International Review of Economics and Finance 15 (2006) 487 – 504<br />
www.elsevier.com/locate/iref<br />
<br />
<br />
<br />
<br />
The effects of monetary policy changes on market interest<br />
rates in Greece: An event study approach<br />
Asimakis Kaketsis a, Nicholas Sarantis b,*<br />
a<br />
Deutsche Asset Management, UK<br />
b<br />
Department of Economics, Finance and International Business, London Metropolitan University,<br />
84 Moorgate, London EC2M 6SQ, UK<br />
Received 20 October 2003; received in revised form 15 September 2004; accepted 29 September 2004<br />
Available online 22 December 2004<br />
<br />
<br />
<br />
Abstract<br />
<br />
The operational procedures of the Bank of Greece underwent major changes during the 1990s. These shifts in<br />
operational strategy made interest rates the main tool of monetary policy for the first time in Greece. This paper<br />
examines the effects of changes in the bank’s operational interest rates on market interest rates at eight maturities<br />
and for different operational regimes. A major feature of our study is the application of the event study<br />
methodology used in finance, which has not been employed in any previous study on this subject. We find that<br />
changes in official interest rates had a significant influence on short-term and intermediate-term rates and that this<br />
relationship was affected by the changes in the bank’s operational procedure.<br />
D 2004 Elsevier Inc. All rights reserved.<br />
<br />
JEL classification: E52; E58; C52<br />
Keywords: Central Bank operational procedure; Monetary policy; Market interest rates; Event study<br />
<br />
<br />
<br />
<br />
1. Introduction<br />
<br />
Since the late 1980s we have witnessed substantial liberalisation of Greece’s financial markets.<br />
Controls on cross-border capital flows have been lifted and restrictions affecting competition and price<br />
<br />
<br />
* Corresponding author. Tel.: +44 20 7320 1464; fax: +44 20 7320 1414.<br />
E-mail address: n.sarantis@londonmet.ac.uk (N. Sarantis).<br />
<br />
1059-0560/$ - see front matter D 2004 Elsevier Inc. All rights reserved.<br />
doi:10.1016/j.iref.2004.09.003<br />
488 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
flexibility in domestic financial markets have almost been completely removed. The experience of other<br />
countries indicates that changes in financial structure can have important implications for the conduct of<br />
monetary policy and a number of them have substantially revised their operating procedures during the<br />
past decade as financial market changes altered the relationship between policy tools and objectives. In<br />
Greece, changes in the monetary framework and financial reform have coincided during the transition,<br />
with the Bank of Greece (Greece’s Central Bank) in effect playing a dual role. It altered its operating<br />
strategy in response to the evolving financial environment, as well as instigating and guiding these<br />
changes. The complex system of controls, which had been in existence since the end of the Second<br />
World War, supported an operating strategy designed to influence the supply of credit, rather than the<br />
price of credit. However with the gradual relaxation of the complex system of controls in the late 1980s,<br />
the Bank of Greece shifted its strategy away from the control of monetary and credit aggregates towards<br />
the use of interest rates as the main tool in the transmission of monetary policy.<br />
This paper examines the effects of the Bank of Greece’s official interest rate on market interest rates at<br />
various maturities over the period 1994–2000, using daily data. The reaction of short-term and long-term<br />
market interest rates to changes in the bank’s official rate provides important information about the<br />
transmission of monetary policy into the money market. But although this relationship has been<br />
investigated in a number of advanced industrial countries, it has not been examined in emerging market<br />
economies undergoing financial liberalisation, like the Greek economy. In addition, Greece is now<br />
member of the European Monetary Union, where the transmission of monetary policy has been the<br />
subject of considerable debate.<br />
To measure the effect of central bank rates on market rates we employ the event study methodology.<br />
Event studies can circumvent many of the problems associated with the time series approach by focusing<br />
on the response of market rates in the days immediately surrounding changes in the intervention rates.<br />
Given rationality in the market place, the effect of an event, such as a change in the operational rate of<br />
the Central Bank, will be reflected immediately in market rates. Thus the impact of a change in a Central<br />
Bank’s intervention rate can be measured using changes in market rates observed over a relatively short<br />
time period. In this way, we can measure the immediate impact that a change has. This information is<br />
important in the conduct of monetary policy.<br />
The pioneering study on the channel between central bank interest rates and market rates using a<br />
similar methodology is that of Cook and Hahn (1989). The authors examine the effect of changes in the<br />
Federal Funds rate on market rates in the United States at various maturities around and on the day of<br />
changes in the Federal Funds rate. Thornton (1998) has also studied the market’s reaction to federal<br />
funds rate changes, but only on the day of the change in the Federal Funds rate. Like Cook and Hahn, he<br />
obtains successively lower values as the maturity increases. Hence, for the short rates, the direct liquidity<br />
effect is the predominant influence, while in the case of longer rates, expectations are more important.<br />
On the other hand, Garfinkel and Thornton (1995) present evidence suggesting that the Federal Funds<br />
rate is a no better indicator of monetary policy than other short term interest rates. Other studies for the<br />
United States include Cook and Hahn (1988), Thornton (1986, 1994), Dueker (1992), Rudebusch (1995)<br />
and Kuttner (2000). Paquet and Pe´rez (1995) carried out a study for Canada and show that changes in the<br />
overnight mostcall rate induce a significant effect on the rates of assets with up to 6 months maturity.<br />
Work has also been done for European countries. Pedersen (1997) reports that changes in the Danish<br />
discount rate are found to have significant effects on market rates and the effect declines with maturity. A<br />
study by the Deutsche Bundesbank (1996) reports similar results for Germany. Neumann and Weidmann<br />
(1998) also investigate the effect of the German discount rate on the overnight rate and find that, post-<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 489<br />
<br />
<br />
unification, the size of this effect not only is reduced, but becomes insignificant. In contrast, Hardy<br />
(1998) finds that German market interest rates responded significantly to changes in the official rates<br />
during the 1990s, and these responses become even stronger when the changes in official rates are<br />
decomposed into anticipated and unanticipated changes. But in line with the evidence for the United<br />
States, Hardy also obtains successively smaller effects as the maturity of assets increases. Buttiglione,<br />
