Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
1
MONETARY POLICY ANNOUNCEMENTS AND STOCK REACTIONS:
AN INTERNATIONAL COMPARISON
Shen Wang and David G Mayes
University of Auckland
Abstract
This article investigates the impact of domestic monetary policy rate announcements on the stock
markets of New Zealand, Australia, the United Kingdom and the euro area, using event-study
methods to identify stock price reactions to the unanticipated/surprise component of announcements.
As Australia and New Zealand did not reach the zero bound we investigate whether there is an
impact from the global financial crisis on stock market reactions that can be distinguished from the
asymmetric reactions to surprises that characterise the business cycle. We find that the euro area
and the UK both show a financial crisis effect but behaviour in New Zealand and Australia does not
change. We conduct robustness checks and explore confounding factors, especially the impact of
guidance from central banks that prepares markets for policy rate changes.
We have two main aims in this article: first to see whether the financial crisis has affected how stock
prices respond to policy surprises. There is some evidence from the UK (Gregoriou et al., 2009) that
stock price responses became significantly positive during the financial crisis, which implies a striking
change in behaviour. We therefore extend the existing literature to Australia and New Zealand because
these two countries did not reach the zero bound for nominal interest rates and, hence used
conventional policy throughout the crisis period. Beyond short run measures to ensure adequate
liquidity, they did not employ quantitative easing or credit easing in addition to interest rate policy. We
also include the UK and the euro area, which did reach the zero bound, as comparators.
The nature of the likely change in behaviour in a crisis is not completely obvious. It is usually
thought that in a crisis people become much more risk averse. This could mean therefore that they
become more sensitive to monetary policy surprises, particularly negative ones. However, it is also
thought that as monetary policy approaches the zero bound it becomes less effective, because people
can see that conventional monetary policy will soon reach its limits. A negative shock could then
simply accelerate the onset of the belief about policy ineffectiveness and hence show a weakened
response in stock prices. As a by-product of this analysis we also get to test whether the experience
2
recorded for the US, the UK and the euro area in normal times can be extended to Australia and New
Zealand.
Secondly, we seek to substantiate the evidence that the response of markets to monetary policy
surprises varies over the course of the business cycle. There is good evidence that monetary policy
responses to asset prices are themselves asymmetric (Mayes and Viren (2011) for the euro area;
D Agostino et al. (2005) for the US) but little in the reverse direction, although Anderson et al. (2007)
find that stock price responses to positive macroeconomic news, including that from interest rates, is
positive in expansions and negative in contractions.1 Simply put, it is normally thought, on the basis of
previous evidence (Bernanke and Kuttner, 2005; Bohl et al., 2008; Bredin et al., 2007a,b; Honda and
Kuroki, 2006 and Wongswan, 2005), that if there is a positive interest rate surprise this will encourage
markets to fear that there is more adverse information available to the central bank than they had
thought existed and hence the stock price response would be negative. However, in uncertain times
such a surprise might lead markets to believe that policy will be more conducive to steady growth in
the future, as the central bank appears more determined to maintain price stability than was previously
thought. Montagnoli and Mayes (2011) for example show that central banks themselves tend to set
policy differently under greater uncertainty.2 The previous discussion of the influence of the global
financial crisis suggests that the reaction of markets may be different in the down and up phases of the
cycle as well as during uncertainty which is usually associated with turning points.
There is extensive evidence that, in addition to affecting inflation and the real economy,
monetary policy has a clear impact on stock prices (and on house prices) (Iacovello and Minetti, 2003,
2008). Since stock prices are forward looking that influence will come through news and monetary
policy surprises. The reaction to news will incorporate the change the central bank is expected to make
in the settings of policy in the light of that same news. Thus when monetary policy decisions are
announced, what will move stock prices is announcements that are different from those expected. All
of the countries in our sample implement a form of inflation targeting, although this is not how euro
area policy is described by the Eurosystem, and try to make their policy predictable. However, they
typically only announce policy decisions at scheduled meetings. Some countries also offer a projection
1 See also Boyd et al. (2005) for an asymmetric stock price response to labour market news. They find a positive response to bad news and expansions and a negative response in contractions. This they argue is because of the expected response of monetary policy. 2 They consider the Czech, Swedish and UK central banks as these have the longest history of recording perceived uncertainty.
3
of how the policy rate might be expected to evolve in the future in the light of current information and
expected future events. In our sample this is only the case in New Zealand.
Although there is wide debate about the appropriateness of reacting to asset price changes,
including stock prices,3 it is clear that monetary policy does indeed also respond to them in practice
(see Mayes and Viren, 2011, for the case of the euro area and Miller et al., 2002 for the US).4 The
relationship is therefore bi-directional. For market participants, changes in monetary policy have
implications for effective investment and risk management decisions. For central banks, an
understanding of the links between monetary policy and asset prices is fundamental, as has been
demonstrated with unwelcome clarity in the present global financial crisis. They need to understand
both how they can influence stock prices and how that influence impacts on inflation and financial
stability. Our analysis here focuses on how stock markets react to policy surprises. To some extent
monetary policy makers do deliberately seek to surprise markets if conventional policy setting does not
appear to shifting expectations as anticipated. For example, in a crisis interest rates might well be
reduced rather further than appears necessary from pre-crisis behaviour, simply to ensure that markets
get the message that the central bank intends to move firmly to head off any prospect of deflation. By
definition such steps are rare or they would get built into what is expected and no longer be a surprise.
They also do not constitute any attempt to move asset prices by some particular amount.
In common with most studies of announcement affects we apply event-study methods (Bernanke
and Kuttner, 2005) as this enables us to identify the behaviour of stock prices around the specific time
of the announcement and to filter out other extraneous sources of price changes. We are somewhat
restricted in our data as we require on the one hand daily stock prices and on the other a sustained
period where a country has applied a similar monetary policy regime and announced its decisions in the
form of a policy interest rate setting. In the case of euro area we are of course limited by the period of
its existence, however, in the case of New Zealand we are more limited than might be expected, as
although it was the earliest adopter of inflation targeting and was very transparent in its decision
making from as early as 1989, most of the early policy setting was indicative, backed up by the threat
of changes in the quantity of overnight money. Although the target was consistently the 90-day
Treasury Bill rate, this was not the instrument and the policy is aptly described Open Mouth
Operations (Guthrie and Wright, 2000; Mayes and Riches, 1996). It is only since April 1999 that New
3 Bernanke and Gertler (2001), Cechetti et al. (2000), Filardo (2000), Goodhart and Hofmann (2000) 4 Rigabon and Sack (2003) show that a rise in the S&P500 index increases the probability of a monetary policy tightening at the next FOMC meeting in the US.
4
Zealand has used the overnight cash rate (OCR) as its explicit policy variable. Similarly the UK has
only been using the Repo rate as its main instrument since 1997. However this gives us 119
observations up until interest rates fell to the zero bound in the present crisis.5
As the global financial crisis is not yet over, more complex changes in behaviour may well
emerge. At this stage, however, monetary policy makers may wish to reflect on whether changes in the
reaction to policy surprises in a crisis have any implications for policy.
In the rest of the article, Section 1 explains the model and the methodology applied. Section 2
considers the issues posed by our data on the four monetary regimes: New Zealand, Australia, the UK,
and the euro area. In Section 3, we discuss the results. Section 4 concludes.
1. The Model and Methodology
Two main approaches have been used to estimate the impact of monetary policy announcements:
event-study (Bernanke and Kuttner, 2005) and identification-through-heteroskedasticity developed by
Rigobon and Sack (2004). In the event-study approach, the returns of stock indices for a short window
of time round the announcement are regressed against the surprise components of policy rate changes.
The regression coefficient measures the magnitude and direction of the response. Expected policy
changes are usually included in the regression in case expectations are not fully acted upon.6 Under the
identification-through-heteroskedasticity approach, the response of asset prices to policy rate changes is
identified based on the increase in the variance of policy shocks that occurs on days of monetary policy
announcements. The identification-through-heteroskedasticity approach is not appropriate here, since it
does not allow us to test the effects of the financial crisis and the business cycle. Hence we follow
Bernanke and Kuttner (2005) in using an event study. By way of reassurance Rosa (2009) suggests, in
a comparison of the two methods, that the event-study approach is to be preferred. However, the
support is not universal, as Kholodilin et al. (2009) argue that there is downward bias in the event-study
approach in the case of the euro area.
An important concern with the event-study method is the problem of endogeneity, namely the
possibility that the policy interest rate decision itself is affected by recent movements in stock prices.7
5 In the euro area interest rates did not fall to zero but effectively reached a lower bound as the ECB was not willing to accept deposits at zero interest rates. 6 Clearly if rational expectations are the basis of the model then the coefficient on the expected policy change will be zero but there is always the possibility of consistent departures from this in practice. 7 See, for example, Rigobon and Sack (2003) and D Agostino et al. (2005). Also see Mayes and Viren (2011) for an asymmetrical response of monetary policy to asset prices/inflation risks.
