# Relationship between monetary policy announcements and the foreign exchange market

### 1 Introduction

Over the last few decades a considerable effort has been spent in order to explain causes of financial market volatility. Many of the studies in this field concentrate on the link between macroeconomic news and foreign exchange volatility. The main purpose of this study is to give insight into foreign exchange rate volatility and how it is affected by monetary policy announcement. I will also investigate how exchange rate volatility changes around monetary policy decision days, and whether there are any pre-announcement effects around interest rate decision meetings. Financial theory and previous empirical work supports hypothesis about pre-announcement effect in different financial markets. (see Jones, 1998; Li and Engle, 1998; French, 1989).

In this study I will empirically investigate the relationship between monetary policy announcements and one of the most important financial markets, the foreign exchange market, by using a generalised autoregressive heteroscesastic (GARCH) framework. I choose currency market, as this market is the biggest (in terms of value traded) and one of the most closely monitored asset markets in the economy and generally regarded as being highly sensitive to political and economic changes.

However, to the knowledge of the author of this paper there is no study which concentrates only on Bank of England interest rate announcements and sterling volatility. Similar research by Malvin et al (2009) examines Bank of England announcement and foreign exchange market; however Melvin focuses only on dollar-sterling volatility, while this paper also looks at the volatility of other main currencies against British pound. (Data used in this paper is more recent and )

Exchange rate movements have an important implication on many issues in international finance.

Volatility is a key variable which permeates most financial instruments and plays a central role in many areas of finance. For example, volatility is crucially important in asset pricing models and dynamic hedging strategies as well as in the determination of options prices. From an empirical standpoint, it is therefore of utmost importance to carefully model any temporal variation in the volatility process. The ARCH model and its various extensions have proven very effective tools along these lines. (ARCH modelling in finance)

The rest of the paper is organised as follows. In Chapter 2 I begin with brief theory of monetary policy in the United Kingdom, some background about MPC meetings and the relationship between monetary policy and exchange rates. Chapter 3 presents short summary of previous empirical work on the volatility effects of macroeconomic news announcements. . Chapter 5 describes data and presents empirical methodology. Chapter 5 contains main empirical findings. Finally, Chapter 6 restates general results and gives an overall conclusion of this study.

### 2. Interaction of Monetary policy and Exchange rates

First to test whether interest rate announcements affect exchange rate volatility it is necessary to look at the brief theory monetary policy in the United Kingdom, how interest rate announcements are made and what economic intuition is behind interest rate and exchange rate movements.

### 2.1. Monetary policy

Monetary policy can be defined as the process by which government, central bank or other money authority tries to achieve a set of objectives that are expressed in terms of macroeconomical variables. While different countries use different monetary policy regimes to target variables such as inflation, real output, employment or exchange rates, the most widely used is inflation targeting, where central banks adjust monetary base in order to achieve target inflation level.

In case of United Kingdom, Bank of England (BoE), central bank of the UK, handles monetary policy. The BoE has two main purposes- monetary and financial stability The Bank sets the official interest rate, on the basis of detailed monthly assessments of trends in the economy, to …………………

Monetary stability -

Financial stability -

Although sterling exchange rate is free-floating and there is no official exchange rate target for British economy since 1992 when UK suspended its membership in the European exchange rate mechanism. The Monetary Policy Committee has occasionally discussed the relative merits and de-merits of intervening in the current markets to influence the external value of the pound

### 2.2. The Monetary Policy Committee

As mentioned above one of the two core purposes of the Bank of England is stable prices and confidence in the currency. Price stability is defined by Government's inflation target (on average of 2% since December 10, 2003) which is announced each year by the Chancellor of the Exchequer in the annual Budget statement. The Bank seeks to meet the inflation target by setting the official interest rate.

