Analyse de l'impact des nouvelles de la BCE sur la volatilité de la structure du terme de taux d'intérêt (document en anglais)

Analysis of ECB’s news impact on the volatility of the interest rate term structure

Financial time series

LIN Yande

BACHIRI Ouafe

Date

04/05/2008

Table of contents

1. Introduction 3

2. Data 4

2.1. Term structure interest rates 4

2.2. News data 4

2.3. Statistical description 4

2.3.1. Non stationary series 5

2.3.2. Stationary series 6

3. General shocks impact 9

3.1. Models Comparison 9

3.2. ARMA(1,1)-Egarch(1,1) 9

3.3. News impact curve 9

4. ECB’s News impact 10

4.1. Model 10

4.2. Regression results with control of heterokedasticity 11

5. Asymmetric impacts of 2 types ECB’s news 12

5.1. Model 12

5.2. Regression results with control of heteroskedasticity 12

5.3. Residual tests 13

6. Limitations 13

7. Conclusion 13

1. Introduction

It is well known that the volatility in markets is highly correlated with central bank’s announcements. This topic is highly covered in literature. For instance, the ECB increased interest rates on October the 6th.The graphic below illustrates the direct reaction of term structure of interest rates at the moment of the announcement.

Figure 1: term structure interest rates in the same day and days before and after announcement of a change of interest rates in October the 6th 2000.

But what’s the impact of such news on the volatility of the interest rates? Sometimes we care more about their volatility than their level when we work on the interest rates derivatives. Therefore the goal of this project is to investigate the impact of ECB’s (European Central Bank) news on volatility of interest rates term structure through econometrics methods.

At first, we will describe the data used in this project. Then we will investigate the shocks’ impact on the volatility of interest rate term structure based on Egarch model. After that, we will focus on the impact of ECB’s news and analysis its specificity. Finally we will decompose the news into two types; increase or decrease of interest rates and analysis separately their impact on the volatility.

2. Data

2.1. Term structure interest rates

We extracted from DataStream the Libor spot interest rates of different maturities (1, 2,3,5,6 months, 1, 2, 10 years) during the period 03/01/2000 to 21/04/2008 in order to analyze the volatility of the interest rate term structure..

2.2. News data

The Governing Council of the ECB sets the key3 interest rates of Deposit facility, Main refinancing operations and Marginal lending facility in the euro area. The table below shows the changes of the interest rate of Main refinancing operations which provide the bulk of liquidity to the banking system. We will incorporate the news on the interest rates in our model presented later.

Main refinancing operations

Date Level Change

2007 13 Jun. 4.00 0.25

14 Mar. 3.75 0.25

2006 13 Dec. 3.50 0.25

11 Oct. 3.25 0.25

9 Aug. 3.00 0.25

15 Jun. 2.75 0.25

8 Mar. 2.50 0.25

2005 6 Dec. 2.25 0.25

2003 6 Jun. 2.00 -0.50

7 Mar. 2.50 -0.25

2002 6 Dec. 2.75 -0.50

2001 9 Nov. 3.25 -0.50

18 Sep. 3.75 -0.50

31 Aug. 4.25 -0.25

11 May 4.50 -0.25

2000 6 Oct. 4.75 0.25

1 Sep. 4.50 0.25

9 June 4.25 0.50

28 Apr. 3.75 0.25

17 Mar. 3.50 0.25

4 Feb. 3.25 0.25

Table 1 : ECB’s announcements of change of directory interest rates of Main refinancing operations. Source : http://www.ecb.eu/stats/monetary/rates/html/index.en.html#data

2.3. Statistical description

2.3.1. Non stationary series

First, let’s focus on one month interest rate series Y0M1. The two pictures below show its statistic characteristics.

Figure 2: evolution of one month Figure 3: The distribution of one month

Interest rates through time Interest rates and classical test statistics

The first graph shows the evolution of the 1-month interest rate through time. It is obvious that this series is not stationary since its level has a certain pattern: the interest rate goes up during the first 1 year then decreases till 2003 and goes up again since 2006. According to J. Bera test its distribution is not Gaussian (Jarque-bera>>6).

Figure 4: Correlogram of one month interest rate series Y0M1

The pattern of correlogram of this kind shows the symptom of non stationary since its autocorrelation is decreasing too slowly and its partial correlation is significantly different from zero for the first order.

Null

...