5
In any case, other variables may have an impact on both the policy interest rate and stock prices,
thereby distorting the estimates of a model that just considered the influence from monetary policy to
prices. Previous studies have addressed the problem by using a short event window of one day or less.
With a short event window, the joint effect of stock prices on monetary policy is minimised as it is very
unlikely that the policy rate decision would be affected by any stock price changes that occurred earlier
during the announcement day.8 The omitted variable problem is also reduced, and any confounding
news release on the announcement days can be controlled for using dummy variables. The ideal
solution is to use high frequency intra-day data9; however, as such data are not available, especially for
New Zealand and Australia, we use an event window of one day.
1.1 Baseline Model
We can express the relationship between monetary policy and stock prices using the following model
(Bernanke and Kuttner, 2005),
rt = a + b PRe t + c PRu
t + Xtd +
t (1)
where rt refers to the one-day return of a stock index on announcement day t, and PR refers to the
policy rate, e denoting the expected change and u the unexpected change. X is a vector of all the other
identifiable factors, other than policy rate changes, which affect the announcement day returns. t is the
announcement day. a, b, c and d are parameters and
is the residual.
Stock prices are forward looking and should therefore be taking account not simply of all the
known factors that will influence returns but all of the expected future events as well. Monetary policy
decisions, which follow an announced timetable, will form part of that expectation. Thus if
expectations are correct the expected change, PRe, should have no observable impact on returns on
the day it is announced. It is only when the policy rate announced is different from that expected that
returns will be affected; i.e. it is the surprise element in policy that moves prices. We should therefore
find that b is not significantly different from zero.
Clearly the way in which expectations are measured will be crucial to the determination of the
surprise. We assume that the price of policy-rate based futures contracts will be a reasonable measure
of what the market expects. Futures are not traded in the policy rates themselves but in closely related
90 day market rates. Therefore we use the 90-day Bank Bill rates instead of the official cash rate (OCR)
8 It would only be where there is an emergency monetary policy committee meeting to handle a rapid threat to financial stability that such a result might occur and that is not a characteristic of our dataset. 9 See Farka (2009)
6
for New Zealand and the cash rate (CR) for Australia.10 In the case of the UK and the euro area we use
the 3-month LIBOR and the 3-month EURIBOR respectively, instead of the Repo Rate/Official Bank
Rate11 and the Main Refinancing Rate12.
The surprise component is calculated as the one-day change in the futures implied rate:13
PRu t = f m, t
f m, t-1 (2)
where f m, t refers to the futures rate on the announcement day (month m, day t) 14. The expected
component of the rate change is then:
PRe t = PR t
PRu t, (3)
except for the euro area where it is calculated as PRe t = PR t+1
PRu t due to the fact that the daily
EURIBOR is released before the policy rate announcement. The drawback of this approach is that if
the expectation is mismeasured then so also will be the surprise. This could result in biased estimates.15
An alternative expectation might be derived from surveys of market analysts shortly before the
announcement, although it is difficult to get consistent data across the whole time period. Jensen et al.
(1996) and Patelis (1997) have also found clear linkages between monetary policy indicators and stock
returns in a different framework. However, these alternative measures would not be high frequency and
measure expectations on the day of the announcement. It will also be difficult to get equivalent
measures from the four countries. Rosa and Verga (2007) show that it is possible to construct an index
from the ECB s use of code words, particularly in their post-decision press conference which give good
predictability of the nature of the next monetary policy decision. Indeed they suggest that the index
they build from these statements and a simple Taylor rule can improve on futures markets as a forecast
of the actual interest rate change if they are combined. This suggests that measures of expectations
through futures might be inefficient. Despite this, Rosa (2008) argues that the Federal Reserve is more
predictable, although it offers far less explanation of its actions than the ECB. Moreover, on Rosa s
data, which cover the period from 1999 through to mid-2006, the Federal Reserve has a greater impact
on the yield curve through monetary policy surprises than does the euro area for any given surprise.
10 Guender and Rimer (2008) spell out the determinants of the 90-day rate and the implementation of monetary policy in New Zealand. 11 The Repo Rate served as the policy rate during 1997-2005; it was replaced by the Official Bank Rate in 2006. 12 See Gregoriou et al. (2009) for the use of LIBOR and Bohl et al. (2008) for the use of EURIBOR. 13 Kuttner (2001) is usually credited with being the first to adopt this approach. 14 The implied futures rate is calculated as 100 minus the daily settlement price. 15 The evidence is somewhat mixed. Chernenko et al. (2004) present evidence that forward and future prices are generally not pure measures of market expectations as they are heavily affected by the presence of risk premia. However, Piazzesi and Swanson (2008) find that, although excess returns on federal funds rate futures in the US have been positive on average and strongly countercyclical, monetary shocks generated from daily futures prices are robust to time-varying risk premia.
7
(We follow up one specific suggestion in Rosa (2008) by seeing whether the indication given by the
Reserve Bank of New Zealand on the path of future interest rate decisions has an impact
see Section
3.2.1.)
Calculating stock returns from the daily price data is straightforward, using continuous
compounding:
r = ln (Pn+1/ Pn) (4)
where Pn is the closing stock price on day n. There is however an element of choice over which index
to use in measuring the overall market, hence in all cases except New Zealand where there is no
obvious substitute covering the data period, we re-estimate using an alternative market index for each
country. This will act as a check on the robustness of the results. The aggregate stock indices are:
NZXALL of New Zealand, S&P/ASX 200 and FTSEAU of Australia, FTSE100 and FTSEUK of the
United Kingdom, and EUROSTOXX50 and EUROSTOXX of the euro area.
1.2 Asymmetry
The baseline model assumes that the response of stock prices to monetary policy surprises does not
vary according to other factors. However, there are several reasons for suggesting that reactions to
monetary policy may not be symmetric over the economic cycle. Economic behaviour is itself not
symmetric over the course of the cycle. Mayes and Viren (2011) show with European data that there
are two main asymmetries. The best known instance is the Phillips curve. In the up phase of the cycle,
falls in unemployment are associated with increasingly large increases in inflation. In the down phase
of the cycle the curve is much flatter and the same proportionate change in unemployment is associated
with little decrease in inflation. Secondly the relationship between employment and output also varies
across the cycle. The falls in employment (rise in unemployment) associated with falling or slowing
growth are larger than the subsequent rises in employment (falls in unemployment) when output
regains the same levels. Thus the economic downturn results in a permanent reduction in employment
compared to output. An alternative explanation would explain the same results in terms of changes in
the behaviour of productivity (as in the real business cycle literature for example).
Basistha and Kurov (2008) and Farka (2009) show that US stocks respond much more strongly
when economic performance is weak (recessions or easing cycles of monetary policy). Mayes and
Viren (2011) also show that the response of monetary to assets prices varies over the course of the
cycle, which might have a further impact on both the estimates and the responsiveness of stock prices
to monetary policy surprises. Surprises to shift asset prices might also be used both as the cycle nears
8
its peak or trough as policy seeks to moderate the cycle. We follow the earlier work in Mayes and
Viren in using a threshold approach to allowing behaviour to vary across the phases of the business
cycle. The nature of the asymmetry may well be more complex but the limited amount of data
precludes many more sophisticated approaches to estimation than this two regime model.
However, a second type of asymmetry which may exist among the stock responses is asymmetry
due to the sign of the surprise rate change. Markets may respond differently when the surprise is
positive rather than negative; for example, if investors are conservative, they may tend to react more
strongly to bad news than to good news. We therefore also test for this type of asymmetry as well. Of
course the two forms of asymmetry may be interrelated. Bad news in the down phase of the cycle may
generate a more heavily downward shift in stock prices than if it were to occur in the up phase. The
response to positive news may be similarly asymmetric but there is no clear prior reason to expect that
the degree of asymmetry should be the same for positive and negative shocks.
To take account of the business cycle effect, we use the threshold regression approach (Teräsvirta
and Granger, 1993; Tong, 1983), which effectively allows all the coefficients to vary between the up
and down phases of the cycle by including the dummy variable CONTRACT, which equals to 1 for all
observations that fall into the contraction periods determined using OECD s business cycle turning
points, and zero otherwise. This variable is also interacted with the expected and surprise monetary
policy changes.
The same approach is used to test for asymmetry in respect to the response to positive rather than
negative surprises. We use negative surprises as the base case and include a dummy variable
POSITSURP, which equals 1 when the surprise is positive and 0 otherwise. These extra variables and
those discussed later in the context of crises and robustness tests form part of the X vector in the
formulation of the relationship shown in (1).