Del Giovane, and Tristani (1997) analyse the impact effects of changes in central bank rates on the term<br />
structure of interest rates in nine industrial countries. They find that central bank rates have a substantial<br />
impact both on short as well as on long rates, with short rates responding similarly across countries,<br />
while the reaction of long rates differs markedly between countries. Dale (1993) examines the impact of<br />
changes in the Bank of England’s bank 1 stop rate on market interest rates at seven different maturities in<br />
the days surrounding these policy changes. Dale’s results suggest that changes in the stop rate lead to<br />
significant responses in market interest rates for maturities of 1 month to 5 years, and that both<br />
anticipatory and learning effects are significant.<br />
An important contribution of our paper is the application of a more sophisticated method to that used<br />
in the above studies. We employ the more established and uniform event study methodology that is<br />
widely used in the field of financial economics (see MacKinlay, 1997). In this literature, most papers<br />
tend to focus on the impact of various events on security returns. However, this methodology has not<br />
been used in previous event studies of money market rates. What has been drawn from this literature is<br />
the methodological framework and the considerations raised from its empirical application to money and<br />
bond market.<br />
The remainder of the paper is organised as follows: in Section 2 we discuss the operational procedures<br />
of the Bank of Greece during the 1990s. In Section 3 we explain the event study methodology employed<br />
in the paper. Section 4 discusses the data. In Section 5 we present an analysis of the empirical results.<br />
Section 6 draws up the conclusions.<br />
<br />
<br />
2. Operational procedures of the Bank of Greece<br />
<br />
During the 1990s, the Bank of Greece underwent two major regime changes (see Annual Report of<br />
the Governor of the Bank of Greece). The first regime describes the 1994–1997 period. Following the<br />
process of gradual financial liberalisation towards the end of the 1980s and early 1990s, combined<br />
with the efforts of achieving the Maastricht convergence criteria, the bank abandoned monetary<br />
aggregates and switched to the operational use of interest rates. The interventions in the interbank<br />
money markets that had begun in 1993 continued in 1994. In March 1994 the Athens Interbank Offer/<br />
Bid Rates were initiated (ATHIBOR/ATHIBID). This step allowed the Central Bank to utilise the<br />
interbank market in order to modernise its operating procedures. Interest rates became the main policy<br />
tool. Throughout this period, the final objective of the bank had been the deceleration of inflation,<br />
with the emphasis of monetary policy shifted from intermediate monetary targets to the protection of<br />
the exchange rate parity. The exchange rate was included for the first time as an intermediate target in<br />
1993, although not on equal footing as the monetary target. In 1994, the exchange rate gained<br />
predominance over monetary aggregate targeting. However, until 1997, both targets were officially<br />
regarded as equally important. In practice, with the establishment of the money market rates with<br />
maturities longer than 1 day at the end of March 1994, the bank obtained a new intermediate target<br />
(exchange rate stability) and a new tool to achieve this target (the interest rate channel).<br />
490 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
From January 1993 until March 1997, the main interest rate used by the bank to conduct the required<br />
adjustments in liquidity was the bid rate in the overnight maturity. However, there were two rates at this<br />
maturity. The bid rate was used to withdraw liquidity and the offer rate was used to inject liquidity to the<br />
market. Interventions using these rates were activated at the bank’s discretion. In this respect the bank’s<br />
system was symmetric.1 The bid rate acted as the bank’s signal of its monetary policy stance. The offer<br />
rate was primarily used to inject liquidity during times of downward pressures on the exchange rate. As<br />
such, use of this rate was seen as temporary.2 Other discretionary interventions in the 14-day and 1-<br />
month maturities in the interbank markets were also used during this period. However, they were used on<br />
a much smaller scale. Their role was confined to acting as supplementary tools to help the bank<br />
withdraw the structural excess liquidity in the Greek money markets.<br />
The second regime lasted from April 1997 until the end of 2000. With the enactment of Law 2548/97<br />
regarding bProvisions Relating to the Bank of GreeceQ and the corresponding amendment to its statute,<br />
the Bank of Greece bacquired a modern institutional framework, compatible with the Treaty on<br />
European Union and the Statute of the European System of Central Banks.Q3 Thus, the Bank of Greece<br />
was granted independence and the operational procedures changed accordingly, with price stability<br />
becoming the primary objective.<br />
On March 27 1997, a two-tranche overnight deposit facility was introduced.4 Additionally, the bank<br />
changed its main monetary policy tool to a weekly repo with a 14-day maturity. This was not used<br />
systematically though, until January 1998. The change occurred in response to a document published by<br />
the European Monetary Institute (1997), where the proposed operating procedures for the new European<br />
Central Bank (ECB) were outlined. In effect the Bank of Greece started implemented the ECB’s<br />
operational procedure after March 27, 1997. Direct interventions in the overnight rate were suspended.<br />
A two-tranche overnight deposit facility was introduced, which represented the floor. The Lombard rate,<br />
which was already in place since 1993, represented the ceiling for the overnight rate.5 Both of these<br />
rates were nondiscretionary. Until January 1998, there were no regular liquidity operations. In January<br />
1998, the 14-day repo, which had been in use as an irregular discretionary instrument, became the new<br />
operational rate of the bank. Operations were conducted weekly. A step closer to compliance with the<br />
ECB system was made during 1998, when a nonregular 91-day operation was introduced. As a result<br />
the bank harmonised its intervention procedures with those in other EU countries and the European<br />
Central Bank, as well as allowing market forces to have a larger influence on money market rates.6<br />
Hence the bank has followed world and European trends in conducting its liquidity management<br />
operation through market operations, with standing facilities being used as a bsafety valve.Q7<br />
1<br />
Symmetric in the sense that facilities existed both to withdraw from and inject liquidity to the markets. This is contrasted to<br />
asymmetric facilities, where the central bank can only do one of the two (see Borio, 1997 for a discussion of this feature of<br />
central banking procedures).<br />
2<br />
This is confirmed by reading the daily reports on the money market in the daily financial newspaper Nafteboriki. (See also<br />
Filippides, Kyriakopoulos, and Moschos (1995).<br />
3<br />
Monetary Policy 1997–1998, April 1998, statement by the governor, Lucas Papademos.<br />
