THE IMPACT OF DEREGULATION ON FINANCIAL
MARKET EFFICIENCY IN SRI LANKA
Arusha Cooray
A thesis submitted in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
School of Economics
The University of New South Wales
November 2000
To Amma, Thaththa and Devoushi
CONTENTS
Page Tables v Figures vii Abstract ix Acknowledgement x 1 Introduction 1 2 The Financial System of Sri Lanka 11 2.1 Introduction 11 2.2 The Financial System and Monetary Policy Prior to Deregulation 12 2.3 Factors Contributing to Financial Deregulation 17 2.4 Financial Deregulation 19 2.4.1 Institutional Reforms 20 2.4.2 Monetary Policy in a Deregulated Environment 38 2.5 Evaluation of Policy Reforms 46 2.5.1 Structure 46 2.5.2 Financial Depth 50 2.5.3 Efficiency 53 2.5.4 Impact on Financial Stability 58 2.5.5 Consequences for the Macro-economy and Macroeconomic Policy 60 2.6 Conclusion 62 3 The Efficient Market Hypothesis 64 3.1 Introduction 64 3.2 Defining Market Efficiency 65 3.2.1 Fama's Definition 66 3.2.2 Further Formalization of the Concept of Efficiency 69 3.2.3 Alternative Definitions of Market Efficiency 71 3.3 Failure of Market Efficiency 74 3.4 Conclusion 77 4 The Expectations Hypothesis of the Term Structure of Interest Rates 78 4.1 Introduction 78 4.2 Interest Rate Deregulation in Sri Lanka 80 4.3 The Expectations Theory of the Term Structure of Interest Rates 84 4.4 The Model 86 4.5 Review of the Empirical Literature 90 4.5.1 Early Empirical Studies 91 4.5.2 Models of Expectation Formation 91 4.5.3 The Yield Spread as a Predictor of Future Interest Rate Changes 96 4.5.4 Deviations From the Expectations Hypothesis 98 4.6 Data 100 4.7 Results 107 4.7.1 The Expectations Hypothesis 107 4.7.1.1 Cointegration 107 4.7.1.2 An Error Correction Model 112 4.7.1.3 OLS Estimation 117 4.7.1.4 Instrumental Variable Estimation 122 4.7.2 The Predictive Content of the Spread 123 4.7.2.1 OLS Estimation 123 4.7.2.2 A VAR Model 125
ii 4.8 Implications 130 4.9 Conclusion 133 5 The Fisher Effect 139 5.1 Introduction 139 5.2 An Overview of the Sri Lankan Experience with Inflation 141 5.3 The Fisher Hypothesis 148 5.3.1 Adaptive Expectations 151 5.3.2 Rational Expectations 152 5.4 Empirical Models 152 5.4.1 The Adaptive Expectations Model 153 5.4.2 Measuring a Rational Expectation of Inflation by its Actual Value 155 5.4.3 Measuring a Rational Expectation of Inflation by its Forecast Value 156 5.4.4 The Inverted Fisher Effect 157 5.5 Previous Studies 161 5.5.1 Fisher’s Findings 162 5.5.2 Adaptive Expectations 163 5.5.3 Rational Expectations and Efficient Markets 165 5.5.4 Empirical Work for Developing Countries 173 5.5.5 Empirical Work on Sri Lanka 175 5.5.6 Deviations from the Fisher Hypothesis 177 5.6 Data 180 5.7 Empirical Results 184 5.7.1 The Adaptive Expectations Model 184 5.7.1.1 Estimation by OLS 186 5.7.1.2 Estimation by Instrumental Variables 187 5.7.2 Rational Expectations with the Actual Rate of Inflation as Proxy for 190 Inflationary Expectations 5.7.2.1 Estimation by OLS 191 5.7.2.2 Estimation by Instrumental Variables 192 5.7.3 Rational Expectations with a Forecast Rate of Inflation for Inflationary 194 Expectations 5.7.3.1 Cointegration 196 5.7.3.2 Estimation by OLS 204 5.7.3.3 Estimation by Instrumental Variables 207 5.8 The Inverted Fisher Effect 210 5.9 Implications 213 5.10 Conclusion 215 6 Exchange Rate Efficiency and Capital Mobility 228 6.1 Introduction 228 6.2 Exchange Rate Behaviour in Sri Lanka 231 6.3 Tests of Market Efficiency 235 6.4 Empirical Models 239 6.4.1 Uncovered Interest Parity 239 6.4.2 Speculative Efficiency 240 6.4.3 Savings–Investment Correlations 242 6.4.4 Consumption–Net Output Correlations 243 6.4.5 Real Interest Rate Linkages 246 6.5 Empirical Work for Developing Countries 248 6.5.1 Tests of Market Efficiency 248 6.5.2 Tests of Capital Mobility 252 6.6 Data 255
iii 6.7 Results 261 6.7.1 Uncovered Interest Parity 261 6.7.2 Speculative Efficiency 271 6.7.3 Savings–Investment Correlations 276 6.7.4 Consumption–Income Correlations 282 6.7.5 Real Interest Rate Linkages 287 6.8 Implications 294 6.9 Conclusion 297 7 Conclusion and Policy Implications 299 7.1 Overall Findings 299 7.2 Policy Implications and Impediments to Reform 302 7.3 Recommendations 312 Bibliography 322
iv
TABLES
Page 2.1 Savings and Investment Ratios 1970–1976 17 2.2 Measures of Economic Performance 1970–1976 18 2.3 Commercial Bank Deposit Mobilization (Rs. Million) 23 2.4 Mobilization of Deposits by The National Savings Bank 1972–1998 (Rs. Million) 24 2.5 Share Market Indicators 34 2.6 Deposit Rates of Commercial Banks (Rates per Annum) 39 2.7 Share of Financial Services to GDP and Investment 47 2. 8 F inancial Depth 51 2.9 Interest Rate Differential 56 2.10 Rate of Foreign Direct Investment Inflows to GDP and to Gross Fixed Capital 57 Formation 2.11 Foreign Participation in Colombo Stock Exchange 58 2.12 Savings and Investment Ratios 61 4.1 Unit Root Tests 1990.1–1998.12 106 4.2 Johansen’s Cointegraton Test for the Three-Month Treasury Yield 110 4. 3 Johansen’ s C ointegraton T est for the S ix-Month T reasury Yield 111 4.4 The Expectations Hypothesis of the Term Structure: OLS Estimates 117 4.5 The Expectations Hypothesis of the Term Structure: Instrumental Variable Estimates 122 4.6 The Spread as a Predictor of the Future Change in Short Rate: OLS Estimation 124 4.7 T he Spread (between the six- and three-month rate) as a Predictor of the Future 126 Change in Short (three month) Rate: A VAR Application 4.8 T he Spread (between the twelve- and six-month rate) as a Predictor of the Future 127 Change in Short (six month) Rate: A VAR Application A4.1 Unit Root Tests 1990.1–1996.12 134 A4.2 The Expectations Hypothesis of the Term Structure: OLS Estimates 136 A4.3 The Expectations Hypothesis of the Term Structure: IV Estimates 136 A4.4 The Spread as a Predictor of the Future Change in Short Rate: OLS 138 Estimates 5.1 Unit Root Tests 183 5.2 The Adaptive Expectations Model: OLS Estimates 186 5.3 The Adaptive Expectations Model: Instrumental Variable Estimates 189 5.4 Rational Expectations with the Actual Rate of Inflation as Proxy for Inflationary 191 Expectations: OLS Estimates 5.5 Rational Expectations with the Actual Rate of Inflation as Proxy for Inflationary 193 Expectations: Instrumental Variable Estimates 5.6 Johansen’s Cointegration Test for the Annual Data 198 5.7 Johansen’s Cointegraton Test for the Quarterly Data 201 5.8 Johansen’s Cointegraton Test for the Monthly Data 202 5.9 Rational Expectations with a Forecast Rate of Inflation as Proxy for Inflationary 206 Expectations: OLS Estimates 5.10 Rational Expectations with a Forecast Rate of Inflation as Proxy for Inflationary 208 Expectations: Instrumental Variable Estimates 5.11 T he Inverted Fisher Effect 212 A5.1 Johansen’s Cointegraton Test for the Quarterly Data Excluding a Drift Term 218 A5.2 Johansen’s Cointegraton Test for the Quarterly Data Including a Drift Term 219 A5.3 Johansen’s Cointegraton Test for the Annual Data Excluding a Drift Term 222 A5.4 Johansen’s Cointegraton Test for the Annual Data Including a Drift Term 223 A5.5 Johansen’s Cointegraton Test for the Monthly Data Excluding a Drift Term 224 A5.6 Johansen’s Cointegraton Test for the Monthly Data Including a Drift Term 225
v A5.7 Likelihood Ratio Tests 226 6.1 Dickey–Fuller Test for Unit Roots for the Levels of the Series 258 6.2 Dickey–Fuller Test for Unit Roots for the First Differences of the Series 260 6.3 Uncovered Interest Parity with Rational Expectations–OLS and SUR Regressions 262 1959–1998 6.4 Pre Float Results 1959–1976: OLS and SUR Estimates 264 6.5 Post Float Results 1977–1998: OLS and SUR Estimates 265 6.6 Uncovered Interest Parity–OLS and SUR Regressions 1990.1– 1998.12 267 6.7 The Dickey–Fuller Unit Root Test for the Residuals of the Cointegration Regression 272 of the Spot Rate on the Forward Rate 6.8 Speculative Efficiency: OLS Estimates 273 6.9 A Test of Speculative Efficiency Employing the Current Spot Rate as a Stationary 275 Deflator: OLS Estimates 6.10 Saving–Investment Correlations 277 6.11 Current Account-Investment Correlations 279 6.12 Saving–Investment and Current Account-Investment Correlations Eliminating 281 Observations from 1977–1980 6.13 Consumption-Income Correlations 284 6.14 Consumption–Income Correlations Relaxing the Assumption of a Constant Real Rate 286 of Interest 6.15 OLS Estimates for Real Interest Rate Linkages 288 6.16 A Partial Adjustment Model for Real Interest Rate Linkages between India and Sri 290 Lanka 7.1 Bank Performance 303 7.2 Determinants of Savings Rates: A Comparison of Trends 306 7.3 Size of the Public Sector in Selected Countries 308
vi FIGURES
Page 2.1 Deposits Mobilized by Commercial Banks 1950–1998 21 2.2 Deposits (Real) Mobilized by Commercial Banks 1952 – 1998 22 2.3 Deposits Mobilized by the National Savings Bank 1972 – 1998 24 2.4 Deposits (Real) Mobilized by the National Savings Bank 1972–1998 25 2.5 Asset Growth of Contractual Savings Institutions as Percentage of GDP 26 2.6 Asset Growth of Long-term Lending Institutions as Percentage of GDP 27 2.7 Bank Branch Expansion–Regional Rural Development Banks 1985–1997 28 2.8 Share Market Capitalization 35 2.9 Private Foreign Capital and Foreign Direct Investment Flows 1970–1998 36 2.10 Ownership of Treasury Bills as at December 1975 41 2.11 Ownership of Treasury Bills as at December 1998 42 2.12 Employment in Financial Services as Percentage of Total Employment 48 2.13 Number of Bank Branches 1950–1998 49 2.14 People per Branch 1950–1998 50 2.15 Total Assets of Financial Institutions as Percentage of GDP 52 2.16 Money Market Instruments (Amount Outstanding) 53 2.17 Intermediation Margin 1970–1998 54 2.18 Interest Rate Volatilities 1990.1–1998.12 59 2.19 Exchange Rate Volatility 1990.1–1998.12 60 4.1 Fixed Deposit Rates 1990.1–1998.12 82 4.2 Call Money Rate 1990.1–1998.12 83 4.3 Treasury Bill Rates in the Primary Market 1990.1–1998.12 84 4.4 Three Month Treasury Bill Rate 1990.1–1998.12 (Spot and Implicit Forward) 101 4.5 Six Month Treasury Bill Rate 1990.1–1998.12 (Spot and Implicit Forward) 102 4.6 Multivariate Dynamic Forecasts for the Level of the T hree Month Rate 115 4.7 Multivariate Dynamic Forecasts for the Level of the Expected Rate 115 4.8 Multivariate Dynamic Forecasts for the Level of the Six Month Rate 116 4.9 Multivariate Dynamic Forecasts for the Level of the Expected Rate 116 4.10 Plot of Cumulative Sum of Recursive Residuals 119 4.11 Plot of Cumulative Sum of Squares of Recursive Residuals 120 4.12 Plot of Cumulative Sum of Recursive Residuals 120 4.13 Plot of Cumulative Sum of Squares of Recursive Residuals 121 4.14 Multivariate Dynamic Forecast for the Spread (between the six- and three-month 129 yields) 4.15 Multivariate Dynamic Forecast for the Spread (between the twelve- and six -month 129 yields) 4.16 Yield Curve for Sri Lanka as at 31.12.98 132 5.1 The Rate of Inflation 1952–1998 142 5.2 Money Supply as Measured by M2 1952–1998 143 5.3 Colombo Consumer Price Index and Import Price Index (Rate of Change) 145 5.4 Exchange Rate–Rupees per Unit of U.S. Dollar 1952–1998 146 5.5 Three Month Treasury Bill Rate 1952–1998 147 5.6 Three Month Treasury Bill Rate and the Rate of Inflation 1952–1998 148 5.7 The Rate of Inflation, Nominal Treasury Bill Rate Real Treasury Bill Rate 210 A5.1 Price Level 1978.1–1996.4 (1952=100) 221 6.1 Interest Rate Differentials–United States 268 6.2 –United Kingdom 269 6.3 –India 269 6.4 –France 270 6.5 –Germany 270 6.6 –Japan 271 6.7 Forward and Spot Exchange Rates 1996.2–1998.12 274
vii 6.8 Investment Ratio, Savings Ratio and Current Account Ratio 1959–1998 281 6.9 Real Interest Rate Linkages: Sri Lanka–U.S. 291 6.10 Sri Lanka–U.K. 292 6.11 Sri Lanka–India 292 6.12 Sri Lanka–France 293 6.13 Sri Lanka–Germany 293 6.14 Sri Lanka–Japan 294 7.1 Net Private Capital Flows (Per cent of GDP) 1992–1996 310 7.2 Growth in Consumption and Investment 1990–1997 311
viii ABSTRACT
The purpose of this study is to investigate the impact of deregulation on financial market efficiency in Sri Lanka. The concept of efficiency used here is due to Fama (1970) who defines an efficient market as one in which prices fully reflect all available information.
Given the significant expansion of Sri Lanka’s financial markets in the post deregulation period, efficiency is investigated in the context of these markets. To this end, the study employs a number of standard tests for market efficiency including; the expectations hypothesis of the term structure, the Fisher hypothesis, uncovered interest parity, speculative efficiency, real interest rate equalization and tests of capital mobility.
Although the overall results presented in this study suggest that Sri Lanka’s financial markets are not fully efficient, the evidence provides significant insight to the performance of these markets. The main policy lesson to be learnt from this analysis is that financial deregulation will not automatically promote market efficiency unless accompanied by positive policy action to reinforce the impact of these reforms. In conclusion therefore, the study makes a number of recommendations which could help to reinforce the impact of financial deregulation on market efficiency.
ix ACKNOWLEDGMENTS
I wish to acknowledge and express my appreciation to the many people who made this research possible. Dr. Glenn Otto, who supervised me in the last twelve months of this research, however, generously gave of his time at all stages of it, deserves much of the credit for the evolution of the thesis toward its present form and content. His invaluable insights on the subject, not only helped to sharpen the focus of this thesis but also my knowledge to a degree that I will always be in his debt. I would also like to express my sincere thanks to my former supervisors, Prof. Ross Milbourne and A/Prof. Bill Rao for their valuable contributions which greatly improved the thesis. I appreciate the suggestions provided to me by Dr. Lance Fischer, Dr. Robert Hill and Dr. Medhi
Monadjemi at my second review. I also wish to thank Dr. Paul-Pezanis Christou, Dr.
Graham Voss, M s. Nadia Blum, Ms. Rebecca Caddy, Ms. Aline Dinel, M s. Sue Nelson and Ms. Sylvana Tomasiello of the School of Economics, UNSW.
I wish to acknowledge with thanks the advice given to me by Dr. Premachandra
Athukorale and Prof. Aruna Seneviaratne and the valuable comments made to me by
Prof. Johannes Juttner (Maquarie).
I gratefully acknowledge the financial support provided to me by the Institute of Social
Studies (ISS), Netherlands, for this research and in particular the assistance of Dr.
Howard Nicholas and Ms.E Mulder. I am also indebted to Prof. W D Lakshman without whose support I would not have received this scholarship. Thanks are also due to Ms.
Buel, Mr. Fernando and Mr. Wijetunge of the UC-ISS Project.
x
This work would not have been possible without the study leave provided to me by the
University of Colombo and the support of Dr. Sarath Vidanagama. Dr. Amala de Silva also deserves special mention.
I am grateful to Mr and Mrs. A S Jayawardene for their support, particularly in the use of the Central Bank, Ms. Monica Jayasekera, Ms Lakshmi De Mel and Lilamani for the provision of data and Nandesena for his assistance. The thesis has also benefited greatly from the input of Dr. Uthum Herath.
I wish to thank my friends Darsono, Punderyk, Ru, Uma and Indrajit for their support .
And last but not least this thesis could not have been completed without the love, support and encouragement of my family; my parents, sister, brothers-in-law, Priyantha and my daughter, Devoushi. Thank you Devoushi for your tolerance, I could not have survived it without you.
xi Introduction
CHAPTER 1
INTRODUCTION
In 1977, Sri Lanka deregulated its financial system through reforms aimed at increasing the role of the market mechanism. The system that prevailed prior to 1977 was characterized by administered interest rates, ceilings on the level of credit, lending to priority sectors and public sector ownership and control of financial institutions. Barriers to trade and international capital movements restricted competition in the local financial market by discouraging the entry of foreign investment.
As the Central Bank of Sri Lanka (1977) stated:
… the rigid control system that had been built up over the years had seriously distorted relative prices, and had dampened private sector incentives; that the public sector which was fostered to fill the vacuum had become wasteful and complacent; and that successive governments had failed to take corrective action but had only resorted to short-term palliatives, which had only compounded the problems.
In 1977, Sri Lanka’s Government recognized the inefficiencies of the interventionist policy and began to place greater reliance on the market mechanism. Financial reform included the introduction of new financial institutions, instruments and markets, the upward revision of interest rates so that they better reflected the demand and supply for credit, the reduction of directed credit, the introduction of prudential regulations of banks and finance companies and the strengthening of debt recovery legislation. The exchange rate was unified and allowed to float in relation to a basket of currencies.1
1 Since 1968 the economy had functioned under a dual exchange rate system.
1 Introduction
The process of deregulating the financial sector in Sri Lanka was aimed at ensuring a more efficient allocation of resources, thereby promoting overall economic performance and social welfare. While efficiency gains in financial intermediation were expected to lead to an improved allocation of resources, the increase in private sector ownership of financial institutions and removal of operational restrictions on foreign banks were expected to foster competition and produce efficiency gains in production. This, however, can only be achieved through a competitive pricing mechanism. An informationally efficient asset market would permit the transfer of surplus funds from lenders to borrowers thereby channeling funds to their most productive uses. If asset prices could be relied upon to reflect new information that the market receives, then they would provide useful signals to savers for constructing their investment portfolios and investors for the efficient utilization of funds. The incomplete dissemination of information, conversely, could lead to inefficiencies in the allocation of financial resources and the creation of monopoly profits, resulting in loss of efficiency in production. This would give rise to questions on the reliability of the pricing mechanism. Pricing in an efficient market would therefore reflect the performance and risk of alternative investments. The role of information has also become more important with the growing globalization of financial markets and greater vulnerability of economies to a broader range of external shocks.
Informational efficiency, therefore, has a direct bearing on allocation and production efficiency through its impact on risk. Sri Lanka provides a particularly interesting case for an investigation of market efficiency in view of the fact that financial sector reforms in Sri Lanka were among the earliest undertaken in South Asia.
2 Introduction
The concept of market efficiency was first developed by Fama (1970), who defined an efficient market as one in which prices fully reflected all available information. In
Fama's words (1970, p.383):
The primary role of the capital market is the allocation of the ownership of the economy's capital stock. In general terms, the ideal is a market in which prices provide accurate signals for resource allocation: that is, a market in which firms can make production and investment decisions, and investors can choose among the securities that represent ownership of the firms' activities under the assumption that security prices at any time 'fully reflect' all the available information. A market in which prices always 'fully reflect' available information is efficient.
In such a market, prices would adjust rapidly and accurately to the arrival of new information so that past information cannot be used to predict future prices. Rational expectations form an important component of the efficient market hypothesis (EMH).
The rational expectations hypothesis asserts that the market utilizes efficiently the information available to it in forming expectations about future outcomes. The efficient utilization of information implies that the market's perception of the probability distribution of future outcomes, conditional on available information, coincides with the actual probability distribution conditional on that information. An implication of rational expectations is that agents will not make systematic errors in forecasting the future.2 The drive for profit will tend to eliminate any opportunities for abnormal gain. Rational expectations therefore constitute an important component of the EM H in that it provides the mechanism by which all available information is reflected in prices.
2 See Begg (1982).
3 Introduction
Testing the EMH, however, has proven to be more difficult than was originally perceived because whether prices fully reflect all available information is not something that is directly observable. Tests of market efficiency therefore require models of expected returns. These expected returns serve as the link between current and future prices. The market can be said to be efficient if the distribution of the current and expected returns correspond with the distribution of future returns. Tests of efficiency therefore become joint tests of the expected returns model and of market efficiency. Thus, rejection of such joint hypotheses could stem from either market inefficiency or the use of an incorrect expected returns model.
The purpose of this study is to address the issue of whether financial deregulation has led to market efficiency in Sri Lanka. To this end, the study employs a number of standard tests for financial market efficiency. These include: first, the expectations hypothesis of the term structure, which tests the hypothesis that forward interest rates are unbiased predictors of future spot rates. Classic contributions to this literature include studies by Fama (1976), Mankiw (1986) and Campbell and Shiller (1987).
Second, the Fisher hypothesis, which evaluates the question of whether nominal interest rates are unbiased predictors of future expected inflation. This question, first addressed by Fisher (1930), stimulated a large number of further studies; see Gibson
(1970), Fama (1975), Nelson and Schwert (1977) Mishkin (1992). Third, uncovered interest parity, which tests the hypothesis that nominal interest rate differentials of financial assets denominated in different currencies are exactly equal to the expected change in exchange rate. Recent studies include those by Taylor (1987), and
4 Introduction
MacDonald and Taylor (1989). Fourth, speculative efficiency, which evaluates the hypothesis that forward exchange rates are unbiased predictors of future spot rates
(see Gweke and Feige (1979), and Hansen and Hodrick (1980)).
Although the EMH has potential relevance for other markets, it has been undertaken primarily in the context of the financial market. This market is important for a number of reasons:3 first, it provides the necessary climate for developing the expertise required for portfolio management and risk bearing. Secondly, it serves as a device in hastening the integration of the traditional or unorganized financial sector with the organized financial sector by replacing traditional assets such as real estate and gold with financial assets. Thirdly, it intensifies the ability and provides the institutional background required for the operation of monetary policy. Finally, it permits foreign exchange gains arising from capital inflows. More importantly, as pointed out by Keane (1983), a feature of this market that distinguishes it from other markets is that all financial assets are reducible to two components, namely, expected return and risk, which permit easy application of the EMH.
Much of the empirical work on security price behaviour has been undertaken with respect to the markets of the developed economies, in particular the U.S.4 There also exist a number of studies which investigate the behaviour of European security markets. While some support is found for the efficiency of these markets, the evidence has not been as clear-cut as that for the U.S. (Solnik (1973) for France,
3 See Drake (1977). 4 See Fama (1970, 1991) for surveys.
5 Introduction
Germany, Italy, the Netherlands, Belgium, Switzerland and Sweden; Jennergren and
Korsvold (1975) for Sweden and Norway; Papaionnou (1984) for Greece). The evidence relating to smaller and less developed security markets is sparse, although the extent of the research has risen over the past few years. Studies have been undertaken by Gandhi, Saunders and Woodward (1980) for the Kuwaiti Stock
Exchange; Elyasiani, Perera and Puri (1996) for the Colombo Stock Exchange; Ayadi and Pyun (1994) the Korean Stock Exchange; and Dawson (1981) the Kuala Lampur
Stock Exchange. The degree of support found for the EMH with respect to these markets has been less favourable in comparison to the evidence for the developed economies, problems arising primarily from the thinness of the markets and the incomplete dissemination of information.
No reported studies to date have examined the relationship between financial deregulation and market efficiency for Sri Lanka in any detail. To date, only two studies have been undertaken on efficiency for Sri Lanka—Samarakoon (1996) and
Elyasiani, Perera and Puri (1996)—and they have confined their investigation to the stock market. Both these studies reject the "random walk" model of stock returns and consequently market efficiency. Given the significant expansion of the money and foreign exchange markets in Sri Lanka in the post-reform period, this study is undertaken in the context of these two markets. In view of the fact that financial markets are highly integrated, evidence to support the EMH could imply that the results for the stock market are an anomaly that requires further investigation.
Rejection, on the other hand, would support the evidence from the stock market.
6 Introduction
The study proceeds as follows. Chapter Two provides an overview of the Sri Lankan financial system, examining the factors leading to financial deregulation and the extent of regulatory reform. An evaluation of the outcome of financial reforms suggests that overall, Sri Lanka’s financial system had gained in terms of width
(measured by the number of financial instruments and markets), depth (measured by the volume of deposits) and resilience (ability to bounce back from a price change).
It is found, however, that despite the significant development of this sector, the process of reform is still far from complete. In conclusion, it is observed that an examination of the efficiency of this sector would be particularly useful in the context of the deregulation policy.
Chapter Three focuses on the definition of market efficiency, examining its derivation and how the concept of efficiency is made operational. The chapter also briefly identifies conditions under which market efficiency may not obtain and their implications for the economy.
Chapter Four evaluates the applicability of the expectations hypothesis of the term structure for Sri Lanka in the context of the Treasury bill market. The chapter also reviews the literature on the expectations hypothesis of the term structure, focusing on the empirical literature and methodological advances. Employing monthly data covering the 1990.1–1998.12 period, the study rejects the hypothesis that the forward interest rate is an unbiased predictor of the future interest rate. An examination of the spread in forecasting the future direction of short rates indicates that although the
7 Introduction
coefficients on the spread variables have the theoretically anticipated sign, they deviate significantly from the theoretically hypothesized value of two; see Mankiw
(1986) and Mankiw and Miron (1986). While the captive nature of the market is cited as a potential cause of the inefficiency, it is also pointed out that the rejection of efficiency could stem from mis-specification of the expected returns model.
Chapter Five examines the relationship between interest rates and inflation for Sri
Lanka and whether this relationship is consistent with the concept of market efficiency. The chapter employs three data frequencies and two approaches for modeling inflationary expectations. While some support is found for the Fisher relationship under both approaches, it is possible to conclude that strongest support is found for the adaptive expectations approach. The backward nature of this model does not support the predictions of the rational expectations and efficient markets hypotheses. In view of the results obtained, a test of the inverted Fisher effect is carried out.5 The high coefficients on the rate of inflation for the post-deregulation period suggest that approximately the entire variation in the rate of inflation is absorbed by the real rate of interest. The results appear to support the existence of an inverted Fisher effect in the post-deregulation period which provides an explanation for the weak support for tests of the direct Fisher effect.
Chapter Six extends the concept of efficiency to the foreign exchange market. The chapter attempts to evaluate the degree to which financial deregulation has
5 The inverted Fisher effect was put forward by Carmichael and Stebbing (1983) in an attempt to explain the lack of empirical support for the conventional Fisher effect.
8 Introduction
contributed to enhanced efficiency of the foreign exchange market. In fulfilling this objective, the chapter tests the theories of uncovered interest parity (UIP), speculative efficiency and real interest rate linkages. Results for UIP point to the complete rejection of interest rate convergence with the coefficients on the interest rate differentials incorrectly signed in most cases. While deviations from interest parity are attributed primarily to the capital controls in place, government intervention in the foreign exchange market is also cited as a possible cause of the rejection. Despite the complete rejection of UIP, there is some evidence of real interest rates linkage. The results for speculative efficiency are not clear- cut with some tests showing support for the hypothesis, while other tests reject the hypothesis.
The chapter also carries out tests of capital mobility as one potential aspect for the failure of tests of market efficiency. These tests include savings–investment correlations advanced by Feldstein and Horioka (1980), and consumption–income correlations developed by Shibata and Shintani (1998). The tests of capital mobility suggest an increase in capital mobility over time. The Shibata and Shintani model is further investigated by relaxing the assumption of a constant real rate of return. The results appear to confirm increased capital mobility. Overall evidence, therefore, points to an increase in capital mobility, suggesting an enhanced role of the exchange rate in the monetary transmission process.
Chapter Seven summarizes the main conclusions, and examines their policy implications for Sri Lanka. The chapter also makes a number of recommendations–
9 Introduction
which include the privatization of the two state-owned commercial banks, extension of the traditional development finance role of development banks to cover the mobilization of private sector resources, revision of the Colombo Consumer Price
Index (CCPI)–that could help to reinforce the impact of deregulation on efficiency.
10 The Financial System of Sri Lanka
CHAPTER 2
THE FINANCIAL SYSTEM OF SRI LANKA
2.1 INTRODUCTION
The financial systems of many less developed countries have undergone radical structural change in the past two decades following regulatory reform and technological innovation.1 The pace and scope of these reforms have varied widely across countries, given the diversity of their financial systems. Financial deregulation, introduced in the context of an overall deregulation programme in Sri
Lanka, was aimed at enhancing efficiency through greater reliance on the market mechanism and greater competition.2 The reversal of some of these liberalization measures, however, has raised concerns as to whether the objectives of deregulation have been satisfied. The purpose of this chapter is to provide an overview of the reform process and an evaluation of the impact of these reforms on the Sri Lankan economy as background for the empirical work in chapters Four to Six.
The chapter is structured as follows. Section 2.2 provides an outline of the Sri
Lankan financial system and monetary policy prior to deregulation. Section 2.3 indicates the factors contributing to deregulation. Section 2.4 reviews the main aspects of financial deregulation, and Section 2.5 assesses the consequences of deregulation. Section 2.6 summarizes the conclusions.
1 Some of these countries include The Philippines, Thailand, Nepal, India, Malaysia, Indonesia, Turkey, Argentina, Chile, Uruguay. 2 These reforms included the revision of the tariff and tax structures, the abolition of price controls and food subsidies and privatization of government-owned enterprises.
11 The Financial System of Sri Lanka
2.2 THE FINANCIAL SYSTEM AND MONETARY POLICY PRIOR TO
DEREGULATION
From the late 1950s until 1977, Sri Lanka pursued a policy of nationalization and government control of the financial sector. An attempt at liberalization in 1968 was reversed in two years, following a rapid reduction in the country’s foreign exchange reserves and deteriorating balance of payments.3 Thus, the financial system that had emerged by the late 1970s was strictly regulated and highly concentrated. By 1977, the Sri Lankan banking system comprised the Central Bank, four commercial banks, the National Savings Bank (NSB), two development finance institutions–the State
Mortgage and Investment Bank and the Development Finance Corporation of Ceylon; a cooperative rural bank for the mobilization of rural savings and provision of rural credit; and a few finance companies–the Insurance Corporation of Sri Lanka, and the
Employees’ Provident Fund (EPF) for the mobilization of long-term financial resources. Apart from two commercial banks, the finance companies owned by the private sector and one development finance institution in which the Government held equity positions, all other financial institutions were owned and operated by the
Government. Public sector dominance of the financial sector produced a system that lacked competition and innovation and was highly oligopolistic in structure.
Sri Lanka provides an example of a country in which commercial banks predominate in the banking sector. The banking system is highly concentrated and oligopolistic
3 These measures included the devaluation of the Rupee by 20% against the pound sterling, the liberalization of non-essential imports, the relaxation of controls on the remittance of dividends, interest and profits, and tax concessions.
12 The Financial System of Sri Lanka
in nature. The two state-owned banks, the Bank of Ceylon and the People’s Bank, owned more than 70% of the total bank advances and deposits of the four commercial banks that existed. The National Savings Bank, set up to service small-scale savers, was used as the institution through which the Government implemented its interest rate policy. Interest rates were low and administratively determined, in an attempt to encourage investment and finance the servicing needs of the Government. The
Development Finance Corporation of Ceylon met long-term development finance needs not met by commercial banks, while the State Mortgage and Investment Bank dealt predominantly in long-term credit for agriculture, industry and housing construction. The Insurance Corporation of Ceylon and the EPF were required to hold their portfolios in government securities, thus limiting the access of funds to the private sector. Finance companies were less restricted and therefore met higher risk- credit demands not provided by commercial banks.
The money market in Sri Lanka consisted of the Treasury bill market, the inter-bank call money market and the foreign exchange market. Between July 1972 and May
1976, the Sri Lanka rupee had been pegged to the Pound Sterling at a rate of Sri
Lanka Rs.15.6 to 1 Pound Sterling. From May 1976, the official exchange rate for the Sri Lanka rupee was pegged to an undisclosed basket of currencies.4 The parities with other currencies was fixed occasionally on the basis of their cross-rates with the
Pound Sterling. Money market activities were confined primarily to transactions in government securities and Treasury bills during this period. The administered
4 See Savundranayagam (1986).
13 The Financial System of Sri Lanka
interest rates deterred the public from participating in money market activity. Hence, about 92% of the Treasury bills issued were held by the Central Bank.5 The call money market was relatively inactive and began to develop only after liberalization.
Foreign exchange market activities were limited to transactions relating to international trade and foreign aid, with the Government as the main participant in this market. The Colombo Stock M arket was the main capital market in Sri Lanka.
However, the absorptive capacity of this market was low and trading limited. The lack of developed money and capital markets prohibited the use of open market operations and indirect controls for operating monetary policy. Therefore, the monetary authorities were compelled to resort to direct controls, including administered interest rates, credit rationing, ceilings and directives to commercial banks to achieve the desired policy objectives.
Monetary Policy in a Regulated Environment
M onetary policy in the period prior to deregulation was aimed at providing balance- of-payment support, directing credit to priority areas and controlling inflationary pressures generated by high fiscal deficits.6 Therefore, monetary policy measures took the form of direct controls, which included administered interests rates, credit rationing, ceilings and directives to commercial banks; and exchange rate controls, which included controls on foreign travel, and capital transfers, margins on letters of credit, the upward revision of import duties and restrictions on credit for imports.7
5 Central Bank of Sri Lanka Annul Report 1975. 6 The deteriorating balance of payments situation resulting primarily from low growth and failure to diversify exports led to the imposition of exchange and import controls. 7 See Colombage (1993) and Jayamaha (1986).
14 The Financial System of Sri Lanka
The Central Bank’s developmental role also required it to channel credit to priority sectors. Hence, monetary policy measures during this period reflected a combination of restraints and directions on the one hand, and concessions and subsidies on the other. The Bank Rate8 and statutory reserve requirement (SRR)9 were used as major monetary policy tools during the period.
The focus of monetary policy during this period was to take corrective action with respect to the weakening balance of payments. This led the authorities to embark on a series of import, exchange and direct controls which in the long run ran counter to the objectives it originally sought to achieve. These controls included margins on letters of credit, restrictions on credit granted for non-essential imports, restrictions on credit to companies (resident and non-resident) registered in Sri Lanka, and surveillance over hire purchase and finance companies. The country’s import expenditure was brought under an Annual Foreign Exchange Budget. Exchange controls were imposed on foreign travel and capital transfers.10 Quantitative restrictions were placed on imports, and export licences were required for most exports. In 1967, the rupee was devalued, and in 1968 a dual exchange rate system was introduced under the Foreign Exchange Entitlement Certificate (FEEC) to provide further support for the continuing deterioration in balance of payments.11 A
8 The rate charged by the Central Bank on advances to commercial banks for their temporary liquidity needs. It affects the money supply through the monetary base. 9 T he minimum cash reserve ratio against deposits to be maintained by commercial banks. It operates through the money multiplier. This ratio is used by the Central Bank to affect the credit creation of commercial banks. 10 See Jayamaha (1985). 11 Essential imports were governed by the official exchange rate while the FEEC rate applied to non- essential imports. The FEEC rate was initially 44% higher than the official exchange rate and was subsequently raised to 55% in 1969.