Since May 1997 (when BoE gained independence) the level of interest rates is decided by a special committee - Monetary Policy Committee (MPC)[1]. The MPC is chaired by the Governor of the Bank of England and consists of nine members (five from Bank of England and four external members appointed by the Chancellor). The MPC meets every month for two-day meeting, usually on the Wednesday and Thursday after the first Monday on each month. The meeting dates for each year are published well in advance of the meetings and are available from Bank of England website.[2]

All throughout the month, the Committee receives extensive briefing on the state of the UK and world economy. This includes briefing meeting (known as pre-MPC meeting), which normally takes place on the Friday prior to each meeting. During this meeting the latest data on the economy as well as analysis of recent trends and relevant issues are provided by senior BoE staff. On the first day of MPC meeting itself, the Committee discusses the most recent economic data. On the second day, each member of the Committee explains their views on what policy should be and The Governor then offers a motion that he believes will result in majority vote and calls for a vote. The decisions are announced at 12 noon immediately after second day meeting. Beginning at 12:15 pm policy is implemented with open market operations.

### 2.3. Monetary approach of exchange rate determination

The monetary approach to exchange rate states that nominal exchange rates are determined by excess supply of money between countries (Islam 2006).

The monetary approach to exchange rates determinations, developed in 1976 by Frenkel and Johnson, argues that exchange rate movements are reflected by disequilibria in money markets.

The monetary approach to exchange rates sees the exchange rate as the price of foreign money in terms of domestic money, determined in turn by the demand and supply of money.

The monetary approach to exchange rates determination provides a monetarist explanation of exchange rate determination.

In economics interest rates are highly correlated with changes in currency values.

According Bank of England, exchange rate shows the relative price of domestic and foreign money and highly depends on both domestic and foreign monetary conditions. An increase in rates of interest in the UK relative to those in other countries would result in an increase in the amount of funds flowing into the UK, as higher interest rate would give investors a higher return in UK assets relative to their foreign-currency equivalents. This should result in an appreciation of the sterling exchange rate against other currencies (as demand for pound increases). However, in practice the relationship between changes in the interest rates and exchange rates is never simple. The exact impact on exchange rates of an official change is uncertain, as it will depend on expectations about domestic and foreign interest rates and inflation, with may themselves be affected by policy change. The exchange rates are influenced both by expectations about future interest rates and unexpected changes in interest rates.

### 3. Review of previous empirical work

4. Empirical Methodology

4.1. Data

To examine the effects of Monetary Policy Committee meetings on exchange rate volatility I used daily observations of nominal exchange rate between UK sterling and Euro, US dollar and Japanese yen. Observations run from 1 January 1999 to 30 March 2010. The beginning of the sample is dictated by the introduction of Euro as an accounting currency in 1 January 1999. The eleven-year sample period includes periods of both high and low interest rates, it also includes recent global economic recession, when financial markets where surrounded by high uncertainty and the magnitude of interest rates cuts was hard to predict. The MPC meets each month twelve times a year, so my sample period contains 135 interest rate decisions[3]. During sample period MPC reduced interest rates 23 times and raised 14. Daily data is justified, (by the nature of the problem, since I am mainly interested in the market positioning before, during and after interest decision meeting and whether there are pre-announcement effects in foreign exchange market.) also by research of Evans and Lyons (2005). They conclude that news in currency markets are fully absorbed only after several days.

All exchange rates were sourced for DataStream and are closing spot rates at 16:00 in London. This time reflects the middle of the “global day” and the time of highest liquidity in the foreign exchange market. MPC meeting dates are taken from the Bank of England[4]. All the statistics and graphs are computed in EViews.

### 4.2. Descriptive analysis

The raw data of all exchange rates are presented in Figures 1, 2 and 3. Where lower curve shows what has happened to sterling value over the last eleven years. It is easy to see that GBP value changed significantly over the sample period.

During first few years of the sample, sterling appreciated against Euro by almost 25%, however, since then pound constantly lost value against European currency and depreciated by 36% since all time high in May 2000. Overall, during sample period sterling fell by more than 20% against euro. Against two other main currencies, yen and dollar, sterling depreciated in the beginning of the sample (by XX% against yen and XX% against dollar), however, after US downturn in 2001, sterling started to appreciate and continued to gain value up until the beginning of global economic crisis in late 2007. After that British pound slumped XX% and xx% against yen and dollar respectively.