4<br />
The limit for the first of the overnight deposit facility is set at 300 billion drachmas. This amount is shared out among<br />
domestic credit institutions dependent on their market share. The second tranche (with a lower interest rate) has no quota.<br />
5<br />
The quota on the facility was gradually raised from 150 billion in 1994 to 480 billion in 1999.<br />
6<br />
See for instance the Report on Monetary Policy by the Bank of Greece 1997–1998.<br />
7<br />
The main difference is that whereas many countries have moved away from standing facilities towards market operations,<br />
Greece has never had an experience of conducting policy using standing facilities, since these were introduced alongside the<br />
discretionary liquidity control measures.<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 491<br />
<br />
<br />
3. Modelling methodology<br />
<br />
The methodology employed in our event study draws upon the work of MacKinlay (1997). This<br />
involves three steps.<br />
<br />
3.1. Defining the event window<br />
<br />
In our case this is the change in the operational rate of the Bank of Greece. In practice, the event<br />
window includes not only the day of the actual change in the operational rate, but is expanded to include<br />
days surrounding the adjustment day.8 The reason for expanding the event window is to capture the<br />
dynamics of the market rate responses in the days surrounding the official interest rate changes. Two<br />
kinds of dynamics are of interest. Firstly, the extent to which the market anticipates changes, and<br />
secondly the degree to which the market responds with a lag (i.e., delayed or learning effects).9 If the<br />
markets anticipate the timing of policy changes it may lead to systematic movements in market rates in<br />
the days leading up to the change. Delayed effects occur when markets take time to digest information.<br />
Dale (1993) points out that significant movements in the interest rates in the day following the change<br />
may indicate a learning process. Such a learning process may be expected to be more pronounced when<br />
the Central Authority does not announce an explicit target level for its policy objective and hence the<br />
markets have to learn about changes in the target. This has direct implications for Greece, where such a<br />
learning process seems quite likely, especially in the period immediately after the change in the<br />
operational procedure and the introduction of more indirect methods of intervention. Prior to the<br />
independence of the bank, the governor was not in the habit of announcing clear and transparent policy<br />
aims as well as targets, certainly not in comparison to more transparent institutions such as the Bank of<br />
England. In view of the above, we follow previous studies in setting an event window of 5 days (2 days<br />
before and 2 days after the policy change).<br />
<br />
3.2. Measuring the effects of the intervention rate<br />
<br />
Two methods have been used in the literature for measuring these effects. Both methods implicitly<br />
assume that there is an association between the magnitude of changes in the intervention rate and the<br />
response of market rates. The first is to run a regression of the type illustrated below:<br />
<br />
DRt ¼ a þ bDðINTERVÞt ð1Þ<br />
<br />
where DR t is the change in the market rate on a particular maturity at time t, and D(INTERV)t is the<br />
respective change in the intervention rate.10 Only days in the event window are considered. Days outside<br />
the event window are not included in the regression. But Dale (1993) notes that the main problem with<br />
Eq. (1) is the implicit assumption that the coefficient of interest, b, is constant. In practice though, the<br />
reaction of market rates to policy change is likely to depend on a whole array of factors, such as current<br />
8<br />
See Cook and Hahn (1989) and Dale (1993).<br />
9<br />
Roley and Sellon (1995) suggest that delayed effects may be the result of revisions in expectations.<br />
10<br />
Dale (1993) and Thornton (1998) suggest that the value of coefficient b in Eq. (1) may vary, depending on whether changes<br />
in the official interest rate are large or small. In our empirical experimentation we tested for this effect, but the results were not<br />
significant.<br />
492 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
market sentiment, the extent to which the policy was anticipated, the information content of the policy<br />
change, external factors,11 and so on. In order for the coefficient b to be unbiased, these missing<br />
variables should be included. However, this is often not possible because the missing explanatory factors<br />
may be qualitative and thus difficult, if not impossible, to identify and measure. As a consequence, b<br />
becomes event specific. In these circumstances, Dale suggests that, in terms of the bivariate regression<br />
(Eq. (1)), the most efficient form of estimation would be to estimate it allowing b to be event-varying.<br />
Unfortunately, in analysing policy rate changes, estimation of this form is limited by the fact that these<br />
types of analysis, typically, have a limited number of observations in their samples. Therefore, although<br />
theoretically robust, in practice, estimation of the bivariate regression is limited to simple OLS on Eq.<br />
(1). This observation leads Dale to propose looking at mean responses in the two interest rates instead.<br />
The response to the change in the market rate is calculated as a percentage of the change in the official<br />
rate. The same method is adopted in our paper.<br />
<br />
3.3. Determining the significance of results<br />
<br />
The method proposed by Dale (1993) is to compare the reaction of rates immediately surrounding<br />
official interest rate changes with those observed across the entire sample. Dale uses a standard t statistic<br />
test for the equality of two means. Such a test though makes strong parametric assumptions. This does<br />
not seem to be entirely adequate for the period under analysis. We introduce the more sophisticated<br />
method outlined in MacKinlay (1997), which has not been used in previous event studies of this type.<br />
The method can be viewed as having the following three steps.12<br />
<br />
3.3.1. Measuring the abnormal effect<br />
To assess the event’s impact we require a measure of the abnormal effect. The abnormal effect for<br />
market rate i, (AC)i , is the actual ex post change of the market rate (DR)i minus the normal change<br />
(NC)i over the event window. The normal change is defined as the change in market rates that<br />
would be expected to occur even if the change in the operational rate did not take place. Ideally, the<br />
abnormal changes in market rates should represent only the effect of the operational rates for the<br />
particular date t. In the general event study literature there are two broad methods for measuring<br />
normal performance. One is to assume away any information and simply use a constant change<br />
model. This assumes that mean change in a given interest rate is constant through time. The second<br />
method is to estimate an econometric model, using suitable conditioning variables for capturing<br />
other influences on market interest rates. It is reasonable to assume that at least two factors apart<br />