15 The Financial System of Sri Lanka
statutory reserve requirement of 50% was imposed on all demand deposits placed in commercial banks from February 1961. This was subsequently increased to 75%.
A two-tier interest rate structure was maintained, one at subsidised rates (below the market rates of interest) and the other at the prevailing market rate, in order to channel credit to priority sectors and restrict credit for non-essential purposes. A policy of “demonitization” was carried out, under which currency notes of denominations Rs.100 and Rs.50 were withdrawn from circulation, with effect from 3
November 1970. This policy was aimed at attracting tax evaders into the banking system.
In the early 1970s, too, monetary policy continued to be dominated by the deteriorating balance of payments. The Bank Rate was raised to 6.5% in 1970, leading to an upward revision of all interest rates. The slow pace at which the economy was growing called for some relaxation in credit controls. Therefore, restrictions on credit to the private sector were relaxed with the view of encouraging private sector participation in economic activity. In 1974, credit restrictions were reintroduced on lending by commercial banks. Restrictions were also placed on
Central Bank accommodation to commercial banks. Lending over and above this rate was subject to a penal rate.
The scope for monetary policy control was continuously impeded, however, by the lack of fiscal restraint. The low interest rate policy, maintained primarily to mitigate
16 The Financial System of Sri Lanka
the Government’s budgetary expenses, had resulted in a low level of financial savings.
This system of regulatory measures was aimed at channelling funds to priority sectors and promoting investment. By the end of the 1970s, the economy was trapped in a vicious cycle of controls and subsidies from which it was difficult to break away without detrimental effects on the economy. It was apparent by then that government regulation of the financial system had inhibited rather than promoted growth.
2.3 FACTORS CONTRIBUTING TO FINANCIAL DEREGULATION
A number of factors, therefore, prompted a change of policy. The excessive controls had resulted in distortion of the market mechanism, thereby discouraging financial savings and investment. Savings and investment recorded a decline from 1970 to 1976.
Table 2.1 Savings and Investment Ratios 1970–1976
1970 1971 1972 1973 1974 1975 1976
Investment Ratio (%) 18.9 17.1 17.3 13.8 15.7 15.6 16.2
Domestic Savings Ratio* (%) 16.7 16.0 16.1 12.5 8.2 8.1 13.9
National Savings Ratio** (%) 15.0 14.7 14.8 11.5 7.5 7.4 13.1
* Domestic Savings as a % of GDP at market prices
** The Domestic Savings Ratio plus net factor income from abroad as a % of market prices
Source: Central Bank of Sri Lanka (1998) “ Economic Progress of Independent Sri Lanka
17 The Financial System of Sri Lanka
Public sector control and ownership of financial institutions had discouraged the private sector from participating actively in economic activity, and by 1976 the public sector had emerged as the main borrower from the financial sector. By 1976, public debt stood at 58.5% of GDP, of which 15% was financed by Treasury bills and 51% by Rupee loans. Furthermore, the Central Bank had emerged as the main source of funds to the economy. The Central Bank held 93% of the total Treasury bill holdings as at the end of 1976. The dominance of the two state-owned commercial banks and restrictions on foreign bank entry had led to the low profitability of commercial banks, while the barriers to trade and international capital movements had led to the creation of an environment not conducive to foreign investment.
An examination of the measures of economic performance, see Table 2.2, suggest that the regulatory policies failed to contribute effectively towards economic development.
Table 2.2
Measures of Economic Performance 1970–1976
1970 1971 1972 1973 1974 1975 1976
GDP Growth Rate 4.3 0.2 3.2 3.7 3.2 2.8 3.0
Inflation 5.9 2.7 6.3 9.7 12.3 6.7 1.2
Unemployment n.a 18.7 n.a 18.3 n.a 19.7 n.a
Current Account -2.6 -1.5 -1.3 -0.9 -3.8 -2.9 -0.2 Deficit as % of GDP
Source: Central Bank of Sri Lanka (1998) “Economic Progress of Independent Sri Lanka”
18 The Financial System of Sri Lanka
Despite the import substitution strategy that the country had been following, the industrial sector failed to contribute effectively towards economic growth and employment. By 1977, the industrial sector accounted for less than 14% of the country’s GDP, 9% of employment and 14% of export earnings.12 The contribution of the industrial sector to GDP and employment record a decline from the levels of
1970.
Therefore, the failure of the closed economic policy in achieving its intended objectives together with the change in the international financial environment towards greater globalization and innovation, deregulation of the financial sector was perceived as a necessary change of course. In July 1977, the United National Party
(UNP) launched an ambitious reform programme, with the aim of promoting greater economic efficiency through enhanced competition and the elimination of distortions in the financial system.13
2.4 FINANCIAL DEREGULATION
The financial deregulation programme as initiated in 1977 constituted the introduction of new financial institutions, instruments and markets, the upward revision of interest rates, the modification of credit and exchange rate controls, relaxation of directed credit, increase in private sector ownership of financial institutions and the introduction of prudential regulations of financial institutions.
12 See Vidanapathirana (1986). 13 The policies of this party were influenced to a great extent by the ideologies of the IMF and World Bank.
19 The Financial System of Sri Lanka
The exchange rate was unified and allowed to float in relation to a basket of currencies.
Two phases can be identified in Sri Lanka’s reform programme. The 1977–1988 period and post-1989 period. The first phase focused on interest rate, exchange rate and banking reform, while the second phase saw more rigorous attempts at stabilization and further relaxation of remaining restrictions on trade and payments.14
The opening of the Colombo Stock Exchange to foreign investors, liberalization of the current account and tax incentives to promote the capital market were among the main efforts at deregulation during the second phase.
2.4.1 Institutional Reforms
In 1979, operational restrictions on foreign banks were removed and, by 1989, nineteen foreign banks had set up in Sri Lanka. The establishment of foreign bank branches was expected to facilitate the inflow of foreign capital and introduce an element of competition to the domestic banking sector. Four new private commercial banks commenced operations during the 1987–1995 period, raising the number of domestic private commercial banks to six. This period also witnessed an expansion in the number of commercial bank branches and volume of deposits mobilized by commercial banks. The number of branches, which stood at 165 in 1970 and 298 in
1976, increased to 489 in 1980 and 1037 by the end of 1998. An important development in branch bank expansion was the opening of branches by the Bank of
14 Foreign aid was made conditional on stabilization and further liberalization: see Dunham and Kelegama (1994).
20 The Financial System of Sri Lanka
Ceylon in India and Pakistan in 1985. The value of deposits, which had grown at an annual average rate of 15.5% in the 1960s and 30.6% from 1970–1976, grew at an annual average rate of 47.6% from 1977–1988 and 36.2% from 1989–1998, see
Figure 2.1.
Figure 2.1
Source: Central Bank of Sri Lanka Annual Reports
This growth was also significant in real terms, particularly in the case of time and savings deposits, as demonstrated by Figure 2.2. Time and savings deposits, which had grown at an annual average rate of 6% in real terms over the 1970–1976 period, grew at an annual average rate of 13.1% in the 1977–1998 period.
21 The Financial System of Sri Lanka
Figure 2.2
Source: Computed from Central Bank of Sri Lanka Annual Reports.
It is evident from the above figure that there was also a change in the composition of deposits mobilized by commercial banks in favour of interest-bearing deposits. This resulted from the interest rate reforms of 1977.
22 The Financial System of Sri Lanka
Table 2.3
Commercial Bank Deposit Mobilization
(Rs. Million)
Year Demand % Savings % Time % Total % Deposits Deposits Deposits 1950 733.9 90.5 26.2 3.2 50.9 6.3 811.0 100.00
1955 745.4 81.8 68.2 7.5 97.3 10.7 910.8 100.00
1960 939.1 60.8 223.9 14.5 382.8 24.7 1545.8 100.00
1965 939.1 60.8 223.9 14.5 382.8 24.7 545.8 100.00
1970 1225.5 51.2 534.8 22. 0 633.7 26.5 2394.0 100.00
1975 1882.4 52.1 947.3 26.3 780.8 21.6 3610.5 100.00
1980 6687.8 38.7 2509.6 14.5 8093.9 46.8 7,291.3 100.00
1985 11,465.4 25.2 1,155.2 24.5 22,837.4 50.3 45,458.0 100.00
1990 29,890.0 22.7 43,601.5 33.2 57,897.1 44.1 131,388.6 100.00
1995 38,951.5 17.0 87,227.0 38.1 102,751.1 44.9 228,929.6 100.00
1998 57,232.0 16.6 157,325.0 45.6 130,504.0 37.8 345,061.0 100.00
Source: Central Bank of Sri Lanka Annual Reports
In comparison to commercial banking, savings banks progressed relatively slowly in
Sri Lanka. One of the main weaknesses of savings banks in Sri Lanka is the lack of diversity in the investment portfolio of the National Savings Bank. This results primarily from the requirement to invest a large proportion of its funds in government securities. Despite the fact that there has been an increase in the volume and change in the composition of deposits mobilized by the National Savings Bank in favour of fixed deposits following the interest rate reforms, this institution has failed to effectively contribute towards raising the level of national savings.
23 The Financial System of Sri Lanka
Table 2.4
Mobilization of Deposits by The National Savings Bank 1972–1998 (Rs. Million)
End of Savings Fi xed Savings Premium NSB Total Period Depos i ts Depos i ts Certificates Savings Pension Savings Bonds Scheme 1972 913 53 67 – – 1,033
1975 1,414 243 113 – – 1,770
1980 2,146 2,759 39 7 3 4,954
1985 3,701 9,480 8 9 31 13,229
1990 9,527 13,529 3 9 36 23,104
1995 20,942 36,726 3 17 68 57,756
1998 32,880 50,383 3 869 145 84,280
Source: Central Bank of Sri Lanka Annual Reports
Figure 2.3
Source: Central Bank of Sri Lanka Annual Reports
24 The Financial System of Sri Lanka
Figure 2.4
Source: Computed from Central Bank of Sri Lanka Annual Reports
While the growth in savings deposits reflects a decline in real terms immediately following the reforms, and then a very gradual rise from about 1983, fixed deposits indicate a sharp rise immediately following the reforms and then a decline through the late 1980s. This could be attributed to the rising rates of inflation experienced in the latter part of the 1980s. From the 1990s, there appears to be a reversal in this trend.
Contractual savings institutions in Sri Lanka are of two types: insurance companies and compulsory savings institutions. The Insurance Corporation of Sri Lanka, which was set up in 1961 and nationalized in 1964, had the monopoly of life insurance until
1979. In 1979, the National Insurance Corporation was set up in order to introduce
25 The Financial System of Sri Lanka
an element of competition to the insurance business. In 1985, the public sector monopoly of insurance business was relaxed and private sector firms were permitted to operate in this field. By 1998, there were seven insurance companies operating in
Sri Lanka. Until 1982, the Employee’s Provident Fund (EPF) was the only compulsory savings institution in Sri Lanka. In 1982, the Employee’s Trust Fund
(ETF) was set up to provide retirement benefits to employees and encourage employee participation in the ownership of private sector enterprises. These institutions indicate significant growth during the post-reform period as measured by the ratio of assets to GDP: see Figure 2.5.
Figure 2.5
Source: Central Bank of Sri Lanka (1998); “ Economic Progress of Independent Sri Lanka”
In 1979, the National Development Bank was set up to supplement the work of the
Development Finance Corporation of Ceylon. This met long-term credit requirements for the promotion of industrial, agricultural and commercial activity.
26 The Financial System of Sri Lanka
The National Housing Development Authority and the Housing Development
Finance Corporation were set up to provide credit for housing and implement government policy on housing. For provision of insurance coverage to the export sector, the Sri Lanka Export Credit Insurance Corporation was established in 1979.
Figure 2.6
Source: Central Bank of Sri Lanka (1998); “ Economic Progress of Independent Sri Lanka”
In 1985, the Central Bank of Sri Lanka became directly involved in the provision of rural credit by setting up Regional Rural Development Banks (RRDBs). These banks have expanded significantly over time, as illustrated by Figure 2.7. By 1997, the number of RRDBs had risen to 17, with a branch network of 176. These banks have contributed significantly to mobilizing rural savings by introducing attractive savings schemes and providing mobile banking services to unbanked and underbanked areas.
A distinct feature of the rural credit policy in 1996 was the emphasis given to micro-
27 The Financial System of Sri Lanka
financing as a method of promoting self-employment among the unemployed, particularly the educated youth. In 1998, the business of 8 RRDBs was vested in three Regional Development Banks, brining the number of RRDBs down to 9 from
17 and the number of branches to 96.
Figure 2.7
Source: Central Bank of Sri Lanka Annual Reports
Financial deregulation led to a rapid growth in the number of finance companies. The lack of proper supervision and regulation led to the closure of a number of these companies in the mid-1980s. After the institution of stringent regulations to ensure the viability of these companies, the number in operation declined significantly. By the end of 1998, there were 25 finance companies in operation.
28 The Financial System of Sri Lanka
Liberalization also led to the creation of number of specialized financial institutions.
Merchant banks were set up to cater to the needs of a growing business community.
By the end of 1998, there were 10 merchant banks operating in Sri Lanka. The need for long-term financing by entrepreneurs led to the creation of venture capital companies whose function is to provide long-term credit to small companies undertaking new enterprises. Leasing companies were set up to provide leasing facilities. There were five leasing companies by the end of 1998. There has been a shift in the portfolio of leasing companies from hire purchase to lease financing in the recent past. The development of this industry is expected to help long-term capital formation in the country. With the development of an active money market, money broking firms entered the market. Their main function is to act as brokers in the inter-bank call money market, leading to increased efficiency in money market activities. A Primary Dealer System was set up in order to develop an active market for government securities. Primary Dealers are given the exclusive right to purchase government bills and bonds when they are first auctioned by the Government, and sell these to investors at a later stage.
It is evident from the above that financial deregulation led to an increase in size of the financial sector.
Money, Foreign Exchange and Capital Markets
With institutional growth, financial markets and instruments began to acquire greater significance. The introduction of a number of new markets and instruments was
29 The Financial System of Sri Lanka
observed during this period to supplement the role of financial institutions as savings mobilisers. There was significant development of the money, foreign exchange and capital markets following the reforms of 1977.
Money Market
This period witnessed a notable expansion of the Treasury bill market. Discount and rediscount windows were opened by the Central Bank in 1981, providing impetus to the development of an active secondary market for Treasury bills. In 1986, weekly primary auctions for Treasury bills commenced, permitting these yields to respond to market conditions, thus increasing the attractiveness of these instruments to investors.
With the objective of developing an active secondary market for Treasury bills, participation in primary market auctions was restricted to accredited primary dealers from 1992. Moreover, the introduction of Treasury bills with maturities of 6 and 12 months in 1989 have provided a better portfolio option to investors and facilitated further development of this market.
In 1993, the Central Bank set up a Repurchase Market (repo) for Treasury bills with a view to stabilizing the lower end of the call money market. This has been successful in preventing call rates from falling very low. A Reverse Repurchase Market (reverse repo) was set up in 1995 to stabilize the upper end of the call market. This, however, has not been as successful as the repo market. The growth of the Treasury bill market and its diversity resulted in the creation of a market-determined benchmark rate for short-term interest rates. It also led to an expansion in short-term money market
30 The Financial System of Sri Lanka
activity, which in turn permitted the more intensive use of open market operations.
The rate of interest also began to be used more effectively as an instrument of policy.
Liberalization also saw an increase in call market activity. Currently, the average daily volume of transactions is around Rs.3000–4000 million.15 This market, however, is thin and therefore very volatile. Large players are in a position to influence the market. Interest rate volatility has been greatly reduced with the introduction of the repo and reverse repo markets.
The high degree of government intervention and high intermediation costs of the two state-owned commercial banks had previously prevented a term structure from developing. The post-deregulation period, however, saw the introduction of a number of financial instruments designed to promote the development of the short end of the money market. These included certificates of deposit, commercial paper, and
Treasury bonds. Certificates of deposit, introduced in 1981, succeeded in attracting a large proportion of “black money” into the banking system. In 1993, the private sector, for the first time, began issuing instruments by the name of commercial paper.
This is a negotiable short-term instrument that enables corporate borrowers to raise funds in the money market. These instruments have grown from a negligible amount to Rs.4.25 billion by 1995. The yield rates determined at the weekly Treasury bill auctions serve as the benchmark rate for the commercial paper rates.
15 Central Bank of Sri Lanka (1998); “ Economic Progress of Independent Sri Lanka”.
31 The Financial System of Sri Lanka
Foreign Exchange Market
The foreign exchange market in Sri Lanka comprises the domestic market, consisting of the Central Bank of Sri Lanka and authorised foreign exchange dealers, and the offshore market after 1977.
Deregulation entailed the unification of the exchange rate at an initially depreciated rate of 46% and the introduction of a managed float. The Central Bank commenced quoting daily rates for six major currencies–the US dollar, the Deutsch Mark, Franc,
Yen, UK pound and Indian Rupee. The US dollar was the intervention currency. In
1982, the Central Bank limited its quotation of rates to the intervention currency and permitted commercial banks to determine cross-rates for other currencies based on market conditions. In 1990, this practice was abandoned and the Central Bank commenced announcing daily buying and selling rates for the US dollar.
Commercial banks were authorized to set up Foreign Currency Banking Units
(FCBUs) in 1979 to promote offshore banking in Sri Lanka. These units could deal in foreign currency units with non-residents, approved residents and Board of
Investment (BOI) enterprises. The total assets of the FCBUs rose from Rs.596 million at the end of 1979 to Rs. 20,458 million in 1983 and Rs.103,465 million by the end of 1998.16 Moreover, a Non-Resident Foreign Currency (NRFC) account scheme was introduced in 1978 to facilitate inward remittances from Sri Lankans. In
1991, a Resident Foreign Currency (RFC) scheme was introduced, permitting
16 See Central Bank of Sri Lanka Annual Reports.
32 The Financial System of Sri Lanka
residents to open accounts in designated currencies with a minimum balance of
$US500.
From 1991, money changers were authorised to engage in foreign exchange transactions apart from the Central Bank and commercial banks. This was aimed at minimising price distortions in the domestic foreign exchange market. An inter-bank market for forward exchange transactions was introduced in 1983. However, the volume of transactions in this market is very small, currently amounting to about
$US8–10 million per day.17 Forward rates can only be obtained for periods of less than six months.
Capital Markets
Share Market
Although share market activity in Sri Lanka dates back to 1896, it was stagnant over a period of twenty years following the Government's policy of nationalization and controls, beginning in the late 1950s. An important feature in financial-sector development in the post-1977 era was the promotion of an institutional framework for the development of the capital market. In 1982, the Colombo Stock Exchange Ltd. was set up to facilitate trade in shares. The Capital Development and Investment
Company was established in 1983 to provide equity financing, and the Security
Council in 1987 to regulate the activities of the stock market. By the end of 1998,
238 companies were listed on the Colombo Stock Exchange.
17 Central Bank of Sri Lanka (1998); “ Economic Progress of Independent Sri Lanka”
33 The Financial System of Sri Lanka
Table 2.5
S hare Market Indicators
M = million 1990 1991 1992 1993 1994 1995 1998
Market Capitalisation (Rs M) 36,880 81,084 64,655 123,789 143,210 106,869 116,660
No. of new issues (number) 1 6 15 12 21 17 6
Total no. of shares issued (M) 11 13 78 42 190 100 28
Number of shares traded (M) 42 79 88 351 506 315 634
Value of shares issued (Rs M) – 326 1,370 1,247 4,859 2,120 349
Value of shares traded (Rs M) 1,563 4,304 4,969 18,579 34,522 11,249 18,130
Price Indices 384 837 605 978 986 663 597 CSE All Share* (as at end Dec.)
CSE Sensitive** 680 1,198 826 1,442 1,438 990 923 (as at end Dec)
* Colombo Stock Exchange All Share Price Index is based on the prices of all listed securities. ** Colombo Stock Exchange Sensitive Price Index is based on a smaller number of well established “ blue chip” securities. Both indices have 1985 as their base year.
Source: Central Bank of Sri Lanka Annual Reports
34 The Financial System of Sri Lanka
Figure 2.8
Source: Central Bank of Sri Lanka Annual Reports
The stock market reflected dynamism and growth during this period with market capitalization as a per cent of GDP rising from 7.3% to 27.4% within the 1986 to
1994 period. Among the measures taken to promote foreign capital into the country were the abolition of the wealth tax on listed company shares, elimination of the 100 per cent transfer tax on share purchases by non nationals and exemption of share transactions from stamp duty. The opening up of the share market to foreign investors in 1992 marked a significant development in share market activity.
Domestic companies were further, permitted to make equity issues abroad on a case by case basis. March 1994, saw a successful launch of the first Euro issue, a Global
Depositary Receipt issue, by a Sri Lankan firm which raised 33 million US dollars. In
35 The Financial System of Sri Lanka
order to improve the efficiency of the market, a Central Depository System was set up in 1991, and in 1997 a fully automated trading system. These measures collectively contributed to increased local participation in share market activity, rising capital inflows and enhanced mobilization of funds for investment projects.
Figure 2.9
Source: Central Bank of Sri Lanka Annual Reports
A decline in foreign capital inflows was observed however, in 1998, following the
East Asian crisis of July 1997, and the nuclear tests carried out by India and Pakistan in May 1998.
Bond Market
The bond market in Sri Lanka remains relatively undeveloped in comparison to the share market. The bond market trades in maturities of over one year and consists of
36 The Financial System of Sri Lanka
government bonds and corporate bonds. The prerequisite for publicly controlled contractual savings institutions to hold a large proportion of their funds in government securities at managed rates, has prevented the development of a secondary market for long-term debt instruments. A few private companies have issued debentures in limited quantities, of which some are listed on the Colombo
Stock Exchange.
In an attempt to develop this market, commercial banks were permitted to issue certificates of deposits with maturities of over four years in 1991. A new instrument termed the Floating Rate Certificate of Deposit, which could be discounted in the secondary market was introduced in 1996. This instrument guaranteed a minimum rate of interest above the Average Weighted Fixed Deposit Rate. In 1997, the
Government introduced the Treasury bond, which was auctioned by the Central Bank on behalf of the Government, permitting market forces to determine the yield rates for this instrument. Treasury bonds with maturities of two, three and four years have been issued, with these rates serving as a benchmark for medium and long-term private debt.
A noteworthy development in the capital market was the creation of Unit Trusts in
1991 to provide portfolio investment management facilities to small investors. At the end of 1998, there were ten Unit Trusts operating in Sri Lanka with funds worth
Rs.2700 million. Therefore, the removal of regulatory restrictions led to a significant
37 The Financial System of Sri Lanka
expansion of the financial sector, as evidenced by the creation of new institutions, markets and instruments.
2.4.2 Monetary Policy in a Deregulated Environment
Greater reliance was placed on the market mechanism in resource allocation in an attempt to enhance the effectiveness of monetary policy. The gradual elimination of interest and exchange controls, curtailing of direct controls, less intensive use of the
Bank Rate as an instrument of monetary policy, the unification of the exchange rate and adoption of a managed float were elements of the new policy. A major element of reform was the deregulation of interest rates. The entire interest rate structure was revised upward in an attempt to stimulate financial savings and channel them to their most productive use. The interest rate on fixed deposits of the National Savings Bank was increased from 7.5% to 12–18%, and on savings deposits from 2% to 8.4%. This consequently led to an increase in the deposit rates of commercial banks from, 5.5% to 7.2% on savings deposits, 6.37% to 8.5% on three-month fixed deposits, 6.75% to
11.5% on six-month fixed deposits and 7.25 to 14.5% on one-year fixed deposits.18
This triggered a rise in the Treasury bill, interbank and lending rates of commercial banks. Treasury bill rates rose from 5% in 1976 to 9% in 1977, call rates from 5%–
8% to 7%–9.5% and lending rates from 8.5%–14% to 13%–19%. The upward revision of interest rates, designed to ensure positive real interest rates, was frustrated by the acceleration of inflation in 1980.
18 See Central Bank of Sri Lanka Annual Report 1977.
38 The Financial System of Sri Lanka
Table 2.6
Deposit Rates of Commercial Banks
(Rates per Annum)
Year Demand Savings Time
3 months 6 months 1 year 2 years
1952 nil 1-2 0.75 0.75 0.75 -
1955 nil 1-2 0.50 0.50 1.0 -
1960 nil 2 2.50 2.50 2.50 -
1965 nil 2-3 0.5-3.25 2.5-3.25 0.5-3.25 -
1970 nil 4.5 4.5-4.75 4.5-4.75 4.5-4.75 -
1975 nil 5.5 6.0-6.75 6.5-7.0 7.0-7.5 7.0-7.5
1977 nil 7.5 8.50 11-12 14-15 -
1980 nil 10-14 12-16 15-18 20 22
1985 nil 10-13 9-17 9-19 12-18 16-18
1990 nil 5.5-14 7-22 8-21.5 11-21 13-20.5
1995 nil 5-13 9.0-18.25 10-17 10-17 12-16
1998 nil 2-10 7-13 7-11.75 9-13 9.56-12
Source: Central Bank of Sri Lanka Annual Reports
The weakening of the distinction between demand and time and savings deposits with
the introduction of new financial instruments and greater capital mobility,
neccessiated the adoption of a broader monetary aggregate for policy purposes.19
Until 1980, it constituted currency and demand deposits held by the public as defined
19 Central Bank of Sri Lanka (1990); “40th Anniversary Commemorative Volume of the Central Bank of Sri Lanka 1950-1990”
39 The Financial System of Sri Lanka
by the Monetary Law Act of 1949. In 1980, the monetary aggregate utilized for policy purposes was redefined with money supply being extended to cover time and savings deposits held by the public as well.
The development of the Treasury bill market enabled open market operations to be employed as an active instrument of monetary policy. Aggressive open-market operations succeeded in significantly reducing government holdings of Treasury bills from 92% in 1975 to 45% in 1990 and 7.46% by 1998: see Figures 2.10 and 2.11.
From 1993, commercial banks became the largest contributor of short-term finance to the Government. Interest rates in the Treasury bill market became increasingly market determined with only minor interventions by the monetary authorities from
1988 onward. The statutory reserve ratio continued to be used as a monetary policy tool during this period. In order to curb the overall liquidity of commercial banks, the reserve ratio on demand deposits was raised from 12% to 14% and the ratio on time and savings deposits from 5% to 6% with effect from June 1981. This was further raised to 16% on demand deposits and 8% on time and savings deposits in 1983. In order to meet the increasing demand for credit by the private sector, the structure of statutory reserve ratios was simplified in 1987 and a uniform ratio of 10% was imposed on demand, time and savings deposits.20 However, in order to restrict the growth in credit subsequently, the statutory reserve requirement was increased and by
1996 the ratio was 15% on all deposits.
20 See Colombage (1993).
40 The Financial System of Sri Lanka
Figure 2.10
Ownership of Treasury Bills as at December 1975 EPF 0. 04% Insuranc e& Financ e Co' s 0. 13% Other Provident Funds Departmental & Other 0. 09% Of f icial Funds 6%
Comme rc ia l b anks 2%
Central Bank 92%
Source: Central Bank of Sri Lanka Annual Report 1975
41 The Financial System of Sri Lanka
Figure 2.11
Source: Central Bank of Sri Lanka Annual Report 1998
One of the main problems confronted by the Central Bank in formulating monetary policy was the lack of a coordinated fiscal policy. A massive investment programme undertaken by the Government in irrigation, power and housing immediately following the reforms exerted pressure on the Government budget, leading to double digit inflation during the post reform period. Capital expenditure rose from Rs.6,614 million in 1978 to Rs.29,534 million by 1985. Although these projects were financed to a large degree by foreign aid, the high level of capital expenditure imposed a considerable strain on the budget. The increased demand for credit by the private sector, added further pressure on the banking system. By 1980, domestic credit had
42 The Financial System of Sri Lanka
grown by 72% over the previous year, money supply by 18.7% and the rate of inflation as measured by the Colombo Consumer Price Index was 26.1%.21 The
Government, therefore, was compelled to resort to action that ran counter to the reforms of 1977. In 1981, the Central Bank introduced a National Credit Plan under which money and credit targets were issued and credit was channelled to priority sectors. The Bank Rate was increased from 12% to 14%, the penal rate structure from 20%-30% per annum to 21%-30%, and the Statutory Reserve Ratio on demand deposits from 12%-14%. M oreover, a secondary Treasury bill market was set up through which the Central Bank engaged in Open Market Operations. In 1982, interest rates were revised downward to encourage private sector participation in economic activity. The Central Bank continued to intervene in the allocation of credit to priority sectors. In 1980, a Medium and Long-term Credit Refinance Fund was set up to provide credit for export-oriented projects at concessionary rates of interest.
In May 1984, a ceiling was introduced on credit to the private sector with the exception of agriculture and exports. Banks were requested to halt forward exchange facilities for non-essential imports and credit to importers beyond the level reached on
23rd March 1984. In 1984 restraints were placed on lending by Finance Companies and Finance Companies were requested to maintain a reserve ratio of 15% against total deposits.
21 Computed from Central Bank of Sri Lanka Annual Report 1980.
43 The Financial System of Sri Lanka
The failure to achieve the desired results made the Central Bank issue its own securities. In June 1984, the Central Bank engaged in Open Market Operations to absorb the excess liquidity that had arisen in the economy due to the favourable prices experienced by tea. By the end of 1985 the rate of inflation had fallen to 1.5%. On the supply side, exporters were instructed to sell their export proceeds to commercial banks with effect from March 1984 to guard against exchange rate fluctuations.
Moreover, the bank introduced an Interest Rebate Scheme to hasten the repatriation of export proceeds and reduce the cost of export credit.
After 1985, the Government's resort to borrowing from the banking sector, became a major destabilizing force. Net credit to the Government by the banking sector rose by
Rs 2 billion in 1986, Rs 4.5 billion in 1987 and Rs 9 billion in 1988. In an attempt to reduce its holdings of Treasury bills, the Central Bank increased the Treasury bill rate from 9.5% to 17%. By 1989, the Bank's holdings of Treasury bills had declined from
88% in January to 57% in December 1989. In 1996, the Treasury bill limit was frozen at Rs. 125,000 to establish fiscal discipline.22 Monetary policy in the early
1990s was restrictive, in an attempt to curb the inflationary pressures generated by the growth in domestic credit and increase in external assets. The Bank Rate was increased from 15% to 17% per annum and the Statutory Reserve Requirement from
10% to 13% on all deposits.
22 See Central Bank of Sri Lanka Annual Report 1996.
44 The Financial System of Sri Lanka
Open M arket Operations were further intensified through the Treasury bill market. In
August 1993, the Statutory Reserve Ratio was further increased to 15% to control credit expansion. Refinance facilities provided under the New Comprehensive Rural
Credit Scheme and Export Credit Refinance Scheme were terminated in 1994 to restrain the growth of reserve money while, restrictions were placed on new credit granted under the Medium and Long-term Credit Fund. As at December 1998 the
Bank Rate remained unchanged at 17% and the Statutory Reserve Requirement stood at 12% on all deposits.
On the external front, several measures were taken to promote greater integration of the Sri Lankan economy with the rest of the world. Among these were the rationalization and simplification of investment procedures, the removal of the 100 per cent transfer tax on share transactions with non nationals, relaxation of exchange controls, introduction of Share Investment External Rupee Accounts and abolition of the capital gains tax on the proceeds of sale of shares. These measures culminated in the opening of the share market to foreign investors in 1992 and liberalization of the current account in 1994.23
Despite the move towards more market oriented policy measures, direct controls continue to exist in the form of margin requirements and credit ceilings. Successful monetary policy management has been continually hampered by the lack of a coordinated fiscal policy.
23 See Hettiarachchi and Herath (1995).
45 The Financial System of Sri Lanka
Although the Government is aware that successful monetary policy management would, finally depend on a reduced budget deficit, it is not conceivable given the rising interest payments on debt, escalating defense expenditure and commitment to the provision of subsidies.
2.5 EVALUATION OF POLICY REFORMS
Financial deregulation generated substantial change in the financial sector, manifest in its size, structure and operational efficiency. This section surveys the economic consequences of financial reform.
2.5.1 Structure
Regulatory reform in Sri Lanka led to a significant expansion in the size of the financial sector as evidenced by the rising share of financial services to GDP and investment, see Table 2.7. While the ratio of financial services to GDP has increased from 1.3% in the pre-deregulation period to 7.6% in 1998, indicating increased monetization; investment in the financial sector has risen from 7.8% to 26.9% in the same period, reflecting a higher absorption of resources by this sector.
46 The Financial System of Sri Lanka
Table 2.7 Share of Financial Services* to GDP and Investment
Financial Sector Contribution to GDP Investment in Financial Services
Per cent of Gross Capital Formation
______
1970-1976 1.3 7.8
1977 1.6 10.0
1978 2.1 9.9
1979 2.5 9.2
1980 2.9 7.9
1981 3.1 10.4
1982 3.9 12.2
1983 3.7 11.9
1984 3.4 11.9
1985 3.8 14.7
1986 4.2 16.1
1987 4.2 16.2
1988 4.4 17.8
1989 4.6 19.2
1990 4.6 18.5
1991 4.9 19.3
1992 5.4 20.2
1993 6.1 21.8
1994 6.8 22.8
1995 7.2 25.2
1996 7.1 26.7
1997 7.5 27.5
1998 7.6 26.9
* Includes banking, insurance and real estate
Source : Computed from Central Bank of Sri Lanka Annual Reports
47 The Financial System of Sri Lanka
The increase in the size of the financial sector is also reflected in the rise of the financial sector’s share in total employment, see Figure 2.12.
Figure 2.12
Note: Includes banking, insurance and real estate Source: Department of Census and Statistics (1998) “Statistical Profile of Sri Lanka” and Labour Force Survey 4th Quarter 1998
The expansion in the bank branch network provides further evidence of the increase in financial services and overall access to credit: see Figure 2.13.
48 The Financial System of Sri Lanka
Figure 2.13
Sources: Central Bank of Sri Lanka Annual Reports and Hettiarachchi (1986)
The growth in the number of bank branches, consequently led to a notable decline in banking density: see Figure 2.14.24
24 Banking density is defined as the population per branch.
49 The Financial System of Sri Lanka
Figure 2.14
Sources: Central Bank of Sri Lanka Annual Reports, Central Bank of Sri Lanka (1998) “Economic Progress of Independent Sri Lanka” and Hettiarachchi (1986)
2.5.2 Financial Depth25
Both non-financial and financial intermediaries have benefited from the increased access to funds and broader range of instruments. Apart from increasing the access to credit, innovations designed to encourage savings among target groups, including children, families, teachers, together with the liberalization of bank interest rates led to a significant expansion in deposit mobilization. (See Tables 2.3 and 2.4, and
Figures 2.1 to 2.4.)
25 With the work of McKinnon (1973) and Shaw (1973), there developed a large literature on the role of finance in economic development. This was subsequently confirmed by Galbis (1977), Fry (1978, 1980, 1995), Mathieson (1980) and more recently by Choe and Moosa (1999), among others.
50 The Financial System of Sri Lanka
Table 2.8
Financial Depth
1970-1976 1977-1988 1989-1998
Ratio of M1 to GDP 14.1 14.1 11.5
Ratio of M2 to GDP 22.0 28.7 31.1
Source: Computed from Central Bank of Sri Lanka Annual Reports
These measures led to a growth in financial depth as measured by the ratio of M2 to
GDP. This ratio records an increase from 22% to 31.1% between the 1970-1976 and
1989-1998 periods. The ratio of M1 to GDP indicates a decline in the post - deregulation period reflective of the shift in asset portfolios of the general public from cash and demand deposits to time and savings deposits.
The marked increase in total assets of financial institutions as illustrated by Figure
2.15, is further indication of the growth in depth of this sector. Financial assets as a proportion of GDP recorded a substantial rise from 64.3 in 1970 to 128.4 by 1996.