1.3. The impact of the global financial crisis
Although crises could be treated simply as just a business cycle with a deeper trough, there are reasons
for suggesting that they engender a quantitatively different response. Crises engender fear and an
element of panic that may be absent in a more gentle recession where financial variables do not form
part of the problem in themselves. The global financial crisis has had widespread negative impacts on
financial markets, hence changes in responses to monetary policy can be expected. The crisis effect has
documented by Gregoriou et al. (2009) for the UK, where there is a dramatic shift in stock price
responses, from significantly negative during the pre-crisis period to highly positive during the crisis.
9
They argue that the finding highlights the inability of monetary policy-makers to reverse, via interest
rate cuts, the negative trend observed in stock prices after the onset of the credit crisis. This change in
sign would be quite a dramatic departure for stock price reactions.
However, due to the severity of the global financial crisis, some central banks have been faced
with the zero bound problem in exercising monetary policy. This is seen most obviously in the UK s
Official Bank Rate. From March 2009 onwards, the OBR has remained at 0.5%. The euro area has
been similarly constrained since May 2009. Once the policy rate has reached, or is close to, the zero
bound, the behaviour of the stock market can be expected to be different. Traditional models, such as
the event-study approach used here, can only reflect upside changes, since no further downward
adjustment is possible in either market expectations or the policy rate. Therefore, the zero bound
period also contributes to the difference in behaviour observed during the crisis.
An obvious extension in the crisis period would be to try to take account of the impact of
quantitative easing. However there is no obvious mapping of the quantitative changes and the interest
rate changes so we would not be able to compare crisis and non-crisis periods directly.16 It is also not
quite clear how the futures market would behave in the two periods. Futures can reflect the existence of
what are effectively negative interest rates.
The euro area s response to the crisis has been a little different from that of the Bank of England
but it also effectively hit the zero bound in that any further interest rate reduction would have removed
any remuneration on deposits. However, a further facet which might make the responses during the
financial crisis period different is that central banks have been responding simultaneously to threats to
financial stability and price stability, yet their range of monetary tools to do so is limited and such
measures will have impacts on both objectives. This may therefore increase the uncertainty about what
the measures are intended to achieve and indeed about what they may achieve. This in turn therefore
may alter the response of stock prices to monetary policy innovations in the crisis period. In the case of
the euro area, interest rates were raised in the first part of 2011 (outside the data period) by the
minimum 25 basis points but without withdrawing the liquidity measures that are assisting European
commercial banks and effectively the troubled governments of Greece, Portugal and Ireland since it is
government bonds that the ECB takes as collateral. There is thus no neatness to the policy regime,
which complicates our ability to investigate the period.
16 Clearly we might be able to construct some sort of expectational variable concerning the size of quantitative easing on each announcement day but it is not immediately apparent how that or the surprise would be translated into interest rate space.
10
Since neither Australia nor New Zealand came close to the zero bound nor did they have to do
anything much in the way of extraordinary measures other than a temporary guarantee for new
wholesale borrowing by the banks, we are able to see whether behaviour was different in the financial
crisis from other periods.
To account for the effect of the financial crisis on the stock responses, we effectively divide our
sample into two by introducing a dummy variable CRISIS, which equals to 1 for the crisis period and 0
otherwise. CRISIS is interacted with the expected change and surprise change variables in equation (1),
to form two additional dummy variables in the new regression. Hence all of the parameters are
permitted to be different in the two periods. We tried some experimentation to see how the crisis period
should be defined, as its intensity varied and in some regions the full force of the crisis did not come
through until the collapse of Lehman Brothers in September 2008. We tried altering the onset date over
the plausible range and found that the collapse of Lehman Brothers made the best threshold date for the
euro area, whereas in the UK it makes sense to date it from August 2007 and the problems with
Northern Rock. For Australia and New Zealand, the different dates lead to similar results, so we choose
the earlier date of August 2007. However the impact of the Lehman Brothers collapse acts so much as a
shock that the ensuing month needs to be treated as a special event, with an extra effect over and above
that of the rest of the crisis. What is particularly interesting is that it is not the zero bound period which
distinguishes behavior in the case of the UK and the euro area but the whole of the crisis period, even
when interest rates were positive and the full extent of the problems not anticipated.
The collapse of Lehmans is not the only special event in the data period as the collapse of the
dotcom bubble in 2002 also led to a short run disturbance in stock markets. We therefore account for
these special events by including two further dummy variables, D2002 and DLehman, to account for
the effects of the dot-com bubble burst and the collapse of Lehman Brothers respectively. DLehman
is equal to 1 between 15 September and 3 December 2008 and 0 otherwise,17 and D2002 is equal to 1
in August 2002 and 0 otherwise.18
1.4 Robustness Checks
17 The 3 December date is that used by Gregoriou et al. (2009), which we maintain for comparative purposes. One might wish to alter the length of period as it is difficult to ascribe any specific event as terminating the episode. Fortunately the estimates are not very sensitive to variation over a two month period round this date. An alternative would be to include a measure of abnormal spreads in the estimation itself in an attempt to get a data driven view of the extent of the abnormal period. 18 For Australia, there is no observation for August 2002, so the D2002 variable is not required.
11
We also need to filter out any consistent or identifiable events that might have affected behaviour in
order to get better determined estimates. There is one obvious example in the case of New Zealand. On
many occasions the interest rate announcement is released at the same time as a Monetary Policy
Statement which normally contains a statement about the direction of probable future changes in the
policy rate (see Section 3.2.1 for a more detailed explanation of New Zealand s monetary policy).
While such statements are only a description of what is likely to be needed should events actually
follow the lines suggested in the forecast and analysis, which is published along with the policy
announcement, they will still have an influence on financial markets view of the future. These
influences are likely to be systematic according to sign of the projected changes. We therefore add
three dummy variables to the regression which hold the value unity in the event that there is a
contemporaneous Statement indicating positive, negative or zero interest rate changes in the future. As
in other examples these variables are interacted with PRu t.
These variables will only be approximate as there is a prospective path for interest rates and not
simply an indication of what the next change will be. The ECB goes to a great deal of trouble to
prepare markets for the next interest change through a number of code words, in the main related to the
term vigilant . If the word is not mentioned then a change in interest rates is not likely at the next
meeting. There are other indications released through the press conference that the President of the
ECB holds on the same day as the interest rate announcement.19
Another factor which may affect the outcome of the regressions is inflation targeting. The
practice involves a central bank steering the current inflation rate towards a preset long-term target rate.
Since monetary policy is a major channel for achieving the inflation target, the presence of an inflation
target is likely to alter the financial markets perceptions of monetary policy. Of the four countries in
this study, New Zealand, the UK and Australia have all adopted inflation targets, while the ECB has
been less specific in its approach to inflation, seeking to keep inflation below but close to 2% a year
over the medium term. Our Australian data allow us to test for the impact of the introduction of an
explicit inflation target in 1994 as the Reserve Bank of Australia was announcing its interest rate
decisions before then. The test takes the same form as our other tests for different responses under
different conditions through a dummy variable PRE-IT, which equals to 1 for observations before
19 Rosa and Verga (2007) offer a coherent attempt to identify the various code words and their implication for the euro area policy stance.
12
August 1994 and 0 otherwise, in the regression equations for Australia. PRE-IT is interacted with both
PRe t and PRu
t to form two additional variables in the revised regressions.20
2. Data and Sample
Our period of investigation is limited by when central banks have published monetary policy decisions.
It is only relatively recently that central banks have had an announced schedule of meeting dates for
decision making accompanied by a clearly announced decision at a specific time. It is still the case that
only some explain that decision on the announcement day. The end date simply reflects the timing of
the analysis and is hence arbitrary.
2.1 Policy Rate Announcements
Information on policy rates is on the websites of the Reserve Bank of New Zealand, the Reserve Bank
of Australia, the Bank of England and the European Central Bank. The respective times at which the
rate announcements are made are: 9:00am for New Zealand s OCR, 2:30 pm (Eastern time) for
Australia s CR, 12:00 noon for the UK s Repo Rate or Official Bank Rate, and 1:45 pm (CET) for the
euro area s Main Refinancing Rate. The sample periods for the announcements are listed in the
appendix.
As mentioned above, we use the market-based daily 90-Day Bank Bill, 3-month LIBOR and 3-
month EURIBOR rates instead of the actual policy rates in the analysis in order to match with the
futures contracts. The 90-Day Bank Bill rates for NZ and AU are found on the respective reserve
banks
websites. The data source for the EURIBOR is Datastream, while the daily LIBOR rates come
from the British Bankers Association.21 The daily LIBOR is announced at around the same time as the
UK policy rate announcements, while the EURIBOR rates are announced at 11 am (CET), before the
ECB s policy rate announcements.
Observations are generated by every policy announcement, even though the most common
outcome is no change. A no change decision can represent a surprise just as much as a change and
hence there will be unexpected and expected changes on each occasion.