Overall, it is easy to see that against all three currencies most of the value British pound lost in the late part of the sample. Since (the beginning of global economic recession) xxx xxx sterling depreciated against euro, yen and dollar by xx%, Xx% and xx% respectively.

Exchange rate returns are shown at the top of each figure. Following the standard practise, returns are calculated by following equation:

where, rt is exchange rate returns on day t; st - spot exchange rate at day t; and st-1 - spot exchange rate at day t-1. Several features of returns are noteworthy.

Firstly, even though exchange rates sometimes increase and sometime decrease, all return series are centered around zero. Descriptive statistics, presented in Table X, also shows that means of the all three returns are very close to zero. It is -0.02% per annum for EUR/GBP, -0.025% for JPY/GBP and -0.008% for USD/GBP. The standard deviation is lowest for EUR/GBP and corresponds to annualized volatility of 8%[5], it is slightly higher for USD/GBP - 10%, 13% for JPY/GBP. This illustrates that exchange returns are almost unpredictable.

Secondly, we can see that there is volatility clustering; the amplitude of returns is changing aver time. For all three currencies there are some periods of high volatility and other periods of low volatility. Amplitude of the returns was quite low in early 2007, however, amplitude dramatically increased with bankruptcy of Lehman Brothers (xxx 2008) and stayed relatively high up until now.

Thirdly, in Table X we can that all currency pairs have excess kurtosis. If returns are normally distributed, then the kurtosis should be equal to three. However, kurtosis, which measures magnitude of the extremes, is substantial at 8.07 for JPY/GBP, 4.96 for USD/GBP and 3.44 for EUR/GBP suggesting that return series have fat tails. Fat tails also can be seen by plotting quantile plots for all return series (see Appendix X). if returns are normally distributed quantile plots should lie on a straight lines, and have s-shape if magnitude of extremes is higher than normal.

Mean

Standard Deviation

Kurtosis

Skewness

EUR/GBP

-0.00008

0.00497

3.44

-0.28

JPY/GBP

-0.00009

0.00830

8.07

-0.69

USD/GBP

-0.00003

0.00597

4.96

-0.06

### Table 1: Summary statistics for daily exchange rate returns

To sum up, exchange rate return series can be characterized by the following features: unpredictable returns, volatility clustering and fat tails. In the next parts of this chapter I will briefly introduce models which were designed precisely to model volatility of the returns with features described above and also I will specify the model to test how volatility differs around monetary policy meeting days.

### 4.3. Modeling volatility

Many approaches of modeling and forecasting volatility have been proposed in the literature; however the most popular models belong within a general class of autoregressive conditional heteroskedasticity (ARCH) models originally introduced by Engle (1982). These models specify time-varying conditional variance of the returns by using previous period information. The simplest example of ARCH process is the ARCH (1) specification. Where the distribution of the return for period t, conditional an all previous returns is defined as:

rt | rt-1, rt-2, … ~ N(μ, ht ) (1)

with:

ht = ω + α ( rt-1 - μ )2 (2)

where ht is conditional time-varying variance, μ - mean of the returns, ω and α volatility parameters. We can see that volatility of the return solely depends on weighted averages of previous squared forecast errors[6] (Taylor book).

Since introduction of ARCH model many specifications have been proposed by many researchers. Still probably the best known and most widely used is generalized ARCH (GARCH) from Bollerslev (1986). The success GARCH model can be explained by few reasons. First, it is easy to estimate model as it includes only four parameters. Second, GARCH explains main stylized facts of financial market daily returns. Third, volatility forecasts using GARCH specification usually have similar accuracy to forecast generated form more complicated specifications (Taylor book).