from the intervention rate could influence changes in money market rates. One is developments in<br />
the foreign exchange market and the other is domestic liquidity conditions.13 During our experi-<br />
mentation stage, we tried to estimate a regression of changes in market rates on various proxies for these<br />
factors. But although we used many combinations of these variables, we could not find a significant<br />
relationship.<br />
As a result, we used the constant-change model to evaluate the normal change. The advantage of this<br />
method is that it is simple and in certain cases has good large sample properties. Brown and Warner<br />
<br />
11<br />
Such as pressures on the exchange rate target.<br />
12<br />
A good analysis of this testing procedure can also be found in Campbell, Lo, and MacKinlay (1997), Chap. 4.<br />
13<br />
See, for example, Cable and Holland (1999).<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 493<br />
<br />
<br />
(1980, 1985) find that it often yields results similar to those of more sophisticated models. In any case, in<br />
the absence of any suitable conditioning information, one may have to use this method to ascertain the<br />
impact of events. The normal change model is given by:<br />
<br />
DRiv ¼ li þ giv ð2Þ<br />
<br />
where l i is the normal change that is estimated using the estimation window, and v=1,. . .,V measures the<br />
number of daily observations in the estimation window. Generally the event window itself is not<br />
included in the estimation period to prevent the event from influencing the estimates of the normal<br />
change model. The estimation window that will be used in this study includes the days surrounding the<br />
5-day event window.<br />
The sample variance of the abnormal changes estimated using the estimation window (the period<br />
outside the event window) is obtained by:<br />
<br />
1 XV<br />
r2giv ¼ ðDRiv li Þ2 ð3Þ<br />
V 1 v¼1<br />
<br />
The abnormal change for market rate i on the event day s, (AC)is , is calculated as the difference<br />
between the actual change and the normal change on the event day (day of change in the operational<br />
rate):<br />
<br />
ACis ¼ DRis li ð4Þ<br />
<br />
3.3.2. Aggregation of abnormal changes<br />
In order to draw overall inferences for the event of interest, the abnormal change observations must be<br />
aggregated across the days within the event window. Let s2, s1, s+1, and s+2 represent the 4 days<br />
surrounding the day, (s), of the operational rate change. We then define ACin as the cumulative abnormal<br />
change for market rate i on the nth operational change, given by:<br />
<br />
X<br />
sþ2<br />
ACin ¼ ACis ð5Þ<br />
s2<br />
<br />
Similarly, the average abnormal change on the nth event window is:<br />
<br />
PP 1 X<br />
sþ2<br />
ACin ¼ ACis ð6Þ<br />
5 s2<br />
<br />
Since we are interested in evaluating the significance of the effect of policy rates on market rates as a<br />
whole, we also need to aggregate across event windows. Assuming N events, we calculate the<br />
cumulative abnormal change for market interest rate i across all policy changes:<br />
<br />
X<br />
N<br />
CACi ¼ ACin ð7Þ<br />
n¼1<br />
494 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
The average cumulative abnormal change for each market interest rate i across all operational rate<br />
changes is given by:<br />
PPP 1 XN<br />
PP<br />
CAC i ¼ ACin ð8Þ<br />
N n¼1<br />
3.3.3. Hypothesis testing<br />
Under the null hypothesis, H0, that the event has no impact on market rates (i.e., the abnormal change<br />
is zero), the distribution properties of the abnormal changes can be used to draw inferences. Following<br />
MacKinlay (1997), we assume that the abnormal changes are distributed normally with a mean of zero. It<br />
is also assumed that they are independently and identically distributed. Under these assumptions, the<br />
abnormal changes can be aggregated over the days within each event window and across event windows<br />
to yield distributional assumptions for the average cumulative abnormal change.<br />
Hence the null hypothesis H0 can be tested using the statistic:<br />
PPP<br />
CACi a<br />
ffi f N ð0; 1Þ<br />
H ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð9Þ<br />
PPP<br />
varðCACi Þ<br />
PPP <br />
The traditional method used for measuring var CACi is the sum of the estimation period residual<br />
variances of the abnormal changes (Eq. (3)) (see Brown & Warner, 1980). But as Boehmer, Musumeci,<br />
and Poulsen (1991) argue, this method ignores the event-induced variance. Changes in policy can have<br />
in principle two kinds of influences on money market interest rates. One is an effect on the mean of the<br />
abnormal changes. The other is on the variance of the abnormal changes. Dale (1993) tests only for a<br />
mean effect. In null hypothesis H0, either a mean or a variance effect will reject the null of no significant<br />
impact of policy changes on market rates. However, when analysing the policy impact, we are interested<br />
in testing for a mean effect on the abnormal changes. Thus we must expand the null hypothesis to allow<br />
for a variance effect by the event. If changes in the operational rates of the Bank of Greece cause a<br />
variance effect, then a measure based on r g2 iv(Eq. (3)) is not a consistent estimator of the variance of the<br />
event window. Boehmer et al. suggest to eliminate the reliance of the null hypothesis on the estimates<br />
obtained from the estimation window and to rely instead on an estimate of the variance of cumulative<br />
abnormal changes. Based on simulations presented in their paper, the authors argue that the proposed<br />
adjustment results in equally powerful tests when the null is false and appropriate rejection rates when it<br />
is true. Therefore, to take account of the potential variance effect, we use the following estimator for the<br />
variance of the average cumulative abnormal changes:<br />
PPP 1 XN<br />
PP<br />
varðCACi Þ ¼ ½ACin ACin 2 ð10Þ<br />
N ðN 1Þ n¼1<br />
<br />
Moreover, as mentioned in Boehmer et al. (1991), the above methodology works best when the constant-<br />
change model is used, as in the present paper.<br />
<br />
3.4. Measuring daily significance<br />
<br />
A problem with the above method is that it may create a bias in the determination of the response of<br />
market rates. The reason is that the above method assesses the significance of the event window as a<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 495<br />
<br />
<br />
whole. However, it does not scrutinise the individual responses in specific days within the event window.<br />
In particular, although the change in the policy rate may have significant effects across the 5-day event<br />
window, the influence may not be statistically significant on all the days in the event window. For<br />
instance, the response 2 days before the operational rate change may be significant, but the response 1<br />
day after may not. The above methodology can be easily adjusted to test whether the coefficients on<br />
particular days within the event window are significant. In order to do this, we aggregate only across<br />
event windows, not within event windows. Thus Eq. (4) is replaced with:<br />
ACiqn ¼ DRiq li ð11Þ<br />
where q=s2, s1, s; s+1, and s+2.<br />
The average cumulative abnormal change in market rate i on the individual days within the event<br />