20 There is some debate over what is the appropriate date for the introduction of inflation targeting in Australia. We follow Bernanke et al. (1999) although it could be argued that it is the exchange of letters Governor Ian Macfarlane and Treasurer Peter Costello in August 1996 that constitutes the formal introduction. The earliest reference is in a speech by the then Governor, Bernir Fraser, in March 1993 (Fraser, 1993). 21 We are grateful to Geoffrey Wood for providing these data
13
2.2 Stock Prices
The daily closing prices of all the stock indices included in this study are obtained from Datastream.
Although we aimed to have two market indices for each country, the NZXALL is the only usable index
for New Zealand since the only other alternative, the NZX50, has too few observations. For Australia,
we include the ASX/S&P200 index, whose start date of June 1992 is slightly later than that of the Cash
Rate announcements. The prices for the other market indices, FTSEAU of Australia, FTSE100 and
FTSEUK of the UK, and EUROSTOXX50 and EUROSTOXX of the euro area, are all available for the
entire date range of the respective policy rate announcements. Clearly, our sample period for each
country is determined by the length of the shortest series.
2.3 Futures Contracts
For calculating the rate surprises, we chose four futures contracts based on the following criteria: first,
the futures contract must be directly based on either the policy rate or a close substitute; and second,
the futures price data must be available since the first policy rate announcement. There are two
potential futures contracts that satisfy the first criterion but not the second
the New Zealand 30 Day
Official Cash Rate Futures, first traded in 2006, and the Australian 30 Day Interbank Cash Rate Futures,
first traded in 2003. Instead of these futures contracts, we use the NZ and AU 90-day Bank Bill futures,
since the 90-day bank bill rates closely follow the policy rates. For the UK and the euro area, we use
the readily available 3-month LIBOR and 3-month EURIBOR futures. All futures prices are obtained
from Datastream.
2.4 Business Cycles
We use the business cycle turning points data provided in the statistics section of the OECD s website
in determining the contraction periods. The data are found under the title OECD Composite Leading
Indicators: Reference Turning Points and Component Series , and are available for all the OECD
countries and major areas. We use the US National Bureau of Economic Research s method to
determine the business cycle phases
the contraction phase is from a peak to a trough, and the
expansion phase is from the trough to the next peak. A cycle is from a peak (trough) to the next peak
(trough). The date of each trough is included in a contraction phase, while the date of each peak is
included in an expansion phase. In our dataset, there are approximately 2.5 cycles for NZ, 3 cycles for
both AU and UK, and 2 cycles for the euro area.
14
2.5 Sample Period
We include as many rate announcement observations as possible for each stock index, while taking into
account the restrictions imposed by the start dates of the individual indices. The main sample period for
each country is: 21/04/1999
26/02/2010 for NZ, 23/01/1990
26/02/2010 for AU, 10/06/1999
26/02/2010 for UK22 and 4/03/1999
26/02/2010 for the euro area. The only index with a sample
period variation is the ASX/S&P200 of Australia, which has a sample period of 08/07/1992
26/02/2010.
2.6 Descriptive Statistics
It is immediately clear from Table 1 that there is large variation in the number of rate announcements
made by each country, reflecting both the start date and the frequency of monetary policy meetings.
Although the Australian sample has the earliest start date among the four countries, it contains the
fewest announcements, due to the fact that the RBA did not announce zero rate changes until 2007. The
ECB has made the most announcements due to its high frequency of meetings. The standard deviations
of the rate surprises show that the AU surprises are the most volatile, and both the UK and euro area
surprises have low volatility. A possible explanation of this is that, as small open economies, Australia
and New Zealand are more open to external shocks. Two other possibilities are that the UK and euro
area markets are either better at predicting policy rate changes, or are consistently biased in their
predictions.
<Insert Table 1 about here>
A comparison of the stock returns standard deviations (last two rows of Table 1) shows that, as is
expected, the volatility in returns on event days is consistently higher than that on the days preceding
the event in each of the four countries. We follow Kholodilin et al. (2009), in considering this single
adjacent period as the comparator. The euro area indices have the highest return volatility on both event
and pre-event days, while the NZXALL index has the lowest. This could be a reflection of the large
geographical and economic coverage of the euro area indices. The AU and UK indices have very
similar standard deviations on both event and pre-event days.
2.7 Outliers
22 The first two years of Repo Rate announcements are excluded due to insignificant reactions. See Section 3.2.3 for a detailed discussion.
15
The crisis period is inevitably a minority of the period under examination but it represents between
11% and 38% of the interest rate decisions taken by each authority. The zero bound period is shorter
and represents 9% of the decision for the UK and 6% for the euro area. Interest rate increases and
decreases have roughly the same frequency over the period for Australia, the euro area and NZ and the
business cycle phases are also well matched for the UK and NZ. The degree of discrepancy in the other
cases is by no means large enough to suggest that the results are due to just a few observations. There
are therefore sufficient observations to get reasonably determined coefficients on the interaction terms
in the equations. Nevertheless to ensure that it is not extreme or unusual observations that are
generating the results, estimates have also been made with such outliers removed.
3. Results and Discussions
3.1 Main Results
If we apply only the most basic regression, where the effects of asymmetry and the crisis are ignored,
only the New Zealand and Australian stock markets display significant reactions to policy rate changes
(Table 2). The NZXALL index and the FTSEAU index both react negatively to surprises, although an
insignificant reaction is observed for ASX200 which has a slightly shorter sample period than FTSEAU.
The size of the NZXALL and FTSEAU surprise coefficients imply an average reaction of about 0.9%
and 0.28% respectively to an unanticipated 25-basis-point rate increase. These reactions are smaller
than that for the US which, according to Bernanke and Kuttner, is about 1%. Surprisingly, we also find
significant negative coefficients for the expected change component.23 This is a departure from theory,
and could be due to the expected rate changes not being fully acted upon prior to the announcement day.
In contrast, the overall responses of UK and euro area market indices to both components are
insignificant. This is largely because, in these regions, behaviour is clearly different in the crisis period
from normal times (Table 3, Panel A). Separating out normal times gives more conventional responses.
<Insert Table 2 about here>
Nevertheless it is worth pursuing this issue of why expected measures may not work as well as
theory predicts. Rosa and Verga (2008) show that the press conference held by the ECB President after
the announcement of the monetary policy decision modifies the market s reaction to the announcement.
23 We had expected that negative coefficient on the expected change would disappear or at least be insignificant when fuller specifications of the model were estimated. However, while significance levels do fall, the expected change term continues to play a role in many of the regressions. The simplest explanation would be that the particular choice of expectations is poorly specified or that people do not fully act on their expectations.
16
Further information, mainly in the form of key phrases or code words, helps the market get a better
understanding of which way interest rates may move in future, which has an immediate impact on the
current rate. Thus since we are measuring surprises on a daily basis and not a tick by tick basis as Rosa
and Verga do, perhaps we are not getting such a clean measure of the surprise from the interest rate
announcement itself. The central banks vary in the amount of information they produce at the time of
the interest rate announcement. The Reserve Bank of New Zealand produces an extensive statement at
the time and the Governor holds a press conference. The Bank of England only reveals the detailed
reasoning with a lag. The euro area as just noted holds an immediate press conference. We would
therefore expect differences between the countries.
As Rosa and Verga (2007) point out, there is some evidence (Paizzesi and Swanson, 2008) that
excess returns on interest rate futures, in the US at any rate, do fluctuate with the economic cycle. In
this case, this effect would need to be removed before the impact of monetary policy could be
evaluated properly. However, the US does not form part of our sample and we are not aware of
evidence that this result is found in the countries that we do study.
While the NZ and AU stock market responses are not significantly affected by the financial crisis,
the crisis effect found by Gregoriou et al. (2009) is present in the euro area as well as in the UK. Prior
to the crisis, all of the market indices of the four countries/regions (including the ASX200 of AU) react
significantly negatively to monetary policy surprises. However, the reactions of the UK and euro area
indices to both expected and surprise components become positive during the crisis period
rather than
serving its original purpose of stimulating the market, a surprise rate cut causes even more pessimism
about economic conditions. The NZ and AU results are in line with expectations since the central banks
of these two countries did not find it necessary to reduce their policy rates to the zero bound, and the
Australian economy was never in recession. Also as expected, the extreme effect of the collapse of
Lehman Brothers in the crisis and the abnormal impact of the collapse of the dotcom boom are
reflected by highly significant negative coefficients of the DLehman and D2002 variables, evident for
all countries except Australia.