51 The Financial System of Sri Lanka
Figure 2.15
Source: Central Bank of Sri Lanka (1998) “Economic Progress of Independent Sri Lanka”
Despite the fact that liberalization has contributed to an appreciable growth of the money market, particularly the Treasury bill market, there is scope for further development. Treasury bills outstanding rose from Rs. 67,968 million to Rs. 119,996 million from 1990 to 1998, while interbank transactions outstanding rose from
Rs.1988 to Rs.4272 in the same period. However, the combined share of other money market instruments outstanding to GDP averaged only about 2% in 1998 in comparison to 12% for Treasury bills suggesting the need for promotion of greater participation in money market activity.
52 The Financial System of Sri Lanka
Figure 2.16
Source: Central Bank of Sri Lanka Annual Reports and Central Bank of Sri Lanka Socio Economic Data 1998
2.5.3 Efficiency
Operational Efficiency
One measure of operational efficiency of a banking system is the intermediation margin between the lending rate and the deposit rate. Despite the fact that financial reform is expected to enhance the operational efficiency of the banking sector through increased competition, the average interest margin of commercial banks widened from 1.32 in the pre deregulation period to –2.48 in the first phase of deregulation and
53 The Financial System of Sri Lanka
then declined to –1.88 in the second phase of deregulation.26 The negative interest margin could be attributed primarily to the high intermediation costs of the two state owned commercial banks, the Bank of Ceylon and the People’s Bank, which own
59% of the total assets of the banking system. Continued pressure by the
Government to finance unviable projects has led to a high volume of non-performing loans by these two banks. By 1996 the ratio of non-performing loans to total advances stood at 18.9% for the Bank of Ceylon and 11.8% for the People’s Bank.27 Widening intermediation margins could also be in part explained by rising inflation. In order to maintain real profits at the same level, the intermediation margin has to rise with rising inflation. Figure 2.17 indicates, however, that the spread has been narrowing from about 1993.
Figure 2.17
Source: Computed from Central Bank of Sri Lanka Annual Reports
26 The lending rate employed is the minimum rate charged by commercial banks on unsecured loans and overdrafts and the deposit rate is the maximum rate offered by commercial banks on 12-month deposits. 27 See IMF Staff Country Report 1997.
54 The Financial System of Sri Lanka
Productivity gains in financial services are also reflected in lower costs required for processing financial transactions, in particular staff costs. Staff costs in the two state- owned commercial banks, the Bank of Ceylon and the People’s Bank respectively have fallen only marginally from 16% and 22% in 1990 to 14% and 20% in 1998.
Productivity gains are also illustrated by the expansion of services provided by financial institutions. The growth in the automated teller machine network from 79 in
1993 to 180 by June 1997, introduction of telebanking whereby customers are able to conduct their banking business over the telephone, extension of banking hours indicate increases in productivity. The establishment of the Sri Lanka Automated
Cheque Clearing House (SLACH) which provides automated cheque clearing facilities has contributed significantly to the enhanced efficiency of banking services.
By 1996, the SLACH was processing an average of 109,000 cheques per day. The introduction of an interbank payment system in 1993 for carrying out high volume transactions and the implementation of the Society for Worldwide Interbank Financial
Telecommunications (SWIFT) in 1994 for the electronic transfer of funds have also contributed to increased operational efficiency.
Allocative Efficiency
Experience suggests that the use of discriminatory credit controls could inhibit overall efficiency by diverting resources from higher to lower productivity uses. The elimination of distortions in the allocation of investment constitutes one of the most important resource allocation gains from financial deregulation. As the removal of direct controls on interest rates is expected to eliminate distortions in the provision of
55 The Financial System of Sri Lanka
funds thereby improving the allocation of investment, a measure of allocative efficiency is provided by the interest rate differential between the representative bank lending rate and the unregulated money market rate. Compared below are the average Bank Rate lending rate with the unregulated 3 month Treasury bill rate. In the absence of distortions, bank lending rates are expected to be slightly above money market rates, indicating higher risk. A negative difference between the two rates on the other hand, is reflective of the degree of distortions from interest rate controls.28
Table 2.9
Interest Rate Differential
1970-1976 1977-1988 1989-1998
Bank Rate less Three Month Treasury Bill Rate 1.53 -0.54 3.6
Source: Computed from Central Bank of Sri Lanka Annual Reports
The interest rate differential appears to have increased from a negative level in the first phase of deregulation to a positive level in the second phase indicative of increased efficiency.
Notable improvements in the capacity of financial markets to channel funds to their most productive use has come from increased capital mobility. It is argued that
28 See Edey and Hviding (1995).
56 The Financial System of Sri Lanka
greater financial openness permits savings to be pooled and allocated globally
through the movement of capital among countries in response to opportunities for real
investment, thereby leading to the international equalization of interest rates. This
issue is empirically investigated in Chapter Six.
The removal of restrictions on external transactions, and the opening of the Colombo
Stock Exchange to foreign investors, led to a significant inflow of private foreign
capital which paved way for international diversification of portfolio holdings by
institutional investors, see Table 2.10. The potential for international financial
diversification has not been fully realized however, given the capital controls.
Table 2.10
Rate of Foreign Direct Investment Inflows to GDP and to Gross Fixed Capital
Formation
1970–1976 1977–1988 1989–1998
Percentage share of inward FDI flows to GDP 0.01 0.71 0.91
Percentage share of inward FDI flows to gross
Fixed capital formation 0.07 2.61 3.37
Percentage share of private foreign inflows to
GDP -0.17 0.99 1.36
Percentage share of private foreign inflows to
Gross fixed capital formation -1.02 3.63 4.91
Source: Computed from Central Bank of Sri Lanka Annual Reports
57 The Financial System of Sri Lanka
Further evidence of the growing international diversification of portfolio holdings by
institutional investors is provided by the non-national participation to total turnover in
the Colombo Stock Exchange. Foreign investor participation in the Colombo Stock
Exchange in 1998 however was low due to the contagion effects of the East Asian
financial crisis. Foreign participation as a percentage of total turnover was only 35%
in 1998 as against 54.75% in 1996.
Table 2.11
Foreign Participation in Colombo Stock Exchange
1991 1992 1993 1994 1995 1998
Foreign participation as
% of total turnover 38.0 41.9 51.6 60.0 59.5 35.0
Source: Computed from Central Bank of Sri Lanka Annual Reports
2.5.4 Impact on Financial Stability
There is a widespread belief that financial markets have in general become
progressively more volatile with financial deregulation. Despite these concerns, there
is little evidence that there has been any persistent increase in volatility. The absolute
monthly volatilities in Treasury bill rates and exchange rate are illustrated in Figures
2.18 and 2.19. The figures suggest that interest rate reform and the adoption of a
managed float have led to greater sensitivity of interest and exchange rates to market
58 The Financial System of Sri Lanka
sentiment and hence a higher degree of volatility. There is no evidence to suggest a general trend increase in volatility as a result of deregulation. In fact, the Sri Lankan financial system proved to be fairly resilient to the East Asian currency crisis of 1997.
This could be attributed in part however, to the capital controls in place.
Figure 2.18
Volatilities are calculated on the basis of absolute monthly percentage changes
Source: Computed from Central Bank of Sri Lanka Annual Reports
59 The Financial System of Sri Lanka
Figure 2.19
Volatilities are calculated on the basis of absolute monthly percentage changes
Source: Computed from Central Bank of Sri Lanka Annual Reports
2.5.5 Consequences for the Macro-economy and Macreconomic Policy
Implicit is the assumption that enhanced financial deepening is positively related to the real sector of an economy. Therefore, there is reason to believe that financial deregulation might have had a favourable impact on savings and investment. The higher rates of savings and investment experienced during the post-deregulation period provide evidence in justification of this.
60 The Financial System of Sri Lanka
Table 2.12
Savings and Investment Ratios
1970-1976 1977-1988 1989-1998
Investment Ratio 16.4 25.1 24.3
National Savings Ratio 12.0 15.8 18.7
Domestic Savings Ratio 13.1 13.9 15.2
Source: Computed from Central Bank of Sri Lanka Annual Reports
It is also argued that the combined efforts of deregulation and innovation would help in the management of public debt by improving the marketability of government securities and increasing the range of instruments that can be issued. Although deregulation has to some degree permitted this, it has not been successful in promoting fiscal discipline in Sri Lanka.
A consequence of financial deregulation has been the decline in importance of monetary aggregates as macroeconomic indicators. Increased capital mobility and currency substitution, greater competition in banking and the adoption of new financial instruments have led to problems of interpreting financial information. A major challenge resulting from financial deregulation has been the movement of capital between countries in response to interest rate differentials impeding the pursuit of an independent monetary policy by the authorities. A related issue concerns the increasing degree of currency substitution, whereby debts and savings in
61 The Financial System of Sri Lanka
local currency in some developing countries are being replaced by debts and savings in foreign currency. This has restricted government control of money supply and undermined the effectiveness of monetary policy. Thus, the freedom to pursue independent policy decisions are becoming severely constrained in such an environment. Despite the potential risks of financial deregulation, there is no persuasive evidence that the overall effectiveness of macro economic policies has deteriorated.
2.6 CONCLUSION
The purpose of this chapter was to review the extent of regulatory reform in the financial sector and to assess the effects of policy reforms. Financial deregulation in
Sri Lanka was expected to enhance efficiency by promoting competition and eliminating constraints on resource allocation. An evaluation of the outcome of financial reforms suggest that overall, Sri Lanka’s financial sector has gained in terms of width (measured by the number of financial instruments and markets), depth
(measured by the volume of deposits) and resilience (ability to bounce back from a price change). Despite the significant development of the financial sector, the process of reform in Sri Lanka is still far from complete. Therefore, as the process of deregulation is likely to continue, the question of whether financial reform has effectively contributed to enhanced efficiency of this sector is one that needs to be addressed. Hence, this study attempts to investigate empirically whether the recent experience of banking deregulation has been associated with an increase in market
62 The Financial System of Sri Lanka
efficiency in Sri Lanka. In view of this, the following chapter goes on to define the concept of efficiency in support of the empirical work that follows.
63 The Efficient Market Hypothesis
CHAPTER 3
THE EFFICIENT MARKET HYPOTHESIS
3.1 INTRODUCTION
An efficient market as defined by Fama (1970) is one in which “prices always fully reflect available information”. An efficient market, therefore, is one that correctly processes and evaluates all available information. Accordingly, this implies that the market adjusts rapidly and accurately to new information so that prices always incorporate available information. The assumption of rational expectations which constitutes an important component of the efficient market hypothesis (EMH) ensures that the market will exploit all available information to take advantage of any perceived profit opportunities. Thus, an efficient market would not permit agents to earn above normal profits.
An important implication associated with the EMH relates to the allocation of scarce capital resources. The allocation of capital resources to their most productive use can only achieved in the presence of an efficient pricing mechanism. The efficient dissemination of information should ensure that capital is allocated to projects that yield the highest expected return with necessary adjustment for risk. Thus pricing in an efficient market has the potential for affecting the allocation of an economy’s savings and investment. It is for this reason that the concept of market efficiency is important.
64 The Efficient Market Hypothesis
In order to place this research in perspective this chapter aims to provide a discussion of the concept of market efficiency, examining conditions under which it might not obtain. The chapter is structured as follows. Section 3.2 examines the derivation of the term efficient and reviews the definitions of efficiency that have been proposed focusing specifically on Fama’s definition. Section 3.3 identifies conditions under which market efficiency may not obtain. Section 3.4 outlines the conclusions.
3.2 DEFINING MARKET EFFICIENCY
The phrase "efficient market" in the context of securities markets was first used by
Fama, Fisher, Jensen and Roll in 1969. An efficient market was defined by them as "a market that adjusts rapidly to new information". The literature on efficient markets prior to Fama, Fisher, Jensen and Roll was closely linked with random walk models.1 The random-walk hypothesis asserts that successive price changes are identically distributed and independent random variables. In an efficient market, the information contained in past prices is fully and instantaneously reflected in current prices. Hence, the opportunity for any abnormal gain on the basis of the information contained in historical prices is eliminated. Market efficiency would then imply that successive price changes are independent. Most of these early studies supported the random-walk behaviour of speculative prices:
Kendall (1953), Roberts (1959), Alexander (1961), Cootner (1964) and Fama
(1965), among many others. It was not until the work of Samuelson (1965) and
1 The first test of the random walk model was by Bachelier in 1900.
65 The Efficient Market Hypothesis
Mandelbrot (1966) that there was a shift in emphasis from describing the statistical properties of security price behaviour to providing an economic explanation for this observed behaviour. None of the above-mentioned studies, however, attempted to express efficiency in terms of an economic model.
3.2.1 Fama's Definition
The first formal representation of the efficient market hypothesis can be attributed to Fama (1970) in his survey of the literature on efficient capital markets. He declared that (1970, p.383) " to a large extent the empirical work in this area preceded the development of the theory. The theory is presented first here in order to more easily judge which of the empirical results are most relevant from the viewpoint of theory." He defined an efficient market as one in which " prices fully reflect available information." Fama presented this in the form of a "fair game" model.2 Using the same notation as Fama,
zj,t+1 = i j,t+1 - E (ij, t+1| φt ) (1)
E(zj, t+1| φ t ) = 0 (2)
ij,t+1 = the realized return on security j in period t+1, where return is defined as the percentage change in security price adjusted for dividends received
E(ij,t+1| φt) = the expected return on security j in period t+1 contingent upon the information set φt.
2 The "fair game" properties of the model are implications of the assumptions that, i) the conditions of the market equilibrium can be stated in terms of expected returns and ii) the information φ is fully utilized by the market in forming equilibrium expected returns and therefore current prices.
66 The Efficient Market Hypothesis
φt = the information set assumed to be fully reflected in prices in period t zj,t+1 = excess return on security j in period t+1
According to Fama, the expected value of the excess return is zero in an efficient market.
Sufficient conditions for market efficiency as given by Fama (1970, p.387) are: no transaction costs in trading securities; all information is costlessly available to all market participants; and all agree on the implications of current information for the current price and distributions of future prices of each security.
These conditions, however, were only sufficient conditions for market efficiency but not necessary. He pointed out that transaction costs would not lead to market inefficiency as long as all available information was embodied in transactions.
Prices would then fully reflect available information. Similarly, the market would be efficient if sufficient numbers of investors had ready access to available information. Likewise, heterogeneous expectations did not imply market inefficiency unless there were investors who could consistently outguess the market. That is, information did not alter the expected return, E(ij,t+1| φt ) = E(ij,t+1).
In equilibrium, the expected return conditional on information φt was equal to the unconditional expected return.
"Fair game" models in the theory of efficient markets was first identified by Mandelbrot (1966) and Samuelson (1965).
67 The Efficient Market Hypothesis
Fama further identified three types of market efficiency namely, weak, semi-strong and strong.3 Weak form market efficiency takes place when all information embodied in past prices is reflected in current prices. Tests of weak form efficiency are associated with the random walk behaviour of prices. The theory of random walks in prices constitutes the hypotheses that successive price changes are independent, and that successive prices changes are identically distributed.
The random walk hypothesis can be formally represented by:
f(ij,t+1| φt) = f(ij,t+1) (3)
The above equation implies that the conditional marginal probability distribution of an independent random variable are identical, and the density function f(.), must be the same for all t.
When prices reflect all publicly available information markets are said to be efficient in the semi-strong form. Publicly available information is so large and diverse that tests of semi-strong efficiency examine the response of the market to different types of public information. These tests involve the reaction to stock splits, dividend, merger announcements, inflation etc. For example, if investors earn abnormal returns following a stock split announcement the market would not be efficient in the semi-strong form. Strong form market efficiency occurs when prices reflect all information, public and private. Problems when testing for strong form efficiency arise due to the unobservable nature of private information. Tests of this form involve examining the profits made by insiders and the performance of
3 This was first suggested by Harry Roberts (1959).
68 The Efficient Market Hypothesis
professionally managed institutional portfolios.4 If insiders make abnormal profits the market is not efficient in the strong form.
3.2.2 Further Formalization of the Concept of Efficiency
Fama's (1970) definition of efficient markets subsequently led to extensive debate, which has continued to date. Rubinstein (1975), Le Roy (1976), Grossman (1976),
Jensen (1978), Grossman and Stiglitz (1980), Beaver (1981), Latham (1986), among several others, contributed to this debate. Criticism was leveled at it on the grounds that the terms “fully reflect” and “information available” were ambiguous. Moreover, the fair game model as given by equation (1) was also criticized for expressing efficiency entirely in terms of expected values of the return distribution. As expected prices are unobservable, market efficiency was not empirically testable on the basis of this definition.
In response to this criticism, Fama (1976) revised his definition of efficiency by defining a density function of future prices contingent upon the information set available at time t.
Let
φt = the information set available at time t relevant to the determination of the price of the security at t
m φ t = the subset of φt which is used by participants in the capital market
4 Defined by the Securities Exchange Act of 1934 as directors, managers and owners of 10% or more of any equity class of securities of that firm.
69 The Efficient Market Hypothesis
m fm( i1t,. . . int | φ t ) = the joint probability density function for security prices at
m time t conditional on φ t f (i1t,. . . int | φt ) = the true joint probability density function for security prices implied by φt then:
m φ t = φt (4) the market uses all available information.
m fm( i1t,. . . int | φ t ) = f(i1t,. . . int |φt ) (5)
Equation (5) indicates that the market understands the implication of the available information for the joint distribution of returns. The joint probability density
m function for security prices at time t conditional on φ t is then compared with the true joint probability density function for security prices implied by φt. Any differences between the two is evidence of market inefficiency. According to
Fama, equations (4) and (5) imply that the market is aware of all available information and uses it correctly.
Fama recognized the need for an expected returns model to establish the link between present prices and the density function of future prices, in order to make this definition empirically testable. Therefore, empirical tests of market efficiency became joint tests of the expected returns model and of market efficiency.
70 The Efficient Market Hypothesis
3.2.3 Alternative Definitions of Market Efficiency
A number of alternative definitions on efficiency were subsequently put forward in an attempt to resolve the ambiguities in Fama’s definition. These definitions attempted to tighten the concept of efficiency as put forward by Fama. These extensions (see Latham 1986, Rubinstein 1975) demonstrated that the equilibrium being contemplated in efficiency should explain asset prices and asset allocations as well.
Accordingly, Rubinstein (1975) saw the failure to incorporate portfolio choice as a major shortcoming of Fama’s definition. Rubinstein (1975, p.812) stated:
In a perfect and competitive economy composed of rational individuals with homogenous beliefs about future prices, by any meaningful definition present security prices must fully reflect all available information about future prices.
Rubinstein attempted to address this problem by incorporating heterogenous beliefs in a model of efficiency.
Beaver (1981, p.28) noted that the term “information set” in Fama’s definition was unclear. He defined market efficiency in terms of an information distribution.
a securities market is efficient with respect to signal φt if and only if t he configuration of security prices {ijt} is the same as it would be in an otherwise identical economy (i.e with an identical configuration of preferences and
endowments) except that every individual receives φt as well as φjt.
Beaver further extended this definition to the information system which generated these signals. He termed this information system efficiency as distinct from signal efficiency. Although this eliminated the ambiguity of the term information set,
71 The Efficient Market Hypothesis
Beaver pointed out that time lags in information dissemination could lead to market inefficiencies under this definition.
The non-incorporation of transaction costs was further seen as a limitation of
Fama’s definition. A definition which attempted to resolve this was that of Jensen
(1978,p.96): “a market is efficient with respect to information set φt if it is impossible to make economic profits by trading on the basis of information set φt.”
Economic profit here refers to the return net of all costs including storage and transport costs.
This definition was criticized by Ball (1988), who pointed out that it failed to address the main issue of how the market behaves. This definition implied that the efficiency of the market was an increasing function of regulatory barriers to the imposition of trading rules, which diverged from the efficiency argument.
Latham (1986) highlighting the shortcomings in the definitions put forward by
Rubinstein (1975), Fama (1976), Jensen (1978) and Beaver (1981) proposed a new definition of the market efficiency hypothesis. He criticized Fama's and Jensen's definitions on the basis of their inability to distinguish abnormal returns from normal returns to risk bearing. He pointed out that Beaver's definition was unclear in that it did not contain the "subset property" an essential property of informational efficiency. The property was that efficiency with respect to an information subset inevitably indicated efficiency with respect to any subset of that
72 The Efficient Market Hypothesis
information set. That is, if a market was efficient in the semi strong form it was efficient in the weak form. Thus, according to him, "security markets are considered efficient with respect to information set φt, if and only if revealing φt to all agents would change neither equilibrium prices nor portfolios." Therefore, this definition included the subset property. Efficiency with regard to φt denoted efficiency with regard to any subset of φ.
These alternative definitions led to a distinction between two schools of thought.
The Chicago or empirical school largely associated with Fama, and the Berkeley or information economics school associated with Rubinstein, Sharpe, Beaver and
Latham. While the information economics group begins with the individual and then model the aggregate in terms of a group of individuals and their response to prices, Fama employs the market as using the available information and setting prices.
Despite the numerous criticisms leveled at the efficient market hypothesis as presented by Fama, it still remains the most widely used concept of efficiency as basis for empirical work in finance. Ball (1988) criticized the information approach on the grounds that tighter definitions of the concept of efficiency were not necessary for the interpretation of evidence. He pointed out that these models did not appear to identify the difference in properties of information from the properties of markets. Therefore, Fama's (1976) definition of efficiency, which is
73 The Efficient Market Hypothesis
the generally used concept in empirical work, is the preferred concept of efficiency for the purpose of this study.
3.3 FAILURE OF MARKET EFFICIENCY
The previous section examined alternative definitions of efficiency. The work by
Kihlstrom and Mirman (1975), Grossman (1976), Grossman and Stiglitz (1980) and Jordan (1983), among others, directed attention to conditions under which market efficiency may not be achieved. There are several conditions under which market efficiency may not attain, among them, the monopolistic access to information by insiders, the incomplete dissemination of information, information costs and the thinness of markets.
The studies of Finnerty (1976), Seyhun (1986), Ippolito (1989) showed that monopolistic access to information by insiders led to the generation of monopoly profits and hence market inefficiency. However, Kihlstrom and Mirman (1975) and Jordan (1983) established conditions under which price expectations could converge. Given Baynesian price expectations, Kihlstrom and Mirman (1975) showed that inside information was accurately disseminated to outsiders leading to convergence in price expectations.
Grossman (1976) and Grossman and Stiglitz (1980) pointed out that if prices reflected too much information, the incentive to produce too much information was lost. Grossman (1976) showed that every individual was uninformed and the
74 The Efficient Market Hypothesis
equilibrium price reflected all the information in the market. Therefore, an informationally efficient price system summed up diverse information perfectly but in the process of doing so, eliminated the private incentive for collecting information. Thus, Grossman showed that the price system could be maintained only if there was incomplete dissemination of information. If this occurred, some traders want to know the reason for price movements, but then, observing prices only, would not be sufficient, as prices would not fully reflect all information. The
Grossman and Stiglitz (1980) model consisted of two types of investors, informed and uninformed. The price system transmitted information from the informed to the uninformed. The price system, however, did not fully reflect all information in equilibrium, for if it did, the incentive for individuals to acquire information would not arise. While optimal behaviour was perceived as a necessary condition for ensuring that prices reflected available information, it was not sufficient, according to Grossman and Stiglitz.
The lack of efficiency in a number of smaller markets is attributed to the thinness
(low volume of trading) of these markets. Gandhi, Saunders and Woodward
(1980) in a study of the Kuwaiti stock market, Papaioannou (1984) the Athens
Stock Exchange and Elyasiani, Perera and Puri (1996) the Colombo Stock
Exchange attributed the lack of efficiency to the thinness of the markets.
Jennergren and Korsvold (1975) further pointed out that information dissemination in a thin market might be slow in comparison to that of larger markets.
Inefficiency could also result from the high degree of government intervention in
75 The Efficient Market Hypothesis
these markets. Ayadi and Pyun (1994) applying the Lo-MacKinlay variance ratio test to the Korean Stock Exchange, found that evidence rejected the random-walk hypothesis when the disturbance terms were assumed to be homoscedastic.
However, when heteroscedasticity was assumed, evidence supported the random- walk hypothesis. Evidence of heteroscedasticity in the market was attributed to government intervention in the stock market.
To summarize the above, market efficiency may not obtain due to a number of reasons. One is the incomplete dissemination of information. If some investors do not possess the information known to others, prices may not be efficient. Another cause of prices failing to reflect information is the inability to estimate accurately the implications of new information. For instance, the implications of sudden shocks on stock prices is difficult to assess as they are unanticipated. Another factor giving rise to price distortions is the existence of information costs. As pointed out by Grossman and Stiglitz, the cost of gathering information could give rise to abnormal returns and hence market inefficiency. The monopolistic access to information by insiders is another potential cause of market inefficiency. One of the main causes of market inefficiency in many developing countries is the monopolistic access to information by the Government. When the Government and the public have different sets of information, each side is inclined to use the information it holds exclusively to its advantage at the expense of the other.
Moreover, the security markets of many developing countries are thin impeding effective information flows. Although the degree to which these imperfections
76 The Efficient Market Hypothesis
lead to market inefficiency is unknown, they have the potential for distorting prices.
If some or all of these factors are operative, it is reasonable to suspect that they might be reflected in the dependence of prices. The dependence of price changes would be suggestive of the existence of opportunities for abnormal gain. To the extent that the markets pricing mechanism is inefficient, the resulting allocation of savings and investment would also be inefficient. It is possible to conclude therefore that the incomplete dissemination of information could lead to inefficiencies in the allocation of financial resources and the creation of monopoly profits resulting in loss of efficiency in production.
3.4 CONCLUSION
The purpose of this chapter was to examine the concept of efficiency, focusing specifically on Fama’s definition. Despite the shortcomings in Fama’s definition, it still remains the most widely used definition of efficiency as basis for empirical work. Therefore, the following chapters carry out some standard tests for financial market efficiency on data for Sri Lanka employing Fama’s definition as the benchmark for efficiency.
77 The Expectations Hypothesis of the Term Structure of Interest Rates
CHAPTER 4
THE EXPECTATIONS HYPOTHESIS OF THE TERM
STRUCTURE OF INTEREST RATES
4.1 INTRODUCTION
This chapter seeks to examine the implications of market efficiency for the behaviour of interest rates in Sri Lanka. In fulfilling this objective the chapter, reviews the literature on the expectations hypothesis of the term structure focusing on the empirical results and methodological advances; and examines the empirical validity of the expectations hypothesis of the term structure for Sri Lanka and its implications.
The expectations hypothesis postulates a formal relationship between long and short-term interest rates. Specifically, the theory leads to the prediction that the long rate is an average of current and expected future short rates. This theory is important for a number of reasons. First, it is related to the concept of market efficiency. Two implications of the efficient market hypothesis are that, the forward rate is an unbiased predictor of future spot rates and that this predictor cannot be improved by using any currently available information. Further, there has been considerable empirical evidence to support the view that the term structure is a useful predictor of future economic variables: the term structure has significant power in forecasting the future direction of short-term rates, Fama and Bliss
(1987), Hardouvelis (1988); inflation, Mishkin (1990), Fama (1990) and real
78 The Expectations Hypothesis of the Term Structure of Interest Rates
economic activity, Estrella and Hardouvelis (1991), Hu (1993), Harvey (1988).
Second, this hypothesis is important in explaining the transmission mechanism of monetary policy. The expectations theory of the term structure implies that monetary policy affects long-term rates by influencing short rates and altering market expectations of future short rates.
M uch of the empirical work on the expectations hypothesis of the term structure is concerned with the markets of the developed countries. The evidence relating to smaller and less developed markets is considerably thinner. Studies have been carried out for Taiwan by Lin (1999); Israel, Zaken (1998); Mexico, Gonzalez,
Spencer and Walz (1999); and Spain, Perez-Rodriguez, Saez and Murillo (1997).
No attempt has been made to date to examine the expectations hypothesis of the term structure for Sri Lanka. Therefore, this chapter aims to test the expectations hypothesis of the term structure for Sri Lanka in the context of the Treasury bill market.
The chapter is organized as follows. Section 4.2 examines interest rate deregulation in Sri Lanka. Section 4.3 outlines the expectations theory of the term structure. Section 4.4 presents the hypotheses to be investigated. Section 4.5 surveys the previous literature. Section 4.6 describes the data. The study is carried out in the context of the Treasury bill market, which exhibited rapid growth during the post-deregulation era. It covers the period 1990.1-1998.12. Section 4.7
79 The Expectations Hypothesis of the Term Structure of Interest Rates
evaluates the empirical results. Section 4.8 examines the implications of the findings and Section 4.9 summarizes the main conclusions.
4.2 INTEREST RATE DEREGULATION IN SRI LANKA
Interest rate policy, formed the core of Sri Lanka's financial deregulation programme as initiated in 1977. The entire interest rate structure was revised upward in an attempt to ensure positive real interest rates which reflected market conditions. The movement towards a more flexible interest rate policy however, was continuously impeded by the lack of a coordinated fiscal policy. The lack of fiscal restraint during the post reform period was a prime deterrent towards achieving price stability. High inflation ranging from 10.8% to 26.1% during the
1978-1984 period, did not contribute towards the maintainence of this interest rate policy. The Central Bank consistently intervened in the determination of interest rates for purposes of keeping the cost of government borrowing down. As a result, deposit rates and Treasury bill rates were not permitted to respond to market forces to the degree required, implying that the money market was not hitherto fully competitive. On the lending front, rates remained administratively determined because long-term lending institutions were, and still remain either solely or jointly owned by the Government of Sri Lanka.
Despite the imperfections of Sri Lanka's deregulation policy it led to the development of the money market. The Treasury bill market developed in size as the instruments traded increased in number. Moreover, the introduction of a
80 The Expectations Hypothesis of the Term Structure of Interest Rates
secondary Treasury bill market in 1981 and the creation of a repurchase market led to the rapid expansion of this market. With the expansion of the Treasury bill market, the Treasury bill rate began to be used as a benchmark rate for other short- term rates in the money market. The growth in short-term money market activity permitted interest rates to move more freely. Hence a notable feature of Sri Lanka's interest rate policy since 1987, has been it's market oriented nature with only minor interventions by the Government.
By 1990, interest rates at the short end of the maturity spectrum were effectively determined by market forces. Call money rates, Treasury bill, Treasury bond, all commercial bank lending and deposit rates, including the Average Weighted Prime
Lending Rate (AWPR) and the Average Weighted Deposit Rate (AWDR) are currently market-determined. Interest rates at the long end of the maturity spectrum are still managed by the Government. Nevertheless, they have been adjusted more flexibly in line with market trends. These include, the Government security (Rupee Loan) rates, the lending and deposit rates of the National Savings
Bank and the lending rates of long-term credit institutions, viz., the State Mortgage and Investment Bank, the Development Finance Credit Corporation, the National
Housing Authority and the National Development Bank.
Despite the fact that interest rates at the short end of the maturity spectrum have been liberalized, a number of problems arise when attempting to use these rates to
81 The Expectations Hypothesis of the Term Structure of Interest Rates
carry out tests of market efficiency. Fixed deposit rates lack flexibility (see Figure
4.1).
Figure 4.1
Note: Mid points of interest rate bands have been used
Source: Central Bank of Sri Lanka Annual Reports and Monthly Bulletins
While call rates are more variable (see Figure 4.2), they are only available for a 24- hour maturity. Weekly, fortnightly and monthly call rates have been computed and published but these are weighted averages of the daily rates. The same problem arises with respect to prime lending rates. The monthly figures that are reported are weighted averages of the estimated weekly rates. Finally, the two-year Treasury
82 The Expectations Hypothesis of the Term Structure of Interest Rates
bond is a relatively new instrument introduced in 1997 and is still gaining popularity. Thus, the only instrument with regularly spaced maturities and flexible rates are Treasury bills. M arket efficiency is therefore, investigated in the context of the Treasury bill market.1 Monthly Treasury bill rates with maturities of three and six months spanning the period 1990 to 1998 are employed. Figure 4.3 displays the movement of Treasury bill rates for the period under study.
Figure 4.2
Note: Mid points of interest rate bands have been used Source: Central Bank of Sri Lanka Annual Reports
1 This market is employed in the work of Roll (1970), Fama (1976), Mishkin(1982), Jones and Roley (1983), among others.
83 The Expectations Hypothesis of the Term Structure of Interest Rates
Figure 4.3
Source: Central Bank of Sri Lanka
4.3 THE EXPECTATIONS THEORY OF THE TERM STRUCTURE OF
INTERES T RATES
The term structure of interest rates is the functional relationship between yield to maturity and term to maturity. The term structure relationship can be expressed in the form of a yield curve with years to maturity plotted on the horizontal axis and yield to maturity on the vertical axis. This curve reflects the expected future course of interest rates. The expectations theory of the term structure was one of the earliest of the theories advanced to explain the shape of the yield curve and reason
84 The Expectations Hypothesis of the Term Structure of Interest Rates
for change. The idea that long-term rates are an average of future short-term rates dates back to Irving Fisher (1930) and was further developed by Hicks (1939) and
Lutz (1940). The expectations theory of the term structure of interest rates is based on the following assumptions. One, profit maximization; two, risk neutrality; three, accurate forecasting of future short-term rates; four, homogenity of all bonds in all respects except in the term to maturity; and five, no transactions costs or taxes.
As investors seek to maximize their expected return over the holding period, they will drive the long-term rates, to the level of the future short-term rates. Therefore, in equilibrium, the current long-term interest rate should equal the market expectation of the average level of current and future short-term rates. It follows from this that the spread between current long and short rates should predict in the correct direction of future changes in short-term rates.
According to this hypothesis, if the short-term rates are currently low but expected to rise in the future, the yield curve will be upward sloping. Similarly, if they are expected to fall in the future, the curve will be downward sloping. If short-term rates are expected to remain constant, the yield curve will be horizontal. If investors expect short-term rates to first rise, and then fall to much lower levels, the yield curve will be humped. The pure expectations theory has strong implications for policy, it implies that the monetary authorities cannot affect the term structure without influencing expectations.
85 The Expectations Hypothesis of the Term Structure of Interest Rates
4.4 THE MODEL
The expectations hypothesis as presented below is based on Tease (1988). In a two-period context, the yield on a two-period security is expressed by,
It = µ + λ it + (1-λ ) E tit+1 (1)
where
It = per period yield on a two period security in period t it = the yield on a one period security in period t
λ = 0.5
µ = constant risk premium which is equal to zero under the pure expectations hypothesis
Et = expectations operator conditional upon information at time t.
Equation (1) states that the return on a two period security equals half the sum of the current one period security return and the expected one period security return in period t+1 plus a constant risk premium.
Assuming rational expectations, the expected one period yield can be expressed as
i t+1 = E tit+1 + vt+1 (2)
where vt+1 is white noise orthogonal to information available at time t
Substituting (2) into (1) yields,
It = µ + λit + (1-λ ) it+1 + εt+1 (3)
86 The Expectations Hypothesis of the Term Structure of Interest Rates
where εt+1 = -(1-λ)vt+1
Incorporating rational expectations into equation (1) leads to a number of testable implications of the expectations hypothesis. Dividing equation (1) by (1-λ) gives,
-1 (1-λ) It - E t it+1 = a + bit (4)
where a = µ/(1-λ) and b = λ/(1-λ). Assuming (2), equation (4) can now be estimated as,
-1 (1-λ) It - it+1 = a + bit - vt+1 (5)
Equation (5) maintains that the expected one period return on a two-period security equals the current one period security rate plus a risk premium. Under the null hypothesis b should be unity and vt+1 should be uncorrelated with information in period t.
A further implication of (3) is provided by the expected spread. If the current long rate exceeds the current short rate, short rates are expected to rise above the current long rate. Dividing both sides of equation (3) by 1-λ and rearranging gives a testable form of this hypothesis,
it+1 - It = a + b(It -it ) +vt+1 (6) where a = -µ/(1-λ) and b = λ/(1-λ).
The null hypothesis tests for b=1.