GARCH model gives weights to three different variance forecast: (1) constant variance that corresponds to the long run average, (2) forecast made in previous period, and (3) new information that was unavailable when forecast about previous variance was made. (Engle (nobelprize)). The standard GARCH (1, 1) specification assuming conditional normal distributions:

,

and

where pt is spot exchange rate, rt - returns, zt - standardized residuals[7]. All other notations is as above.

### 4.4. Model Specification and Hypothesis

Given that market participants knows when monetary policy meetings are held and when decisions about interest rates are announced, I am going to examine whether there are any systematic patterns in euro-sterling, yen-sterling and dollar-sterling exchange rate movements around monetary policy meeting days in the UK. To examine the behavior of volatility around interest decision days separate dummies for one day prior, one day after and the actual announcement day are incorporated into the conditional variance equation of a GARCH (1, 1) (equation X). Now the model takes the following specification:

where D1 is a dummy variable which takes the value of unity on one day before scheduled MPC meetings and zero elsewise, D2 is a dummy variable which takes the value of unity on days of scheduled MPC meetings and zero elsewise and D3 is a dummy variable which takes the value of unity on one day after scheduled MPC meetings and zero elsewise. All other notations is as above in GARCH specification.

As mentioned in Chapter X, previous empirical work finds that there are pre-announcement effects in financial market volatility. In his research of treasury market, Jones (1998) described the “calm before the storm” effect, where he find that conditional volatility is lower in the days leading up to the announcements of major economic data and higher on the announcement day itself. Similar results reported in many other financial markets. (Li and Engle, 1998; French, 1989). To the extent that market participants see MPC meeting days as major economic news, the link between monetary policy decisions and market volatility will be studied in the foreign exchange market.

Formally, I will test null hypothesis that dummy variables do not affect exchange rate volatility against alternative hypothesis that dummy variables have significant impact on exchange rate volatility:

H0 : δk = 0

H1 : δk ≠ 0

where k refers to different dummy variables for different days.

Also in this study integrated GARCH (1, 1) model will be used. Engle and Bollerslev (1986) defined IGARCH (1, 1) as a special form of general GARCH (1, 1), where volatility process has a unit root.

Mathematical expression of the IGARCH (1, 1) looks exactly like regular GARCH (1, 1):

{equatiosn} (X)

however the coefficients of ARCH (α) and GARCH (β) terms in conditional variance equation must sum to one:

{equatiosn} (X)

If I fail to reject the null hypothesis of unit root in conditional exchange rate volatility then I will disavow the general GARCH regression and use the integrated GARCH with the same hypothesis and parameters as in GARCH (1, 1).

### 5. Empirical Results and Analysis

5.1. Diagnostic tests

(gal ci aglima trumpa sakini apie kas kodel)

Test for Stationarty in returns

The time series is said to be stationary if for all values it is true that: (1) mean and (2) variance is constant over time, (3) the value of covariance between two values depends only on the length of time between the two values, and not on the actual times at which the variables are observed (Hill 2001). Time series which don't have these properties are known as non-stationary or unit root.

The use of non-stationary time series data in regression can lead to significant results form unrelated data. Exchange rates as well as many other macroeconomic and financial time series are often found to be non-stationary. However, econometric model used in this study involves only exchange rate returns, which usually tend to be stationary.

Augmented Dickey - Fuller test was used to test for stationarity is all currency pairs (Dickey and Fuller (1979)). The null hypothesis that returns have a unit root was rejected for all three currency pairs at 1% level. The test results can be found in Appendix X.

Test for Stationarty in Volatility

Next, all three currency pairs will be tested for unit root in volatility. The equation (X) in chapter X can be rewritten as:

{Equation} (X)

and hence

{Equation} (X)

whenever expectations are finite, as z(t-1) is independent of h(t-1) (Taylor book). So GARCH (1, 1) process is stationary if and only if α + β < 1 ( Kazkastai).