<Insert Table 3 about here>
The crisis effect in the UK and the euro area intensifies as the policy rates approach the zero
bound (Table 4). During the zero bound period, the reactions of UK and euro area market indices to
both components are positive and extremely large, especially for the euro area. The sizes of these
coefficients no longer allow any practical interpretation other than that the markets tumbled even faster
after a negative shock. Compared to the zero-bound period, the pre-zero-bound crisis period has a
17
much smaller, although still significant, effect on the responses. Nevertheless it is clear that the change
from normal behaviour occurs around the time of the onset of the crisis in August 2007 rather than
simply later when the zero bound was approached.
<Insert Table 4 about here>
Although the Australian stock price responses are not significantly affected by the financial crisis,
they are significantly procyclical under the business cycle model. The AU market responses to both
expected and surprise rate changes are insignificant during expansions, but significantly negative
during contractions (Table 3, Panel B). (When the pre-inflation-targeting period is separated out, the
response in expansions becomes positive, as discussed in section 3.2.2.) This is similar to the US
results of Basistha and Kurov (2008) which show a stronger response to the surprise rate changes
during recessions. Basistha and Kurov argue that the presence of cyclicality is a reflection of the credit
channel of monetary policy, and that the underlying cause of the difference in response between
expansions and contractions is procyclical fluctuations in both the availability of banks loans and the
credit worthiness of firms. However, there is no evidence for such cyclicality in the stock responses of
the NZ, UK and euro area.
There is also some evidence from our sample that stock responses can differ depending on the
sign of the surprise. Unlike the experience of Bernanke and Kuttner (2005) in finding no evidence of
asymmetry in the response to positive as opposed to negative surprises in the US, we find a stronger
reaction to positive surprises in the case of the NZ stock market. Table 5 shows that the response of the
NZXALL index to positive rate surprises is larger in magnitude than the response to negative surprises
under all regression models. The difference becomes more significant and clearer if we take account of
the confounding from announcements of probable future rate changes (Table 6). This suggests that NZ
investors are generally conservative and are more willing to adjust to bad news than they are to good
news. However, the conservatism hypothesis is not supported by the results of the other countries,
which do not indicate any clear asymmetry due to the sign of the surprise.
<Insert Table 5 about here>
Thus for all countries in our sample we see changes in stock price responses according to
different conditions: for the UK and the euro area it is in the financial crisis, in Australia it is across the
business cycle and in New Zealand it is according to the sign of the policy surprise.
3.2 Explanation of Some of the Anomalies
3.2.1 Projections of future rate changes
18
New Zealand is unique among our sample of countries in that the Reserve Bank of New Zealand s
announcements frequently contain information about the likely direction and approximate timing of
future rate changes.24 For example, the 18 August 1999 news release states: The Reserve Bank today
left the Official Cash Rate (OCR) unchanged. However, it indicated that an increase before the end of
the year is increasingly likely. We therefore augment our analysis by including the nature of the
projected possible rate changes in our explanatory equation. While the path described can be more
complex we focus on simply whether the initial indication for future decisions is for an increase, a
decrease or maintenance of the present setting. This acts as a complement to the work of Rosa and
Verga (2007, 2008) and Rosa (2008) where they show that there is some predictive power in the use of
code words by the ECB as it tries to make sure that markets are not surprised by policy decisions. Put
differently, the announcement effect, and hence the surprise, are effectively moved forward by one
month to the time of the release of the code words. This complicates the analysis. We have not here
attempted to extend Rosa and Verga s dataset on the ECB, which assigns values of +1, 0 and -1
according to the preponderance of coded phrases, to our data period, however, ECB pronouncements
have been used to construct indicator variables in a different context Montagnoli and Mayes, 2011).
Our tests show that an indication of a probable future change has a significant effect on stock
prices. The NZ market responds to a future positive change or a future negative change in the same
direction as the change, while there is very little reaction to an indication of future zero rate change
(Table 6). These responses contrast with the clear negative reaction to a contemporaneous surprise. In
the case of a contemporaneous positive surprise, the impacts of the surprise and a future increase are
similar in size but opposite in sign. Hence if they occur together there would be little net impact on
stock prices. In comparison, the effect of future negative changes on the contemporaneous response is
larger in magnitude; however, it is only significant during the crisis and business cycle contractions,
unlike the impact of future positive changes, which is unaffected by economic conditions.
<Insert Table 6 about here>
The positive reaction to a probable future rate increase could be explained if such a change
signals both projected growth in the economy and strong credibility in the Reserve Bank of New
Zealand s inflation fighting credentials. The signal may be even stronger if it were to occur during the
crisis period (which is not the case in our dataset). Similarly, a probable future rate cut is perceived as
an indicator of weakening economic conditions, especially during the crisis and contractions, when
24 Other countries publish a similar description of likely future interest changes if expected circumstances do not change. Norway and Sweden are perhaps the best known examples.
19
markets are generally pessimistic. When economic conditions are favourable, a future rate cut is given
a more neutral interpretation and treated similarly to a zero future rate change.
3.2.2 The introduction of full inflation targeting in Australia
We find in the ASX200 index a significant change in stock price response to monetary policy after the
adoption of an explicit inflation target (Table 7). While the index s pre-inflation-targeting reaction to
the surprise component is significantly negative, this becomes insignificant after the adoption of the
target, except when the business cycle effect is taken account of. After the separation of the expansions
and contractions in the inflation targeting period, we observe that the ASX200 reacts positively to both
components in expansions and negatively in contractions. These responses suggest that, when inflation
targeting is taking place, the possible effects of rate surprises on inflation become an important driver
of market reactions.
<Insert Table 7 about here>
During expansions, a positive surprise may be welcomed as an extra effort to combat inflation;
while a negative surprise is seen as unfavourable in an already booming economy, when inflation-
reducing measures are much anticipated. As a result we observe a positive response. During
contractions, however, investors become more risk averse. With the slowed economic growth, inflation
is no longer a primary concern; the detrimental affects of positive surprises now exceed the benefits,
and the theoretical negative reactions expected by theory are restored.
The inflation targeting effect is not evident in the FTSEAU index which has an earlier sample
start date, although the two indices are very similar in terms of the asymmetry and crisis effects.
3.2.3 The impact of the choice of sample period on the results for the UK
For our UK sample, we have chosen the start date as June 1999, the same as Gregoriou et al. (2009),25
even though the Repo Rate announcements began in June 1997. We obtained similar results as
Gregoriou et al. under the crisis model, as their reported coefficients for expected and surprise changes
are -8.17 and -6.52 respectively. When the first two years of data are included, however, we observe
insignificant stock responses to both expected and surprise components (Table 8). This shows that our
UK results, and those reported in previous papers, are somewhat sensitive to the choice of sample
period. In particular, the results cannot be extended to the first two years of Repo Rate announcements.
25 We thank Alberto Montagnoli for helpful comments and suggestions about the data.
20
<Insert Table 8 about here>
There are various possible explanations for the insignificant UK stock reactions during the first
two years. The newly established Monetary Policy Committee might have been less consistent in
setting the interest rates, or participants in the stock market may have been less able to derive correct
expectations of the policy rate changes. There is some evidence for the second explanation from our
data. If the market is less able to anticipate policy rate changes, then the volatility of the surprise
component during the first two years should be higher. This is indeed the case -- the volatility during
the first two years, as measured by the standard deviation, is a little over 9 basis points, as compared to
an average of 7 basis points during the years that followed.26
4. Conclusion
This study explores the responses of aggregate stock price indices of New Zealand, Australia, the UK
and the euro area to monetary policy rate announcements. Similar to previous studies, we find
significant negative stock price reactions to monetary policy surprises. We contribute several new
findings to the literature. First, the financial crisis effect identified by Gregoriou et al.(2009) for the
UK is also present in the euro area stock market. Whereas the pre-crisis reactions are significantly
negative, the UK and euro area responses to both expected and surprise rate change components
become positive during the crisis. This effect is amplified during the zero bound period. In contrast, the
New Zealand and Australian stock responses remain negative during the crisis. This is consistent with
the fact that the NZ and AU policy rates did not reach the zero bound.
Second, the Australian stock market response is significantly procyclical. The responses of both
ASX200 and FTSEAU to the rate change components are stronger (more negative) in business cycle
contractions than in expansions. According to Basistha and Kurov (2008), the cyclicality in response is
attributable to the credit channel of monetary policy. Furthermore, we find clear evidence of a change
in response after the onset of full inflation targeting (August 1994) in the ASX200 index. While the
index reacts negatively to surprise rate changes in the pre-inflation-targeting period, the overall
response from August 1994 onwards is insignificant
in expansions, the response is positive,
consistent with the extra inflation reducing effort of a surprise rate increase or the inflation increasing
effect of a surprise rate cut. In contractions, inflation is presumably no longer a primary concern, and
the theoretical negative reactions are restored.