87 The Expectations Hypothesis of the Term Structure of Interest Rates
The empirical validity of the expectations hypothesis implies that forward rates should be optimal forecasts of future spot rates.2 That is, estimation of the following equation,
it+1 = a0 + bft+1 + εt +1 (7)
should yield parameter values of a0=0 and b=1, where ft+1 represents the expected or forward rate at time t+1 (forward rate is used synonymously with expected rate), and it+1 denotes the spot rate at time t+1. In the pure expectations theory, the entire variation in the forward spot differential is due to the expected change in the spot rate. There are no term premia. Therefore, under the pure expectations hypothesis the evaluation of equation (7) should yield estimates of a0=0 and b=1. If b=1 there is perfect forecasting of the future spot rate. A positive a0 implies the existence of a term premium. It should be noted that as forward rates cannot be observed at the beginning of the investment period, they are inferred from two spot rates. The three-month forward rate for instance, is calculated by using the six-month and three-month spot rates.
2 In a two period context, the following relationship exists between It, it and ft+1 for securities of over one year; 2 (1+It) = (1+it)(1+ ft+1) The implicit forward rate (expected rate) is hence given by, 2 (1+It) - 1 = ft+1
(1+it) Note well that for securities of less than one year an adjustment needs to be made to the annualised yields (see footnote 5).
88 The Expectations Hypothesis of the Term Structure of Interest Rates
The Predictive Content of the Spread
An implication of the expectations hypothesis of the term structure is to see if the spread between the long and short rates accurately reflect changes in the short rate.3
Following Mankiw and Miron(1986) in their analysis of the predictive power of the spread, the relation between long and short rates is expressed by,
It = µ + ½ (it + Eti t+1) (8)
Equation (8) can be rewritten as,
Eti t+1 – it = -2µ + 2 (It –it) (9)
The theory relates the expected change in the short rate (Eti t+1 – it) to the slope of the yield curve (It –it). A test of the predictive content of the spread involves testing the rationality of this forecast. Formally,
i t+1 = Eti t+1 + vt+1 (10)
the realized future short rate is expressed as the sum of the expectation and a forecast error, where vt+1 is orthogonal to information available at time t. Equation
(9) is now expressed as,
it+1 – it = α + β(It – it) + vt+1 (11)
∆i t+1 = α + β St + vt+1 (12) where ∆i t+1 = it+1 – it
St = It – it (the spread)
Mankiw and Miron show that α = -2µ and β = 2.
3 See Mankiw (1986) and Mankiw and Miron (1986) for tests of the forecasting ability of the spread of expected short-term rates.
89 The Expectations Hypothesis of the Term Structure of Interest Rates
4.5 REVIEW OF THE EMPIRICAL LITERATURE
Early studies of the expectations hypothesis consisted largely of testing the forecasting power of forward rates.4 Evidence over the 1930s to 1950s period pointed to the rejection of the expectations hypothesis on the basis of the inaccuracy of the forecasts, which led Meiselman (1962) to suggest that the focus should be on the specification of a proper expectations formation mechanism rather than on the accuracy of the predictions. This consequently led to the development of many expectations formation models. The incorporation of rational expectations into the term structure literature was a significant advancement in this area.
Debate also arose, with respect to the existence of term premia in term structure models. By the early 1970s, a general consensus was reached on the existence of term premia. Then the question arose as to whether it varied over time or not. The hypothesis that term premia are zero for all maturities is attributed to Fisher (1930) and Lutz (1940). The original theory of the term structure according to which term premia are zero is now termed the pure expectations theory. A number of versions of the expectations theory of the term structure have been put forward since. Much of the literature on the expectations theory of the term structure deal with less restrictive versions which assume the existence of time varying term premia. Over the years, the hypothesis tested and methodology employed have changed considerably. Some of the major developments are examined in this section.
4 See Macaulay (1938), Hickman (1942) and Culbertson (1957).
90 The Expectations Hypothesis of the Term Structure of Interest Rates
4.5.1 Early Empirical Studies
Early work on the expectations hypothesis revolved around examining the forecasting power of forward rates. Perhaps the earliest empirical study of significance on the impact of interest rate expectations was that of M acaulay
(1938). Using call money and 90 day loan rates, Macaulay observed that the time money seasonal showed signs of predicting the call money seasonal. This constituted evidence in favour of current spot rates predicting in the future direction of forward rates. Similarly, Hickman (1942) comparing forward rates with realized short-term rates for the 1935-1942 period, found that current spot rates yielded better predictions than did forward rates. Culbertson (1957) in a regression of the long-term rate on the spot rate employing three month Treasury bonds and one week Treasury bills found that short rates were in general lower and more volatile than long rates despite the fact that they moved in the same direction. Culbertson’s study established that the substitutability between assets of different maturities was limited, articulating the market segmentation hypothesis.
4.5.2 Models of Expectation Formation
Meiselman’s Error Learning Hypothesis
In an attempt to resolve the apparent contradiction between theory and empirical evidence, Meiselman (1962) presented an important specification of the manner in which expectations of forward rates were formed in the expectations model of the term structure. M eisleman in his error learning hypothesis postulated that expectations were revised according to the size of the error between the forecast of
91 The Expectations Hypothesis of the Term Structure of Interest Rates
the spot rate given by the forward interest rate and the actual spot rate. Employing annual data on corporate bonds for the period 1901–1954 he found that the correlation coefficients between changes in the forward rates and the forecasting error ranged from 0.95 to 0.59. The coefficients on the forecasting error were positive ranging from 0.70 to 0.21 and observed to decline with maturity. This was interpreted by M eiselman as evidence in support of the expectations hypothesis.
Subsequently, a number of tests were carried out on the error-learning hypothesis.
Results, however, proved to be inconclusive due to problems encountered in obtaining accurate data.
The Modigliani-Sutch Preferred Habitat Theory
As in the Meiselman model, the specification of an expectations-forming mechanism formed the core of the Modigliani and Sutch (1966) model. This theory assumed that agents had preferred habitats, but were willing to switch in response to interest rate differentials if compensated by sufficiently high term premia. The hypothesis implied that term premia if present, would be a function of the term, but not necessarily monotonic. The yield differential between long and short rates was represented as a function of a moving average of past short rates weighted according to a lag structure. Estimating this model with quarterly data for two periods running from 1952.1–1961.4 and 1952.1–1966.1, Modigliani and
Sutch found the regression coefficients to be of the right size and sign providing evidence in support of the preferred habitat theory.
92 The Expectations Hypothesis of the Term Structure of Interest Rates
Rational Expectations, Efficient Markets and the Expectations Hypothesis
The 1970s saw the integration of the theories of rational expectations and efficient markets in the term structure literature. Two implications of the efficient market hypothesis are: the forward rate is an unbiased predictor of the future spot rate and that this predictor cannot be improved by using currently available information.
The efficient market hypothesis assumes that agents are rational in the sense of
Muth (1961). Thus tests of the efficient market hypothesis are conditional upon the satisfaction of the assumption of rational expectations. This was in direct contrast to the expectations forming mechanisms proposed by Meiselman (1962) and
Modigliani and Sutch (1966). While the Meiselman (1962) and Modigliani and
Sutch (1966) models were based on backward looking expectations, the rational expectations approach was based on forward looking expectations. The incorporation of rational expectations implied that, the forecast error for short rates was uncorrelated with any linear combination of information in the current information set.
Roll (1970) was the first to integrate Samuelson's (1965) martingale model with the term structure of interest rates. Formulating a static equilibrium equation for forward rates, Roll showed that the forward rate applicable to a fixed future date, less a liquidity premium followed a martingale sequence. Employing weekly U.S.
Treasury bill rates for the periods, October 1949 through December 1964 and
March 1959 to December 1964, the pure expectations hypothesis was rejected on the basis of means and non-parametric covarience tests. The parametric covarience
93 The Expectations Hypothesis of the Term Structure of Interest Rates
tests however did not reject the theory. Evidence pointed to the existence of positive forward premia that tended to increase with maturity, however, not monotonically.
Subsequently a number of studies were carried out along the lines of the study by
Roll. While the studies of Fama (1976), Jones and Roley (1983), Phillips and
Pippenger (1976,1979), Park (1999) investigated the joint hypothesis of rational expectations and the expectations hypothesis of the term structure within a framework of an efficient markets model, the work of Sargent (1979), Mishkin
(1978), Pesando (1978) carried out explicit tests for rationality and the expectations hypothesis. Fama (1976), examining the predictive power of forward rates in forecasting the future direction of spot rates, concluded that once adjustment was made for variation through time in expected premia, forward rates were unbiased predictors of future spot rates, consistent with the predictions of the efficient market hypothesis. Jones and Roley (1983), on the contrary, found that while the joint hypothesis of rational expectations and the expectations of the term structure could not be rejected in a simple regression of the six-month Treasury bill rate on the three-month Treasury bill rate, the incorporation of other variables, including risk, unemployment, and Treasury bill supplies, foreign holdings of Treasury bills resulted in a fall in the coefficient on the short rate from 0.97 to - 0.63, suggesting that expectations were not rational or that time-varying term premia existed.
Phillips and Pippenger (1976, 1979), in a comparison of the Modigliani-Sutch with the efficient market hypothesis, rejected the Modigliani-Sutch model in favour of
94 The Expectations Hypothesis of the Term Structure of Interest Rates
the efficient market hypothesis. They found that the addition of lagged short-term rates did not contribute significantly to the explanatory power of the model. More recently, Park (1999) investigated the efficiency of the Canadian Treasury bill market in the context of a vector error correction representation of a vector autoregressive model for the 1961.1-1998.8 period. Results indicated that forward and spot rates were cointegrated providing support for the efficient market hypothesis.
Some support for the martingale model was found by Mishkin (1978), Pesando
(1979, 1981) and Sargent (1979) among others. Sargent (1979) in a test of the restrictions implied by the rational expectations hypothesis employing a vector autoregression of long and short rates found that evidence did not permit rejection of the null hypothesis. Using 90 day Treasury bill and long-term bond rates for
Canada for the 1957.1-1979.1 period, Pesando (1981) similarly found evidence in support of the martingale model. In comparing these results with three sets of recorded forecasts, he observed that while survey forecasts of short-term rates were superior to the predictions of the martingale model, their forecasts of long-term rates were inferior. Using annual data on corporate bonds and conditional expectations estimated from AR(2) and IMA(1,1) models, Nelson (1972) discovered that the random walk assumption of interest rates was better approximated by the IMA model.
95 The Expectations Hypothesis of the Term Structure of Interest Rates
Sargent (1972) examining the implications of rational expectations in the expectations hypothesis of the term structure over the 1950-1966 period, utilizing
Treasury bill and government bond rates for the US, found that evidence rejected the martingale property of forward rates. Campbell and Shiller (1987) employing a vector autoregressive framework comprising of two elements, long rates and short rates, and data spanning the period 1959–1983, found evidence to be inconsistent with the rational expectations hypothesis. The model however, appeared to explain the behaviour of long rates in response to short rates and their own lagged values.
4.5.3 The Yield Spread as a Predictor of Future Interest Rate Changes
A number of studies on the term structure of interest rates incorporating rational expectations have involved examining the power of the spread in forecasting the future direction of interest rates. Mankiw (1986), examining the power of the spread in predicting the future direction of long and short rates over the 1961.1–
1984.4 period for the US, Canada, UK and Germany, found that the estimated coefficients on the spread variable for long rates were in the range of –0.01 and –
0.11, predicting in the opposite direction to that hypothesized by theory. Forecasts of changes in short rates nonetheless, were found to be consistent with theory.
Shiller (1979) using a variety of data frequencies and samples for the US and the
UK, similarly found that the coefficients on the spread were negative and significantly below their hypothesized value of unity. He attributed this to excess volatility in long-term rates. Consistent with the findings of M ankiw (1986) and
Shiller (1979), Shiller, Campbell and Schoenholtz (1983) observed that the yield
96 The Expectations Hypothesis of the Term Structure of Interest Rates
spread between three and six-month Treasury bill rates were poor predictors of future changes in interest rates. They concluded that “the simple theory that the slope of the term structure can be used to forecast the direction of future changes in interest rates seem worthless”.
Mankiw and Miron (1986), using data at the short end of the maturity spectrum from 1890 to 1979, observed that the predictive power of the spread was sensitive to the policies of different monetary regimes. They found that prior to the establishment of the Federal Reserve System in 1915, the predictive power of the yield curve was higher than after. They attributed the low power of the spread after the founding of the Federal Reserve to its commitment of stabilizing interest rates.
Fama (1984) obtained more successful results in respect of the predictive power of the yield spread. Using one-to-six-month U.S Treasury bill rates over the 1959-
1982 period, he obtained positive coefficients for the spread coefficient for all maturities except one. Notable was the observation that the forecasting power of the spread appeared to decline with the increase in time horizon.
M ore recently, M acDonald and Speight (1991), examined the predictive power of the spread for four countries including the US, the UK, Canada, Belgium and
Germany, adopting a bivariate autoregressive system. Using data spanning the period 1964.1–1986.4 and Treasury bill and long-term bond rates, they found mixed support for the expectations hypothesis. While results for the UK and the
US appeared to be consistent with theory, it was rejected for the other countries.
97 The Expectations Hypothesis of the Term Structure of Interest Rates
Employing a VAR methodology, Hardouvelis (1994) examined the predictive content of the spread for the G7 countries. Evidence showed that while long rates moved in the direction predicted by thoery for France and Italy, it moved in the opposite direction to that predicted by the expectations hypothesis with the coefficients on the spread ranging from –0.29 to –0.576 for Canada, Japan,
Germany and the UK. This was explained by way of an additive white-noise error on long rates by Hardouvelis. The use of instrumental variables reversed the sign of the slope coefficients, significantly improving the coefficient estimates. For the
US, however, the theory was decisively rejected by the data. The use of instruments did not improve the results.
4.5.4 Deviations from the Expectations Hypothesis
The empirical rejection of the expectations hypothesis has been attributed primarily to time-varying term premia and over- or under-reaction of long rates to short rates.
An extension of the expectations hypothesis allowing for uncertainty was first advanced by Hicks in 1939. This concurs with the importance of both expectations and risk aversion in expectations formation. Initially known as the liquidity preference theory, it is now regarded an extension of the expectations hypothesis.
M ore recently, it has been recognized that term premia vary over time. The existence of time-varying term premia is supported by the findings of Fama (1984),
Mankiw (1986), Campbell (1986), Engle, Lilien and Robins (1987), among others.
98 The Expectations Hypothesis of the Term Structure of Interest Rates
Fama (1984), regressing the premium on changes in the spot rate, observed that there was considerable evidence of time-varying term premia. The slope coefficients in the premium regressions for the overall period were observed to be more than three standard errors from zero. Similarly, M ankiw (1986) noted that the reason for the yield spread predicting in the opposite direction to that hypothesized by theory was the existence of time-varying term premia that was positively correlated with the spread. Employing an ARCH model to estimate time-varying term premia, Engle, Lilien and Robins (1987) concluded that risk premia so calculated explained rejections of the expectations hypothesis. Longstaff (1990) found that the average expectations hypothesis term premia for holding periods of one year were in the range of –8 to –90 basis points for bonds of maturities of 5 to
30 years. These results supported the findings of Fama and Bliss (1987). He, however, concluded that time-varying term premia need not necessarily be interpreted as evidence against the expectations hypothesis; rather, it stressed the need for new approaches of evaluating term structure models.
Another possible source of the rejection of the efficient market hypothesis is the over- or under-reaction of long rates to short rates. Shiller (1981), Shiller,
Campbell and Schoenholtz (1983), Mankiw and Summers (1984) interpreted the yield spread’s ineffectiveness in forecasting the future changes in interest rates as evidence of under-reaction of long rates to current information. Shiller, Campbell and Schoenholtz (1983), found that the three-month forward rate rose by 8.8 basis points immediately following a one billion dollar money stock surprise, while the
99 The Expectations Hypothesis of the Term Structure of Interest Rates
spot rate was anticipated to rise by 26.6 basis points, implying market under- reaction to money announcements.
Currently, the expectations hypothesis is often tested in conjunction with a number of other theories: capital- and asset-pricing models (Cox, Ingersoll and Ross 1981,
1985); and arbitrage-pricing models (Vasicek 1977, Brennan and Schwartz 1979,
Beja 1979). The term structure has also been evaluated as a transmission mechanism for marcroeconomic policy (Blanchard 1981, Turnovsky and Miller
1984, Turnovsky 1989). More recently, theories of the expectations hypothesis of the term structure suggest that there is an equilibrium relationship among interest rates of different maturities. This raises the question of whether the long-run movement of the term structure of interest rates is driven by some common trend.
The rest of this chapter is therefore devoted to examining long-term trends in interest rates using unit root and cointegration techniques.
4.6 DATA
M onthly Treasury bill rates with maturities of 3 and 6 months spanning the period
1990–1998 have been have been utilized for the study. The 3, 6 and 12-month spot
100 The Expectations Hypothesis of the Term Structure of Interest Rates
rates have been used to compute 3 and 6 month forward rates for Treasury bills.5
Figures 4.4 and 4.5 illustrate the movement of the forward and spot rates for the period under study.
Figure 4.4
Three Month Treasury Bill Rate 1990.1 – 1998.12 (Spott and Implicit Forward t-1/t)
Source: Spot rate–Central Bank of Sri Lanka; Forward rate computed (see footnote 5)
5The spot rates for 6 and 3-month Treasury bills are .1728 and .1694 respectively for 1990.1. The implicit forward rate for the 3 month Treasury bill for 1990.1 is computed as follows (see Jones and Roley 1983),
ft = {[1+( 180/365).It]/ [1+(90/365).it]-1}.(365/90)
ft = {[1+( 180/365)*.1728]/ [1+(90/365)*.1694]-1}.(365/90)
ft = [(1.0852)/(1.0418)-1]*4.1
ft = 0.178080
101 The Expectations Hypothesis of the Term Structure of Interest Rates
Figure 4.5
Six Month Treasury Bill Rate 1990.1 – 1998.12 (Spott and Implicit Forwardt-1/t)
Source: Spot rate–Central Bank of Sri Lanka; Forward rate computed (see footnote 5)
The frequency of the data used is monthly and the interest rates are for 3 and 6 month horizons. The use of overlapping observations can result in a moving average error structure of order two and order five for the 3 and 6 month rates respectively, see Hansen and Hodrick (1980) and Pope and Peel (1989). Newey and West (1987) provide a method by which to correct for this problem. Therefore, the Newey and West estimator for autocorrelated disturbances is employed to compute the appropriate estimated covarience matrix and hence standard errors for the estimated OLS parameters, in the presence of a covarience matrix whose structure is in conformity with the nature of the overlapping data employed. See
Table 4.3.
102 The Expectations Hypothesis of the Term Structure of Interest Rates
Unit Root Tests
M any economic time series are non-stationary. The use of standard inference procedures for the estimation of non-stationary time series could lead to spurious results.6 Therefore, it is useful to first examine the time series properties of the variables in question.
The most widely used test for unit roots was first presented by Fuller (1976) and
Dickey and Fuller (1979). These tests are commonly referred to as Dickey-Fuller
(DF) tests. The simplest form of the DF unit root test is based on the estimation of the following data-generating process:
Xt = X t-1 + εt (13) where Xt denotes the variable in question and εt represents white noise with zero mean.
The DF test entails testing of the hypothesis, that the coefficient of Xt- 1 is equal to
1. Therefore, the regression equation,
Xt = ρX t-1 + εt (14) is run for ρ=1 or alternatively,
6 These regressions usually exhibit a high R2 but low Durbin Watson (DW) statistic. Granger and Newbold (1974) point out, the high R2 could indicate correlated trends rather than a good fit, while the low DW statistic could reflect autocorrelated residuals.
103 The Expectations Hypothesis of the Term Structure of Interest Rates
∆Xt = βXt-1 + εt (15) for β=0, where ∆Xt = Xt - Xt-1.
The null hypothesis of a unit root or β=0, is tested against the alternative of β<0 or stationarity of the series. If the data displays a unit root, that is β=1, the series is non-stationary and the standard students t distribution cannot be used. Dickey and
Fuller have compiled τ statistics on the basis of Monte Carlo experiments for this purpose. If the computed τ statistic is less than the critical value in absolute terms, the null hypothesis of a unit root is not rejected.
Once it is established that the series are non-stationary, the next step is to test for the order of integration of the series. A series which is differenced once to achieve stationarity, is integrated of order 1, denoted by I(1). If Xt ~ I(1) then ∆ Xt
~ I(0).
In general, a time series that is differenced d times to achieve stationarity, is said to be integrated of order d. Denoted by I(d). If d=0, the resulting series is stationary or a I(0) series. While a I(0) series has a mean and will often return to the mean value, a I(1) series will rarely return to the mean value. Similarly, the variance of a
I(0) series is constant while the variance of a I(1) series rises indefinitely over time.
2 For example, the random walk given by equation (11) has variance v( Xt) = tσ .
104 The Expectations Hypothesis of the Term Structure of Interest Rates
Equation (15) can be extended to include a constant and time trend. With a drift term it is expressed as:
∆Xt = β0 + β1 Xt-1 + εt (16) where β0 represents a drift parameter, and with a trend term,
∆Xt = β0 + β1Xt-1 + β2t + εt (17)
where β2 denotes a trend parameter.
Note that the DF distribution is not invariant to the inclusion of a drift term and time trend. Equation (18) is estimated both with and without a trend in mean.
Starting with a maximum lag length of twelve for the ADF test, lags were eliminated if they were found to be insignificant on the basis of a t test. If omission resulted in autocorrelation the lags were retained. The trend term is omitted from the first differences of the series as it was shown to be insignificant on the basis of a F test.
105 The Expectations Hypothesis of the Term Structure of Interest Rates
Table 4.1
Unit Root Tests 1990.1 1998.12
Without Time Trend in Model
Variable k LM ADF it (3 months) 0 1.56 -2.04 ft (3 months) 0 1.67 -2.54 it (6 months) 0 1.02 -2.08 ft (6 months) 0 2.86 -2.08
S1t 0 1.30 -8.13***
S2t 2 8.72 -3.93***
∆it (3 months) 0 2.49 -10.31***
∆ft (3 months) 0 2.10 -11.75***
∆it (6 months) 0 1.61 -10.41***
∆ft (6 months) 0 1.54 -11.95***
W i th Ti m e Tr e n d i n Mo de l
Variable k LM ADF it (3 months) 0 3.21 -2.88 ft (3 months) 0 1.79 -3.38* it (6 months) 0 1.67 -2.90 ft (6 months) 0 2.82 -3.17*
S1t 0 1.19 -8.17***
S2t 2 9.81 -5.89***
Note: Significance levels without trend are : 10%, -2.58: 5%, -2.90 and 1%, -3.51 With trend 10%, -3.16; 5%, -3.46; 1%, -4.07 (Davidson and MacKinnon 1993) 2 th A sixth order autoregressive model is used. The χ .05 statistic for 6 order serial correlation in the residuals with 6 degrees of freedom is 12.59 *, **, *** significant at the 10%, 5% and 1% levels respectively.
106 The Expectations Hypothesis of the Term Structure of Interest Rates
The results in Table 4.1 appear to suggest that it and f t are non-stationary in levels.
The null hypothesis of a unit root in the first differences is clearly rejected for both
7 series. Evidence appears to suggest, however, that the spread, St, is I(0). The inclusion of a trend term does not change the order of integration of the data series.
It is reasonable, therefore, to conclude that it and ft are I(1) and St is I(0). The non- stationarity of it and ft provide support for the use of cointegration to test for a long-run relationship between the variables.
4.7 RESULTS
4.7.1 The Expectations Hypothesis
The expectations hypothesis given by equation (7) is tested in this section employing the Johansen (1988) and Johansen and Juselius (1990) procedure.
4.7.1.1 Cointegration
The Johansen (1988) and Johansen and Juselius (1990) procedure is employed to test for a long-run relationship between the variables. This approach is preferred to the Engle-Granger (1987) technique because it has better asymptotic properties than the latter and therefore yield more robust estimates.8 More important perhaps is that this approach permits identification of all cointegrating vectors within a given set of variables. An additional advantage of this technique is that it permits direct hypothesis tests on the variables entering the cointegrating regression.
7 Note that the stationarity of the spread is a necessary condition for the expectations model to hold as the latter implies that the series for long-term and short-term rates should be cointegrated with a cointegrating parameter of –1 (see Campbell and Shiller 1987). 8 See Phillip and Ouliaris (1990).
107 The Expectations Hypothesis of the Term Structure of Interest Rates
Johansen and Juselius propose a maximum likelihood estimation approach for the estimation and evaluation of multiple cointegrated vectors. Johansen and Juselius
(1990) consider the following model:
Let Xt be an nx1 vector of I(1) variables, with a vector autoregressive (VAR) representation of order k,
Xt = Π1 Xt-1 + …. + Πk Xt-k + υ +et (19) t= 1, 2,…….T where υ is an intercept vector and et is a vector of Gaussian error terms.
In first difference form equation (19) takes the following form,
∆Xt = Γk- 1 ∆Xt-k+1 + …. + Π Xt-k + υ +et (20) where
Γi = - ( I - Π1 - …Πi ) , for i= 1, ….. , k-1 and
Π = - ( I - Π1 - ……- Πk)
Π is an nxn matrix whose rank determines the number of cointegrating vectors among the variables in X. If matrix Π is of zero rank, the variables in Xt are integrated of order one or a higher order, implying the absence of a cointegrating relationship between the variables in Xt. If Π is full rank, that is, r=n, the variables in Xt are stationary; and if Π is of reduced rank, 0 Π=αβ' where α and β are nxr matrices, with r the number of cointegrating vectors. 108 The Expectations Hypothesis of the Term Structure of Interest Rates Hence, although Xt itself is not stationary, the linear combination given by β'X is stationary. Johansen and Juselius propose two likelihood ratio tests for the determination of the number of cointegrated vectors. One is the maximal eigenvalue test which evaluates the null hypothesis that there are at most r cointegrating vectors against the alternative of r+1 cointegrating vectors. The maximum eigenvalue statistic is given by, λmax = - T ln (1 - λr+1) (21) where λ r+1,….λn are the n-r smallest squared canocial correlations and T= the number of observations. The second test is based on the trace statistic which tests the null hypothesis of r cointegrating vectors against the alternative of r or more cointegrating vectors. This statistic is given by λ trace = -T Σ ln (1 - λi) (22) In order to apply the Johansen procedure, a lag length must be selected for the VAR. A lag length of one is selected on the basis of the Akaike Information Criterion (AIC).9 9 2 The AIC is computed as: AIC(k) = ln|Σk| + (2 p k)/ n , where Σ is the residual covariance matrix; p, the number of variables in the system; n, the number of observations and k the order of lag in the VAR. 109 The Expectations Hypothesis of the Term Structure of Interest Rates Critical values for both the maximum eigenvalue and trace statistics for 3 and 6 month Treasury bills are reported in Tables 4.2 and 4.3 respectively, where r denotes the number of cointegrating vectors. Table 4.2 Johansen’s Cointegration Test for the Three-Month Treasury Yield A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it ft Intercept Null Alternative Statistic 95% Critical Value r = 0 r = 1 50.0958 15.8700 r <= 1 r = 2 4.1494 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r >= 1 54.2472 20.1800 r <= 1 r >= 2 4.1494 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it ft Intercept 1 - 0.02438 0.012748 -0.17343 (-1.0000) (.52285) (-7.1134) D. Test of Parameter Restrictions LR 95% Critical Value b=1 5.25 3.84 110 The Expectations Hypothesis of the Term Structure of Interest Rates Table 4.3 Johansen’s Cointegration Test for the Six-Month Treasury Yield A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it ft Intercept Null Alternative Statistic 95% Critical Value r = 0 r = 1 37.5158 15.8700 r <= 1 r = 2 3.5134 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r >= 1 41.0291 20.1800 r <= 1 r >= 2 3.5134 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it ft Intercept 1 -0.04537 0.03891 - 0.10844 (-1.0000) (.85762) (-2.3900) D. Test of Parameter Restrictions LR 95% Critical Value b=1 8.59 3.84 Panels A of Tables 4.2 and 4.3 report the maximal eigenvalue test of the null hypothesis that there are at most r cointegrating vectors against the alternative of r + 1 cointegrating vectors. The null hypothesis that there are no cointegrating vectors (r=0) against the alternative of one cointegrating vector (r=1), yields a test 111 The Expectations Hypothesis of the Term Structure of Interest Rates statistic of 50.09 for the three month yield and 37.51 for the 6 month yield, suggesting that there is at least one cointegrating vector. The null hypothesis of r<=1 against r=2 is rejected for both yields, suggesting the existence of an unique cointegrating vector. Panel B reports the trace statistics of the null hypothesis that there are at most r cointegrating vectors against the alternative of more than r cointegrating vectors. The null of r=0 against r>=1 is rejected while the null of r<=1 against r=2 cannot be rejected for both data series, implying again the existence of an unique cointegrating vector. Panels C of tables 4.2 and 4.3 report the estimated cointegrating vector. The coefficients in parenthesis are normalized on it. The coefficients on the forward rate for the 3 and 6 month yields are, respectively, 0.52 and 0.86. Despite the fact that the coefficients are significantly different from zero, the Likelihood Ratio (LR) test statistics for the restriction that the coefficient on the forward rate is unity is rejected for both yields (see panels D). Hence, although the cointegration tests appear to suggest that forward and spot rates are driven by a common trend, the LR statistics reject the hypothesis that forward rates are optimal predictors of future spot rates. 4.7.1.2 An Error Correction Model The fact that the time series are cointegrated provides support for the use of an error correction mechanism (ECM) representation between the variables in order to examine their short-run dynamics. The ECMs for it and ft can be expressed as: 112 The Expectations Hypothesis of the Term Structure of Interest Rates k k (23) ∆it = α1 + ∑ βj ∆it-j + ∑ γj ∆ft-j + φ1 (it-1 - γ ft-1) + ϖ1t j =1 j =1 k k (24) ∆ft = α2 + ∑ βj ∆it-j + ∑ γj ∆ft-j + φ2 (it-1 - γ ft-1) + ϖ2t j =1 j =1 where φ(it-1 - γ ft-1) represents the error correction term and ϖ1t and ϖ2t are white noise. The ECM’s are reported below. Error Correction Models 3-month Treasury bill rate ∆it = - 0.15 - 0.08 ∆it-1 + 0.05 ∆ft-1 - 0.02 ( i-0.523f )t-1 (0.48) (0.15) (0.07) (0.15) 2 2 2 2 χ sc = 8.59 χ ff = 3.89 χ n = 183 χ hs = 0.07 ∆ft = - 0.26 - 0.11 ∆it-1 + 0 .03 ∆ft-1 - 1.3 ( i – 0.523f )t-1 (1.05) (0.33) (0.14) (0.33) 2 2 2 2 χ sc = 8.52 χ ff = 1.46 χ n = 329 χ hs = 0.03 6-month Treasury bill rate ∆it = - 0.07 + 0.10 ∆it-1 - 0.06 ∆ft-1 - 0.33 (i – 0.858f)t-1 (0.23) (0.13) (0.11) (0.13) 2 2 2 2 χ sc = 7.73 χ ff = 0.18 χ n = 184 χ hs = 0.02 ∆ft = - 0.11 - 0.01 ∆it-1 - 0 .14 ∆ft-1 - 0.32 (i - 0.858f )t-1 (0.27) (0.15) (0.13) (0.15) 2 2 2 2 χ sc = 5.49 χ ff = 9.9 χ n = 157 χ hs = 2.80 Standard errors reported in parenthesis 113 The Expectations Hypothesis of the Term Structure of Interest Rates The error correction terms in all except the first equation are statistically significant, with the estimated coefficient on the three-month forward rate suggesting that approximately the entire discrepancy between the forward and spot rate is corrected within a month. The coefficients on the 6 month yield suggest that a third of the discrepancy is corrected in each month. Diagnostic tests have been performed for serial correlation, functional form misspecification, normality of residuals and heteroscedasticity. The LM statistics for 12th order serial correlation in the residuals are to be compared with the 5% critical value of 21.03. In each case, the data support the assumption of serial independence. Ramsey’s (1969) RESET test statistics for functional form misspecification are to be compared with the 5% critical value of 3.84. It is observed that the LM statistics do not permit rejection of the functional misspecification hypothesis. The Jarque-Bera (1980) test for the normality of residuals indicates a non-normal distribution for the disturbance terms in all equations. A large value for this test could be generated by the existence of outliers. The presence of outliers more than two standard errors away from the mean could give rise to a distribution with fatter tails than a normal distribution. All equations, support the assumption of homoscedasticity on the basis of a LM test. 114 The Expectations Hypothesis of the Term Structure of Interest Rates T he exist ence of an error-correct ion model suggest s some Granger causality in the system which, in turn, implies that the error-correction model might be a useful forecasting tool. Figure 4.6 Figure 4.7 115 The Expectations Hypothesis of the Term Structure of Interest Rates Figure 4.8 Figure 4.9 116 The Expectations Hypothesis of the Term Structure of Interest Rates 4.7.1.3 OLS Estimation An examination of the data for a shorter time period (see appendix to this chapter) indicates that the time series are stationary. Shiller and Perron (1985) point out that the power of these tests depends on the span of the data rather than on the sample size. They believe that the data would tend to display a unit root if observed more frequently. Another problem is that asymptotic critical values could be misleading when applied to finite samples. Due to the difficulty therefore in drawing conclusive inferences on the parameter estimates, this section estimates the model using the more conventional OLS and instrumental variable (IV) methods of estimation. Table 4.4 presents results for the OLS regressions of the spot rates on the forward rates for the period 1990.1–1998.12. Table 4.4 The Expectations Hypothesis of the Term S tructure: OLS Estimates it = a0 + bft + εt 2 Instrument a0 b Wald:b=1 Wald:a0=0 Wald:a0=0; b=1 R DW 3 month TB 2.93 0.80 54.14*** 47.33*** 56.80*** 0.90 1.4 (0.42) (0.03) (0.50) (0.03) 6 month TB 4.34 0.69 181.63*** 123.62*** 261.95*** 0.90 0.98 (0.39) (0.02) (0.55) (0.03) Note: Standard errors reported within parenthesis. The second term in parenthesis is the adjusted standard error corrected by the Newey West (1987) procedure. *** significant at the 1% level. 117 The Expectations Hypothesis of the Term Structure of Interest Rates Consistent with the findings of the cointegration tests, the results point to the statistical rejection of the null hypothesis that b=1 at the 1% level of significance for both maturities. The null hypothesis that a0=0 is also rejected for both yields, suggesting the existence of constant term premia. A joint test of the null hypothesis that a0 =0and b=1 is further rejected, pointing to the statistical rejection of the expectations theory of the term structure. The R2 of the regression equations, however, are 0.90, suggesting high explanatory power of the equations. There is some evidence of serial correlation in the residuals for the 6 month Treasury yield on the basis of the DW statistic. It was pointed out in section 4.6 that the use of overlapping data could result in a moving average error structure of orders two and five for the 3 and 6 month yields respectively. M ore specifically, the standard errors of the coefficients estimated by OLS depend upon the assumption that the observations on the dependent variable are homoscedastic and uncorrelated. The overlapping data used here is not in accord with this specification. Therefore, the Newey and West (1987) procedure is employed to compute the appropriate estimated covariance matrix and hence standard errors for the estimated OLS parameters. This adjustment to the standard errors of the OLS coefficients yields the adjusted standard errors reported in parenthesis below the unadjusted standard errors in Table 4.4. Standard diagnostic tests are difficult to perform, given the overlapping nature of the data. Therefore, tests of structural stability are carried out. 118 The Expectations Hypothesis of the Term Structure of Interest Rates Structural Stability The behaviour of the model as the sample is recursively increased can be evaluated using the cumulative sum of recursive residuals (CUSUM ) and the cumulative sum of the recursive residuals squared (CUSUM SQ) tests of parameter stability.10 The CUSUM test enables identification of systematic changes in the regression coefficients, while the CUSUMSQ test enables detection of unanticipated or random changes in the regression coefficients. Figures 4.10 to 4.13 illustrate the CUSUM and CUSUMSQ plots for the 3 and 6 month maturities. Figure 4.10 Three-Month Maturity 10 This was proposed by Brown , Durbin and Evans (1975). 119 The Expectations Hypothesis of the Term Structure of Interest Rates Figure 4.11 Figure 4.12 S ix-Month Maturity 120 The Expectations Hypothesis of the Term Structure of Interest Rates Figure 4.13 Figures 4.10, 4.11 and 4.12 indicate acceptance of the null hypothesis that the regression equation is correctly specified at the 5% level of significance. The CUSUM plot for the six-month yield, however, deviates from the boundaries of the 95% confidence interval for the last two observations, providing evidence of some misspecification or structural break. The CUSUM and CUSUMSQ statistic confidence intervals are based on the underlying assumption of normality. These statistics are therefore sensitive to outliers and non-normality in the disturbance term. Thus structural instability on the basis of this test could result from an outlying observation even if the observation has negligible leverage in the sample. 