window is given by:<br />
PPP 1 XN<br />
CACiq ¼ ACiqn ð12Þ<br />
N n¼1<br />
Using the results from Eq. (12), we can then calculate the significance levels for individual days using<br />
the variance estimator (Eq. (10)), which takes into account the variance effect that policy changes have<br />
on market rate changes.<br />
<br />
<br />
4. Estimation period and data<br />
<br />
As explained in Section 2, the Bank of Greece implemented two distinct operational procedures during<br />
our sample period. Hence our study will cover two estimation periods corresponding to the different<br />
operational procedures. The first procedure spans the March 1994–March 1997 period. As is to be<br />
expected, the interbank market did not function properly in the first year. This is because the dere-<br />
gulation of the financial sector was not yet complete.14 Thus, the study will examine the effects of the<br />
intervention rate over the period May 1995–March 1997. The second sample investigates the reaction of<br />
money market rates to the official interest rate changes over the April 1997–April 2000 period.<br />
We use daily data obtained from the Bank of Greece. For the money market rates we use the overnight<br />
and 1-, 2-, 3-, 6-, 9-, and 12-month Athibor rates. These consist of averages of the money market rates<br />
quoted by banks. For the second part of the sample (April 1997–April 2000), a long-term rate has also<br />
been included. Unlike the earlier period, by 1998, the government bond rates were set by market forces.<br />
The rates were no longer administratively set and banks were not legislatively obliged to hold<br />
government paper.<br />
As has been explained in Section 2, the bid rate will be taken as the operational rate of the Bank of<br />
Greece during the May 1995–March 1997 period. With the exception of two observations, all the 19<br />
changes in the bid rate that occurred in this period are used. On November 21, 1995 there were serious<br />
problems with the prime minister’s health and this caused a disproportionately large response that is an<br />
<br />
<br />
14<br />
In particular, the rate on government bonds was still administratively set. Furthermore, in March 1994, full liberalisation in<br />
capital movements had not yet occurred.<br />
496 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
outlier. The second observation to be excluded is the market change on December 13. The reason is that<br />
it stands between two operational rate changes, on December 11 and 15. Hence using it will be double<br />
counting, since it would represent both a reaction 2 days after the change on December 11, and an<br />
anticipation 2 days before the change on December 15.<br />
Between April and October 1997, the operational rate was the two-trance system. During this period,<br />
it was changed three times. Between November 1997 and January 1998, the Asian crisis intervened. The<br />
drachma came under strong pressure, owing to substantial capital outflows at the end of October 1997,<br />
after the outbreak of monetary turmoil in the Asia markets. To support its exchange rate policy, the Bank<br />
of Greece used direct measures. Namely, it imposed a surcharge of 0.4% per day on the increase in the<br />
debit balances of commercial banks’ current accounts with the Central Bank. This is equivalent to an<br />
annual interest rate of 170%. Liquidity was also provided via the 14-day maturity. After the main impact<br />
of the crisis subsided, the bank initiated the 14-day reverse repo as the main operational tool.15 The first<br />
two open market operations were conducted on January 5 and 14, respectively. Both represented changes<br />
using 14 days as the operational rate. However, during this period, there was substantial exchange rate<br />
instability. This was due to two reasons: firstly, the rekindling of the Asian crisis and, secondly,<br />
uncertainty over the EMS entry of Greece (both concerning the date and the central parity of the drachma<br />
in entry). Therefore, these dates are not considered indicative of the effects of monetary policy and,<br />
consequently, are not included in the estimation. The sample includes the other 24 changes that occurred<br />
in this period.<br />
<br />
<br />
5. Empirical results<br />
<br />
5.1. Mean responses<br />
<br />
In Greece, market interest rate responses to operational rate changes display huge variability. A look<br />
at the plots of the proportional series for the seven money market interest rates across the 43 policy<br />
changes, in Figs. 1 and 2, suggests that the responses are event specific. The plots of the proportional<br />
series highlight the sharp fluctuations in the reaction of market interest rates across the different policy<br />
changes. This would make regression analysis using Eq. (1) unwise. Therefore, we will follow Dale<br />
(1993) in considering instead mean changes in the two interest rates. In the examination of the impact of<br />
changes in the Bank of Greece’s intervention rate, we shall calculate the change in the market rate as a<br />
percentage of the change in the intervention rate. Hence, a resulting figure that is greater than 100%<br />
implies that markets overreact. Conversely, a figure less than 100, but still positive, implies a partial<br />
reaction by the market. Finally, a figure less than zero implies a reaction by the market in the opposite<br />
direction to the change in the intervention rate.<br />
Plotting each proportional official rate change individually gives an idea of the variance in the<br />
market’s reaction and the extent that such reactions are event specific. The results for Greece support<br />
much variation in market reactions. Negative reactions or large positive reactions are not uncommon.<br />
Negative reactions may occur when the change was already anticipated by the market, but the actual<br />
change was smaller than what the market had anticipated. This would lead to a market correction on the<br />
actual day of change that will appear as a negative reaction empirically. Large positive reactions could<br />
15<br />
Interventions in the repo definition.<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 497<br />
<br />
<br />
D1M(t) D2M(t)<br />
2.5 D3M(t) D6M(t)<br />
D9M(t) D12M(t)<br />
<br />
2<br />
<br />
<br />
1.5<br />
<br />
<br />
1<br />
<br />
<br />
.5<br />
<br />
<br />
0<br />
<br />
<br />
-.5<br />
<br />
<br />
-1<br />
<br />
<br />
<br />
0 5 10 15 20 25 30 35 40<br />
Policy Changes<br />
<br />
Fig. 1. Proportional responses: event day responses. D1M—change in the 1-month athibor; D2M—change in the 2-month<br />
athibor; D3M—change in the 3-month athibor; D6M—change in the 6-month athibor; D9M—change in the 9-month athibor;<br />
D12M—change in the 12-month athibor.<br />
<br />
<br />
<br />
occur, for instance, if the interest rate change generates expectations of further changes in the future.<br />
Taking mean changes at each maturity provides information of how the markets react in general to<br />
interest rate changes.<br />
The mean responses for the two sample periods are reported in Tables 1–4. For both periods we notice<br />
large anticipation responses. Learning responses are also substantial for the first period, but not for the<br />
second sample period. Looking at cumulative responses, there seems to be an overreaction in the first<br />