26 Results for other countries are not so sensitive to the choice of the specific data period.
21
Third, we show that the NZ stock market responds more strongly to positive surprises than
negative ones, which lends support to the investor conservatism hypothesis. However, no evidence of a
similar asymmetry is found for the other countries. NZ is unique in our sample in announcing
conditional probable future interest rates changes at the same time as the current policy decision.
Indications of these probable future rate changes are also found to have a significant effect on the NZ
market. A probable future rate increase has a positive effect on the contemporaneous response of the
NZXALL index, while a probable future rate cut has a negative effect. In the case of a future zero
change there is little reaction. Interestingly, the reaction to future rate cuts is only significant during the
crisis and contractions, indicating that a further cut is only regarded as bad news when markets are
generally pessimistic. In contrast, a future rate increase is viewed as a favourable signal regardless of
economic conditions.
Last, our test of an extended sample period for the UK shows that the market response to
monetary policy is insignificant during the first two years of Repo Rate announcements. Hence the
conclusions drawn from the UK sample in our study and those in the existing literature cannot be
applied to this period.
Taken together therefore our results show that while there are some similarities between the US
and Australia, the euro area, New Zealand and the UK in the response of stock prices to monetary
policy surprises, there are also important differences. There are some signs of asymmetry both across
the economic cycle and depending on the sign of the surprise but Australia and New Zealand, which
did not hit the zero nominal interest rate bound in the global financial crisis, do not show a change in
behaviour unlike the euro area and the UK. However, the global financial crisis is not over and the
addition of further data points could lead to different conclusions.
22
References
Anderson, Torben G., Tim Bollerslev, Francis X. Diebold, and Clara Vega, 2007, Real-time price
discovery in global stock, bond and foreign exchange markets, Journal of International
Economics, vol. 73(2), pp. 251 277.
Basistha, Arabinda and Alexander Kurov, 2008, Macroeconomic cycles and the stock market s reaction
to monetary policy. Journal of Banking and Finance, vol. 32(12), pp. 2606-2616.
Bernanke, Ben and Mark Gertler, 2001, Should central banks respond to movements in asset prices?,
American Economic Review, vol. 91(2), pp. 253-7.
Bernanke, Ben S., and Kenneth N. Kuttner, 2005, What explains the stock market s reaction to federal
reserve policy? Journal of Finance, vol. 60(3), pp. 1221-1257.
Bernanke, Ben, Tim Laubach, Frederic Mishkin and Adam Posen, 1999, Inflation Targeting: Lessons
from the International Experience, Princeton NJ: Princeton University Press.
Bredin, Don, Stuart Hyde, Dirk Nitzsche and Gerard O Reilly, 2007a, European monetary policy
surprises: the aggregate and sectoral stock market response, International Journal of Finance and
Economics, vol. 14(2), pp. 156-171.
Bredin, Don, Stuart Hyde, Dirk Nitzsche and Gerard O Reilly, 2007b, UK stock returns and the impact
of domestic monetary policy shocks, Journal of Business Finance and Accounting, vol. 34(5-6),
pp. 872-888.
Bohl, Martin T., Pierre L. Siklos, and David Sondermann, 2008, European stock markets and the ECB's
monetary policy surprises, International Finance, vol. 11(2), pp. 117-130.
Boyd, John H., Jian Hu and Ravi Jagannathan, 2005, The stock market s reaction to unemployment
news: why bad news is usually good for stocks. Journal of Finance, vol. 60(2), pp. 649-672.
Cecchetti, Stephen, Henk Genburg, John Lipsky and Sushil Wadhwani, 2000, Asset Prices and
Monetary Policy, Geneva Report on the World Economy, no.2, London: CEPR.
Chernenko, Sergey V., Krista B. Schwarz and Jonathan H. Wright, 2004, The information content of
forward and futures prices: market expectations and the price of risk, FRB International Finance
Discussion Paper 808, Board of Governors of the Federal Reserve System.
D Agostino, Antonello, Luca Sala and Paolo Surico, 2005, The fed and the stock market, Computing in
Economics and Finance 2005 Conference Paper 293, Society for Computational Economics.
23
Farka, Mira, 2009, The effect of monetary policy shocks on stock prices accounting for endogeneity
and omitted variable biases, Review of Financial Economics, vol. 18(1), pp. 47-55.
Filardo, A. J., 2000, Monetary policy and asset prices, Federal Reserve Bank of Kansas City Economic
Review, vol. 85(3), pp. 11-37.
Fraser, Bernie, 1993, Some aspects of monetary policy, Reserve Bank of Australia, Bulletin, April,
pp.1-7.
Goodhart, Charles and Boris Hofmann, 2000, Asset prices and the conduct of monetary policy, mimeo,
LSE.
Gregoriou, Andros, Alexandros Kontonikas, Ronald MacDonald and Alberto Montagnoli, 2009,
Monetary policy shocks and stock returns: evidence from the British market. Financial Markets
and Portfolio Management, vol. 23(4), pp. 401-410.
Guender, Alfred and Oyvinn Rimer, 2008, The implementation of monetary policy in New Zealand:
what factors affect the 90-day bill rate?, North American Journal of Economics and Finance,
vol.19(2), pp. 215-34.
Guthrie, Graeme and Julian Wright, 2000, Open Mouth Operations, Journal of Monetary Economics,
vol. 46(2), pp. 489-516.
Honda, Yuzo and Yoshihiro Kuroki, 2006, Financial and capital markets responses to changes in the
central bank s target interest rate: the case of Japan, The Economic Journal, vol. 116(513), pp.
812 842.
Iacoviello, Marco and Raoul Minetti, 2003, Financial liberalization and the sensitivity of house prices
to monetary policy: theory and evidence, Manchester School, vol. 71(1), pp. 20-34.
Iacoviello, Marco and Raoul Minetti, 2008, The credit channel of monetary policy: Evidence from the
housing market, Journal of Macroeconomics, vol. 30(1), pp. 69-96,
Jensen, Gerald R., Jeffrey M. Mercer, and Robert R. Johnson, 1996, Business conditions, monetary
policy and expected security returns, Journal of Financial Economics, vol. 40(2), pp. 213-237.
Kholodilin, Konstantin, Alberto Montagnoli, Oreste Napolitano and Boriss Siliverstovs, 2009,
Assessing the impact of the ECB's monetary policy on the stock markets: a sectoral view,
Economics Letters, vol. 105(3), pp. 211-213.
Kuttner, Kenneth N., 2001, Monetary policy surprises and interest rates: Evidence from the Fed funds
futures market, Journal of Monetary Economics, vol. 47(3), pp. 523 544.
Mayes, David G. and Brendon Riches, 1996, The Effectiveness of Monetary Policy in New Zealand,
Reserve Bank of New Zealand Bulletin, vol. 59(1), pp. 5-20.
24
Mayes, David G. and Matti Viren, 2011, Asymmetry and aggregation in the EU, Baingstoke: Palgrave-
Macmillan.
Miller Marcus, Paul Weller and Lei Zhang, 2002, Moral hazard and the US stock market: analysing the
Greenspan Put, Economic Journal vol. 112(478), pp. C171-86.
Montagnoli, Alberto and David G. Mayes, 2011, Uncertainty and Monetary Policy, forthcoming in
P.Siklos and J-E Sturm (eds) Central Bank Communication, Decision Making and Governance,
Cambridge MA: MIT Press.
Patelis, Alex D., 1997, Stock return predictability and the role of monetary policy, Journal of Finance,
vol. 52(5), pp. 1951-1972.
Piazzesi, Monika, and Eric T. Swanson, 2008, Futures prices as risk-adjusted forecasts of monetary
policy, Journal of Monetary Economics, 55(4), pp. 677-691.
Rigobon, Roberto and Brian Sack, 2003, Measuring the response of monetary policy to the stock
market, The Quarterly Journal of Economics, 118(2), pp. 639-669.
Rigobon, Roberto, and Brian Sack, 2004, The impact of monetary policy on asset prices, Journal of
Monetary Economics, 51(8), pp. 1553-1575.
Rosa, Carlo, 2008, Talking less and moving the market more: is this the recipe for monetary policy
effectiveness? Evidence from the ECB and the Fed , CEP Discussion Paper no. 855, LSE.
Rosa, Carlo, 2009, The validity of the event-study approach: evidence from the impact of the Fed's
monetary policy on US and foreign asset prices, Economica Early View.
Rosa, Carlo and Giovanni Verga, 2007, On the consistency and effectiveness of central bank
communication: evidence from the ECB, European Journal of Political Economy, vol. 23(2), pp.
146-175.
Rosa, Carlo and Giovanni Verga, 2008, The impact of central bank announcements on asset prices in
real time , International Journal of Central Banking, vol. 4(2), pp. 175-217.
Teräsvirta, Timo and Clive Granger, 1993, Modelling Nonlinear Economic Relationships, Oxford:
Oxford University Press.