121 The Expectations Hypothesis of the Term Structure of Interest Rates 4.7.1.4 Instrumental Variable Estimation OLS estimation of models that incorporate expected variables could lead to a simultaneous equation bias: see McCallum (1976), Pesaran (1987a). More specifically, expected and spot rates might be driven by the same underlying forces, so that the expected rate is correlated with the error term. Instrumental variables (IV) can be used to obtain consistent estimates of the forward rate when it is known to be correlated with the error term. Therefore, the model is re-estimated employing instrumental variables to account for errors in variables with it-3, i t-4, ft-3 and ft-4 as instruments for 3 month Treasury bills, and it-6, i t-7, ft- 6 and ft-7 as instruments for 6 month Treasury bills where it represents the spot interest rate and ft the forward interest rate. Table 4.5 The Expectations Hypothesis of the Term S tructure: Instrumental Variable Estimates it = a0 + bft + εt 2 Instrument a0 b Wald:b=1 Wald:a0=0 Wald:a0=0; b=1 GR LM 3 month TB 2.67 0.82 7.83*** 6.64*** 12.99*** 0.48 10.06 (1.04) (0.06) (1.11) (0.06) 6 month TB 6.04 0.59 113.48*** 88.30*** 164.38*** 0.28 28.26 (0.64) (0.03) (2.02) (0.11) Instruments used are : it-3, it-4, ft-3 and ft-4 for three month yields it-6, it-7, ft-6 and ft-7 for six month yields Standard errors reported within parenthesis. The second term in parenthesis is the adjusted standard error corrected by the Newey West (1987) procedure. GR2 denotes the Generalized R2 for IV regressions proposed by Pesaran and Smith (1994). *** significant at the 1% level. 122 The Expectations Hypothesis of the Term Structure of Interest Rates Confirming the previous evidence, the individual restrictions that b=1 and a0=0 are rejected for both maturities. The joint hypothesis that b=1 and a0=0 is also rejected at the 1% level, pointing to the statistical rejection of the expectations hypothesis of the term structure. Hence, overall results reject the hypothesis that the forward rate is an unbiased predictor of the future spot rate consistent with the findings under OLS. Rejection of the expectations hypothesis points to some inefficiency in the Treasury bill market. The existence of a long-run relationship between long and short rates, however, suggest s that long and short rat es are driven by some common fact ors. 4.7.2 The Predictive Content of the Spread 4.7.2.1 OLS Estimation Motivated by the work of Mankiw (1986) and Mankiw and Miron (1986), this section examines the predictive content of the spread (the difference between long and short rates). Equation (12) is now estimated using data over the 1990.1– 1998.12 period. 123 The Expectations Hypothesis of the Term Structure of Interest Rates Table 4.6 The Spread as a Predictor of the Future Change in Short Rate: OLS Estimation ∆i t+1 = α + β St + ϖt+1 Variable α β (yield spread) R2 DW ∆i t (3 months) -0.11 0.27 0.03 2.00 (0.15) (0.29) ∆it (6 months) -0.27 0.36 0.03 2.05 (0.18) (0.22) Standard errors reported within parenthesis The coefficients on the spread variables have the theoretically anticipated sign and thus predict in the correct direction of changes in short rates. However, the coefficient values are significantly different from the theoretically hypothesized value of two. The R2 of 0.03 for the two equations imply that the spread exhibits little predictive power. There is no evidence of serial correlation in the residuals on the basis of the DW statistics. 124 The Expectations Hypothesis of the Term Structure of Interest Rates 4.7.2.2 A VAR Model In order to account for the discrepancy of the estimated slope coefficient from its hypothesized value, a vector auto-regressive (VAR) methodology as employed by Campbell and Shiller (1987) is adopted to estimate the predictive content of the spread. As opposed to the single equation technique employed before, the VAR is a simultaneous equation model in that the variables are considered to be endogenous and jointly determined. It is believed that the spread will have additional explanatory power for predicting the future change in spot rates, given information other than the history of ∆it. A VAR representation of order 2, for ∆ it and St would involve estimating regressions of the form, ∆ it = α1 + ψ1 ∆it-1 + ψ2 ∆it-2 + γ1 St-1 + γ2 St- 2 + v1t (21) St = α2 + φ1 ∆it-1 + φ2 ∆it-2 + δ1 St-1 +δ2 St-2 + v2t (22) An implication of this representation is that the spread must Granger cause changes in short-term rates. Tables 4.7 and 4.8 present summary statistics for the VAR application. 125 The Expectations Hypothesis of the Term Structure of Interest Rates Table 4.7 The Spread as a Predictor of the Future Change in Short Rate: A VAR Application (1) (2) 2 lag VAR ∆it S t ∆it-1 0.02 -0.02 (0.10) (0.04) ∆it-2 0.01 -0.04 (0.10) (0.04) St-1 0.32 0.19 (0.26) (0.10) St-2 -0.13 -0.05 (0.26) (0.10) R2 0.02 0.07 LR Tests of block non-causality: 2 St does not Granger cause ∆it; χ (2) = 1.65(0.44) 2 ∆it does not Granger cause St; χ (2) = 2.00(0.37) Note: The spread here denotes the spread between the six and three month rates and the short rate is the three month Treasury bill rate. Standard errors reported within parenthesis 126 The Expectations Hypothesis of the Term Structure of Interest Rates Table 4.8 The S pread as a Predictor of the Future Change in S hort Rate: A VAR Application (1) (2) 2 lag VAR ∆it S t ∆it-1 0.00 -0.05 (0.10) (0.04) ∆it-2 -0.02 -0.04 (0.10) (0.04) St-1 0.36 0.43 (0.28) (0.10) St-2 -0.01 0.11 (0.28) (0.10) R2 0.03 0.30 LR Tests of block non-causality: 2 St does not Granger cause ∆it; χ (2) = 2.45(0.29) 2 ∆it does not Granger cause St; χ (1) = 2.83(0.24) Note: The spread here denotes the spread between the twelve and six month rates and the short rate is the six month Treasury bill rate. Standard errors reported within parenthesis 127 The Expectations Hypothesis of the Term Structure of Interest Rates The order of the VAR is selected on the basis of the Akaike Information Criterion. Results for the three-month Treasury bill rate indicate that the VAR has little predictive power for explaining the variation in short rates. The R2 values for the ∆it and St equat ions resp ect ively are .02 p er cent and .07 p er cent , suggest ing low explanatory power for changes in short rates. However, the null hypothesis that the spread does not Granger cause changes in short rates cannot be rejected at the .44 per cent level of significance. Results for the six-month Treasury bill rate are similar, with some of the estimated 2 coefficients incorrectly signed. The R for the ∆it equation is .03 per cent, indicating little explanatory power for changes in short rates. The hypothesis that the spread does not Granger cause changes in the short rate cannot be rejected at the .29 per cent level of significance while the hypothesis that changes in the short rate does not Granger cause the spread cannot be rejected at the .24 level of significance. In order to investigate the economic significance of the statistical rejection of the expectations hypothesis, Figures 4.14 and 4.15 plot the theoretical spread generated by the VAR and against the actual spread. Surprisingly, the theoretical spread appears to track the actual spread fairly closely in both figures. 128 The Expectations Hypothesis of the Term Structure of Interest Rates Figure 4.14 Note: St is the difference between the 6 and 3 month yields. Figure 4.15 Note: St is the difference between the 12 and 6 month yields 129 The Expectations Hypothesis of the Term Structure of Interest Rates 4.8 IMPLICATIONS An efficient market, as defined by Fama, is one in which security prices fully incorporate all available information. Although the data examined here appears to support the hypothesis that forward and spot rates are cointegrated, it rejects the hypothesis that forward rates are unbiased predictors of future spot rates. A number of possible reasons could account for this. Although short-term interest rates in Sri Lanka have been effectively determined by market forces since 1990, the Government continues to exert pressure through the auction of Treasury bills, the National Savings Bank and the two state-owned commercial banks. The Central Bank influences the level of reserve money through its purchases of Treasury bills at the weekly auctions and the regulations it establishes for its secondary window where the bills are subsequently resold. Until 1995, the rates on savings deposits offered by the National Savings Bank were subject to government approval, which resulted in placing a floor on all deposit rates. Further, the National Savings Bank was mandated to invest a large proportion of its funds in government securities, thus constraining the efficient functioning of the money market. The National Savings Bank Amendment Act of 1995 gave it the authority to purchase government securities directly at the auctions and greater autonomy in the determination of interest rates. Although commercial banks have replaced the Central Bank as the principal holder of Treasury bills, the Government continues to exert an indirect influence on the 130 The Expectations Hypothesis of the Term Structure of Interest Rates market through the two state-owned banks. On occasion, the Central Bank absorbs the residual quantity of Treasury bills auctioned. The secondary market for Treasury bills is still very thin, probably due to the captive nature of the market. Not only is the speed and accessibility of information flows questionable in such a market; the manipulation of the market by large traders is also greater in the case of such markets. Thin markets are further characterized by greater volatility and therefore larger-term premia due to the greater risk involved. M onetary policy management in Sri Lanka remains confined by the lack of depth and competition in money and securities markets. The lack of market driven securities of over one year until March 1997, had prevented the development of a yield curve of over one year which, in turn, posed problems to issuers and investors who had no reference point for long-term securities. Although a large proportion of the Government's financing needs are met by the issue of long-term rupee securities, these are placed with captive sources mainly the National Savings Bank and the Employees’ Provident Fund at administratively determined rates. This has prevented savings institutions from actively participating in the capital market. In March 1997, however, the Central Bank of Sri Lanka began auctioning Treasury bonds, which are envisaged to eventually replace the rupee securities. Treasury bonds rates now serve as a benchmark yield rate for medium- and long-term securities. 131 The Expectations Hypothesis of the Term Structure of Interest Rates Figure 4.16 3-12 months Treasury bill yields 24-48 months Treasury bond yields Source: Central Bank of Sri Lanka Annual Reports The same reasons could account for the failure of the spread to conform to the expectations hypothesis in its forecast of future short rates. The empirical results show, however, that the spread predicts in the correct direction of future short rates. Similarly, the cointegration tests suggest that the Sri Lankan term structure is driven by a common stochastic trend. It is possible that interest rates are influenced by factors such as inflation and/or the budget deficit. Public sector dependence on the financial system has arisen largely due to fiscal imbalance in Sri Lanka. Therefore, unless a concerted effort is made to reduce the fiscal deficit, the servicing requirements of the Government debt will continue to impede the objectives of interest rate reform. Although low interest rates could temporarily 132 The Expectations Hypothesis of the Term Structure of Interest Rates relieve budgetary pressures by providing cheap finance, it would eventually compress the Government revenue base. 4.9 CONCLUSION This chapter examines the applicability of the expectations hypothesis of the term structure in the context of the Sri Lankan Treasury bill market. While the data supports the hypothesis that forward and spot rates are cointegrated, suggesting a stochastic trend in the structure of interest series, the hypothesis that forward rates are unbiased predictors of future spot rates is rejected. It is, however, important to note that interest rates in Sri Lanka have only recently begun to be market- determined and that a term structure is in the process of developing which could perhaps be in part an explanation for the rejection of the expectations hypothesis. Although the results appear to be quite robust, there are a few points that merit attention. The rejection of the restriction that the coefficient on the forward rate is unity need not necessarily be interpreted as evidence against market efficiency. The pure expectations hypothesis consists of a number of limiting assumptions of which zero term premia is one. A number of empirical studies—Fama (1984), Mankiw and Miron (1986), Hardouvelis (1988)—have shown that time-varying term premia could obliterate the predictive ability of the term structure. Thus, the expectations hypothesis is conditional on the fulfilment of the assumptions on which it is based. Rejection of the hypothesis, therefore, need not necessarily result from market inefficiency, but the use of the wrong economic model. 133 The Expectations Hypothesis of the Term Structure of Interest Rates Appendix The empirical finding of the rejection of the unit root hypothesis for a shorter sample period ending 1996.12 led to an investigation of the data for the 1990.1 to 1996.12 period as well. Table A4.1 reports DF tests for unit roots. Table A4.1 Unit Root Tests 1990.1–1996.12 Without time Trend in Model Variable k LM ADF it (3 months) 0 3.18 -3.01** ft (3 months) 0 1.13 -3.60*** it (6 months) 0 1.14 -3.08** ft (6 months) 0 2.70 -3.04** S1t 0 0.60 -8.30*** S2t 0 2.77 -5.27*** With Time Trend in Model Variable k LM ADF it (3 months) 0 1.54 -3.01 ft (3 months) 0 1.18 -3.64** it (6 months) 0 1.34 -3.11 ft (6 months) 0 2.32 -3.34* S1t 0 0.54 -8.88*** S2t 0 2.77 -6.95*** Note: Significance levels without trend are: 10%, -2.58: 5%, -2.90 and 1%, -3.51 With trend 10%, -3.16; 5%, -3.46; 1%, -4.07 (Davidson and MacKinnon 1993) 2 th The χ .05 statistic for 6 order serial correlation in the residuals with 6 degrees of freedom is 12.59 *, **, *** significant at the 10%, 5% and 1% levels respectively. 134 The Expectations Hypothesis of the Term Structure of Interest Rates The results reported in Table A4.1 for the no trend equations, suggest that the series are stationary at the 5% level of significance. The magnitude of the computed DF statistics for the trend equations indicate that the three month forward rate is a I(0) series at the 10% level while it is marginally below the 10% critical value. The results suggest that it is possible to reject the null hypothesis of a unit root when the trend term is omitted from the regression. Thus, a F test has been carried out in order to evaluate the significance of the trend term. The computed F statistics of zero restrictions on the coefficients of the trend term for six month Treasury rates yield values of 4.85 and 5.20 for it and ft respectively, which are below the 5% critical value of 6.73. The computed F statistics for it and ft for the 3 month Treasury yields are 4.55 and 6.98. The overall F test results are biased towards stationarity around a mean. Since the computed F statistic for ft only marginally exceeds the 5% critical value, it is reasonable to conclude that it is I(0). The spread variables clearly reject the null hypothesis of a unit root. In the light of these results, the rest of this analysis is carried out under the assumption that the time series are stationary. 135 The Expectations Hypothesis of the Term Structure of Interest Rates OLS Estimation Table A4.2 The Expectations Hypothesis of the Term S tructure: OLS Estimates it = a0 + bft + εt 2 Instrument a0 b Wald:b=1 Wald:a 0=0 Wald:a0=0; b=1 R DW 3 month TB 3.08 0.82 12.99*** 9.54*** 28.73*** 0.81 1.8 (0.78) (0.04) (2.03) (0.11) 6 month TB 2.80 0.77 23.48*** 10.25*** 194.7*** 0.77 1.1 (0.87) (0.05) (0.55) (0.03) Note: Standard errors reported within parenthesis. The second term in parenthesis is the adjusted standard error corrected by the Newey West (1987) procedure. *** significant at the 1% level. The results appear to corroborate the previous evidence with the hypothesis that b=1 and a0=0 rejected at the 1% level of significance. Instrumental Variable Estimation The results from IV estimation are summarized in Table A4.3. Table A4.3 The Expectations Hypothesis of the Term S tructure: IV Estimates it = a0 + bft + εt 2 Instrument a0 b Wald:b=1 Wald:a0=0 Wald:a0=0; b=1 GR LM 3 month TB -2.51 1.12 0.31 0.43 9.37*** -0.08 4.41 (3.80) (0.22) (2.21) (0.13) 6 month TB -0.18 0.93 0.04 0.00 128.8*** -0.89 10.61 (5.73) (0.30) (4.78) (0.25) 136 The Expectations Hypothesis of the Term Structure of Interest Rates Table A4.3: The Expectations Hypothesis of the Term Structure: IV Estimates notes (cont) Instruments used are: it-3, it-4, ft-3 and ft-4 for three-month yields it-6, it-7, ft-6 and ft-7 for six-month yields Standard errors reported within parenthesis. The second term in parenthesis is the adjusted standard error corrected by the Newey West (1987) procedure. *** significant at the 1% level GR2 denotes the Generalized R2 for IV regressions proposed by Pesaran and Smith (1994). In contrast to the previous results, the restrictions that b=1 and a0=0 yield values well within the acceptance region at the 5% significance level. The joint test that b=1 and a0=0, however, is rejected for both maturities at the 1% level. Interestingly, non-rejection of the hypothesis is obtained for the shorter sample period. Finally, a test of the yield spread is carried out for the shorter sample period in order to see if the results change significantly from the results for the 1990.1 to 1998.1 period. The results are reported in Table A4.4. 137 The Expectations Hypothesis of the Term Structure of Interest Rates Table A4.4 The Spread as a Predictor of the Future Change in Short Rate: OLS Estimates ∆i t+1 = α + β St + ϖt+1 Variable α β (yield spread) R2 DW ∆i t (3 months) -0.05 - 0.26 0.01 2.00 (0.15) (0.29) ∆it (6 months) -0.28 - 0.40 0.03 2.01 (0.23) (0.26) Standard errors reported within parenthesis In contrast to the findings for the longer sample period, the spread appears to predict in the opposite direction to that hypothesized by theory. The results appear to support the findings of Mankiw(1986) for long rates. The overall results appear to reject the expectations hypothesis of the term structure. Therefore the assumed stationarity or non-stationarity of the data series does not appear to make a significant difference to the general conclusion of the statistical rejection of the hypothesis. 138 The Fisher Effect CHAPTER 5 THE FISHER EFFECT 5.1 INTRODUCTION The high inflation experienced by both developed and less developed countries in the 1970s and 1980s brought the Fisher Effect to the forefront of monetary theory. During this period, it became clear that inflation was an important factor in determining short-term as well as long-term interest rates. A rising rate of inflation not only increases the nominal amount of credit required to finance a given level of real expenditure, but through its impact on expectations influences speculative borrowing as well. Hence, inflationary expectations bear importantly on short-term interest rates. The main objective of this chapter is to examine the relationship between interest rates and inflation for the post-deregulation period. Specifically, the following issues are addressed: the relationship between interest rates and the expected rate of inflation, and whether this relationship is consistent with the concept of market efficiency. The relationship between interest rates and inflation, first put forward by Fisher (1930), postulates that the nominal interest rate in any period is equal to the sum of the real interest rate and the expected rate of inflation. This is termed the Fisher Effect. The Fisher Effect is closely related to the concept of efficiency. 139 The Fisher Effect The semi-strong form of market efficiency states that all publicly available information is reflected in interest rates. Therefore, if the market correctly uses all available information about the rate of change in inflation in setting nominal interest rates, then it can be said to be consistent with efficiency in the semi- strong form. Two types of data can be identified when testing for semi strong market efficiency; semi-strong efficiency tests, using macro data such as inflation, money supply, exchange rates and balance of payments, and semi-strong efficiency tests, using micro data such as company-specific announcements. For example, merger, dividend, stock split announcements. Macro data on inflation is used to test for efficiency in this chapter. To date, only two studies have been undertaken on the Fisher effect for Sri Lanka, see Samarakoon (1996) and Payne and Ewing (1997). While the Samarakoon finds some support for the Fisher effect, Payne and Ewing find strong support for the Fisher effect. This study attempts to examine the robustness of these results by employing a variety of data frequencies, samples and expected inflation proxies. The chapter is structured as follows. Section 5.2 provides an overview of the Sri Lankan experience with inflation. Section 5.3 presents the Fisher Hypothesis. Section 5.4 presents the empirical models being tested. Two approaches, namely 140 The Fisher Effect the adaptive expectations and rational expectations approaches, are employed to examine the impact of price changes on the nominal interest rate. Section 5.5 discusses the previous evidence. Section 5.6 describes the data sources and the time series properties of the individual variables. The study makes use of annual, quarterly and monthly data. Section 5.7 reports results of the Fisher hypothesis and evaluates the empirical results. Section 5.8 presents results of the inverted Fisher hypothesis. Section 5.9 details the implications and Section 5.10 summarizes the main conclusions. 5.2 AN OVERVIEW OF THE S RI LANKAN EXPERIENCE WITH INFLATION In the aftermath of economic deregulation, Sri Lanka experienced a steady rise in the rate of inflation, as measured by the Colombo Consumer Price Index (CCPI). Inflation, which averaged around 2.2% in the 1960s, increased to 12.1% in 1978 and then accelerated to 26.1% by 1980. Although due, in part, to the oil price shock of 1979, other contributory factors included the Government's policy of deficit budgeting, increased demand for credit by the private sector, relaxation of price controls, removal of subsidies and continuing exchange rate depreciation. The demand management policies employed by the Government to contain inflation were successful in bringing down the rate of inflation to a single-digit level from 1985–1987.1 The late 1980s, however, saw an acceleration in the rate of inflation again, which reached a peak of 21.5% in 1990. Exchange 1 These policies included credit ceilings and the issue of Central Bank securities. 141 The Fisher Effect depreciation, the upward revision in the guaranteed price of paddy, and rise in fuel prices due to the Gulf crisis contributed to this increase. From 1991 onward, there was a decline in inflation, and by 1998, the rate of inflation averaged 4%. Figure 5.1 records the rate of inflation, as measured by the rate of change in the Colombo Consumer Price Index. Figure 5.1 Source: Central Bank of Sri Lanka Annual Reports One of the many explanations for the continuing increase in price level is the monetarist proposition of the excessive growth in money supply. A consequence of financial deregulation in Sri Lanka has been the growth in all monetary 142 The Fisher Effect aggregates. Money supply as defined by M2, which rose threefold during the 1965–1977 period, rose 23-fold during the 1978–1996 period. See Figure 5.2. Figure 5.2 Source: Central Bank of Sri Lanka Annual Reports The lack of fiscal restraint has been a prime factor underlying this monetary expansion. The large-scale investment programmes, that were primarily deficit- financed, and the eruption of the ethnic crisis in 1983 led to a progressive increase in the size of the budget deficit. The budget deficit, as a proportion of GDP, which averaged around 8.4% in 1977, averaged 24.5% in 1980 and 9.2% by 1998. Monetarists contend that a prime source of the expansion in monetary base is generated by the financing of budget deficits. Given the narrowness of the non- bank financial sector in Sri Lanka and the lack of alternative financial assets to 143 The Fisher Effect money, the authorities are left with little alternative to borrowing from the banking sector. Recent studies on inflation in Sri Lanka—Nicholas (1990), Nicholas and Yatawara (1991), Weerasekera (1992), Rupananda (1994)—suggest the importance of supply-side factors, primarily import prices and the exchange rate in affecting the general level of prices. A notable structural feature of the Sri Lankan economy has been its heavy reliance on international trade. International trade, which accounted for 55% of the GDP at independence, declined considerably during the ensuing period with attempts at import substitution. With the removal of import and exchange controls during the post-liberalization period, imports began to gain greater significance in affecting prices. By 1987, external trade accounted for 57% of the GDP, the difference between the 1950s and post- liberalization period being that imports had come to account for a larger share of the GDP than exports.2 Figure 5.3 plots the rate of change in the Colombo Consumer Price Index (CCPI) against the import price index. The figure indicates a close relationship between the two indices. 2 See Nicholas (1990). 144 The Fisher Effect Figure 5.3 Source: Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF The adoption of a floating exchange rate system, together with the liberalized trade and payment policies, added further pressure on prices. In November 1977, the Sri Lankan rupee was devalued and allowed to float against a basket of currencies. Over the years, the rupee has progressively depreciated against the major currencies. See Figure 5.4. With imports beginning to gain greater importance in price determination, the impact of exchange-rate movements on the rate of inflation has also come to acquire greater significance and is therefore included as a likely explanatory variable in the empirical work that follows. 145 The Fisher Effect Figure 5.4 Source: Economic Progress of Independent Sri Lanka, Central Bank of Sri Lanka; Annual Report Central Bank of Sri Lanka 1998 Hence, the model employed to forecast a price level for Sri Lanka is a modified version of the monetarist explanation of inflation extended to incorporate the influence of import prices and movements in the exchange rate.3 As the Fisher effect is a study of the relationship between the rate of inflation and interests rate, it will be crucial to examine the behaviour of the rate of interest in the period under study. Figure 5.5 illustrates the movement of the three-month Treasury bill rate over the 1952–1998 period. 3 Several studies on inflation have incorporated exchange rates and foreign prices as additional variables- Diz (1970), Lowinger (1978), Nugent and Glezakos (1979), Sheehey (1979), Saini (1984), Moser (1995) and in some cases as primary explanatory variables; Otani (1975), Aigbokhan (1991). 146 The Fisher Effect Figure 5.5 Source: Central Bank of Sri Lanka Annual Reports It is evident from the above figure that interest rates were permitted to respond to market forces only after the reforms of 1977. However, as mentioned in the previous chapter, the full impact of market forces was felt on interest rates only from the late 1980s onward. Figure 5.6 plots the rate of inflation and the three-month nominal Treasury bill rate over the 1952–1998 period. 147 The Fisher Effect Figure 5.6 Source: Central Bank of Sri Lanka Annual Reports While Treasury bill rates do not appear to have responded to the rate of inflation prior to 1977, they appear to have moved in the same direction over the 1977 to 1992 period with the rate of inflation displaying greater variability than the Treasury bill rate. It is observed from the figure that in 1997, the two rates moved very closely together; however, they diverged again in 1998. 5.3 THE FISHER HYPOTHESIS Fisher (1930) hypothesized that the nominal interest rate could be decomposed into two components, a real rate plus an expected inflation rate. Fisher claimed a one-to-one relationship between inflation and interest rates in a world of perfect 148 The Fisher Effect foresight, with real interest rates being unrelated to the expected rate of inflation and determined entirely by the real factors in an economy, such as the productivity of capital and investor time preference. This is an important prediction of the Fisher Hypothesis for, if real interest rates are related to the expected rate of inflation, changes in the real rate will not lead to full adjustment in nominal rates in response to expected inflation. Formally, the nominal return from holding a security is given by (1+it) = (1 + Et-1 rt) (1+ Et-1πt) (1) it = Et-1 rt + Et-1πt + Et-1 rt . Et- 1 πt (2) where it = nominal rate of interest πt = the rate of inflation rt = the real rate of interest E = expectations operator conditional on information at time t-1. The cross-product Et-1 rt . Et-1 πt, is assumed to be negligible and therefore omitted from the equation. Thus the following approximation is obtained: it = Et-1 rt + Et-1πt (3) Assuming a constant real rate of return of r, equation (3) can be rewritten as it = r + Et-1πt (4) The Fisher hypothesis predicts that the coefficient on the rate of inflation equals one. The Fisher effect, however, is a long-run relationship and, therefore, 149 The Fisher Effect inflation could affect real interest rates in the short run.4 During the adjustment process, the real rate will change so that the nominal rate reflects both changes in real rates and inflationary expectations. In the long run, when all adjustments have occurred, the increase in inflation is fully incorporated in nominal interest rates. A problem that arises when testing for the Fisher effect is the lack of any direct measure of inflationary expectations. For this reason, a proxy variable for inflationary expectations must be employed. Over the years, a number of approaches have been used to derive proxies for the expected rate of inflation. The majority of early studies on the Fisher effect used some form of distributed lag on past inflation rates to proxy for inflationary expectations. Models based upon this approach can be found in Cagan (1956), Meiselman (1962), Sargent (1969) and Gibson (1970). With the theory of rational expectations pioneered by Muth (1961), and the theory of efficient markets advanced by Fama (1970), there developed an alternative approach to modeling expectations. Subsequent studies, therefore, saw the incorporation of rational expectations in the formation of expectations. This approach is adopted by Fama (1975), Lahiri and Lee (1979), and Levi and Makin (1979). 4 Only in the case of complete flexibility of wages and prices combined with rational expectations would the adjustment occur instantaneously without any dynamics. 150 The Fisher Effect 5.3.1 Adaptive Expectations One approach to estimating expectations is to use an adaptive expectations approach, as used by Fisher (1930). This method of estimating inflationary expectations assumes that the information available to the market concerning future levels of inflation is contained in a series of current and past inflation rates. In each period, expectations are revised for forecasting errors made in the past period. Formally, Et-1 πt = πt-1 + λ (t-2 πt−1 − πt-1) (5) where πt-1 = the actual price level at t-1 t-2 πt−1 − πt- 1 = the forecasting error in t-1 λ = an adjustment factor 0 ≤ λ ≤ 1 Having some forecast Et-2 πt-1 based upon information available at the end of time t-2, agents examine ex-post how well the forecast predicted the actual value πt-1 and revise their forecast for π at time t by a fraction of the forecasting error at time t-1. This adjustment mechanism, as suggested by the adaptive expectations approach, is appropriate if the rate of inflation is subject to stochastic shocks. However, a limitation of this approach is that it could systematically over-predict or under-predict the rate of inflation in the event of a downward or upward trend in inflation rates. If, on the other hand, forecasts are rational, the prediction error is equal to the random term in the equation being forecast. Since, by assumption, the random term exhibits no pattern, neither will the forecasting error. Therefore, the assumption of rational expectations implies the absence of any pattern in 151 The Fisher Effect forecasting errors, which makes it a more precise measure of expectations formation. 5.3.2 Rational Expectations A second approach to estimating inflationary expectations is to assume rational expectations. Here, the market is assumed to be efficient in the sense of the current rate of inflation embodying all the information relating to future price changes that is in a time series of past inflation rates. This can be expressed by πt = Et-1 πt + wt (6) where wt is the forecast error of inflation consisting of the following properties, Et-1 wt = 0 2 2 Et-1 w t = σ w Et-1 wt wt-j = 0 j = 1,2,…. ∞. It follows from this that the errors of rational forecasts have a mean of zero, they exhibit no pattern, and have a finite variance. The theory of rational expectations asserts that optimising agents do not make systematic errors in expectations formation. The expected rate of inflation deviates from the actual rate only by the random shocks that are not foreseeable at the time expectations are formed. 5.4 EMPIRICAL MODELS The adaptive expectations hypothesis has the advantage of permitting the modeling of expectations purely in terms of past observations of πt without the need to specify the process by which the initial level of expectations is 152 The Fisher Effect determined. However, this backward-looking mechanism of expectations formation gives rise to the possibility of systematic forecasting errors. Rational expectations, on the other hand, assumes that optimising agents employ all available information relating to the process determining πt in forecasting inflation, thus making it a more efficient measure of inflationary expectations. This approach, however, has been criticized on the grounds that tests involving rational expectations are joint tests of the rational expectations hypothesis itself and the rest of the model. Therefore, in the event of rejection, it is not possible to discern the exact cause of the rejection. Due to the relative merits and demerits of both approaches, this study employs both adaptive and rational expectations in the formation of price expectations. Following Attfield, Demery and Duck (1985), two approaches are employed to incorporate rational expectations in the formation of inflationary expectations. 5.4.1 The Adaptive Expectations Model The model to be estimated is given by it = α + β Et-1πt +vt (7) where it = nominal rate of interest α = a constant real rate of return Et-1πt = expected inflation vt = random error term with mean zero Lagging equation (7) by one period and multiplying through by the adjustment factor, λ, gives 153 The Fisher Effect λit-1 = αλ + βλ Et-2 πt-1 + λvt-1 (8) Subtracting (8) from (7), it - λit-1 = α(1 - λ) + β( Et-1πt - λ Et-2 πt-1) + vt – λvt-1 (9) By the definition of adaptive expectations, which assumes that expectations are formed according to the previous period's forecasting error, equation (9) becomes it - λit-1 = α(1 - λ) + β(1 - λ) πt-1 + vt – λvt-1 (10) or it = θ0 + θ1it-1 + β′πt-1 + ωt (11) where θ0 = α(1 - λ) θ1 = β(1 - λ) ωt = vt – λvt-1 0 ≤ λ ≤ 1. In the case of perfect adjustment, with no change in the real rate of interest, λ would be expected to be zero. It should be noted, however, that the error term ωt is equal to vt – λvt-1 a first order moving average or MA(1) process. As it-1 involves vt-1, it is observed that it-1 is correlated with the error term ωt. Estimation by OLS, therefore, could yield inconsistent parameter estimates. To guard against the criticism of the inefficiency of the forecasts, instrumental variables are employed. 154 The Fisher Effect 5.4.2 Measuring a Rational Expectation of Inflation by its Actual Value The model to be estimated was given above as it = α + β Etπt+1 +vt (11a) Rational expectations in the formation of πt+1 implies that the actual value of πt+1 deviates from the expected value Etπt+1 , by a forecast error, πt+1 = Et πt+1 + wt+1 (12) where wt+1 is the forecast error of inflation . Substituting (12) into (11a) yields it = α + β πt+1 + et (13) where, et = vt - βwt+1 The Fisher hypothesis tests for β=1. If β=1, a long-run unit proportional relationship exists between interest rates and inflation. Estimation by OLS, however, could yield inconsistent parameter estimates as the rate of inflation πt+1, is correlated with the composite disturbance term et, which 5 incorporates the rational expectations forecast error wt+1. To the extent that changes in πt+1 are due to changes in wt+1 rather than Et πt+1, it will appear to be unaffected by expected inflation. The larger the changes in actual inflation resulting from a change in wt+1, the more severe will be the bias. This can be resolved by using instrumental variables that are correlated with πt+1 and uncorrelated with et. 5 See McCallum (1976) for a discussion of the problems that arise with respect to this approach. 155 The Fisher Effect 5.4.3 Measuring a Rational Expectation of Inflation by its Forecast Value An alternative approach to the incorporation of rational expectations in the formation of inflationary expectations is to presume that πt is determined by a process. Barro (1977a) employs such a procedure in estimating the predictable component of the rate of growth of the quantity of money. The model to be estimated was given above as it = α + β Etπt+1 +vt If the process determining πt+1 is specified by πt+1 = xt+1 γ + εt+1 (14) where πt+1 = the rate of inflation in period t+1 xt+1 = (1* q) vector of variables determining πt+1 γ = vector of parameters to be estimated εt+1 = normally distributed random error Given the process specified by equation (14), the rational expectation of πt+1 should be Et πt+1 = xt+1 γ (15) where possible variables for xt+1 could include money supply, import prices, real gross national product and exchange rates. Imposing rational expectations yields the two-equation model πt+1 = xt+1 γ + εt+1 156 The Fisher Effect ∧ it = α + β πt+1 +vt ∧ where the predictors πt+1 have been obtained by regressing πt+1 on xt+1. The Fisher hypothesis tests for the restriction that β=1. The correlation observed between the rate of inflation and the disturbance term when using the actual rate of inflation to measure inflationary expectations would not arise under this approach as implied by equation (15). Pagan (1984) has suggested potential problems that could arise with the OLS version of the two step estimator. These problems relate to inconsistent estimates of the variance of β and large asymptotic t statistics of the estimated coefficients. He observed that a consistent estimate of the variance of β could be obtained by applying two-stage least squares to the system. Therefore, estimation is carried out employing both OLS and 2SLS. 5.4.4 The Inverted Fisher Effect Empirical studies on the Fisher effect have failed to produce conclusive results. A number of possible reasons have been put forward among them, wealth effects, tax effects, peso effects. An alternative argument based on the work of Tobin (1965) and Keynes (1936) was put forward by Carmichael and Stebbing (1983) according to which the after-tax nominal rate of interest would remain constant with the after-tax real rate of interest varying one-for-one with the rate of inflation. Carmichael and Stebbing maintain that the lack of empirical support 157 The Fisher Effect for the Fisher hypothesis in its most rigorous form results from the attempt to apply it to yields on financial assets. They suggest that it should be applied to the rate of return on real capital. Given a certain degree of regulation in the financial market and high degree of substitution between regulated and non-regulated financial assets, they argue that the after-tax nominal rate of interest would remain approximately constant, while the after-tax real rate would move inversely one- for-one with the rate of inflation. The use of after-tax rates were justified on the grounds that individual budget constraints were determined by after-tax rather than pre-tax rates of interest. The reaction of economic agents to tax burdens of assets is an important issue at both the level of the individual and the aggregate economy. Therefore this chapter also attempts to see if taxes are important in the determination of the Fisher relationship. The Model Carmichael and Stebbing begin with two identities, the first defining the after-tax nominal interest rate as the after-tax real rate plus the rate of inflation, and the second defining the same relationship as holding between the mathematical expectations of the variables: it = rt + πt (16) Et-1it = Et-1 rt + Et-1πt (17) Where E is the expectations operator conditional on information at time t-1. Assuming that expectations are unbiased yields, 158 The Fisher Effect πt = Et-1πt + ε t (18) where ε t is a the random error term and the after tax nominal interest rate is known at all points in time, Et-1it = i t (19) Combining (16) to (19) yields, rt – Et-1rt = Et-1π t – π t = - ε t (20) The inverted Fisher effect in its expectations form is specified by Et- 1 rt = β0 + Et-1π t + ξ t (21) or equivalently It = β0 + ξ t (22) where β0 is a constant term premium on the financial asset and ξ t is a random 2 error term with zero mean and variance σ ξ. ξ t and ε t are uncorrelated. The inverted Fisher hypothesis can be tested in any one of four ways. As the after-tax nominal rate of interest is assumed to be a constant, it is independent of both Et-1π t and πt . Therefore, equation (22) can be estimated by running either of t he following regressions: it = β0 + β1 Et-1π t + ξ t (23) for β1=0, or it = a0 + β*1 π t + ξ t (24) for β*1=0. 