period, but underreaction in the second period.<br />
<br />
5.2. Overall significance levels<br />
<br />
Having obtained estimates of the response of money market rates to changes in Central Bank official<br />
interest rate changes, one is naturally interested whether these are significant. Following the method<br />
outlined in Section 3.3, we evaluated the significance of the results across the event windows. As Tables<br />
5 and 6 show, the responses across the event windows are significant.<br />
In the first period of analysis, the h statistics strongly reject the null hypothesis that official interest<br />
rate changes had no effect on market rates for all maturities (except for the 12-month rate). In the<br />
second sample period, the variances are considerably larger, although the responses of market rates to<br />
official interest rate changes are still significant (although not as significant as for the first period) for<br />
all maturities except for the overnight and the 10-year rates. There are two potential explanations for<br />
this result in the second sample period. Firstly, the bank’s presence in the market was reduced. The<br />
bank’s interventions were instead through the regular weekly 14-day repo rates. As a result, the<br />
498 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
DOVERN(t)<br />
20<br />
10 DOVERN(t-2)<br />
<br />
0 10<br />
-10<br />
0<br />
-20<br />
<br />
0 10 20 30 40 0 10 20 30 40<br />
5<br />
DOVERN(t-1) 10 DOVERN(t+1)<br />
<br />
0<br />
<br />
-5 0<br />
<br />
-10<br />
<br />
0 10 20 30 40 0 10 20 30 40<br />
DOVERN(t+2)<br />
10<br />
<br />
5<br />
<br />
0<br />
<br />
<br />
0 10 20 30 40<br />
<br />
Fig. 2. All the proportional responses for the overnight rate. t= event day response; t+1 and t+2=delayed responses; t1 and<br />
t2=anticipation effects.<br />
<br />
<br />
variance of market rates increased. Second, this was a period of significant turbulence associated with<br />
the Asian financial crisis.<br />
<br />
5.3. Individual day significance levels<br />
<br />
The above methodology may overestimate the reaction of market rates. Although particular responses<br />
within the event window may be insignificant, this is hidden, since one is testing for the overall<br />
significance. Hence, using the methodology described in Section 3.4, we will measure the day-by-day<br />
significance levels. Tables 7 and 8 report the h statistics regarding the significance levels, which are<br />
computed using the estimate of the variance obtained by using Eq. (10).<br />
There is clearly a difference in the response observed between the two periods. It is evident that in<br />
both periods there is a difference between the reaction of the overnight rate and that of the longer money<br />
market maturities. In the first period, the overnight rate reacts strongly negatively in the days before the<br />
<br />
<br />
Table 1<br />
Mean responses in the days surrounding the policy change: May 1995–March 1997<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month<br />
rate rate rate rate rate rate rate<br />
2 days before 31.07 37.09 36.83 35.27 30.60 41.49 46.9<br />
1 day before 15.47 20.75 21.48 25.14 25.14 19.19 25.49<br />
Day of change 99.95 37.83 40.76 38.32 36.37 32.16 29.56<br />
1 day after 39.96 40.27 39.30 34.66 38.32 43.05 40.46<br />
2 days after 6.30 5.26 6.09 9.68 6.09 11.25 0<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 499<br />
<br />
<br />
Table 2<br />
Cumulative responses in the days surrounding the policy change: May 1995–March 1997<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month<br />
rate rate rate rate rate rate rate<br />
2 days before 31.07 37.09 36.83 35.27 30.60 41.49 46.83<br />
1 day before 15.59 57.83 58.30 60.41 55.74 60.68 72.32<br />
Day of change 84.36 95.66 99.07 98.73 92.11 92.84 101.89<br />
1 day after 124.31 135.94 138.36 133.39 130.43 135.89 142.34<br />
2 days after 130.62 141.20 144.45 143.07 136.52 147.15 142.34<br />
<br />
<br />
<br />
policy change, and subsequently overreacts. In the second period, the large negative reaction occurs on<br />
the event day. Reading through the financial press, it is noted that in the first period the bank used to dry<br />
up the market excessively in the days before policy rate cuts. As a result, the overnight rate jumped up.<br />
This fact may also help explain the significant anticipation effect. In the second period, interventions<br />
occurred only once a week under normal circumstances. Hence on the day the reverse repo was held, the<br />
bank swiped up all the liquidity. Consequently the overnight jumped up to balance the market for high-<br />
powered money. Moreover, looking at the plots of proportional responses in Figs. 1 and 2, it is clear that<br />
after the 17th observation when the second sample begins, the proportional reactions of the overnight<br />
rate are larger and more oscillatory. This conforms with our expectations. Since the bank moved to a<br />
longer maturity in the second period and no longer intervened on a daily basis, it controlled the overnight<br />
rate far less closely.<br />
With regards to the behaviour of the longer maturities, the results on Table 7 suggest that there was a<br />
highly significant anticipation effect in the first period. There are systematic movements in markets<br />
leading up to the change, indicating that policy moves were broadly anticipated over this period. More<br />
than half of the change in the intervention rate was on average anticipated by the markets. By the day of<br />
the change in policy, the change has been fully discounted, since the cumulative change by time t is<br />
between 92% and 99% for maturities 1–6. Moreover, market rates display significant learning effects,<br />
but only for 1 day after the policy change. Similar results were found for the UK (Dale, 1993) and the<br />
United States (Cook & Hahn, 1989). In Greece one would expect these reactions to be pronounced, since<br />
the money markets had only been recently established. Markets lacked experience with market<br />
determined interest rates, and were inexperienced in interpreting the future policy intentions of the<br />
central bank from current policy changes. The results for the 9- and 12-month maturities are similar, but<br />
react less strongly on average. However, they should be viewed with skepticism, since the 9- and 12-<br />
<br />
<br />
Table 3<br />
Mean responses in the days surrounding the policy change: April 1997–April 2000<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month 10-year<br />
rate rate rate rate rate rate rate rate<br />
2 days before 81.81 32.08 13.72 30.53 31.69 21.64 28.22 18.04<br />
1 day before 3.71 26.96 45.89 28.41 26.38 31.50 27.83 8.70<br />
Day of change 86.90 13.80 15.74 18.06 17.50 13.80 8.70 0.22<br />
1 day after 137.29 15.07 7.44 6.18 5.80 6.67 6.18 7.42<br />
2 days after 59.50 1.35 1.26 1.74 3.77 2.03 0.48 2.02<br />
500 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
Table 4<br />
Cumulative responses in the days surrounding the policy change: April 1997–April 2000<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month 10-year<br />
rate rate rate rate rate rate rate rate<br />
2 days before 81.81 32.08 13.72 30.53 31.69 21.64 28.21 18.04<br />
1 day before 78.10 59.03 59.61 58.94 58.07 53.14 56.04 26.74<br />
Day of change 8.80 72.83 75.35 77 75.57 66.94 64.74 26.52<br />
