1970, “Efficient Capital Markets: A Review of Theory and Empirical

Work”, Journal of Finance, 25, Pages 383-417.

31
Fama, E. F., 1976,”Efficient Capital Markets: A Review of Theory and Empirical Work”,

Journal of Financial Economics, Volume 3, Number 4, Pages 361-377.

Fama, E. F., 1990, “Stock Returns, Expected Returns and Real Activity”, Journal of

Finance, 45, pages 1089-1108.

Fama, E. F., 1991, “Efficient Capital Markets II”, Journal of Finance, 46, pages 1575-

1617.

Granger, C. W. J., 1981, “Some Properties of Time Series Data and their Use in

Econometric Model Specification”, Journal of Econometrics, Volume 11, Pages 121-130.

Granger, C. W. J., 1988, “Some Recent Developments in a Concept of Causality”,

Journal of Econometrics, Volume 39, Pages 199-211.

Granger, C. W. J., 1986, “Developments in the Study of Cointegrated Economic

Variables”, Oxford Bulletin of Economics and Statistics, 48, Pages 213-228.

Granger, C. W. J., and Weiss, A. A., 1983, “Time Series Analysis of Error Correcting

Models”, in Studies in Econometrics, Time Series and Multivariate Statistics, New York,

Academic Press, Pages 255-278.

32
Greenbaum, S. I., and Thakor, A. V., 1995, Contemporary Financial Intermediation, The

Dryden Press, Orlando, Pages 97-131.

Hansen, H., and Johansen, S., 1993, “Recursive Estimation in Cointegrated VAR

Models”, No. 1, Institute of Mathematical Statistics, University of Copenhagen.

Hill, R. C., Griffiths, W. E., and Judge, G. G., 2001, Undergraduate Econometrics, John

Wiley and Sons, Second Edition.

Hondroyiannis, G., and Papapetrou, E., 2001, “Macroeconomic Influences on the Stock

Market”, Journal of Economics and Finance, Volume 25, Number 1, Pages 33-49,

Spring.

Huang, R. D., and Kracaw, W. A., 1984, “Stock Market Returns and Real Activity: A

Note”, Journal of Finance, 39, Pages 267-273.

James, C., 1987, “Some Evidence on the Uniqueness of Bank Loans”, Journal of

Financial Economics, Volume 19, Pages 217-235.

Jarque, C. M., and Bera, A. K., 1987, “A Test for Normality of Observations and

Regression Residuals” International Statistical Review, Volume 55, pp. 163-172.

33
Johansen, S, 1988, “Statistical Analysis of Cointegration Vectors”, Journal of Economic

Dynamics and Control, Volume 12, Pages 231-254.

Johansen, S., 1991, “Estimation and Hypothesis Testing of Cointegrating Vectors in

Gaussian Vector Autoregressive Models”, Econometrica, Volume 59, Pages 1551-1580.

Kwan, Chung S., and Shin, Tai S., 1999, “Cointegration and Causality between

Macroeconomic Variables and Stock Market Returns”, Global Finance Journal, 10:1,

Pages 71-81.

Maysami, R C., and Tiong Sim Koh, 2000, “A Vector Error Correction Model of the

Singapore Stock Market”, International Review of Economics and Finance, 9 (2000)

pages 79-96.

MacKinnon, J., 1991, “Critical Values for Cointegration Tests” in R. F. Engle and C. W.

J. Granger (eds), Long Run Economic Relationships, Oxford University Press, Oxford,

pp. 267-276.

Mukherjee, T K., and Naka, A., 1995, “Dynamic Relations between Macroeconomic

Variables and The Japanese Stock Market: An Application of a Vector Error Correction

Model”, Journal of Financial Research, Vol, XVIII, No.2, Pages 223-237, Summer.

34
Pearce, D. K., and Roley, V. V., 1998, “Firm Characteristic, Unanticipated Inflation, and

Stock Returns”, Journal of Finance, 43, Pages 965-981.

Phillips, P. C. B., and Perron, P., 1988, “Testing for a Unit Root in a Time Series”,

Biometrica, Volume 75, pp.335-436.

Ross, S. A., 1976, “The Arbitrage Theory of Capital Pricing”, Journal of Economic

Theory, 13, Pages 341-360.

Sims, C. A., 1988, “Bayesian Scepticism on Unit Root Econometrics”, Journal of

Economic Dynamics and Control”, Volume 12, pp. 463-474.

Viney, C. J., 2000, “Finance Institutions, Instruments and Markets”, McGraw Hill,

Melbourne, Pages 43-45.

Wei, K. C. J., and Wong, K. M., 1992, “Tests of Inflation and Industry Portfolio

Returns”, Journal of Economy and Business, 44, Pages 77-94.

35