159 The Fisher Effect Alternatively, employing (20) to eliminate the unobservable term Et-1rt, equation (21) can be written in a testable form: rt = β0 + β2 Et-1π t + ξ t − ε t (25) where ξ t and ε t are independent random variables. It follows from equation (18) that Et- 1 πt and εt are also independent. Hypothesis tests are carried out for, β2= -1. Substituting equations (18) and (20) into (21) to eliminate the unobservable Et-1π t yields, rt = β0 + β2 *π t + ξ t − (1+β2 *)ε t (26) Carmichael and Stebbing pointed out that, under a test of the null hypothesis for β2 *= -1, the error term ε t would disappear from the right-hand side of the equation, permitting the evasion of problems encountered with respect to errors in variables. Carmichael and Stebbing investigated the Fisher inversion on data from the US and Australia. Employing three-month Treasury bill rates for the US, the hypothesis that the coefficient on the rate of inflation was -1 yielded coefficient values in the range of –1.04 and –1.01 for the periods 1953.1– 1971.2, 1953.1–1965.4 and 1966.1–1978.4, providing strong evidence in support of a Fisher inversion. The Fisher inversion for Australia was tested on commercial bill rates and five-year debenture yields. Evidence was found to be consistent with the inverted Fisher effect with the estimated coefficients on the 160 The Fisher Effect rate of inflation ranging from –0.979 to –1.015 for the commercial bill rate and –1.050 and –1.136 for the five-year debenture yield. 5.5 PREVIOUS STUDIES Two phases can be identified in the development of the empirical literature pertaining to the Fisher hypothesis: first, work that concentrated largely on re- examining Fisher’s work in conforming the long and distributed lag in expectations formation;6 and second, subsequent work, which saw the integration of the Fisher hypothesis with the theories of rational expectations put forward by Muth (1961) and efficient markets developed by Fama (1970).7 With the incorporation of these theories in the Fisher hypothesis, methodological advances involved examining the time series properties of the variables in question.8 The first investigation in this section is of Fisher’s own conclusions and of the literature that concentrated primarily on verifying Fisher’s results. The majority of these studies confirmed Fisher’s findings of a distributed lag structure in expectations formation. The lack of empirical support for the Fisher hypothesis in its most rigorous form led to the development of a number of arguments in an attempt to reconcile these results. These arguments are also briefly surveyed. 6 See Sargent (1969), Gibson (1970, 1972), Friedman and Swartz (1982), Yohe and Karnosky (1969). 7 See Fama (1975), Lahiri and Lee (1979), Levi and Makin (1979) 161 The Fisher Effect 5.5.1 Fisher’s Findings Fisher (1930) hypothesized that the nominal rate of interest was equal to the sum of both the real rate of interest and the expected rate of inflation. He claimed a one-to-one relationship between the rate of interest and expected inflation, with the real rate being independent of the rate of inflation. Assuming that inflationary expectations were formed on the basis of a distributed lag structure, Fisher (1930) examined the relationship between nominal interest rates and the rate of inflation for the U.S and the U.K. Using annual data over the 1890–1927 period for the US, and 1820–1924 period for the U.K, Fisher found that inflationary expectations were not instantaneously reflected in interest rates. For the US, the highest correlation, 0.86, between long-term interest rates and price changes was obtained when the latter was lagged over 20 years, while for the UK, a correlation coefficient of 0.98 was obtained when price changes were spread over 28 years. A study of short-term commercial paper rates in relation to quarterly price movements for the US further corroborated the evidence from correlating long-term interest rates and price changes. Thus, Fisher (1930, p.451) concluded: We have found evidence general and specific … that price changes do, generally and perceptibly affect the interest rate in the direction indicated by a priori t heory. But since foret hought is imperfect , the effect s are smaller than the theory requires and lag behind price movements, in some periods, very greatly. When the effects of price changes upon interest rates are distributed o v er several years, we have found remarkably high coefficient s of correlat ion, t hus indicating t hat 8 See Mishkin (1992, 1994), Wallace and Warner (1993), MacDonald and Murphy(1989), Peng (1995). 162 The Fisher Effect int erest rat es follow price changes closely in degree, t hough rat her dist ant ly in time.9 This subsequently led to a voluminous literature in an attempt to reconcile Fisher’s findings with the theory. The following section will review some of the more important studies. 5.5.2 Adaptive Expectations The work of Sargent (1969), Gibson (1970), Yohe and Karnosky (1969), Lahiri (1976) concentrated primarily on verifying Fisher’s results with respect to the existence of a distributed lag structure in expectations formation. While adopting the basic distributed lag mechanism as that of Fisher in the formation of expectations, the specifications involving the lagged variables differed from the arithmetically declining weights as originally proposed by Fisher. Sargent (1969) and Gibson (1970) employed geometrically declining weights,10 while Yohe and Karnosky (1969) used the Almon lag technique11 in order to avoid problems of multicollinearity. The studies of Sargent and Gibson, based on data from the pre- war period, confirmed Fisher’s findings of a significant distributed lag effect in expectations formation. Gibson, moreover, observed that there appeared to be a cyclical factor in the formation of price expectations, suggestive of a higher- order weighting pattern for past price changes. An important implication that 9 This provided an explanation of the “ Gibson Paradox” which was the high correlation observed between the price level and interest rates during the pre-Word War II period. 10 Also known as the Koyck lag,originally proposed by Koyck (1954). 11 Put forward by Almon (1965), where βi follows a polynomial of degree r in i. 163 The Fisher Effect emerged from his study was that policy action designed to influence interest rates would eventually be felt on price expectations. From the 1960s, there was significant evidence of a shortening of the time lag in expectations formation, as suggested by the studies by Yohe and Karnosky (1969), Gibson (1972) and Lahiri (1976). Yohe and Karnosky found an acceleration in the speed of expectations formation, with the price expectation effect much greater for the 1961–1969 period than the 1952–1960 period. Gibson (1972) similarly observed that there was almost a point-for-point adjustment in nominal interest rates to changes in inflation during the 1959–1970 period, with a time lag of about six months. The results revealed a shorter lag in the formation of expectations and greater impact of expectations on interest rates for the period after 1959, supporting the evidence of Yohe and Karnosky. Lahiri, employing four approaches to estimate inflationary expectations—the weighted, adaptive, extrapolative, and Frenkel’s approach—found that expectations were forming more rapidly in the period after 1960, consistent with the findings of Yohe and Karnosky and Gibson. Gibson’s (1972) model, however, differs in its use of data. To overcome the problem of systematic forecasting errors produced by backward-looking models of expectations formation, Gibson employed survey data published by the Federal Reserve Bank of Philadelphia. While the structural break observed in the 1960s was attributed by Yohe and Karnosky to a shift in the interest rates equation, 164 The Fisher Effect Gibson suggested the possibility of a shift in the formation of the price expectations equation. Lahiri’s findings appeared to support that of Gibson. Therefore, a positive relation between interest rates and inflation with a significant shortening of the time lag in expectations formation from the 1960s onwards is evidenced by these studies. Moreover, the Fisher hypothesis took a different turn during this period in that it began to be integrated with the theories of rational expectations and efficient markets. 5.5.3 Rational Expectations and Efficient Markets The crux of the argument changed with the incorporation of the theories of rational expectations put forward by Muth (1961) and efficient markets developed by Fama (1970) in the Fisher hypothesis. While Fisher argued that past changes in the price level became embodied in the current rate of interest, Fama (1975) argued that future price changes were reflected in the current rate of interest. This was interpreted by him as evidence of an efficient market. Fama’s study, therefore, differed from the models discussed above in its analysis of inflationary expectations. This approach rejected Fisher’s conclusions of a distributed lag structure in the formation of expectations. Instead, it assumed that rational forecasters would use all available information in forming price expectations. Using data for one-mont h Treasury bills to approximate interest rates and the rate of change in the consumer price index to approximate price changes, he tested the joint hypothesis that the U.S Government Treasury bill market was efficient and 165 The Fisher Effect that the real return on one-to-six month Treasury bill was constant within a rational expectations framework. Fama computed sample autocorrelations of the expected change in purchasing power and real return for lags from 1–12 for the period January 1953 to July 1971. The estimated sample autocorrelations of πt were large, indicating that past rates of change in πt contained information about expected future rates of change. The sample autocorrelations of the real return were insignificantly different from zero, consistent with the hypothesis of a constant real return. Tests were also carried out for longer-term maturities for up to six months. Results for all maturites indicated that the market used all the available information about the rate of inflation in setting nominal rates of interest, thus supporting the efficient market hypothesis. Fama's findings were subsequently challenged by Hess and Bicksler (1975), Carlson (1977), Joines (1977), and Nelson and Schwert (1977). Carlson (1977), using Livingston data on the CPI for the period 1953–1971, rejected Fama's findings that short-term interest rates were efficient predictors of subsequent rates of inflation. Carlson introduced a business cycle variable to Fama’s regression equation which was represented by the ratio of employment to population, lagged by six months. With the incorporation of this variable, the coefficient on the interest rate in Fama’s model was found to deviate significantly, which led Carlson to conclude that information about inflation that was not fully incorporated in interest rates was reflected in this ratio. Joines (1977) observed a seasonal pattern in the forecast errors of the rate of price inflation used by Fama 166 The Fisher Effect which he pointed out was inconsistent with the concept of market efficiency leading him to question the accuracy of the price data used by Fama. Nelson and Schwert (1977) and Hess and Bicksler (1975) employed a Box-Jenkins approach to construct a time series predictor of inflation, based on past rates of inflation. The regression of the rate of inflation on the rate of interest and the estimated rate of inflation yielded a non-zero coefficient for estimated inflation, indicating that the forecaster contained information about the rate of inflation not embodied in the rate of interest. With the incorporation of rational expectations and efficient markets in the Fisher hypothesis literature, it was believed that the time series in question should approximate a random walk in an efficient market. The random-walk model requires that changes in past rates of inflation and interest rates be uncorrelated with all prior information. This was is in sharp contrast to the distributed lag effect in expectations formation which implied that inflation rates were highly and positively correlated. Although the studies of Hess and Bicksler (1975), Carlson (1977), Fama and Gibbons (1982) suggested that when expected real returns were assumed to display a unit root, Treasury bill rates were good predictors of inflation, no explicit tests for unit roots were carried out by them. Mishkin (1992), in an attempt to explain why there was strong evidence of a Fisher effect for some periods and not for others, pointed out that a Fisher effect would only appear in samples where inflation and interest rates displayed 167 The Fisher Effect stochastic trends. The reasoning behind this was that when the two series exhibit trends, they would trend together, resulting in a strong correlation between them. This involved determining the univariate statistical properties of the respective time series, namely, inflation and interest rates. Using monthly data from January 1953 through to December 1990, and the Dickey Fuller and Phillips (1987) tests for unit roots, he observed that both the levels of inflation and interest rates contained a unit root. Cointegration tests for a common trend in inflation and interest rates revealed the existence of a long-run Fisher effect, however the absence of a short-run relationship. As predicted, a Fisher effect was observed for the post-war period until October 1979 in which evidence was strongest of stochastic trends in inflation and interests rates. There was no evidence of a trend and therefore a Fisher effect for the pre-war period and October 1979 to September 1982. Studies by Bonham (1991), Jacques (1995) and Wallace and Warner (1993), covering a similar time period, confirmed Mishkin's findings that inflation contained a unit root. Using an expectations model of the term structure, Wallace and Warner (1993) examined the effects of inflation on long as well as short-term interest rates. Applying the Johansen and Juselius (1990) cointegration test to quarterly data from 1948.1–1990.4, they found interest and inflation rates to be I(1) processes in the majority of cases. Cointegration tests provided support for both the Fisher relationship in the short and long term, and the expectations theory 168 The Fisher Effect of the term structure. They could not reject the point-for-point relationship between interest rates and inflation as postulated by Fisher. Bonham (1991), applying the Dickey Fuller test to monthly data from 1955.1–1990.3, found results to be consistent with those of Wallace and Warner (1993). Results provided support for stationarity in the first differences while the null hypothesis of no cointegration could not be rejected at the 5% level for the 1995.1–1986.1 period. Pelaez (1995) tested the Fisher relationship, using both the Engle Granger two-step procedure and Johansen's vector autoregressive error correction mechanism, for the period 1959.1–1993.4. Although results appeared to corroborate previous evidence, with both the rates of interest and inflation displaying unit roots, there was no evidence of a Fisher relationship. This was attributed to the random-walk effect displayed by the ex ante real rate. In contrast, Rose (1988) found inflation to be a I(0) series and interest rates to be a I(1) series. Using annual data for the U.S for two sample periods, 1892–1970 and 1901–1950, he discovered that the null hypothesis of a unit root was rejected for inflation.12 For further verification of results, he also used quarterly data from eighteen OECD countries. The null hypothesis of a unit root was rejected at the 5% level for all eighteen countries, lending support to the results from the annual data on the U.S. He obtained the same result with monthly data for the US for the 12 Four measures of prices, the GNP deflator, consumer price index, implicit price deflator, and wholesale price index and two measures of nominal interest rates, yield on high-grade corporate bonds and short term commercial paper rate were used for this purpose. 169 The Fisher Effect 1947.1–1986.6 period, except for the period following the monetary policy change in October 1979. He concluded that others too—Huizinga and Mishkin (1984)—had found a structural break at this point. As opposed to Rose, Jaques (1995) observed that the interest rate spread contained distinctly different statistical properties from the rate of inflation. Using monthly observations from 1958.12–1991.12, he found inflation to be a I(1) series, while the interest rate spread appeared to be a I(0) series. Hence, while the majority of studies on the U.S appear to suggest a positive relationship between interests rates and inflation, they do not establish a one-to- one relationship as postulated by Fisher. It is useful, therefore, to examine if similar results have been obtained in respect of the Fisher relationship for other countries. Studies on the Fisher effect for samples of OECD countries have been undertaken by Mishkin (1984), Peng (1995) and MacDonald and Murphy (1989). Mishkin (1984), studying real interest rate movements in seven OECD countries for the period 1967.2–1979.2 in the euro deposit market, found a close relationship between nominal interest rates and expected rates of inflation for the U.K, the U.S and Canada. He found, however, that Germany, the Netherlands and Switzerland exhibited a much weaker Fisher effect. Consistent with the results obtained by Mishkin, Peng (1995) found a long-run relationship between interest rates and expected inflation for France, the U.K and the U.S for the 1957–1994 period, 170 The Fisher Effect using the Johansen (1988) and Johansen and Juselius (1990) methodology. Expected inflation was found to have a much weaker impact on interest rates in Germany and Japan. Peng noted that inflation persistence was sensitive to the degree of monetary accommodation, leading to the observation of cointegration between inflation and interest rates over time. He concluded that the strong anti- inflationary policies pursued by the monetary authorities in Germany and Japan had led to less persistent inflation and hence a weaker Fisher effect. Similarly, MacDonald and Murphy (1989) found evidence of a Fisher relationship for the U.S, Belgium, Canada and the U.K for the period 1955 to 1986. They discovered that the null hypothesis of no cointegration could not be rejected for all countries for the entire sample period, indicating the existence of the Fisher relationship. When t he samp le was divided int o fixed and float ing exchange rate regimes, however, some evidence of cointegration was observed for the U.S and Canada during the fixed exchange rate regime. There was no evidence of cointegration for any of the countries under the floating exchange rate regime. Contrary to the findings of Peng, and MacDonald and Murphy, Yuhn (1996) found evidence of a Fisher effect for the U.S, Germany and Japan, but little evidence of it for the U.K and Canada. Results also pointed to the fact that the Fisher effect was not robust to policy changes. As opposed to the results obtained by Mishkin (1992), evidence pointed to a strong Fisher relationship for the 1979.4–1993.2 period, while there was no evidence of a Fisher effect for the 1982.4–1993.2 period. Dutt and Ghosh (1995), examining the validity of the 171 The Fisher Effect Fisher theorem under fixed and floating exchange regimes for Canada, found no support for the Fisher effect for Canada, as did Yuhn. Tests on the Fisher effect for Australia can be found in Mishkin and Simon (1995), Atkins (1989), Olekalns (1996), Hawtrey (1997) and Inder and Silvapulle (1993). Unlike in the case of the U.S, where results appear to be broadly consistent, results for Australia appear to be mixed with only weak evidence in support of a Fisher effect. Mishkin and Simon (1995), using data spanning the period 1962.3–1993.4, found evidence of a long-run Fisher relationship; however, the absence of a short-run effect. Atkins (1989), employing the post-tax nominal bill rate as the dependent variable for Australia, came up with results consistent with the Fisher equation, while Olekalns (1996), using pooled data for the pre- and post-deregulation periods, found only partial adjustment of the interest rate to changes in inflationary expectations. Complete adjustment was found to obtain with the use of only post-deregulation data. Money-supply shocks which affected the real rate were seen as an impediment to full adjustment prior to deregulation. In a similar vein, Hawtrey (1997), using the Johansen methodology, found that there was no evidence of a Fisher effect before financial deregulation, while after deregulation there was evidence of one. Inder and Silvapulle (1993) using the ex post real bill rate as the dependent variable, found that results were inconsistent with the Fisher hypothesis. Thus, while evidence for the U.S seems to be broadly consist ent wit h suggest ion of a Fisher effect , result s for ot her developed nat ions are not so clear-cut. 172 The Fisher Effect 5.5.4 Empirical Work for Developing Countries Empirical work on the Fisher effect for developing countries is sparse. The limited evidence that has accumulated in respect of developing countries is briefly reviewed in this section. Empirical studies on the Fisher effect for the Latin American countries have been undertaken by Phylaktis and Blake (1993), Garcia (1993), Thornton (1996) and Mendoza (1992). An interesting conclusion that emerges from these studies is the consistency in results with significant evidence of a Fisher effect. The same degree of consistency is not observed in respect of other developing countries. Phylaktis and Blake (1993) examined the Fisher effect for three high-inflation economies, namely Argentina, Brazil and Mexico, for the 1970s and 1980s decades. Addressing the specific issue of whether there existed a long-run Fisher relationship, using the techniques of unit roots and cointegration, they found that there existed a long-run unit proportional relationship between nominal interest rates and inflation for the three countries reviewed. They noted that the results were in contrast to the mixed evidence obtained for low-inflation economies, suggesting that agents in high-inflation economies tended to invest more in inflation forecasts and hence have greater incentive to incorporate inflationary expectations in yield returns. Comparing the speed of adjustment of interest rates to unanticipated inflation of these three countries with that of Australia and the US, they found that the high-inflation economies took longer to adjust. However, for all countries, it was found that the speed of adjustment was not a function of 173 The Fisher Effect the absolute level of inflation or inflation rate volatility. Similarly, Garcia (1993), examining the Fisher effect for Brazil for the period 1973–1990, using interest rate data on non-indexed certificates of deposit from a sample of Brazillian banks, found that data was consistent with the Fisher hypothesis. Inflationary expectations were found to explain 99% of the movement in nominal interest rates. Thornton (1996), investigating the existence of a Fisher effect between Treasury bill rates and inflation in M exico for the period 1978–1994, using unit root and cointegration techniques, found that results were broadly consistent with those of Phylaktis and Blake. The likelihood ratio statistic for β=1 could not be rejected at the 5% level, consistent with the existence of a Fisher effect. M endoza (1992), investigating the Fisher effect in the context of the of partial financial indexation mechanism currently operating in Chile, found evidence in support of it. Results showed that indexation facilitated financial intermediation in an inflationary environment and did not necessarily lead to interest rates higher than under a system absent of indexation. Despite the fact that these studies employ different structures, evidence appears to lend strong support for the Fisher hypothesis. The same degree of support is not found in studies with respect to other developing countries. Kim (1989), Ham and Choi (1991), and Nam (1993) evaluated the Fisher relationship for Korea. Using data from 1974.1–1991.2 and vector autoregressive techniques, Nam found that the liquidity effect dominated the Fisher effect in the long run. These results contrasted with previous findings 174 The Fisher Effect by Kim and Ham and Choi, who reported that the Fisher effect dominated the liquidity effect. Zilberfarb (1989), employing survey data for the 1980.1–1988.2 period, examined the significance of the liquidity effect, unanticipated inflation and supply shocks in interest rate determination for Israel. He concluded the liquidity effect and unanticipated inflation had a negative impact on interest rates, while supply shocks had a positive effect on interest rates. 5.5.5 Empirical Work on Sri Lanka Only two studies have been carried out on the Fisher effect for Sri Lanka, and one is in the context of the stock market. Samarakoon (1996) examined the relationship between stock returns and inflation in Sri Lanka, using monthly and quarterly data for the period 1985 to 1996. Using three proxies— contemporaneous inflation, lagged inflation and expected inflation proxied by an ARIMA process—to model the expected rate of inflation, he discovered that both lagged inflation and expected inflation displayed a significant positive relation to stock returns, while contemporaneous inflation exhibited only a weak relation to stock returns. Employing the Johansen (1988) and Johansen and Juselius (1990) procedure, Payne and Ewing (1997) evaluated the Fisher effect for nine developing countries, of which Sri Lanka was one. Data covered the period 1978.1–1995.3 for Sri Lanka. Unit root tests revealed that interest rates and inflation were integrated of order one for all countries. The Johansen and Juselius cointegration approach 175 The Fisher Effect indicated the presence of a long-run relationship between nominal interest rates and inflation for Sri Lanka, Malaysia Singapore and Pakistan. The likelihood ratio statistic to test the null hypothesis of a coefficient of unity for inflation yielded a value of 0.987 for Sri Lanka. Hence, a unit proportional relationship was found to exist between nominal interest rates and inflation for Sri Lanka. A unit proportional relationship was also found for Malaysia and Pakistan, while there was no evidence of a Fisher effect for Argentina, Fiji, India, Niger and Thailand. Payne and Ewing reached the conclusion of a unit proportional relationship between interest rates and inflation, based on quarterly observations spanning the period 1978–1995. As opposed to the study by Payne and Ewing, this essay attempts to examine the robustness of the results extensively, employing a variety of data frequencies, samples and expected inflation rate proxies. Apart from using the actual rate of inflation to proxy the inflationary expectations, an attempt is also made to test the Fisher hypothesis using a forecast rate of inflation (see Section 5.6). The results from the regression, based on the forecast rate of inflation, corroborate the result from the regression of interest rates on the actual rate of inflation. Results appear to be in direct contrast to those of Payne and Ewing. 176 The Fisher Effect 5.4.5 Deviations from the Fisher Hypothesis Despite the positive relationship observed between interest rates and inflation, the majority of empirical studies have not conformed to the Fisher hypothesis in its strictest form. A number of possible explanations have been put forward in an attempt to reconcile the contradictory results obtained in respect of the Fisher hypothesis. Theoretical justification for the partial adjustment was provided by M undell (1963) and Tobin (1965) in terms of a “wealth effect”, and Darby (1975) and Feldstein (1976) in terms of a “tax effect”. Empirical justification was provided by Mishkin (1981, 1984) and Pelaez (1995), among others, in terms of a random- walk effect displayed by the ex ante real rate. An alternative explanation, based on the work of Tobin (1965) and Keynes (1936), was put forward by Carmichael and Stebbing (1983), which stated that the nominal rate remained constant with the real rate of interest moving inversely one-for-one with the rate of inflation. These arguments are briefly reviewed below. Mundell (1963) and Tobin (1965) demonstrated that the nominal interest rate would rise by less than unity in response to a change in inflation through the impact inflation had on the real rate. This implied that inflation led to a fall in real money balances and the resulting decline in wealth led to increased savings bringing downward pressure on real rates. The adjustment in nominal interest rates would therefore be less than one for one with the expected rate of inflation. 177 The Fisher Effect Empirical support for the Mundell-Tobin effect can be found in Woodward (1992) for shorter-term maturities in the U.K indexed bonds market. An alternative argument was put forward by Darby (1975) and Feldstein (1976), who declared that in the presence of taxes on interest income, nominal interest rates would rise by more than unity in response to expected inflation for a given after-tax real rate of interest. The nominal interest rate was predicted to rise at a rate of 1/(1-t) where t was a proportional tax rate on interest income. This argument, however, has had limited success in explaining actual interest rate movements—see Tanzi (1980), Cargill (1977), Carr, Pesando and Smith (1976). The Darby-Feldstein explanation was subsequently modified by Nielson (1981) and Gandolfi (1982) to incorporate capital gains taxation. They found that, while the nominal interest rates rose by more than unity in response to a change in the rate of inflation, it was not as high as that suggested by Darby and Feldstein. In contrast, Peek (1982) found strong evidence in support of a tax-adjusted Fisher effect. A number of recent studies—Mishkin (1984), Rose (1988), Pelaez (1995)— attribute the rejection of the Fisher effect to the non-stationarity of the ex ante real rate. Mishkin (1984), examining real interest rate behaviour in a sample of OECD countries for the 1967.2–1979.2 period, found that the constancy of the real rate was rejected for all seven countries studied. Pelaez (1995), examining a 178 The Fisher Effect longer time period from 1959.1 to 1993.4, came up with similar results for the U.S. An alternative argument is put forward by Carmichael and Stebbing (1983) by what they termed an ‘inverted’ Fisher effect. According to them, given a certain degree of regulation in the financial market and high degree of substitution between regulated and non-regulated financial assets, they argue that the after tax nominal rate of interest would remain approximately constant while the after-tax real rate would move inversely one for one with the rate of inflation. This was subsequently tested by Amsler (1986) and Graham (1988) for the US, and Choudhry (1997) for Belgium, France and Germany. While Graham, using the same time period as Carmichael and Stebbing, found that evidence clearly rejected the Fisher inversion, Amslers tests failed to reject the inverted Fisher hypothesis. Graham, however, found strong evidence in favour of a partial adjustment effect. Choudhry, employing a longer time period, ranging from 1955–1994, found some support for a partial adjustment, nonetheless, little support for a Fisher inversion. Therefore, evidence with respect to the inverted Fisher hypothesis has not been clear-cut. While the importance of taxes in the Fisher effect was previously highlighted in the work of Darby (1975) and Tanzi (1976), the empirical literature has not lent much support for the Darby-Tanzi hypothesis. This has been explained by way of fiscal illusion, tax evasion, tax exempt agents - see Tanzi (1984). Peek (1982) 179 The Fisher Effect however, finds strong evidence supporting the inclusion of income tax effects in the Fisher effect. In comparing tax-adjusted and non-tax-adjusted versions of the Fisher equation he finds that while the tax-adjusted version of the Fisher effect does not get rejected, the non-tax-adjusted version is rejected in four of the six equations tested. The evidence relating to the role of taxes in the Fisher effect although has not been consistent is nevertheless important. 5.5 DATA The empirical analysis makes use of annual, quarterly and monthly data. The annual data covers the period 1952–1998, comprising in sum a total of 47 observations, and the quarterly data covers the period 1978–1998, consisting in sum a total of 84 observations. The monthly data spans the 1990–1998 period, which is the period in which full effect of market reforms was felt on interest rates. The rate of interest used is the three-month Treasury bill rate and inflation is measured using the Colombo Consumer Price Index (CCPI). Using the CCPI to measure the price level results in a number of biases. First, it takes into account only consumer goods. Secondly, the question arises of how representative the basket of consumer goods used to measure the CCPI is, in respect of Sri Lankan consumers. Thirdly, the weights assigned to the goods in the index have remained unchanged over time (1952=100). Despite these shortcomings, the CCPI remains the best indicator of how an individual's basket of consumer goods 180 The Fisher Effect changes in price over time. It also has the advantage of not only being the official measure of inflation but also universally accepted by the community as the country's true measure of inflation. The GDP deflator has the advantage of having a broader coverage and therefore, more closely approximating the general price level of all final goods and services. However, since 1978, the CCPI has moved closely in line with the GDP deflator. Moreover, the GDP deflator does not directly reflect changes in import prices. Therefore, the CCPI has been selected as the measure of inflation.13 Forecasting a rate of inflation by assuming that the price level is determined by a process requires the following variables: money supply, real GNP, the import price index, and exchange rate. Due to the lack of quarterly and monthly data on real GNP, quarterly and monthly series are constructed, subject to the constraint that real GNP grew at a constant rate each quarter/month throughout the year. A similar procedure is used for the construction of quarterly and monthly series for import prices. All data is obtained from the annual reports and monthly bulletins of the Central Bank of Sri Lanka and the International Financial Statistics. Unit Root Tests Prior to performing the empirical tests, it is necessary to establish the order of integration of the variables in question. The ADF test for unit roots is employed 13 The CPI is employed in the work of Fama (1975), Mishkin (1982,1984), Fama and Gibbons (1984), among others. 181 The Fisher Effect to detect the presence of unit roots. A lag length, (k), for the ADF test is selected on the basis of the LM statistic to ensure the absence of serial correlation. Starting with a lag length of 12 for all data frequencies, lags are gradually eliminated if there is no evidence of autocorrelation in the residuals. A joint F test of zero restrictions is carried out in order to see if a trend in mean is required or not. Overall results point to the inclusion of a trend term for the levels of the series. For the first differences, it is insignificant and therefore is omitted. Table 5.1 reports results of unit root tests. 182 The Fisher Effect Table 5.1 UNIT ROOT TESTS Data Frequency Variable k LM ADF Annual 1952-1998 Pt 1 10.42 -1.97 Mt 2 3.75 -2.41 GNPt 0 8.13 -2.13 IMPt 1 6.16 -2.67* ERt 0 0.37 -2.18 it 0 4.32 -2.68* ∆Pt 1 11.14 -2.41 ∆Mt 0 3.47 -3.86*** ∆GNPt 0 6.91 -5.73*** ∆IMPt 0 9.43 -3.85*** ∆ERt 0 2.20 -6.05*** ∆it 0 6.23 -7.51*** Quarterly 1978-1998 Pt 3 9.61 -2.40 Mt 0 12.14 -2.30 GNPt 5 11.88 -2.53 IMPt 8 2.21 -2.94* ERt 0 3.88 -0.32 it 0 8.55 -3.98*** ∆Pt 1 6.79 -4.81*** ∆Mt 11 8.20 -4.58*** ∆GNPt 4 12.06 -2.92** ∆IMPt 7 4.37 -3.01** ∆ERt 0 4.28 -9.23*** Monthly 1990-1998 Pt 2 8.81 -3.24* Mt 5 11.35 -0.30 GNPt 1 0.61 -3.33* IMPt 1 10.79 -2.69* ERt 1 7.33 -2.19 it 0 1.56 -2.04 ∆Pt 2 3.68 -9.33*** ∆Mt 4 6.71 -5.83*** ∆GNPt 0 3.53 -2.73* ∆IMPt 0 7.53 -2.79* ∆ERt 0 6.20 -6.94*** ∆it 0 2.49 -10.31*** Note: Significance levels: 1%, -4.07: 5%, -3.46: 10% , -3.16 with trend and 1%, -3.51 : 5%, -2.90: 10% -2.58 without trend (Davidson and MacKinnon 1998) *, ** and *** indicate significance at the 10%, 5% and 1% levels respectively k refers to the order of the autoregression used to calculate the ADF statistic. A sixth order autoregressive model is used. The χ2 statistic for 6th order serial correlation in the residuals with 6 degrees of freedom is 12.59. 183 The Fisher Effect The results indicate that Mt ,GNPt, and ERt are I(1) in first differences for all data frequencies. However, the unit root tests for, it and Pt do not display the same degree of consistency across all data frequencies. While the results suggest that the rate of inflation is I(1) in the monthly and quarterly data, it is marginally below the 10% significance level in the annual data. In contrast, the Treasury bill rate appears to be stationary at the 10% and 1% levels of significance in the annual and quarterly data, while it appears to be I(1) in the monthly data. Due to the lack of consistency across the data frequencies in the order of integration in interest rates and inflation, the Fisher effect is tested using the conventional methods of OLS and instrumental variables. 5.7 EMPIRICAL RESULTS This section presents the empirical results relating to the Fisher hypothesis, employing the different approaches to modeling inflationary expectations described above. OLS and instrumental variables are employed in estimating the Fisher relationship. 5.7.1 The Adaptive Expectations Model This section attempts to identify a long-run relationship between interest rates and inflation employing OLS. The lack of consistency in the unit root tests for interest rates and inflation for the different data frequencies precludes the use of cointegration. Results for the adaptive expectations model employing OLS are reported in Table 5.2. The annual, quarterly and monthly data used here cover the 184 The Fisher Effect 1952–1998, 1978–1998 and 1990–1998 periods, respectively. The annual observations have been divided into two sub-samples in order to distinguish between the pre- and post-deregulation periods. It was observed in Section 5.4.1 that the error term in ωt in the adaptive expectations model, was equal to vt – λvt-1, resulting in a moving average (MA) error structure of order one. In order to correct for this, a M A(1) correction procedure is employed on the residuals. 185 The Fisher Effect 5.7.1.1 Estimation by OLS Table 5.2 The Adaptive Expectations Model: OLS Estimates it = θ0 + λit-1 + θ1πt-1 + ωt 2 Data Frequency θ0 λ θ1 β = θ1/1-λ R LM Wald β=1 Annual Data 3 month TB 1952- 1998 0.73 0.95 -0.02 -0.40 0.89 0.43 0.37 (0.52) (0.07) (0.07) (0.44) (0.06) (0.07) 1952–1976 0.22 0.93 0.03 0.43 0.91 1.86 1.72 (0.26) (0.11) (0.04) (0.32) (0.12) (0.04) 1977–1998 5.59 0.70 -0.07 -0.23 0.53 0.03 11.15*** (2.01) (0.15) (0.10) (2.16) (0.14) (0.10) Quarterly Data 3 month TB 1978.1 – 1998.4 4.42 0.68 -0.03 -0.09 0.47 7.00 23.52*** (1.11) (0.08) (0.07) (0.97) (0.07) (0.07) Monthly Data 3 month TB 1990.1 – 1998.12 1.24 0.92 -0.04 -0.50 0.83 6.91 2.88 (0.64) (0.04) (0.07) (0.70) (0.04) (0.06) Note: Standard errors reported within parenthesis. The second term in parenthesis is the standard error corrected by a MA(1) procedure. LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly and monthly test statistics should be compared with 3.84, 9.49 and 21.03 which are the χ2 statistics for 1st, 4th and 12th order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. *** significant at the 1% level. 