1 day after 128.48 87.90 82.79 83.18 81.36 73.60 70.93 33.95<br />
2 days after 187.98 89.26 84.05 84.92 85.13 75.63 71.41 35.96<br />
<br />
<br />
<br />
<br />
month maturities were not introduced in the Greek interbank market until the beginning of 1996. Thus,<br />
very few observations were available. Overall, all market rates overreacted to the policy change, except<br />
the 12-month rate.<br />
The second and first periods share similar strong anticipation effects. Again, markets discount more<br />
than half of the operational rate change. However, the difference is that the cumulative response is<br />
smaller. With the exception of the overnight rate, we do not observe any significant learning effects (see<br />
Table 8). This may be for two reasons. Firstly agents had time to get accustomed with market<br />
determined interest rates. Moreover, experience was built in interpreting the central bank moves. Also,<br />
one could argue that the switch to the regular more thinly conducted open market operations as well as<br />
the independence of the central bank made policy more transparent, thus markets did not have to digest<br />
information after the changes in policy. The second is related to the different inflation environment,<br />
with inflation rates being lower in the second period due to the efforts by the government and Bank of<br />
Greece to meet the Maastricht convergence criteria. Another important difference from the first<br />
operational period is that all market rates underreacted, with the significant cumulative changes being<br />
well below 100%.<br />
One can also observe differences along the term structure between the two periods. In the first<br />
period, cumulative responses were fairly uniform across the maturity spectrum (see Tables 2 and 7). In<br />
the second period, however, we observe a pronounced decline in responses along the maturity<br />
spectrum (see Tables 4 and 8). Moreover, we also observe a decline in significance along the maturity<br />
spectrum. This highlights the need to look at individual days within the event window. For instance,<br />
<br />
<br />
Table 5<br />
Statistical results across the event windows: May 1995–March 1997<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month<br />
rate rate rate rate rate rate rate<br />
Normal change (l i )_ 0.003 0.006 0.007 0.009 0.010 0.009 0.009<br />
PPP<br />
CACi 0.3093 0.326 0.322 0.307 0.288 0.137 0.132<br />
rˆ 0.0599 0.1031 0.099 0.0983 0.0874 0.0657 0.0833<br />
h Statistic 5.165* 3.161* 3.252* 3.122* 3.297* 2.084* 1.585<br />
rˆ =The sample standard error obtained using the estimate of the variance of cumulative abnormal changes for market rate i<br />
(Eq. (10)).<br />
h Statistic is computed using rˆ.<br />
* Statistically significant coefficients.<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 501<br />
<br />
<br />
Table 6<br />
Statistical results across the event windows: April 1997–April 2000<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month 10-year<br />
rate rate rate rate rate rate rate rate<br />
Normal change (l i )_ 0.021 0.008 0.008 0.008 0.009 0.010 0.011 0.004<br />
PPP<br />
CACi 0.517 0.372 0.354 0.352 0.340 0.288 0.265 0.140<br />
rˆ 0.592 0.163 0.151 0.149 0.141 0.138 0.137 0.093<br />
h Statistic 0.875 2.282* 2.344* 2.362* 2.411* 2.072* 1.934* 1.505<br />
* Statistically significant coefficients.<br />
<br />
<br />
<br />
<br />
from Table 5, one would conclude by looking at the h statistic that the reaction of the 12-month rate is<br />
not significant. Similar conclusions would be made for the 10-year rate from Table 6. However, these<br />
would be misleading. By checking individual significance levels, we can obtain the significant<br />
component of the reaction, reported in Tables 7 and 8. For the 10-year rate for instance, the significant<br />
element in the response over the event window is one quarter of the change in the operational rate,<br />
down from one-third. Similarly, one could conclude that the 12-month rate in the first sample period<br />
underreacted to changes in the operational rate. However, both for the 12-month as well as for the 9-<br />
month maturities there are few observations available, so the results should be interpreted with<br />
caution.<br />
How do the cumulative reactions obtained for Greece compare to those reported for other countries?<br />
In the first sample period, cumulative responses were fairly uniform across the maturity spectrum, which<br />
contrasts with the evidence for other countries. But in the second period, the cumulative response is<br />
closer to the response observed in the United States (Cook & Hahn, 1989; Thornton, 1998), Germany<br />
(Deutsche Bundesbank, 1996; Hardy, 1998), the UK (Dale, 1993), and Denmark (Pedersen, 1997). We<br />
observe that the common empirical finding of a pronounced decline in responses along the maturity<br />
spectrum is also occurring in Greece.<br />
After entering a lower inflation environment as well as a period of higher credibility towards the end<br />
of the 1990s, the responses of the Greek money market rates seem to match the other countries in the<br />
European Union. This is in line with Buttiglione et al.’s (1997) empirical findings which suggested that<br />
the inflation environment, the credibility of the central bank, as well the state of public finances were<br />
factors found to be closely related to differences and similarities in money market responses to Central<br />
<br />
<br />
<br />
Table 7<br />
h Statistics, indicating the individual day significance in the days surrounding the policy change: May 1995–March 1997<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month<br />
rate rate rate rate rate rate rate<br />
2 days before 6.17* 3.77* 3.85* 3.63* 3.41* 2.92* 2.27*<br />
1 day before 2.83* 2.11* 2.24* 2.64* 2.90* 0.99 1.00<br />
Day of change 19.83* 4.11* 4.59* 4.26* 4.45* 2.11* 1.50<br />
1 day after 7.76* 4.40* 4.41* 3.81* 4.72* 3.05* 2.24*<br />
2 days after 0.84 0.23 0.28 0.61 0.18 0.19 0.52<br />
* Statistically significant coefficients.<br />
502 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
Table 8<br />
h Statistics, indicating the individual day significance in the days surrounding the policy change: April 1997–April 2000<br />
Overnight 1-month 2-month 3-month 6-month 9-month 12-month 10-year<br />
rate rate rate rate rate rate rate rate<br />
2 days before 2.11* 3.53* 1.89* 3.63* 4.14* 2.77* 3.78* 3.77*<br />
1 day before 0.23 3.22* 6.08* 3.68* 3.57* 4.39* 3.83* 1.79*<br />
Day of change 3.17* 1.78* 1.99* 2.33* 2.36* 1.80* 0.97 0.27<br />
1 day after 4.05* 1.77* 0.42 0.26 0.34 0.54 0.45 1.40<br />
2 days after 2.44* 0.02 0.038 0.01 0.26 0.00 0.30 0.30<br />
* Statistically significant coefficients.<br />
<br />
<br />
<br />
Bank rates. In the second sample period, the Greek economy displayed greater convergence to its EU<br />
partners on these factors.<br />
<br />
<br />
6. Conclusions<br />
<br />
This paper has examined an important aspect of the monetary transmission mechanism in Greece<br />
during the transition period of the 1990s, when the operational procedures of the Bank of Greece<br />
underwent a number of major changes. The principal objectives of the research were: first, to provide an<br />
analytical account of the main features of the transition from a system of direct monetary controls to<br />
more indirect methods of conducting monetary policy, where the operating strategy is designed to<br />