Tong, Howell, 1983, Threshold Models in Nonlinear Time Series Analysis, New York: Springer Verlag.
Wongswan, Jon, 2005, The response of global equity indexes to U.S. monetary policy announcements,
Journal of International Money and Finance, vol. 28(2), pp. 344-365.
25
Appendix: Sample Periods for Policy Rate Announcements
Country Start Date Description of Start Date End Date
NZ 21/04/1999
first published OCR policy decision
26/02/2010 AU 23/01/1990
first published monetary decision
UK 06/06/1997
first announcement after intro. of Repo Rate
euro area 4/03/1999 first published monetary decision
26
Table 1 Descriptive Statistics
This table reports selected descriptive statistics for policy rate surprises and returns of equity market indices of New Zealand, Australia, the United Kingdom and the euro area. Sample period is: 21Apr99
26Feb10 for NZ, 08Jul92
26Feb10 for ASX200, 23Jan90
26Feb10 for FTSEAU, 10Jun99
26Feb10 for UK, and 4Mar99
26Feb10 for euro area. Same as Kholodilin et al. (2009), we define non-event days as the days preceding the event
days. NZ Index
AU Indices
UK Indices
euro area Indices
NZXALL ASX200
FTSEAU
FTSE100
FTSEUK
EUROSTOXX50
EUROSTOXX
Number of events in sample: policy rate announcements 88 57 69 130 130 167 167 Standard deviation of rate surprise, basis points 10 15 19 7 7 6 6 Standard deviation of equity return on event days, % 0.80 1.33 1.29 1.29 1.27 1.78 1.60 Standard deviation of equity return on nonevent days, % 0.74 1.07 1.08 1.10 1.08 1.44 1.33
Table 2 Baseline Regression Results
This table reports the results of the baseline regression, where the indices' returns are regressed against the expected and surprise rate change components. Sample size is: 88 for NZXALL, 57 for ASX200, 69 for FTSEAU, 130 for FTSE100 and FTSEUK, and 167 for EUROSTOXX50 and EUROSTOXX. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively.
NZ Index
AU Indices
UK Indices
euro area Indices
Regressor NZXALL ASX200 FTSEAU FTSE100 FTSEUK EUROSTOXX50 EUROSTOXX Intercept 0.000 0.000 0.000 -0.002 -0.002 -0.001 -0.001
(0.230) (0.093) (0.121) (1.492) (1.544) (1.023) (1.111) Expected rate change -4.787** -2.417** -3.577*** -3.073 -3.177 -8.057 -5.685
(2.475) (2.065) (3.791) (1.107) (1.133) (1.631) (1.369) Surprise rate change -3.694*** -0.601 -1.127** 2.188 2.137 -0.225 1.156
(5.076) (0.575) (2.580) (0.519) (0.511) (0.064) (0.382)
R 2 0.193 0.020 0.079 0.076 0.080 0.046 0.041 Adjusted R2 0.174 -0.017 0.051 0.061 0.065 0.034 0.030
27
Table 3
The Crisis and Business Cycle Effects
This table reports the results of the crisis effect and business cycle effect regressions for each market index. Sample size is: 88 for NZXALL, 57 for ASX200, 69 for FTSEAU, 130 for FTSE100 and FTSEUK, and 167 for EUROSTOXX50 and EUROSTOXX. Crisis is defined as August 2007 onwards for NZ, AU and UK, and end of September 2008 onwards for EA. Contraction is determined using OECD business cycle turning points. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively.
Panel A: The crisis effect
NZ Index
AU Indices
UK Indices
euro area Indices
Regressor NZXALL ASX200 FTSEAU FTSE100 FTSEUK EUROSTOXX50 EUROSTOXX Intercept 0.000 0.000 0.000 -0.001 -0.001 0.000 0.000
(0.668) (0.045) (0.054) (0.588) (0.646) (0.068) (0.062) Expected rate change -6.015** -2.417 -3.690*** -6.884*** -7.059*** -12.117*** -9.337***
(2.374) (1.510) (3.847) (3.968) (4.035) (3.971) (3.797) Surprise rate change -3.424*** -1.265* -1.415*** -5.130** -5.112** -3.901** -2.188
(3.123) (1.862) (3.793) (2.180) (2.196) (2.347) (1.620) Expected rate change * CRISIS 2.753 4.699 6.280 13.993*** 14.321*** 35.324*** 31.618***
(0.814) (0.784) (1.105) (3.638) (3.696) (3.278) (3.393) Surprise rate change * CRISIS 0.144 5.697 6.403 10.268*** 10.479*** 37.933** 33.874**
(0.086) (0.827) (0.941) (2.914) (2.991) (2.057) (2.095) dLehman -0.029*** -0.008 -0.008 -0.056*** -0.055*** -0.034*** -0.033***
(17.922) (0.542) (0.542) (14.582) (14.241) (2.799) (2.961) d2002 -0.004*** - - -0.050*** -0.048*** -0.048*** -0.038***
(4.258) - - (42.170) (42.801) (39.170) (35.592)
R2 0.353 0.070 0.133 0.436 0.438 0.267 0.263
Adjusted R 2 0.305 -0.021 0.064 0.408 0.411 0.240 0.236
28
Table 3, continued
Panel B: The business cycle effect
NZ Index
AU Indices
UK Indices
euro area Indices
Regressor NZXALL ASX200 FTSEAU FTSE100 FTSEUK EUROSTOXX50 EUROSTOXX Intercept 0.001 0.000 0.000 -0.001 -0.001 0.000 0.000
(0.948) (0.123) (0.174) (0.984) (1.056) (0.276) (0.418)
Expected rate change -4.114** 1.682 1.442 -3.341 -3.519 -9.243** -8.557** (2.289) (0.869) (0.976) (0.961) (1.027) (2.569) (2.430)
Surprise rate change -3.451*** 2.008 1.805 -5.768 -5.649 -0.553 -0.480 (3.677) (1.234) (1.637) (1.575) (1.592) (0.249) (0.237)
Expected rate change * CONTRACT -0.597 -5.023** -5.430*** -0.146 -0.025 1.474 3.635 (0.188) (2.002) (2.924) (0.032) (0.005) (0.246) (0.681)
Surprise rate change * CONTRACT 0.296 -3.519* -3.258** 5.263 5.106 -2.908 -0.741 (0.209) (1.775) (2.498) (1.318) (1.302) (0.779) (0.232)
dLehman -0.030*** -0.011 -0.012 -0.049*** -0.048*** -0.050*** -0.046*** (22.223) (0.763) (0.777) (9.968) (9.935) (5.403) (5.325)
d2002 -0.005*** - - -0.048*** -0.046*** -0.048*** -0.038*** (6.173) - - (39.688) (38.473) (36.698) (33.531)
R2 0.347 0.067 0.125 0.410 0.408 0.220 0.216
Adjusted R 2 0.298 -0.025 0.056 0.382 0.379 0.191 0.186
29
Table 4
Effect of Zero Bound Period on UK and EA Responses
This table reports the effect of the zero bound period on the response of UK and euro area stock markets to expected and surprise rate changes. Only one index per country is shown as the alternative index has very similar results. Zero bound period is defined as March 2009 onwards for the UK and May 2009 onwards for the euro area. Pre-ZB crisis period is defined as Aug 2007
Mar 2009 for the UK and Sep
2008
May 2009 for the euro area. Sample size is 130 for FTSE100 and 167 for EUROSTOXX 50. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively. Regressor FTSE100
EUROSTOXX50
Intercept -0.001 -0.001 0.000 0.000
(1.032) (0.780) (0.000) (0.224) Expected rate change -4.472** -6.901*** -7.689* -12.089***
(2.143) (3.885) (1.803) (3.945) Surprise rate change -3.415* -5.171** -2.308 -3.896**
(1.674) (2.166) (1.010) (2.327) Expected rate change * PreZBCRISIS - 10.890*** - 34.348**
- (3.782) - (2.269) Surprise rate change * PreZBCRISIS - 9.113*** - 37.013
- (2.823) - (1.138) Expected rate change * ZB 39.052*** 41.485*** 252.770*** 259.388***
(2.607) (2.655) (3.571) (3.706) Surprise rate change * ZB 25.729* 27.531* 230.906*** 234.590***
(1.677) (1.730) (3.541) (3.644) dLehman -0.055*** -0.053*** -0.049*** -0.035*
(11.931) (17.509) (5.352) (1.740) d2002 -0.049*** -0.049*** -0.048*** -0.049***
(39.008) (41.741) (39.399) (39.159)
R2 0.444 0.467 0.236 0.280
Adjusted R 2 0.417 0.431 0.207 0.244
30
Table 5 Asymmetrical Effect of Positive Surprise
This table reports the results of regressions which isolate the effect of positive surprises. Sample size is: 88 for NZXALL, 69 for FTSEAU, 130 for FTSEUK, and 167 for EUROSTOXX50. Only one index is shown per country as the alternative index has similar results. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively.