For all three currency pairs the estimated sum of ARCH and GARCH coefficients is very close to one, which is consistent with previous empirical work with high frequency data. However, since estimated standard errors of persistence are fairly small, it is necessary to perform hypothesis test.

The null hypothesis that volatility has unit root will be tested against alternative hypothesis that volatility is stationary. For GARCH (1, 1) model the null and alternative hypothesis is stated as:

H0: φ=1, where φ is persistence parameter and equal to α + β,

H1: φ <1.

Wald test was used to test hypothesis for stationarity in all exchange rate volatility. Lee and Hansen (1994) and Limsdaine (1995) showed that robust Wald test can be used to test unit root hypothesis and does not need to be modified in the GARCH (1, 1) context. The null hypothesis that volatility process is non-stationaty (unit root) was rejected for JPY/GBP and USD/GBP at 5% level, yet I fail to reject unit root for EUR/GBP exchange rate volatility process. The test result can be found in Appendix X.

### 5.2. Results

In this section I will present regression results for all currencies. The significance on monetary policy meeting and surrounding days differ depending on currency.

For EUR/GBP exchange rate data (for Full report see Appendix X)

As mentioned above I fail to reject null hypothesis of unit root in volatility, so for this exchange rate IGARCH (1, 1) model was used. Exhibit X provides estimated results for volatility equation. In the case of monetary policy surrounding days only exact MPC announcement day is significant at 1 % level, other days (before and after the meeting) are statistically significant only at 10% level.

The coefficient X is positive, which suggest that euro/sterling is more volatile on interest rate decision day, yet he coefficients x and x, are negative, so currency is less volatile before and after the MPC meeting this is consistent with “calm before storm” …. However, the significance of days before and after contradicts claim that foreign exchange markets …..

The magnitude of the coefficient indicates that the conditional variance increases by XX% on the announcement days. Other dummies (MPC-1 and MPC+1) a indicates that on the day before meeting an after conditional variance decreases XX% and XX% respectively.

All in all, euro/sterling exchange rate volatility is typically greater on monetary policy announcement days and lower one day before and after.

For JPY/GBP exchange rate data (for Full report see Appendix X)

I reject null hypothesis of unit root in volatility, thus GARCH (1, 1) coefficient were estimated. Exhibit X provides estimated results for volatility equation. In this case, only the day before interest rate decision is statistically significant. The coefficient X is positive, which suggest that yen/sterling is more volatile before interest rate decision day. This is opposite to the result discovered for euro/starling and contradicts with “calm before storm” …… This maybe can be explained by the difference in time between Europe and Japan, however daily data is taken at 18:00 London time when liquidity in Foreign exchange market is highest, thus further research is necessary to explain this difference.

For USD/GBP exchange rate data (for Full report see Appendix X)

I reject null hypothesis of unit root in volatility, thus GARCH (1, 1) coefficient were estimated. Exhibit X provides estimated results for volatility equation. For this currency pair, none of the coefficients before dummy variables are statistically significant. This suggest that dollar/starling volatility is unaffected by monetary policy announcements in the UK.

### 5.3. Interpretation of results

Few important conclusions can be drawn from estimation results provided above Firstly, between all currencies pairs considered in this paper there are no consistency in how exchange rate volatility changes around monetary policy announcements in the UK.

### 6. Conclusion

6.1. Possible extensions

Father studies could look at intraday data of exchange rates and how exchanger ate volatility reacts to monetary policy on a …. basis.

More complex study could use expectations about laukiama monetary policy decision, and how depending on expectations market participants position themselves before interest rate announcements.

[1] For institutional background on the MPC and the monetary policy process, see Bean (1999).

[2] www.bankofengland.com

[3] Excluding special meeting (18 September 2001) due to reaction of financial markets to the terrorist attacks on September 11, 2001.

[4] www.bankofengland.co.uk

[5] Annualized volatility, σ= σD (261)1/2, where σD - standard deviation of daily logarithmic returns, scaling constant (261) is the average number of returns per annum in the time series.

[6] Residual at time t is : et = rt +μ