186 The Fisher Effect The adjustment factor, λ, is in the range of 0.95 to 0.68, suggest ing the importance of an error-learning mechanism in expectations formation. The results for the annual data, however, indicate a decline in λ from 0.93 in the pre- deregulation period to 0.70 in the post-deregulation period, suggesting a shortening of the time period in expectations formation. The estimated values of θ1 are negative in every period except the 1952-1976 giving rise to negative β coefficients for all periods except the pre-deregulation period. The Wald test statistics for a test of the null hypothesis that β equals one is not rejected for the monthly data, annual data covering the full sample and the annual data covering 2 the pre-deregulation period. The R for the regression equations are in the range of 0.91 to 0.47, suggest ing high exp lanat ory p ower of t he regression equat ions, and the LM statistics indicate the absence of first, fourth and twelfth-order serial correlation for the annual, quarterly and monthly observations respectively. The results however, should be interpreted with caution because the application of OLS to models with lagged dependent variables give rise to biased estimates. In order to correct for this the next section estimates the model employing instrumental variables. 5.7.1.2 Estimation by Instrumental Variables The adaptive expectations model involves a regression of it on it-1 which is of the autoregressive form. As it-1 incorporates vt-1 (see Section 5.4.1), and vt – λvt-1 is equal to ωt, it-1 is correlated with the error term ωt. Estimation by OLS, therefore, 187 The Fisher Effect could yield inconsistent parameter estimates. In order to avert this, the instrumental variables estimation procedure is also employed. 188 The Fisher Effect Table 5.3 The Adaptive Expectations Model: Instrumental Variable Estimates it = θ0 + λit-1 + θ1πt-1 + ωt 2 Data Frequency θ0 λ θ1 β = θ1/1-λ GR LM Wald: β=1 Annual Data 0.43 0.70 0.32 1.1 0.81 0.46 0.06 3 month TB (0.77) (0.18) (0.24) 1952- 1998 (0.44) (0.06) (0.07) 1952–1976 0.41 0.93 0.01 0.14 0.94 1.58 6.69*** (0.16) (0.07) (0.03) (0.31) (0.12) (0.04) 1977–1998 3.08 0.44 0.42 0.75 0.25 0.00 0.08 (5.23) (0.34) (0.58) (2.58) (0.15) (0.11) Quarterly Data 5.10 0.66 -0.16 -0.47 0.43 2.81 7.21*** 3 month TB (1.26) (0.09) (0.18) 1978.1–1998.4 (0.97) (0.07) (0.07) Monthly Data 1.35 0.92 -0.10 -1.25 0.69 6.62 2.12 3 month TB (0.72) (0.05) (0.11) 1990.1–1998.12 (0.70) (0.04) (0.07) Instruments Annual Data : Constant, πt-2, πt-3,it-2, it-3 Quarterly Data: Constant, πt-2, πt-3,it-1, it-2 Monthly Data: Constant, πt-2, πt-2, πt-3,πt-4, πt-5 , πt-6, ,πt-7 it-2, it-3 , it-4, it-5, it-6, it-7 Note: Standard errors reported within parenthesis. The second term in parenthesis is the standard error corrected by a MA(1) procedure. LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly and monthly test statistics 2 st th th should be compared with 3.84, 9.49 and 21.03 which are the χ (.05)statistics for 1 , 4 and 12 order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. GR2 denotes the Generalized R2 for IV regressions proposed by Pesaran and Smith (1994). Pesaran and Smith show that under certain assumptions the GR2 is a better measure of the goodness of fit for models estimated by IV than R2. *** significant at the 1% level. 189 The Fisher Effect The adjustment factor is in the range of 0.93 to 0.44, confirming the importance of a distributed lag structure in expectations formation as in the OLS regression estimates. The β coefficient for the annual data covering the full sample is 1.1 indicating full adjustment of interest rates to inflation within an year. Moreover, the restriction that β=1 cannot be rejected for the annual data covering the full sample and the annual and monthly data covering the post deregulation period. While the β coefficients take on negative values in the regressions based on the quarterly and monthly data, the adjustment factor is high even for these frequencies. 5.7.2 Rational Expectations with the Actual Rate of Inflation as Proxy for Inflationary Expectations This section estimates the Fisher equation with the actual rate of inflation as proxy for inflationary expectations. As in the previous section, estimation is carried out using OLS and IVs. 190 The Fisher Effect 5.7.2.1 Estimation by OLS Table 5.4 reports results of the hypothesis tests employing OLS. Table 5.4 Rational Expectations with the Actual Rate of Inflation as Proxy for Inflationary Expectations: OLS Estimates it =α + β πt+1 + e 2 Data Frequency α β Wald : α=0 Wald:β=1 LM R Annual Data 3 month TB 1952 – 1998 3.35 0.65 10.71*** 10.21*** 15.72 0.45 (1.02) (0.11) (1.45) (0.16) 1952 – 1976 2.26 0.26 55.04*** 113.9*** 12.98 0.38 (0.32) (0.07) (0.40) (0.09) 1977 – 1998 15.13 -0.07 50.90*** 46.25*** 9.39 0.01 (2.12) (0.15) (1.11) (0.04) Quarterly Data 3 month TB 1978.1 – 1998.4 13.51 -0.02 1052*** 1626*** 42.39 0.009 (0.42) (0.03) (0.39) (0.01) Monthly Data 3 month TB 1990.1 – 1998.12 15.60 0.007 2436*** 5523*** 89.17 0.003 (0.32) (0.01) (0.33) (0.01) Note: standard errors reported within parenthesis. The second term in parenthesis is a Newey-West standard error that allows for a MA(1) error process. LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly and monthly test 2 st th statistics should be compared with 3.84, 9.49 and 21.03 which are the χ (.05) statistics for 1 , 4 and 12th order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. *** significant at the 1% level 191 The Fisher Effect The results in Table 5.4 indicate that the coefficients on the rate of inflation are significantly below the hypothesized value of unity except in the case of the annual data for the full sample. The β coefficient falls from 0.26 in the pre- deregulation period to –0.07 in the post-deregulation period for the annual data, while the monthly and quarterly data denote much less support for the Fisher hypothesis than the annual data. The Wald test statistics for a test of the null hypothesis that the coefficients on the rate of inflation is one, are above the 1% level of significance for all data frequencies. The coefficients on the intercept t erms are fairly high in all t he regression equations, suggest ing the exist ence of inflation premia. The R2 for all regressions are low, indicating low explanatory power of the regression equations. Overall, the data lend only weak support for the Fisher effect, when the actual rate of inflation is proxy for inflationary expectations. A Newey-West procedure is used to correct for first order serial correlation in the error term. 5.7.2.2 Estimation by Instrumental Variables Since OLS will be inconsistent if the expected rate of inflation is correlated with the disturbance term, estimation is also carried out using instrumental variables. 192 The Fisher Effect Table 5.5 Rational Expectations with the Actual Rate of Inflation as Proxy for Inflationary Expectations: IV Estimates it =α + β πt+1 + e 2 Data Frequency α β Wald : α=0 Wald : β=1 LM GR Annual Data 3-month TB 1952 – 1998 -2.48 1.45 1.14 2.62 2.79 0.81 (2.33) (0.28) (1.51) (0.17) 1952 - 1976 1.51 0.53 7.37*** 10.78*** 3.71 0.62 (0.56) (0.14) (0.43) (0.11) 1977 - 1998 30.65 1.34 5.15*** 4.63*** 1.09 0.30 (13.51) (1.09) (8.27) (0.69) Quarterly Data 3-month TB 1978.1 1998.4 15.28 -0.16 247*** 221*** 38.3 0.07 (0.97) (0.08) (0.39) (0.01) Monthly Data 3-month TB 1990.1 – 1998.12 15.49 0.02 1628*** 1719*** 83.2 0.005 (0.38) (0.02) (0.33) (0.01) Instruments Annual Data: Constant, πt-1, it-1 Quarterly Data: Constant, πt-2, it-2 Monthly Data: Constant, πt-1, πt-2, πt-3,πt-4, πt-5 , πt-6, it-1, it-2, it-3, it-4, it-5, it-6 Note: standard errors reported within parenthesis. The second term in parenthesis is a Newey- West standard error that allows for an MA(1) error term. LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly 2 and monthly test statistics should be compared with 3.84, 9.49 and 21.03 which are the χ (.05) statistics for 1st, 4th and 12th order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. GR2 denotes the Generalized R2 for IV regressions proposed by Pesaran and Smith (1994). *** significant at the 1% level. 193 The Fisher Effect In contrast to the OLS estimates, the IV estimates appear to yield better results. There is a one-to-one adjustment of interest rates to inflation in the annual data covering the full sample and post-deregulation period. While there is evidence of an increase in efficiency, with the coefficient on inflation rising from 0.53 to 1.34 between the pre- and post-deregulation periods, the restriction that β=1 cannot be rejected for the annual data comprising the full sample. The coefficient on inflation for the quarterly data are incorrectly signed suggesting that changes in inflation might have a negative impact on the rate of interest over the shorter term. While both the adaptive and rational expectations approaches appear to yield better results under IV estimation, it is difficult to reconcile the somewhat ambiguous results which indicate evidence in support of the Fisher hypothesis under both approaches for the full sample. Therefore, for further verification, the Fisher effect is also estimated with a forecast rate of inflation to proxy inflationary expectations. 5.7.3 Rational Expectations with a Forecast Rate of Inflation as Proxy for Inflationary Expectations In order to obtain a rational expectations model incorporating the Fisher equation, it is necessary to specify a long-run model for the price level. Prior to specifying a model, it would be useful to briefly review the theories that have been advanced to explain the increase in the general level of prices in order to place the 194 The Fisher Effect model in perspective. One is the monetarist theory of inflation which in its simplest form states that the rate of inflation varies positively with the rate of change in money supply and negatively with the rate of change in real income. Studies based upon this approach can be found in Harberger (1963), Vogel (1974), Aghveli and Khan (1977), Rao, Fahimuddin and Bajpai (1997). An alternative explanation is provided by the Structuralists, who maintain that changes in the aggregate price level result from autonomous or structurally induced changes in costs of production and relative prices. Among the major sources of inflation are wages and import prices. The limited applicability of the monetarist model in an open economy context has led to the development of a number of models incorporating cost-push factors in the monetarist specification. Nugent and Glezakos (1979) and Lowinger (1978) have extended the monetarist model to incorporate exchange rates; Otani (1975) and Saini (1984) import prices; and Moser (1995) exchange rates and rainfall. Following the third approach, the price level is assumed to be a function of money supply, real GNP, import prices and exchange rates.14 Starting off with the levels of the series due to loss of information incurred in data differencing,15 the following model is specified: Pt = γ1 + γ2 M t + γ3 GNP + γ4 IMPt + γ5 ERt + wt (27) where Pt = price level 14 The choice of variables is determined on the basis of the nature of the Sri Lankan economy (see section 5.2) and the availability of data. The lack of an overall wage rate index precludes the use of wages in the model. 15 Data differencing eliminates the trend or long term movement in variables. 195 The Fisher Effect M t = money supply (M 2) GNPt = real Gross National Product IMPt = import price index ERt = official exchange rate Rupee/US Dollar wt = random error term 5.7.3.1 Cointegration First, the existence of a long-run relationship among the variables is established on the basis of the Johansen (1988) and Johansen Juselius (1990) test. Given a long-run relationship between the variables, the information in the error term of the cointegrating vectors is used to specify a dynamic error correction model. The optimal error correction models are then used to forecast a rate of inflation. Testing for cointegration requires that the variables in question be integrated of the same order. The unit root tests presented in Table 1 indicate that all series are I(1). In the case of the annual data, the exchange rate is consistently incorrectly signed and therefore is omitted from the cointegrating regression equation. This is not surprising in the light of the fact that exchange rates were fixed over the 1952 to 1977 period. The cointegrating regression equations, corresponding to the quarterly and monthly data that cover the period in which exchange rates are floating, lend some support for the inclusion of the exchange rate. The real GNP variable was consistently insignificant for all data frequencies and has therefore been dropped from the analysis. 196 The Fisher Effect In order to apply the Johansen procedure, a lag length must be selected for the VAR. The lag length of 1 is selected for all data frequencies on the basis of the Akaike Information Criterion (AIC). Johansen and Juselius have put forward two likelihood ratio tests for the determination of the number of cointegrated vectors. One is the maximal eigenvalue test, which evaluates the null hypothesis that there are at most r cointegrating vectors against the alternative of r+1 cointegrating vectors. The maximum eigenvalue statistic is given by: λmax = - T ln (1 - λr+1) where λ r+1,….λn are the n-r smallest squared canocial correlations and T= the number of observations. The second test is based on the trace statistic, which tests the null hypothesis of r cointegrating vectors against the alternative of r or more cointegrating vectors. This statistic is given by λ trace = -T Σ ln (1 - λi) Cointegration tests are reported in Tables 5.6, 5.7 and 5.8, where r denotes the number of cointegrating vectors. Estimates of inflation for monthly, quarterly and annual data are computed for purposes of comparison, with results measured by the actual rate of inflation. 197 The Fisher Effect Table 5.6 Johansen’s Cointegration Test for the Annual Data A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: Pt Mt IMPt Intercept Null Alternative Statistic 95% Critical Value r = 0 r = 1 115.6185 22.0400 r <= 1 r = 2 13.8711 15.8700 r <= 2 r = 3 7.2619 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r >= 1 136.7516 34.8700 r <= 1 r >= 2 21.1331 20.1800 r <= 2 r = 3 7.2619 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector Pt Mt IMPt Intercept 1 0.21206 -0.04745 -0.10085 -0.66058 (-1.000) (0.22377) (0.47559) (3.1150) 2 0.90293 -0.72563 0.26225 1.0416 (-1.000) (0.80364) (-0.2904) (-1.1536) Panel 1 reports the maximum eigenvalue test of the null hypohthesis that there exist at most r cointegrating vectors against the alternative of r+1 cointegrating vectors. Starting with the null hypothesis of r=0 against the alternative of r=1, the computed test statistic exceeds the 95% critical value, indicating the existence of 198 The Fisher Effect one cointegrating vector. The null hypothesis of r<=1 against r=2 cannot be rejected. Panel 2 reports the trace test of the null hypothesis that there are at most r cointegrating vectors against the alternative that there are more than r cointegrating vectors. Both the null of r=0 against the alternative of r>=1 and the null of r<=1 against r >= 2 are rejected, suggesting that there are at most two cointegrating vectors. Panel 3 presents the estimated cointegrating vectors. The coefficients in parenthesis are normalized on the CCPI. In the first cointegrating vector, money supply and import prices have the expected signs, while in the second cointegrating vector, the coefficient on import prices is incorrectly signed. Given these results, and in view of the fact that the existence of two cointegrating vectors is rejected by the maximal eigenvalue test, the first cointegrating vector is taken as the preferred estimate of the forecast rate of inflation. Tables 5.7 and 5.8 present the computed test statistics for the quarterly and monthly data. For the monthly data, the maximal eigenvalue and trace statistics indicate the existence of two cointegrating vectors at most. For the quarterly data, however, while the eigenvalue statistic indicates that there are two cointegrating vectors, the trace statistic suggests the existence of three cointegrating vectors at most. However, as this statistic is only marginally above the 95% critical value 199 The Fisher Effect and the eigenvalue test indicates the existence of two cointegrating vectors, only two are selected. 200 The Fisher Effect Table 5.7 Johansen’s Cointegration Test for the Quarterly Data A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: Pt Mt IMPt ERt Intercept Null Alternative Statistic 95% Critical Value r = 0 r = 1 129.8505 28.2700 r <= 1 r = 2 41.7300 22.0400 r <= 2 r = 3 15.8506 15.8700 r <= 3 r = 4 5.1358 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r >= 1 192.5668 53.4800 r <= 1 r >= 2 62.7163 34.8700 r <= 2 r >= 3 20.9864 20.1800 r <= 3 r = 4 5.1358 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector Pt Mt IMPt ERt Intercept 1 -1.1032 0.25756 0.44833 0.57328 0.64141 (-1.0000) (0.23347) (0.40640) (0.51966) (0.58142) 2 0.31497 -0.21678 0.030261 -0. 10981 0.66964 (-1.0000) (0.68855) (-0.09607) (0.34863) (-2.1260) 201 The Fisher Effect Table 5.8 Johansen’s Cointegration Test for the Monthly Data: A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: Pt Mt IMPt ERt Intercept Null Alternative Statistic 95% Critical Value r = 0 r = 1 189.1660 28.2700 r <= 1 r = 2 33.9095 22.0400 r <= 2 r = 3 9.9444 15.8700 r <= 3 r = 4 1.5337 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r >= 1 234.5536 53.4800 r <= 1 r >= 2 45.3876 34.8700 r <= 2 r >= 3 11.4781 20.1800 r <= 3 r = 4 1.5337 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector Pt Mt IMP t ERt Intercept 1 -0.51937 0.17741 0.12884 0.13958 0.60476 (-1.0000) (0.34159) (0.24807) (0.2687) (1.1644) 2 -2.1506 0.80098 -0.12445 1.2067 2.0050 (-1.0000) (0.37245) (-.05787) (0.5611) (0.9323) The results presented in Tables 5.7 and 5.8 suggest that there are two cointegrating vectors linking prices, money supply, import prices and the exchange rate. While the estimated coefficients in the first cointegrating vector 202 The Fisher Effect for both data frequencies have the expected signs, imports are incorrectly signed in the second cointegrating vector. If a cointegrating relationship exists among a set of variables, then a dynamic error correction representation also exists. Using the first cointegrating vector for each data frequency, dynamic error correction models are estimated for forecasting a rate of inflation. Starting with a lag length of one for the annual data, and four for the quarterly and monthly data, insignificant variables are eliminated on the basis of the t statistic, as greater efficiency may be gained by eliminating insignificant coefficients. The optimal error correction models are then used to forecast a rate of inflation for each data frequency. The optimal error correction models are as follows, where ECt-1 is the error correction term: Error Correction Models Annual Data ∆Pt = 0.02 + 0.11 ∆IMPt-1 + 0.59∆Pt-1 - 0.08ECt-1 (1.93) (2.95) (5.78) (-1.19) 2 2 2 2 χ sc = 5.30 χ ff = 0.47 χ n = 3.76 χ hs = 3.42 Quarterly Data ∆Pt = 0.004 + 0.16 ∆Mt-1 + 0.42 ∆ Pt-2 + 0.18 ∆ IMPt-3 - 0.34ECt-1 (0.87) (1.75) (4.62) (3.92) (-5.10) 2 2 2 2 χ sc = 7.55 χ ff = 0.02 χ n = 4.53 χ hs = 1.04 Monthly Data ∆Pt = 0.003 + 0.17 ∆Mt-2 + 0.42 ∆ Pt-1 - 0.32ECt-1 (1.24) (1.39) (4.15) (-4.66) 2 2 2 2 χ sc = 21.43 χ ff = 0.32 χ n = 1.35 χ hs = 0.11 t statistics reported in parenthesis 203 The Fisher Effect The error correction models for all data frequencies appear to perform well. In each case, the error correction term is of the correct sign. The error correction terms in the quarterly and monthly models are highly significant, suggesting that approximately a third of the entire change in the quarterly/monthly rate of inflation is corrected by the end of the first quarter/month. The lagged rates of inflation appear to be important in affecting inflation on an annual, quarterly and monthly basis. Import prices are significant in the annual and quarterly models, implying that changes in import prices are important in influencing inflation over longer time intervals, while money supply is significant in the quarterly and monthly models, suggesting that money supply is more important in affecting prices over shorter time intervals. The diagnostic tests for functional form misspecification, heteroscedasticity, normality suggest that the models are well- specified. The χ2 statistics for serial correlation in the annual, quarterly and monthly models are to be compared with the critical values of 3.84, 9.48 and 21.03, with 1, 4, and 12 degrees of freedom respectively. There appears to be some evidence of serial correlation in the annual model. These models are then used to forecast a rate of inflation that is used to estimate the Fisher effect. 5.7.3.2 Estimation by OLS Having established a long run relationship between prices and the variables believed to influence prices in Sri Lanka, an error correction framework was used above to compute forecast rates of inflation for each data frequency. The Fisher 204 The Fisher Effect effect is estimated in this section using these forecast rates of inflation as proxy for inflationary expectations. 205 The Fisher Effect Table 5.9 Rational Expectations with a Forecast Rate of Inflation as Proxy for Inflationary Expectations: OLS Estimates it =α + β πt+1 + e 2 Data Frequency α β Wald: α=0 Wald: β=1 LM R Annual Data 3 month TB 1952 – 1998 4.51 0.16 20.95*** 687 14.32 0.37 (0.99) (0.03) (0.98) (0.01) 1952 – 1976 2.74 0.03 75.03*** 4357*** 19.93 0.15 (0.32) (0.01) (0.34) (0.01) 1977 – 1998 11.52 0.08 55.51*** 569*** 7.66 0.19 (1.54) (0.04) (1.06) (0.02) Quarterly Data 3 month TB 1978.1 – 1998.4 14.14 -0.06 720*** 765*** 34.28 0.03 (0.52) (0.04) (0.44) (0.02) Monthly Data 3 month TB 1990.1 – 1998.12 16.71 0.01 3103*** 2988*** 54.68 0.006 (0.30) (0.02) (0.33) (0.02) Note: standard errors reported within parenthesis. The second term in parenthesis is a Newey-West standard error that allows for a MA(1) error process. LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly and monthly test 2 st th th statistics should be compared with 3.84, 9.49 and 21.03 which are the χ (.05)statistics for 1 , 4 and 12 order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. *** significant at the 1% level 206 The Fisher Effect The results reported in Table 5.9 do not appear to support the existence of a long term Fisher effect. The β coefficient which is significantly different from unity, rises marginally between the pre and post-deregulation periods for the annual data. The β coefficient for the quarterly model is incorrectly signed, while the coefficient for the monthly model, although of the correct sign, is insignificant. A test of the null hypothesis that β=1 is rejected for all samples and data frequencies, suggesting the complete absence of a Fisher effect. 5.7.3.3 Estimation by Instrumental Variables The results using IV estimation are presented in Table 5.10. 207 The Fisher Effect Table 5.10 Rational Expectations with a Forecast Rate of Inflation as Proxy for Inflationary Expectati ons: IV Esti mate s it =α + β πt+1 + e 2 Data Frequency α β Wald: α=0 Wald: β=1 LM GR Annual Data 3 month TB 1952 – 1998 -1.72 0.46 0.40 24.14*** 0.52 0.91 (2.68) (0.11) (1.94) (0.24) 1952 – 1976 1.51 0.16 2.23 121*** 3.45 0.87 (1.01) (0.08) (0.47) (0.09) 1977 – 1998 0.74 0.39 0.00 2.50 0.65 0.23 (13.44) (0.39) (5.15) (0.51) Quarterly Data 3 month TB 1978.1 – 1998.4 15.31 -0.15 456.47*** 410*** 23.11 0.10 (0.72) (0.06) (2.37) (0.21) Monthly Data 3 month TB 1990.1 – 1998.12 16.72 0.02 1513*** 857*** 50.48 0.01 (0.42) (0.03) (0.33) (0.02) Instruments Annual Data: Constant, πt-1, it-1 Quarterly Data: Constant, πt-2, it-2 Monthly Data: Constant, πt-1, πt-2, πt-3,πt-4, πt-5 , πt-6, it-1, it-2, it-3, it-4, it-5, it-6 Note: standard errors reported within parenthesis. The second term in parenthesis is a Newey-West standard error that allows for a MA(1) error process. 208 The Fisher Effect Table 5.10 – notes (cont) LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly and monthly test statistics should be compared with 3.84, 9.49 2 st th th and 21.03 which are the χ (.05) statistics for 1 , 4 and 12 order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. GR2 denotes the Generalized R2 for IV regressions proposed by Pesaran and Smith (1994). *** significant at the 1% level. The results from estimating the Fisher equation using IVs are similar to the results obtained under OLS. The β coefficient rises between the pre and post- deregulation period as in the OLS estimates, however, here a test of the hypothesis that β=1 in not rejected for the post-deregulation period in the annual data. Given the lack of consistency in results in the rational expectations models, it is possible to conclude that best support is found for the adaptive expectations model. The high adjustment factor in the adaptive expectations models, further confirm the importance of an error-learning mechanism in expectations formation. Figure 5.7 graphs the nominal and real Treasury bill rates and the rate of inflation for the 1952–1998 period. The failure of the rate of inflation to get incorporated instantaneously into the rate of interest is evident from the graph. While the nominal Treasury bill rate exhibits little variation in the 1950s and 1960s decades, the real rate fluctuates due to the variation in rate of inflation. It appears that the high nominal rates of the post-deregulation period are clearly not sufficient to compensate for the rising rate of inflation. Therefore, real rates are negative for most of the period between 1978–1985. 209 The Fisher Effect Figure 5.7 Source : Central Bank of Sri Lanka Annual Reports. Real rate calculated as i - π. A probable explanation to the slow adjustment on the part of the market in reacting to inflation could lie in the fact that inflation has an indirect influence on the nominal rate through its impact on the real rate. Therefore, the following section goes on to examine the relationship between the real rate and the rate of inflation. 5.8 THE INVERTED FISHER EFFECT This section examines the possibility of an inverted Fisher effect as a probable explanation for the results here obtained. Table 5.11 presents results from 210 The Fisher Effect estimating equation (25) employing annual, quarterly and monthly data covering the 1952–1998, 1978.1–1998.4 and 1990.1–1998.12 periods. The ex post after- tax real rate of interest is calculated using the formula, rt= (1-t)i – π, where t represents the tax rate on income. The CCPI is used to approximate the rate of inflation as in the previous section. The annual observations are split into two samples for purposes of comparison between pre- and post-deregulation periods. 211 The Fisher Effect Table 5.11 The Inverted Fisher Effect rt = β0 + β2 *π + ξ − (1+β2 *)ε 2 Data Frequency β0 β2* Wald: β0=0 Wald: β2*=-1 LM R Annual Data 3 month TB 1952 - 1998 3.29 -0.30 12.23*** 51.76*** 20.68 0.17 (0.94) (0.10) (1.04) (0.09) 1952 - 1976 2.24 -0.70 61.24*** 22.22*** 9.98 0.85 (0.29) (0.06) (0.30) (0.06) 1977 - 1998 12.12 -0.83 45.81*** 1.46 11.47 0.64 (1.79) (0.14) (0.94) (0.03) Quarterly Data 3 month TB 1978.1 – 1998.4 13.26 −1.00 1034.0*** 0.01 41.30 0.95 (0.41) (0.03) (0.39) (0.01) Monthly Data 3 month TB 1990.1 – 1998.12 15.51 -0.99 2414.6*** 0.78 90.70 0.98 (0.31) (0.01) (0.33) (0.01) Note: standard errors reported within parenthesis. The second term in parenthesis is a Newey- West standard error that allows for a MA(1) error process. LM denotes the Lagrange Multiplier version of Godfrey’s (1987a, 1987b) test statistic for serial correlation. The annual, quarterly and monthly test statistics should be compared with 3.84, 9.49 and 21.03 2 st th th which are the χ (.05)statistics for 1 , 4 and 12 order serial correlation in the residuals with 1, 4 and 12 degrees of freedom respectively. *** significant at the 1% level The coefficients on the rate of inflation for the post-deregulation period are in the range of –0.83 and –1.00, suggesting that approximately the entire variation in the 212 The Fisher Effect rate of inflation is absorbed by the real rate of interest. The Wald test statistics for a test of the null hypothesis that β=-1 cannot be rejected for the post-deregulation period for all data frequencies. The results appear to support the existence of an inverted Fisher effect in the post-deregulation period. This provides an explanation for the low coefficient estimates on expected inflation obtained for the direct Fisher effect. 5.9 IMPLICATIONS While the results appear to lend some support for both the rational and adaptive expectations models over the full sample, caution must be exercised in interpreting the results of the rational expectations model, due to the significant evidence of serial correlation in the residuals. It is possible, therefore, to conclude that best support is found for the adaptive expectations model, suggesting a slow learning process on the part of the market in forming expectations. It appears that even after the reforms of 1977, the market has been slow in reacting to inflation. The evidence in support of the adaptive expectations approach, confirm the backward-looking expectations established in the previous chapter. An important implication that arises from these findings is that the market is likely make systematic mistakes under-predicting or over-predicting the actual rate of inflation for many periods hence. A likely explanation for this observed phenomenon could lie in the deficiencies of the Colombo Consumer Price Index, which is the official indicator for measuring inflation in Sri Lanka. Based on an outdated basket (1952 = 100) and dominated by 213 The Fisher Effect staple foods (a weight of 61%) many of which are administered, it tends to understate the true rate of change in the general price level. Given that the prices of several food items are administered, the index is also affected by government action. Moreover, the CCPI is not representative of price changes in the entire economy as it covers only the Colombo metropolitan area and is biased towards the lower end of the income scale. Further, important present-day consumption items such as gasoline are not included in the index. Hence, it is possible that under-estimation of the true rate of inflation has led agents to systematically under-predict the actual rate of inflation. M oreover, real rates are calculated by adjusting the nominal rate to account for current inflation. The use of the CCPI in the computation of the real rate of interest may also overstate real yields, giving rise to the high correlation observed between real yields and inflation in Sri Lanka, as evidenced by the inverted Fisher effect. Therefore, to the extent which the CCPI does not provide an accurate depiction of the trend in general price level, monetary policy formulated on the basis of these prices will prove inefficient. This would also mislead foreign investors who compare national interest rates in making investment decisions. Due to the lack of quarterly and monthly data on real GNP and import prices, quarterly and monthly series are constructed subject to the constraint that real GNP grew at a constant rate each quarter/month throughout the year in the rational expectations model estimated by a forecasting model. A problem that arises when constructing data series in this manner is that it could give rise to model specification 214 The Fisher Effect errors. The estimated coefficients could be biased in the presence of measurement error in the explanatory variables, which perhaps provides a likely explanation for the lack of support for the rational expectations model estimated by inflation forecasts. Mishkin (1984) showed that one way of discerning the information content of nominal rates on the tightness of monetary policy was to examine the degree of correlation between nominal and real interest rates. As evidenced by Figure 5.7, nominal and real rates are not tightly linked in Sri Lanka, and therefore nominal rates could prove to be a misleading indicator of the stance of monetary policy. The non- existence of a Fisher effect over shorter intervals suggests that monetary policy could be used to influence real interest rates over the shorter term. Moreover, the failure of inflationary expectations to get incorporated into nominal interest rates could provide the incentive for the Government to run debt-financed fiscal deficits. 5.10 CONCLUSION The main objective of this chapter was to test for efficiency of the Treasury bill market with respect to inflationary expectations for the post-deregulation period. Evidence points to the statistical rejection of semi-strong form efficiency of the Treasury bill market in instantaneously responding to publicly available information on inflation in predicting nominal interest rates. Support for the Fisher effect, however, is found under the adaptive expectations approach. While this approach is entirely backward-looking, it allows the possibility of systematic forecasting errors 215 The Fisher Effect for many periods hence. The sub-optimal use of information does not, therefore, support the predictions of the rational expectations/efficient markets approach. 216 The Fisher Effect Appendix An Attempt at Replicating a Previous Study on the Fisher Effect for Sri Lanka In the preceding analysis, inflation was found to be stationary while interest rates were below the 1% critical value in the quarterly model. The order of the autoregression for the Dicky Fuller test was selected on the basis of the LM statistic by testing down from a twelve-lag system to ensure that the residuals were white noise. The stationarity of the series did not permit the use of cointegration tests. OLS and IV regressions of the rate of interest on the rate of inflation provided no support for the Fisher hypothesis for the quarterly data. In contrast, using quarterly data over the 1978.1–1995.3 period and the Johansen Juselius technique, Payne and Ewing (1997) found a one-to-one relationship between nominal interest rates and inflation for Sri Lanka. An attempt at replicating their results show that selection of the order of the autoregression for the Dickey Fuller unit root test on the basis of the Schwartz Baynesian Criterion leads to non-rejection of the null hypothesis of non-stationarity for both inflation and interest rates. Further differencing show these series to be I(1). The application of Johansen’s cointegration test on the data, excluding a drift term, gives rise to a unit proportional relationship between interest rates and inflation. See panel 3 of Table A5.1. The use of the same procedure with the inclusion of a 217 The Fisher Effect drift term, however, leads to implausibly large, incorrectly signed coefficients, signifying the absence of a Fisher effect. See Table A5.2. Table A5.1 Johansen’s Cointegration Test for the Quarterly Data Excluding a Drift Term A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it πt Null Alternative Statistic 95% Critical Value r = 0 r = 1 21.6398 11.0300 r <= 1 r = 2 0.20735 4.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r = 1 21.8473 12.3600 r <= 1 r = 2 0.20735 4.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it πt 1 -0.014322 0.017264 (-1.0000) (1.2054) 218 The Fisher Effect Table A5.2 Johansen’s Cointegration Test for the Quarterly Data Including a Drift Term A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it πt Null Alternative Statistic 95% Critical Value r = 0 r = 1 24.0179 15.8700 r <= 1 r = 2 6.0770 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r = 1 30.0949 20.1800 r <= 1 r = 2 6.0770 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it πt Intercept 1 0.0032799 0.018681 -0.26358 (-1.0000) (-5.6958) (80.3646) The above results indicate the sensitivity of the estimates to alternative model specifications and the order of the autoregression in the Dickey Fuller test. Therefore, the question arises as to the basis upon which the appropriate lag length for the Dicky Fuller test should be determined. Selection of the order of the autoregression, according to the Schwartz Baynesian criterion, leads to the selection of a lag length of 9, which is far in excess of the number of lags required 219 The Fisher Effect to whiten the residuals. This yields a Dickey Fuller test statistic of –2.39 for the rate of inflation in levels as obtained by Payne and Ewing. Testing down from a 12 lag system and selecting the appropriate lag length, according to the LM statistic, on the other hand, requires only 1 lag to whiten the residuals. Choi (1994) has pointed out that in the event of a moving average error structure, the long autoregressions used in the Dicky Fuller test to account for serial correlation biases the OLS estimates towards non-rejection of the unit root hypothesis. Next, the question arises as to the significance of the drift term in estimating the relationship. Johansen and Juselius (1990) showed that the inclusion of a drift term was closely related to the concept of whether the levels of the data displayed a trend or not (Johansen and Juselius, 1990, p.206). Figure A5.1 illustrates the price level for the period under study. A F test of zero restrictions for the drift term yields a test statistic of 4.34 which is above the 5% critical value of 3.07, pointing to the incorporation of a trend term in the cointegrating regression. 220 The Fisher Effect Figure A5.1 Source: Central Bank of Sri Lanka Annual Reports Assuming that inflation and interest rates are I(1), the Johansen approach is applied to the annual and monthly data. For the annual data, where an intercept is not included, the results are as follows: 221 The Fisher Effect Table A5.3 Johansen’s Cointegration Test for the Annual Data Excluding a Drift Term A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it πt Null Alternative Statistic 95% Critical Value r = 0 r = 1 17.1120 11.0300 r <= 1 r = 2 0.012126 4.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r = 1 17.1242 12.3600 r <= 1 r = 2 0.012126 4.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it πt 1 -0.027946 0.034025 (-1.0000) (1.2175) 222 The Fisher Effect Table A5.4 Johansen’s Cointegration Test for the Annual Data Including a Drift Term A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it πt Null Alternative Statistic 95% Critical Value r = 0 r = 1 17.5119 15.8700 r <= 1 r = 2 2.0182 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r = 1 19.5301 20.1800 r <= 1 r = 2 2.0182 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it πt Intercept 1 -0.025166 0.034194 -0.037162 (-1.0000) (1.3588) (-1.4767) 223 The Fisher Effect Table A5.5 Johansen’s Cointegration Test for the Monthly Data Excluding a Drift Term A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it πt Null Alternative Statistic 95% Critical Value r = 0 r = 1 44.7232 11.0300 r <= 1 r = 2 0.55607 4.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r = 1 45.2793 12.3600 r <= 1 r = 2 0.55607 4.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it πt 1 -0.002982 0.004556 (-1.0000) (1.5281) 224 The Fisher Effect Table A5.6 Johansen’s Cointegration Test for the Monthly Data Including a Drift Term A. Cointegration LR Test Based on Maximal Eigenvalue of the Stochastic Matrix List of Variables included in the cointegrating vector: it πt Null Alternative Statistic 95% Critical Value r = 0 r = 1 44.8523 15.8700 r <= 1 r = 2 4.2285 9.1600 B. Cointegration LR T est Based on Trace of the Stochastic Matrix Null Alternative Statistic 95% Critical Value r = 0 r = 1 49.0808 20.1800 r <= 1 r = 2 4.2285 9.1600 C. Estimated Cointegrated Vectors in Johansen Estimation (Normalized in Brackets) Vector it πt Intercept 1 0.001308 -0.004551 0.026888 (-1.0000) (3.4785) (-20.550) While the results for the annual data appear to support a long-run relationship between interest rates and inflation with and without an intercept term, results for the monthly data are similar to the quarterly results. Therefore, a likelihood ratio statistic is employed to test for the restriction that the coefficient on the rate of inflation is equal to unity. A joint test for α=0 and β= 1 is also carried out on the equations with an intercept term to examine its significance in the cointegrating regressions. 225 The Fisher Effect Table A5.7 Likelihood Ratio Tests Data Frequency LR (β =1) LR (α = 0; β=1) Annual (no intercept ) 16.18 (with intercept) 13.43 16.58 Quarterly (no intercept) 19.53 (with intercept) 1.72 21.91 Monthly (no intercept) 28.52 (with intercept) 1.45 28.65 The likelihood ratio statistic for β=1 is a chi square distribution with 1 degree of freedom; and for α=0 and β=1 a chi square distribution with 2 degrees of freedom. The results from this test are totally inconsistent with the cointegrating relationships observed above. The restriction that β=1 is rejected in the cointegrating regression equations without an intercept term which yield β coefficients close to unity, while it is accepted in the cointegrating regressions with an intercept which produce implausible results. Moreover, the log-likelihood ratio statistics for the joint restriction that α=0 and β=1 are above the chi squared critical value of 5.99, suggesting the significance of an intercept term. Cointegration can only be applied in the case where individual time series are integrated of the same order. It was pointed out at the beginning that the interest rate and inflationary series were not integrated of the same order and therefore did not require cointegration tests, which probably is the reason for the ambiguity in 226 The Fisher Effect results. It can therefore be concluded that Payne and Ewings' results lack robustness. 