The Effects of Central Banks’ Rate Change Patterns on Financial Market Variables

Abstract

This study examines whether different patterns of change to the benchmark interest rates of central banks are associated with their contributions to variances in the forecast errors of three financial market variables: the long-term interest rate, the foreign exchange rate, and the stock market index. On average, the central bank’s interest rate accounts for approximately 20% of the variance in each variable. We find that the total range of changes is more important than the frequency of changes. The panel regression shows that the range and frequency of policy rate changes is positively associated with the volatility of long-term interest rates but no association with the volatility of stock prices and exchange rates. These results suggest that small and frequent adjustments of policy rates are desirable for reducing the volatility of interest rates. The panel VAR represents interest rate channel is a more important than exchange rate and stock price channel.

KEY WORDS:

• Monetary policy
• benchmark interest rate
• financial market variables
• transmission mechanism

JEL CLASSIFICATION:

• E58
• E44

Acknowledgements

The views expressed herein are solely those of the authors and do not necessarily represent the views of the Bank of Korea or Korea Development Institute. We are grateful to two referees for insightful comments and suggestions.

Notes

1 Taylor (1995) also presents the financial market prices framework of the transmission mechanism. He argues that measurement problems have forced econometric modelers away from the quantity of credit and foreign exchange and toward the prices of these items.

2 Foreign exchange rate represents the value of each country’s currency to US dollar. In case of the United States, dollar index, or the value of the US dollar to the basket of major six foreign currencies, is used instead.

3 Considering our short sample period, monthly data are preferable to quarterly data to secure a sufficient degree of freedom for a VAR analysis.

4 The models specified by Equations (2) and (3) are estimated by cross-section data from 24 countries. The dependent variable used in each regression is a single unique value for each country. Therefore, all independent variables such as X, Y are also cross section data. Total number of observations is equal to 24, the total number of countries used in this paper.

5 We used both conventional t-statistics and bootstrapping interval to check the robustness of statistical inference. Significant estimated coefficients based on t-statistics are located outside the bootstrapping confidential range supporting null hypothesis.

6 Unlike Equations (2) and (3), independent variables X, Y are time series data in Equations (4) and (5). They stand for the range and frequency of policy rate changes during each year.

7 Our sample originally includes 24 nations and the study period is mainly from 1999 to 2007. However,we use 18 countries (India, Philippines, Slovakia, Switzerland, Taiwan, Thailand are excluded) for balanced panel regression due to lack of economic activity or inflation data in some countries. Number of observation is nine for each country because annual data are used for panel regression.

8 In Equation (5), MR, FR, Gini, Adv are cross-section data while other independent variables are panel data.

9 We include Gini coefficients as an explanatory variable in the regression equations (5). The more equal an economy is the more people can participate in the financial markets. Thus, financial markets may respond more to shocks or events.

10 The fixed effect model doesn’t include cross-sectional independent variables such as MR, FR, Gini, Adv in Table 13 because pooled OLS including cross-section data can be regarded as a fixed-effect model. In Table 13, the random effect model including cross-sectional variables can be interpreted as the result of reflecting the fixed effect in different ways.

11 We confirmed the robustness of the regression by replacing variables of business conditions, such as industrial production and retail sales, with the GDP. The results of estimating Equations (4) and (5) using GDP are similar to the results estimated by using industrial production or retail sales.