Regressor NZXALL
FTSEAU
FTSEUK
EUROSTOXX50
Intercept 0.002 0.002 -0.001 0.000 -0.001 -0.001 -0.001 -0.001
(1.525) (1.656) (0.523) (0.135) (1.141) (1.056) (0.489) (0.796) Expected rate change -6.116** -4.554** -3.289*** 1.452 -7.378*** -3.667 -11.546*** -7.927*
(2.643) (2.491) (2.979) (0.965) (4.472) (1.068) (3.942) (1.793) Surprise rate change -1.924 -1.941* -1.896*** 1.875 -7.356** -6.382 -5.697** -2.352
(1.231) (1.897) (2.653) (0.897) (2.014) (1.448) (2.165) (0.831) Surprise rate change * POSITSURP -3.283 -2.927 1.862 -0.078 3.434 0.878 3.831 4.328
(1.563) (1.594) (1.219) (0.041) (0.786) (0.242) (0.830) (0.963) Expected rate change * CRISIS 2.748 - 6.994 - 16.328*** - 34.966*** -
(0.821) - (1.245) - (3.673) - (3.249) - Surprise rate change * CRISIS -0.598 - 6.572 - 12.782*** - 37.722** -
(0.364) - (0.973) - (2.852) - (2.073) - Expected rate change * CONTRACT - -0.107 - -5.464*** - 0.231 - 0.538
- (0.034) - (2.742) - (0.049) - (0.088) Surprise rate change * CONTRACT - -0.379 - -3.315* - 5.584 - -3.180
- (0.305) - (1.797) - (1.371) - -(0.820) dLehman -0.027*** -0.028*** -0.008 -0.012 -0.058*** -0.049*** -0.035*** -0.050***
(16.308) (21.229) (0.532) (0.771) (10.764) (7.823) (2.901) (5.739) d2002 -0.004*** -0.005*** - - -0.048*** -0.046*** -0.048*** -0.048***
(4.410) (5.568) - - (39.123) (38.578) (39.780) (35.919)
R2 0.369 0.358 0.140 0.125 0.441 0.408 0.270 0.223
Adjusted R2 0.313 0.302 0.057 0.041 0.409 0.374 0.238 0.189
31
Table 6
Confounding Factors in NZ Regressions
This table reports the revised regression results for the NZXALL index, after taking into account the confounding factors of probable future rate changes. Column (a) is the baseline regression, column (b) identifies the crisis and business cycle effects, and column (c) tests for asymmetry according to the sign of surprise. The sample contains 88 observations. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively.
NZXALL
Regressor (a) (b) (c) Intercept 0.000 0.000 0.000 0.001 0.001 0.001
(0.223) (0.353) (0.577) (1.624) (1.530) (1.606) Expected rate change -5.507** -5.993** -4.085** -5.545** -6.043** -4.568**
(2.344) (2.289) (2.361) (2.328) (2.533) (2.539) Surprise rate change -3.570*** -3.818*** -4.102*** -1.966 -2.196 -2.438**
(4.168) (3.063) (3.815) (1.636) (1.325) (2.054) Surprise rate change * POSITSURP - - - -4.459** -3.774* -3.432*
- - - (2.037) (1.909) (1.897) Surprise rate change * FUTPOSIT 2.750** 2.767** 3.234*** 3.804*** 3.404*** 3.707***
(2.543) (2.247) (2.942) (3.409) (3.288) (3.828) Surprise rate change * FUTNO -4.703 -4.184 -3.536 -3.683 -3.648 -2.956
(0.666) (0.585) (0.454) (0.514) (0.499) (0.389) Surprise rate change * FUTNEG -11.165* -4.453 -5.215 -11.080* -4.446 -5.274
(1.796) (0.996) (1.410) (1.955) (1.008) (1.430) Expected rate change * CRISIS - 2.202 - - 2.132 -
- (0.546) - - (0.524) - Surprise rate change * CRISIS - 0.816 - - 0.063 -
- (0.437) - - (0.036) - Expected rate change * CONTRACT - - -1.065 - - -0.510
- - (0.301) - - (0.145) Surprise rate change * CONTRACT - - 1.022 - - 0.320
- - (0.671) - - (0.249) dLehman - -0.024*** -0.024*** - -0.022*** -0.022***
- (5.967) (6.043) - (5.062) (5.224) d2002 - -0.004*** -0.005*** - -0.004*** -0.005***
- (3.932) (5.929) - (4.295) (5.625)
R2 0.300 0.375 0.376 0.331 0.395 0.392
Adjusted R2 0.257 0.303 0.304 0.281 0.317 0.313
32
Table 7 The Effect of Inflation Targeting
This table reports the revised regression results for the Australian market indices, showing the effect of the pre-inflation-targeting variable on the response coefficients. Column (a) is the baseline regression, column (b) identifies the crisis and business cycle effects, and column (c) tests for asymmetry according to the sign of surprise. Sample size is 57 for ASX200 and 69 for FTSEAU. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively.
ASX200 Index
FTSEAU Index
Regressor (a) (b) (c) (a) (b) (c) Intercept 0.000 0.000 0.000 0.000 0.000 0.000 0.000 -0.001
(0.057) (0.079) (0.185) (0.062) (0.171) (0.048) (0.207) (0.377) Expected rate change -2.194 -2.024 4.475** 2.826 -2.519* -2.478 2.082 2.142
(1.570) (0.916) (2.460) (1.207) (1.874) (1.105) (1.500) (1.564) Surprise rate change -0.478 -1.144 3.403** 4.436 -0.558 -1.296** 2.060** 1.008
(0.422) (1.515) (2.529) (0.561) (0.523) (2.019) (2.028) (0.340) Surprise rate change * POSITSURP - - - -1.897 - - - 1.253
- - - (0.237) - - - (0.393) Expected rate change * PRE-IT -1.911 -1.997 -7.126*** -0.797 -1.724 -1.697 -1.606 -2.158
(1.489) (0.969) (3.299) (0.902) (1.265) (0.771) (1.219) (1.054) Surprise rate change * PRE-IT -3.918*** -3.396** -6.509*** -7.553** -0.856 -0.084 -0.156 -0.019
(3.103) (2.523) (5.599) (2.144) (0.688) (0.112) (0.126) (0.014) Expected rate change * CRISIS - 4.301 - - - 5.067 - -
- (0.682) - - - (0.829) - - Surprise rate change * CRISIS - 5.565 - - - 6.283 - -
- (0.790) - - - (0.909) - - Expected rate change * CONTRACT - - -7.605*** -7.446 - - -5.174*** -4.582**
- - (3.521) (1.118) - - (3.030) (2.029) Surprise rate change * CONTRACT - - -4.862*** -5.633 - - -3.348** -2.567
- - (3.101) (0.947) - - (2.210) (1.104) dLehman - -0.008 -0.011 -0.011 - -0.008 -0.012 -0.012
- (0.537) (0.742) (0.766) - (0.534) (0.777) (0.772)
R2 0.023 0.073 0.084 0.076 0.085 0.135 0.129 0.130
Adjusted R 2 -0.052 -0.059 -0.046 -0.078 0.028 0.036 0.029 0.014
33
Table 8 Effect of Extended UK Sample Period
This table reports variations in the results of UK market indices under the crisis model, when an extended sample period of Jun 1997
Feb 2010 is used. The main sample contains 130 observations for both
indices; the extended sample contains 154 observations. Parentheses contain t-statistics, calculated using Newey-West heteroskedasticity-consistent estimates of the standard errors. ***, **, * indicate statistical significance at the 1, 5, 10 % level, respectively.
Main Sample (Jun99 - Feb10)
Jun97 - Feb10
Regressor FTSE100 FTSEUK FTSE100 FTSEUK Intercept -0.001 -0.001 -0.001 -0.001
(0.588) (0.646) (1.116) (1.162) Expected rate change -6.884*** -7.059*** -3.699 -4.010
(3.968) (4.035) (1.350) (1.512) Surprise rate change -5.130** -5.112** -1.713 -1.888
(2.180) (2.196) (0.587) (0.677) Expected rate change * CRISIS 13.993*** 14.321*** 10.457** 10.944**
(3.638) (3.696) (2.303) (2.438) Surprise rate change * CRISIS 10.268*** 10.479*** 6.733* 7.144*
(2.914) (2.991) (1.682) (1.832) dLehman -0.056*** -0.055*** -0.055*** -0.054***
(14.582) (14.241) (13.954) (13.731) d2002 -0.050*** -0.048*** -0.048*** -0.047***
(42.170) (42.801) (29.232) (29.963)
R 2 0.436 0.438 0.344 0.352 Adjusted R 2 0.408 0.411 0.317 0.325