influence the price of credit and markets have an important say. Second, to investigate the transmission<br />
process between the Bank of Greece’s operating interest rate instruments and the market interest rates at<br />
various maturities, by applying the event study methodology used in the field of finance.<br />
Our event study results suggest that changes in the official interest rates exert a significant influence<br />
on short-term and intermediate-term market interest rates, and that this relationship was affected by the<br />
changes in the Bank of Greece’s operational procedures during the 1990s. This is reflected in both the<br />
relative strength of anticipation and learning responses of market rates to policy changes, and in the<br />
responses across the maturity spectrum.<br />
It seems that the Greek money markets anticipate the bank’s moves and discount changes in official<br />
rates quickly. This indicates that markets quickly adjusted to a market based system where the central<br />
bank guides the markets through signals, rather than direct actions. We also found significant learning<br />
responses for the first part of our sample period, but not for the second period associated with the latest<br />
operational procedure adopted by the Bank of Greece. These empirical findings appear to suggest that<br />
increased policy transparency may have speeded up the transmission process. Importantly, in the second<br />
period, the response of Greek market interest rates seems to be closer to the response observed in other<br />
more advanced industrial countries. The common empirical finding of a pronounced decline in responses<br />
along the maturity spectrum is also observed in Greece.<br />
Our study could be extended in a number of useful ways. First, one could investigate the links<br />
further down the monetary transmission mechanism. However Greece is reaching the end of the<br />
transition process and another operational regime change might well apply as a result of entrance in<br />
the EMU. Second, it would be interesting to apply the event study methodology to other countries<br />
going through a similar process of changes in the operational procedures of their Central Bank, and in<br />
A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504 503<br />
<br />
<br />
particular to the Central and Eastern European Transition economies, which aim to join the European<br />
Union.<br />
<br />
<br />
Acknowledgements<br />
<br />
This paper was presented at the 2003 annual conference of the European Economics and Finance<br />
Society, University of Bologna, Bologna, Italy, May 2003. We are grateful to the participants of this<br />
conference and to two anonymous referees for their helpful comments. The usual disclaimer applies.<br />
<br />
<br />
References<br />
<br />
Annual Report of the Governor of the Bank of Greece, various issues.<br />
Boehmer, E., Musumeci, J., & Poulsen, A. B. (1991). Event-study methodology under conditions of event-induced variance.<br />
Journal of Financial Economics, 30, 253 – 272.<br />
Borio, C. E. V. (1997, March). Monetary policy operating procedures in industrial countries. Bank for International Settlements<br />
Working Papers, 40.<br />
Brown, S., & Warner, J. (1980). Measuring security price performance. Journal of Financial Economics, 8, 205 – 258.<br />
Brown, S., & Warner, J. (1985). Using daily stock returns: The case of event studies. Journal of Financial Economics 14, 3 – 21.<br />
Buttiglione, L., Del Giovane, P., & Tristani, O. (1997, April). Monetary policy actions and the term structure of interest rates:<br />
A cross-country analysis. Temi di Discussione-Servizio Studi della Banca d’Italia, 306.<br />
Cable, J., & Holland, K. (1999). Regression vs. non-regression models of normal returns: Implications for event studies.<br />
Economics Letters, 64(1), 81 – 85.<br />
Campbell, J. Y., Lo, A. W., & MacKinlay, A. C. (1997). The econometrics of financial markets. Princeton University Press.<br />
Cook, T., & Hahn, T. (1988). The information content of discount rate announcements and their effect on market interest rates.<br />
Journal of Money, Credit and Banking, 20, 167 – 180.<br />
Cook, T., & Hahn, T. (1989). The effect of changes in the federal funds rate target on market interest rates in the 1970s. Journal<br />
of Monetary Economics, 24, 331 – 351.<br />
Dale, S. (1993). The effect of changes in the UK rates on market interest rates since 1987. Manchester School, LXI (Suppl.),<br />
76 – 94.<br />
Deutsche Bundesbank (1996, October). The response of money market rates and short term bank interest rates to changes in<br />
central bank rates, Deutsche Bundesbank Monthly Report.<br />
Dueker, M. J. (1992, July/August). The response of market interest rates to discount rate changes. Federal Reserve Bank of St.<br />
Louis Review, 78 – 91.<br />
European Monetary Institute (1997, January). The single monetary policy in stage three: Specifications of the operational<br />
framework, European Monetary Institute, 9.<br />
Filippides, A. A., Kyriakopoulos, P. E., & Moschos, D. M. (1995). Bank of Greece monetary policy instruments. Bank of<br />
Greece Economic Bulletin, 6.<br />
Garfinkel, R., & Thornton, D. L. (1995). The information content of the federal funds rate: Is it unique? Journal of Money,<br />
Credit and Banking, 27, 838 – 847.<br />
Hardy, D. C. (1998). Anticipation and surprises in central bank interest rate policy. IMF Staff Papers, 45, 647 – 671.<br />
Kuttner, K. N. (2000). Monetary policy surprises and interest rates: Evidence from the Fed funds futures market. Federal<br />
Reserve Bank of New York Staff Paper.<br />
MacKinlay, A. C. (1997, March). Event studies in economics and finance. Journal of Economic Literature XXXV, 13 – 39.<br />
Nafteboriki, daily financial newspaper, various issues. The newspaper archive from July 1996 onwards is available on<br />
www.nafteboriki.gr, in Greek.<br />
Neumann, M. J. M., & Weidmann, J. (1998). The information content of German discount rate changes. European Economic<br />
Review, 42, 1667 – 1682.<br />
504 A. Kaketsis, N. Sarantis / International Review of Economics and Finance 15 (2006) 487–504<br />
<br />
<br />
Paquet, A., & Pe´rez, T. (1995). La Re´action du Marche´ Financier Face a` Diffe´rentes Sources de Signal de la Politique Mone´taire<br />
au Canada. Canadian Journal of Economics, 28, 808 – 821.<br />
Pedersen, S. L. (1997). Market reactions to changes in the Danish discount rate. Danmarks Nationalbank Monetary Review, 3rd<br />
Quarter.<br />
Report on Monetary Policy of the Bank of Greece, various issues.<br />
Roley, V. V., & Sellon, G. H. (1995). Monetary policy actions and long term interest rates. Federal Reserve Bank of Kansas City<br />
Economic Quarterly, 80(4).<br />
Rudebusch, G. D. (1995). Federal Reserve interest rate targeting, rational expectations, and the term structure. Journal of<br />
Monetary Economics, 35, 245 – 274.<br />
Thornton, D.L. (1986, August/September). Discount rate and market interest rates: Theory and evidence. Federal Reserve Bank<br />
of St. Louis Review, 5 – 21.<br />
Thornton, D. L. (1994). Why do T-bill rates react to discount rate changes? Journal of Money, Credit and Banking, 26, 839 – 849.<br />
Thornton, D.L. (1998, November/December). Tests of the market’s reaction to federal funds rate target changes. Federal<br />
Reserve Bank of St. Louis Review, 27 – 36.<br />