227 Exchange Rate Efficiency and Capital Mobility CHAPTER 6 EXCHANGE RATE EFFICIENCY AND CAPITAL MOBILITY 6.1 INTRODUCTION The purpose of this chapter is twofold: to evaluate the degree to which financial deregulation has contributed to enhanced efficiency of the foreign exchange market, and to examine the extent to which the assumption of capital mobility is supported by the data. In fulfilling the first objective, the chapter evaluates the applicability of three alternative hypotheses. The first is the theory of uncovered interest parity (UIP) which states that nominal interest rate differentials of financial assets denominated in different currencies are exactly equal to expected changes in exchange rates. The second is the theory of speculative efficiency, which postulates that forward exchange rates are unbiased predictors of future spot rates. Third is the hypothesis that real interest rates are equalized across countries. An examination of real interest rate linkages is expected to provide a comprehensive perspective on the overall degree of financial market efficiency. In satisfying the second objective, the empirical validity of two tests is examined. The first is the test advanced by Feldstein and Horioka (1980), which examines the correlation between domestic savings and investment. This hypothesis predicts that with greater capital mobility, the link between domestic savings and investment should weaken because the level of investment in a country need not be constrained by the level of domestic savings. 228 Exchange Rate Efficiency and Capital Mobility The second is a measure of international capital mobility, suggested by Shibata and Shintani (1998), based on the correlation between a country’s consumption and net output. Here consumption changes are uncorrelated with predictable changes in net output under conditions of perfect capital mobility. Direct tests of market efficiency on less developed countries based on yield returns are sparse. The lack of empirical work in this area on LDCs can be attributed in part to the lack of data. The absence or undeveloped nature of forward exchange markets in several of these countries has precluded the testing of covered interest parity. Tests of uncovered interest parity have often involved indirect tests that rely on a maintained hypothesis of covered interest parity due to the unobservable nature of future spot rates. In this case, the forward rate should act as an unbiased predictor of the future spot rate—the speculative efficiency condition. The formulation of this hypothesis assumes risk-neutrality and rational expectations. Direct tests of uncovered interest parity have involved testing the interest differential as an unbiased predictor of the rate of depreciation, given the assumptions of rational expectations and risk-neutrality. However, expectations could be irrational and agents risk averse. Therefore, rejection of market efficiency need not necessarily be evidence against the hypotheses of market efficiency, but rational expectations or risk-neutrality. Moreover, interest and exchange rates in developing countries are often subject to administrative controls, giving rise to problems in interpreting results of uncovered int erest p arit y . 229 Exchange Rate Efficiency and Capital Mobility With the advent of floating exchange rates and changing pattern of capital inflows experienced by less developed countries in the recent past,1 the efficiency of foreign exchange markets has become an important consideration. The efficiency of a country’s foreign exchange market has an important bearing on the optimality of its resource allocation. Crucial is the assumption of capital mobility for the satisfaction of market efficiency. Therefore, tests of capital mobility can provide a useful additional source of information in explaining deviations from market efficiency. Contrary to expectations, empirical tests on capital mobility have consistently failed to validate theoretical predictions.2 Feldstein and Horioka found a high correlation between domestic savings and investment for a sample of sixteen OECD countries, which implied that there were significant imperfections in the international capital market. This subsequently gave rise to two further questions: the persistence of a strong positive correlation between savings and investment for developed countries,3 and a higher savings investment correlation for developed rather than developing countries.4 To date, no attempt has been made to test for efficiency of the foreign exchange market in Sri Lanka. The two studies of market integration that have been undertaken for Sri Lanka have limited their investigation to indirect tests of 1 Studies on capital mobility for less developed countries have been undertaken by Haque and Montiel (1991), Dooley and Mathieson (1994), Kuen and Song (1996), among others. 2 Feldstein and Horioka (1980), Feldstein (1983), Dooley, Frankel and Mathieson (1987), Bayoumi(1990), Golub(1990) among others, found a significant positive correlation between savings and investment. 3 See Feldstein (1983) and Penati and Dooley (1984). 4 See Fieleke (1982), Summers (1985), Dooley, Frankel and Mathieson (1987), Wong (1990). 230 Exchange Rate Efficiency and Capital Mobility integration.5 This chapter, therefore, aims at measuring the degree of efficiency of the Sri Lankan foreign exchange market, employing both direct and indirect measures of efficiency. The rest of this chapter is organized as follows. Section 6.2 provides a brief outline of exchange rate behaviour and capital controls in Sri Lanka. Section 6.3 details the tests of market efficiency. Section 6.4 presents the empirical models. Section 6.5 presents the empirical evidence that has gathered in respect of less developed countries. Section 6.6 presents the data. Section 6.7 evaluates the empirical results relating to tests of market efficiency and capital mobility. Section 6.8 considers the implications of the results, and section 6.9 summarizes the conclusions. 6.2 EXCHANGE RATE BEHAVIOUR IN SRI LANKA In November 1977, the Sri Lankan exchange rate was unified; the rupee was devalued by 46% and a managed float was adopted with the aim of making the exchange rate an active policy instrument. The Central Bank commenced quoting the daily rates for six major currencies (the US dollar, Deutsche Mark, Franc, Yen, UK pound and Indian rupee) and only intervened in the market in order to even out undue fluctuations. In 1982, the Central Bank limited its quotation of rates to the intervention currency, the US dollar, and permitted commercial banks to determine the cross-rates for other currencies based on market conditions. This practice was 5 Dooley and Mathieson (1994), and Hague and Montiel (1990) have carried out capital mobility tests for Sri Lanka using the framework developed by Edwards and Khan (1985). 231 Exchange Rate Efficiency and Capital Mobility abandoned in 1990 and the Central Bank commenced announcing daily buying and selling rates of the US dollar against the Sri Lankan rupee for transactions with commercial banks within margins of 2%. Today, Sri Lanka’s exchange rate is largely market-determined, with only minor interventions by the Government. Several measures were taken during the post-liberalization period to encourage foreign direct investment into the country. An investment promotion zone was set up to attract foreign investment on projects producing for export, and a statutory authority, the Greater Colombo Economic Commission (GCEC), now named the Board of Investment (BOI), was set up to develop the infrastructure and manage the zone. The projects located in the zone were entitled to a number of fiscal incentives, including 100% foreign ownership in investment projects, tax holidays and preferential tax rates of 15% for 15–20 years, duty-free import of machinery and raw material and unrestricted repatriation of profits. As at the end of 1996, the total potential capital investment in these projects stood at Rs.631,588 million, of which 83% was foreign investment, while realized investment in BOI industries was Rs.91,622 million, of which foreign investment accounted for 67%.6 On the exchange front, exchange controls on travel and study abroad were relaxed. Commercial banks were permitted to set up Foreign Currency Banking Units (FCBUs) in 1979, with the aim of developing an offshore market in Sri Lanka. In 6 See Central Bank of Sri Lanka (1998),”Economic Progress of Independent Sri Lanka”. 232 Exchange Rate Efficiency and Capital Mobility 1980, commercial banks were permitted to open non-resident foreign currency (NRFC) accounts. Sri Lankans employed abroad and non-nationals could maintain NRFC accounts in designated foreign currencies. From 1991, residents were permitted to operate resident foreign currency (RFC) accounts with a minimum balance equivalent to $US500 in designated currenceis. The capital account, however, was not liberalized even after the reforms of 1977. The attempts at reform slowed down significantly after 1982, due to fiscal imbalances. In 1990, liberalization measures were accelerated once again, focusing on the removal of restrictions on trade and payments to create an environment conducive to private-sector investment. By March 1994, the current account was fully liberalized. After 1990, capital controls were relaxed to some degree with respect to equity investment. By 1992, foreign equity participation of 100% was permitted. These measures were aimed at promoting increased integration of the Sri Lankan economy with the rest of the world. Despite the progress made since 1977 in liberalizing capital account transactions, restrictions continue to apply over capital movements. Controls are more pervasive with respect to capital outflows rather than inflows, reflecting the authorities' concern at risk of capital flight. Restrictions relate in particular to transactions in government securities and debt instruments. Foreign investment in government Treasury bills, bonds and securities is prohibited, while foreign participation is not 233 Exchange Rate Efficiency and Capital Mobility permitted in the government debt market. Private foreign capital has access only to certain specified types of investment.7 There also remain restrictions on access to foreign funds by Sri Lankan nationals. Local enterprises other than those in free trade zones do not have unlimited access to foreign capital. Similarly, Sri Lankan commercial banks do not have unlimited access to international capital and money markets. From 1995, however, they were permitted to obtain foreign loans of up to 5% of their capital and reserves. This was increased to 15% in 1997 with the approval to grant foreign currency loans to non- BOI exporters.8 Commercial banks are also permitted to grant foreign currency loans from their foreign currency banking units to non-BOI exporters for importation of inputs required for export. Controls also remain on long-term capital movement in Sri Lanka, particularly with respect to foreign ownership of real estate. Therefore, capital controls still remain a pervasive feature of Sri Lanka’s financial system. The authorities have adopted a gradual approach to dismantling restrictions with respect to capital account transactions for fear of undermining macroeconomic stability. Needless to say, capital controls are frequently cited as causing deviations from interest parity. Nonetheless, as the elimination of capital controls continues, the impact of these developments on the degree of efficiency of the foreign exchange market has become an important consideration. 7 These include banking, finance and plantations. Foreign investment is not permitted in sectors such as money-lending, pawn-broking and fishing. 8 See Central Bank of Sri Lanka (1998), “Economic Progress of Independent Sri Lanka”. 234 Exchange Rate Efficiency and Capital Mobility 6.3 TESTS OF MARKET EFFICIENCY A number of different approaches have been suggested for measuring foreign exchange market efficiency. These tests can be broadly classified into two groups: tests based on the convergence of yield rates, which include covered interest parity, uncovered interest parity, and real interest parity; and tests based on the magnitude of capital flows, such as savings-investment and consumption-income correlations. Given the purpose of this chapter, the preferred measure of efficiency based on yield returns is the narrower concept of covered interest parity (CIP), which emphasises the integration of financial markets. The absence of forward rate data for the period prior to 1996 for Sri Lanka has led to the selection of uncovered interest parity (UIP). UIP is the proposition that nominal interest rate differentials of assets denominated in different currencies is exactly equal to the expected rate of change in the exchange rate. Tests of UIP can be found in Cumby and Obstfeld (1981, 1984), Taylor (1987b), and MacDonald and Taylor (1989). Given CIP which postulates that nominal interest rate differentials between assets denominated in different currencies is exactly equal to the corresponding forward discount/premia, UIP is consistent with the hypothesis that the forward rate is an optimal predictor of the future spot rate—the speculative efficiency condition. This hypothesis maintains that covered interest parity and uncovered interest parity coincide when the forward rate is exactly equal to the expected spot rate. This will only hold under the assumptions of rational expectations and risk-neutrality. Due to the unobservable nature of future spot rates, tests of UIP have often involved indirect tests 235 Exchange Rate Efficiency and Capital Mobility incorporating CIP into the maintained hypothesis. This approach is employed by Gweke and Feige (1979), Hansen and Hodrick (1980), Fama (1984), among others. An alternative approach to measuring market efficiency is provided by the less stringent measures of capital mobility. Feldstein and Horioka (1980) put forward a test of capital mobility based on the correlation between a country’s level of domestic savings and investment. They argue that, with greater capital mobility, the level of investment in a country need not be constrained by the level of domestic savings, as any discrepancy can be financed by foreign savings. It follows from this that the correlation between domestic savings and investment is zero with perfect capital mobility, and that savings equals investment in the case of capital immobility. Using data from 1960–1970, Feldstein and Horioka ran a regression of the investment ratio on the savings ratio for a cross section of 16 OECD countries. The regression was also run with the sample period divided into three sub-samples. The coefficient on savings was in the range of 0.94 and 0.83 for the 4 sample periods examined, pointing to the conclusive rejection of perfect capital mobility. Contrary to theoretical predictions, data revealed almost a one-to-one increase in the domestic savings ratio in response to an increase in the domestic investment ratio. Feldstein (1983), extending the sample period to cover the 1974–1979 period, found support for the previous findings of Feldstein and Horioka, with the coefficients on the savings ratio ranging from 0.78 to 0.99 for all the sample periods studied. Their findings were subsequently confirmed by many others—Dooley, Frankel and 236 Exchange Rate Efficiency and Capital Mobility Mathieson (1987), Penati and Dooley (1984), Frankel (1986), Bayoumi (1990), Golub(1990), and Kim (1993).9 Vredin and Warne (1991) and Krol (1996), however, found some support for the theory. The savings investment relationship was examined from a different perspective by Sachs (1981), who defined the difference between savings(S) and investment(I) as the current account balance. According to him, investment had a negative impact on the current account balance under conditions of capital mobility because higher domestic investment would lead to greater international borrowing and hence a higher current account deficit. Regressing the current account balance (CA) on the investment ratio, Sachs found a significant negative relationship between the current account and investment ratios for a cross-section of 14 OECD countries for the 1960–1979 period. In regressions of ∆(CA/GNP) on ∆(I/GNP) and ∆(S/GNP), respectively, for the period 1968–1979, the regression coefficient on the change in investment rate was –0.61, while the estimated coefficient on the change in the savings rate was –0.34, establishing a significant negative correlation between investment and the current account balance. 10 These findings were in contrast to those of Feldstein and Horioka. 9 This puzzle has been explained by way of institutional and legal restrictions (Feldstein and Horioka 1980), population growth, income growth, terms of trade shocks (Obstfeld 1986, Summers 1988); non-traded consumption goods, immobile factors of production (Frankel 1986, Murphy 1986, Wong 1990); government policy (Summers 1988, Bayoumi 1990). 10 See Sachs (1981, Table 14, p.250). 237 Exchange Rate Efficiency and Capital Mobility M ore recently, Shibata and Shintani (1998) put forward a measure of capital mobility based on the correlation between a country’s consumption and net output. They employ the permanent income model of Campbell and Mankiw (1989, 1990, 1991). The intuition underlying this model is that under conditions of perfect capital mobility, changes in consumption should be uncorrelated with predictable changes in net output. Estimating the model for a sample of 11 OECD countries, they concluded that capital mobility appeared to be greater in countries that had previously maintained capital controls than in those which had not. Tests of market efficiency are inevitably joint tests of risk-neutrality and rational expectations. In view of this problem in interpreting results of market efficiency, tests of capital mobility can provide a useful additional source of information in explaining deviations from market efficiency. Tests of capital mobility, however, will only be valid in the presence of real interest rate convergence, as savings and investment in an economy depend upon the real rather than nominal rate of interest. The equality of real interest rates suggests that domestic monetary authorities have no control over their real rates in pursuing domestic stabilization policies. Therefore, in order to shed further light on tests of capital mobility, the hypothesis that real interest rates are equal across countries is examined. Moreover, the equality of real interest rates is also closely connected to the basic parity conditions, as will be seen in the following section. 238 Exchange Rate Efficiency and Capital Mobility 6.4 EMPIRICAL MODELS 6.4.1 Uncovered Interest Parity The theory of covered interest parity (CIP) postulates that nominal interest rate differentials are exactly offset by the corresponding forward discount or premia ensuring the absence of profitable arbitrage opportunities at equilibrium. This relationship can be expressed by: Ft/St = (1+ it ) / (1+ it*) (1) where Ft = the forward exchange rate St = spot exchange rate it = the domestic interest rate and it* = the foreign interest rate Taking the natural logarithm of both sides gives: log Ft – log S t = log (1+ it) – log (1+ it*) (2) The logarithm of (1 + i) is approximately equal to i. Hence, the right-hand side of equation (2) is approximately equal to it – it*. If ft = log F t and st = log St, equation (2) may be written as: ft –st = it - it* (3) A more restrictive version of interest parity is given by the uncovered interest parity (UIP) condition, which states that nominal interest rate differentials of financial assets denominated in different currencies, is exactly equal to the expected change in exchange rate. That is: Ets t+1 – st = it –it* (4) 239 Exchange Rate Efficiency and Capital Mobility where Ets t+1 = expectations formed at time t for the period t+1 Rational expectations imply that the spot rate realized at time t+1 will differ from the expected spot rate by a random error term with zero mean, s t+1 = Et st+1 + v t+1 (5) The expectational error v t+1 = s t+1 – Et s t+1 is uncorrelated with information known in period t at the time of expectation formation. Replacing Et s t+1 in equation (4) with s t+1 – v t+1 and shifting v t+1 to the right-hand side yields: s t+1 – st = ∆s t+1 = α0 +α1 (i –i*) t + v t+1 (6) 11 Direct tests of UIP involve testing for α0=0 and α1=1. 6.4.2 Speculative Efficiency Given CIP, equation (4) implies that forward rates are unbiased predictors of future spot rates, ft = Ets t+1 (7) This is termed speculative efficiency. An alternative approach to evaluating market efficiency is provided by the speculative efficiency condition. 11 Equation (6) has been tested extensively using different econometric techniques. Cumby and Obstfeld (1981), Loopesko (1984)—error orthogonality tests; Taylor (1987b)—vector autoregression analysis; Karfakis and Parikh (1994), Bhatti and Moosa (1995)—cointegration analysis. The majority of findings, however, point to the rejection of UIP. Support for the model has been found with the inclusion of other variables: transaction costs (Frenkel and Levich 1975, 1977); risk premia (Fama 1984, Mark 1985, Cumby 1988); exchange risk and political risk (Aliber 1973, Dooley and Isard 1980). 240 Exchange Rate Efficiency and Capital Mobility The speculative efficiency hypothesis states that, in an informationally-efficient market, the forward rate should be an unbiased predictor of the future spot rate. Following Tease (1988) this condition is expressed by: Ets t+n = ft (8) where ft = forward rate in period t for payment at period t+n Ets t+n = the expected spot rate in period t+n conditional on information at time t The conditional forecast error is given by: εt+n = s t+n – f t (9) Given (8) and (9) the expected forecast error is: Etεt+n = Et[st+n – ft] = 0 (10) The speculative efficiency condition implies that the expected value of the conditional forecast error is zero. Provided εt+n has a zero mean, α in equation (11) should be insignificantly different from zero. ε t+n = α + υt (11) εt+n should be uncorrelated with past information. Any correlation between information in period t and εt+n would indicate that the forward rate is not an unbiased predictor of the future spot rate and hence be inconsistent with market efficiency. If the market is efficient and agents are risk-neutral, the forward rate should be an accurate predictor of the future spot rate. A test of speculative efficiency, therefore, involves carrying out a regression of the form, st+n = α + β f t + υ t (12) 241 Exchange Rate Efficiency and Capital Mobility The restrictions imposed by the speculative efficiency hypothesis are that α=0 and β=1.12 A test of this hypothesis can alternatively be carried out by running the following regression, st+n - st = α + β ( f t – st) + υ t (13) for α=0 and β=1. 6.4.3 S avings-Investment Correlations Feldstein and Horioka (1980) asserted that, with perfect capital mobility, there should be no relation between a country’s domestic savings rate and domestic rate of investment. Testing this hypothesis involves running a regression of the following form: (I/Y)t = α + β (S/Y)t (14) where I = gross domestic investment S = gross domestic savings Y = gross domestic product With perfect capital mobility, there should be no systematic relationship between domestic savings and investment. Therefore, a test of this model entails testing for a zero coefficient on the savings ratio. In the case of complete capital immobility, the value of β should take on a value of unity. 12 See Gweke and Feige (1979), Hansen and Hodrick (1980), Hakkio (1981), Fama (1984), Levy and Nobay (1986) for tests of this form. 242 Exchange Rate Efficiency and Capital Mobility A different approach to the savings-investment correlation was taken by Sachs (1981), who defined the difference between savings and investment as the current account balance. This involves regressing the current account balance on the rate of investment and testing for a negative coefficient on the investment ratio. (CA/Y)t = a + b (I/Y)t (15) 6.4.4 Consumption-Net Output Correlations Employing the permanent income approach of Campbell and Mankiw (1989, 1990, 1991) Shibata and Shintani put forward a test of capital mobility based on consumption–net output correlations. Their model is restated below. Assuming a world interest rate of r, a country’s budget constraint is given by, A t+1 = (1+ r) At + Y t – Ct – I t –G t = (1 + r) A t + X t – Ct (16) where Y = gross domestic product, I = investment, C = private consumption, G = government expenditure, A = foreign asset holdings and X = Y – I – G = the country’s net output. From the national income accounting identity it follows that CAt = rA t + X t – C t (17) where CA = the current account. The two polar cases of perfect international capital mobility and financial autarky have been considered. Given a quadratic utility function and equality between the 243 Exchange Rate Efficiency and Capital Mobility consumer discount rate and world interest rate, optimal consumption in the case of perfect capital mobility is given by: p i C t = r { A t + ( 1/ 1+ r) Σ ( 1/ 1+ r) Et X t+i } (18) Taking the first differences of equation (18) yields, p i ∆C t = r / 1+ r Σ (1/ 1 + r) ( Et – E t-1 ) X t + I (19) where ( Et – E t-1 ) X t + i denotes changes in expectations between periods t-1 and t. If expectations were rational, p ∆C t = e t (20) where et is a rational forecast error orthogonal to information available at time t -1. From equations (16), (17) and (18), the optimal current account can be expressed as, i CAt = - (r /1+r) Σ (1/1+r) Et (X t+i – Xt) (21) i = - Σ (1 / 1+ r) Et (∆ X t+ i) (22) Thus, both the current account and consumption are determined by future expectations of net output. In the case of financial autarky, a country’s consumption is constrained by its current net output, a Ct = X t (23) This implies that the trade balance TBt = CAt – rAt (24) is zero so that domestic saving is equal to domestic investment. 244 Exchange Rate Efficiency and Capital Mobility Aggregate consumption in a case of capital mobility between these two polar cases is given by, p a p Ct = (1- λ) C t + λ Ct = (1 – λ) C t + λ Xt (25) where λ represents a coefficient of capital mobility. It takes on a value of zero in the case of perfect capital mobility and a value of one in the case of capital immobility. Taking the first differences of equation (25) yields, p a ∆ Ct = (1- λ) ∆ C t + λ ∆ Ct = (1 – λ) et + λ ∆ Xt (26) Shibata and Shitani estimate equation (25) and test for λ =0. Shibata and Shintani assume that the real rate of interest is constant. Michener (1984), however, points out that consumption could appear sensitive to income due to variation in real interest rates through time, despite the inter-temporal optimization by agents in the absence of borrowing constraints. The study, therefore, also investigates the model permitting for changes in the real rate of interest. Relaxing the assumption of a constant real interest rate the model can now be written as:13 ∆ Ct = (1 – λ) [ et + δrt ] +λ ∆ Xt (27) The existence of a statistically significant real interest rate could imply that the ex ante real interest rate is associated with the growth rate of consumption. 13 See Campbell and Mankiw (1990, 1991). 245 Exchange Rate Efficiency and Capital Mobility 6.4.5 Real Interest Rate Linkages Following Mishkin (1984), the real rate of interest for a country is given by, r`t = it – Et-1πt (28) where it = the nominal rate of interest πt = the rate of inflation r`t = the real rate of interest E = expectations operator conditional on information at time t-1 The real rate defined above, which is more precisely the ex ante real rate, is unobservable and, therefore, it is necessary to employ the ex post real rate which is defined as, rt = it – πt = r`t – (πt - Et-1πt) = r`t - et (29) where rt = the one period ex post real rate at time t πt = the actual rate of inflation et = πt - Et-1πt = the forecast error of inflation A critical assumption underlying this model is the rationality of expectations, which implies that the forecast error of inflation, et, is unforecastable. Hence, r`t = Et-1 rt (30) The equality of real interest rates across countries implies that, rt – r`t = Et-1(rt – r*t) = 0 (31) 246 Exchange Rate Efficiency and Capital Mobility where the asterisk denotes a foreign variable. Equation (31) suggests that the ex post real rate differential between countries is unforecastable given any information at time t-1. A test of real interest rate equality can be carried out by running the following regression, r = a + br* (32) where, r = real interest rate Sri Lanka r*= real interest rate of foreign country In (32) a =0 and b=1 implies complete real interest rate convergence, while b=0 denotes absence of interest rate linkage. Partial interest rate linkage is implied by 0< b<1. Mishkin (1984) has further shown that the equality of real interest rates is closely related to the interest parity and speculative efficiency conditions. The covered interest parity condition was given by equation (3) as, ft –st = it - it* The ex ante version of purchasing power parity (PPP) is expressed by Et-1(πt – π*t – (st – st-1)) = 0 and the speculative efficiency condition by ft = Et-1st Combining these three equations gives the UIP condition, 247 Exchange Rate Efficiency and Capital Mobility Et-1 (it – i*t – (st – s t-1 )) = 0 Subtracting the PPP condition from the above equation yields, Et-1 (rt – r*t) = 0 = r`t – r`*t which is the condition for real interest rate equality. Hence deviations from any of the above conditions could lead to the rejection of the hypothesis of the equality of real interest rates. 6.5 EMPIRICAL WORK FOR DEVELOPING COUNTRIES This section examines the empirical evidence that has gathered in respect of less developed countries. A large number of less developed countries are characterized by the absence of forward exchange and eurocurrency markets. For many countries where liberalization is a new policy option, controls have obstructed the progress of extensive forward exchange and offshore deposit markets. Therefore, for many of these countries, tests of foreign exchange market efficiency have involved the less stringent measures of capital mobility. The review of the empirical work is hence divided into two sections. First, the literature on tests of market efficiency is reviewed, and second, the literature on tests of capital mobility is surveyed. 6.5.1 Tests of Market Efficiency Tests of market efficiency based on the integration of money and capital markets can be found in Chinn and Frenkel (1994), Kuen and Song (1996) and Woo and Hiramya (1996). Chinn and Frenkel (1994), investigating covered interest rate differentials for a sample of countries around the Pacific rim for the 1982.9–1992.3 period, found 248 Exchange Rate Efficiency and Capital Mobility that the slope coefficient for all countries, including Hong Kong, Malaysia and Singapore, to be significantly different from zero, suggesting substantial financial market integration. Including Korea and Taiwan in the sample of countries, a test of uncovered interest parity indicated that the null hypothesis of β=1 was rejected in every case. They concluded that, despite evidence of a decline in interest rate differentials, the region had still to achieve interest rate convergence. In an attempt to find out how the linkage in Singapore had changed over time, Kuen and Song (1996) used the band and variability of deviations from covered and uncovered interest parities. Consistent with the findings of Chinn and Frenkel, deviations from CIP were found to be small, except for a short period in the 1980s. The results for UIP, in contrast, indicated that deviations from interest parity were decreasing and less variable over time. By 1990, it was observed that the deviations were negligible, suggest ing almost p erfect capit al mobility. Woo and Hiramya (1996), investigating monetary autonomy in the presence of capital flows for Indonesia, Malaysia and Singapore, found, on the contrary, that monetary autonomy existed despite the open capital accounts maintained by these countries. The negative coefficients on the interest rate differential for Japan and Malaysia over the 1973.3–1993.3 and 1973.3–1992.3 periods, respectively, rejected the hypothesis of the interest rate differential being an unbiased predictor of future exchange rate changes. They attributed the monetary autonomy in these countries to risk premia created by the lack of access to foreign funds and government intervention in exchange rate activities. The positive coefficient for Indonesia was attributed to the 249 Exchange Rate Efficiency and Capital Mobility predictable depreciation of the exchange rate by approximately 3%–5% each year by the Government. Incorporating covered interest parity into the maintained hypothesis, an indirect test of uncovered interest parity in the form of the speculative efficiency condition was carried out for Singapore by Blejer and Khan (1983). Using monthly data from 1976.6 –1981.3, they found that the coefficient on the forward rate was not significantly different from unity, implying that the forward rate contained all the information about the future spot rate. It was observed that the inclusion of lagged variables of the forward rate did not significantly alter the results, verifying efficiency of the forward exchange market. The results appeared to support the evidence of Kuen and Song. A large literature on developing countries is also devoted to evaluating the economic effects of capital controls on deviations from interest parity. An attempt to measure the effectiveness of controls was made by Dooley and Isard (1980). Their study was based on Aliber’s (1973) reinterpretation of the interest parity theorem. Aliber identified between exchange risk and political risk as factors giving rise to deviations from interest parity. Deviations from interest parity were attributed to exchange risk when assets were denominated in different currencies, and political risk when assets were issued in different countries under different legal jurisdictions.14 Studies based upon the impact of controls on deviations from interest parity can be found in 14 See Dooley and Isard (1980). 250 Exchange Rate Efficiency and Capital Mobility Frankel and MacArthur (1988), Melvin and Schlagenhauf (1985), and Speigel (1990). In a study of the impact of political and currency premia on real interest rate differentials for 25 countries, including developed and developing countries (with and without capital controls), for the 1982–1987 period, Frankel and MacArthur (1988) found that deviations from covered interest parity were much smaller in countries with open capital accounts (Canada, Germany, Hong Kong, Singapore) than in those with restrictions (France, Mexico, Baharain, Greece and South Africa). They attributed this deviation to expected real depreciation for the countries with open capital markets, and a combination of both factors for the countries with less open capital accounts. Building upon the framework of Dooley and Isard (1980), Spiegel (1990) examined the impact of capital controls on deviations from interest parity. As opposed to the positive impact of capital controls on deviations from interest parity, as predicted by Dooley and Isard, Speigel observed that the Mexican capital control policy had a negative but temporary impact on deviations from interest parity. The negative impact was found to abate in approximately six months. The implications for policy were mixed with results, suggesting that the Mexican capital control policy was successful for a limited period of time in lowering deviations from interest parity. Using a dynamic factor analysis model to estimate a country risk index for Mexico for the 1978–1984 period, Melvin and Schlagenhauf (1985) found the country risk index to be positively associated with the Mexican 251 Exchange Rate Efficiency and Capital Mobility capital controls and policy changes of 1982, consistent with the predictions of Dooley and Isard. Studies have also been carried out on examining the impact of other factors such as transaction costs, taxes and other practices on deviations from interest parity. Lizondo (1983), using the Frenkel Levich (1975) framework analyzed, the deviations from covered interest parity between dollar-denominated assets and peso- denominated assets, and the extent to which transaction costs accounted for those deviations. He found that transaction costs accounted only for a very small percentage of the deviations from parity that could be explained by way of regulations on forward market operations and taxes on foreign exchange gains. Employing the Frenkel and Levich (1975) framework, Khor and Rojas-Suarez (1991) examined the link between interest rate and exchange rate expectations, and perceptions on default risks associated with Mexico’s external debt. While the deviations from CIP were small, the results for UIP were less conclusive, pointing to a lack of complete credibility (peso problem) in exchange rate policy. 6.5.2 Tests of Capital Mobility The studies reviewed above emphasise the need for well-developed forward markets to control for expected exchange rate changes. A number of developing countries, however, are characterized by the complete absence or lack of well-developed forward markets. In such instances, a proxy must be used for expected exchange rate 252 Exchange Rate Efficiency and Capital Mobility changes. This has led to the development of alternative models based on UIP, and tests based on the magnitude of capital flows. Tests of the Feldstein-Horioka model have been undertaken by Dooley, Frankel and Mathieson (1987) and Frankel and MacArthur (1988) on data from a number of developed and developing economies, and by Wong (1990) on data from a sample of developing economies including Sri Lanka. While the studies by Dooley et al. and Frankel and MacArthur point to a strong association between domestic savings and investment for the economies with relatively open capital accounts, they find only a weak correlation between the two for developing economies that rely heavily on foreign aid to finance their current account deficits. The study by Wong tends to find some support for the Feldstein-Horioka predictions. He finds however, that the saving–investment relationship is significantly affected by the non-traded goods sector. He argues that, even in the event of perfect capital mobility, if domestic residents consume traded as well as non-traded goods, an increase in the consumption of non-traded goods could lead to an increase in the production and therefore an increase in the investment in the non-traded goods sector. The existence of a systematic relationship between savings and investment, therefore, could be explained by way of the existence of a non-traded goods sector, according to Wong. The studies of Edwards and Khan (1985), Haque and Montiel (1990, 1991), and Dooley and Mathieson (1994) use a variant of the UIP theorem to test for capital mobility. Edwards and Khan (1985) proposed a test for interest rate determination in 253 Exchange Rate Efficiency and Capital Mobility LDCs wherein the currently existing interest rate represented a weighted average of the UIP interest rate and the domestic market clearing rate in a completely closed economy. Using data for Singapore and Colombia, they demonstrated the relative importance of interest parity and domestic monetary factors in determining national interest rates for Colombia (which still maintains capital controls), while for Singapore (which is a relatively open economy), only open economy factors appeared to matter. Building upon this framework, Haque and Montiel (1991) estimated a model for 15 LDCs, of which Sri Lanka was one. They found that of the 15 countries studied, 10 displayed high capital mobility over the 1969–1987 period. Sri Lanka fell into this category, with a coefficient of 0.63 for the index of capital mobility. Four displayed partial capital mobility, and one, India, capital immobility. From this, they concluded that the ability of these countries to pursue an independent monetary policy was limited, despite the maintenance of capital controls. Dooley and Mathieson (1994) used a modified version of the Haque-Montiel model to investigate the degree of international capital mobility for a sample of developing countries, including Sri Lanka, Indonesia, Korea, Malaysia, Myanmar, the Philippines and Thailand. Consistent with the findings of Haque and Montiel, the coefficient of capital mobility for Sri Lanka indicated a high degree of openness. A high level of capital mobility was observed for all countries except Myanmar for the 1965–1990 period. This line of research, however, does not take into account the convergence of yield rates, which are more rigorous tests of market integration. Moreover, these tests rely 254 Exchange Rate Efficiency and Capital Mobility on the money demand function to estimate the unobserved domestic market clearing rate, which could lead to inaccuracies not encountered with tests of capital mobility such as the savings investment correlations or consumption income correlations. Therefore, this study improves upon previous studies of market integration on Sri Lanka by employing both tests based on yield convergence and tests based on the magnitude of capital flows. 6.6 DATA Exchange Rates Uncovered interest differentials are examined for six countries, namely the US, UK, India, Japan, Germany and France, with Sri Lanka as the “home” country. The countries have been selected on the basis of the currencies to which the Sri Lanka rupee is linked. Tests are carried out employing two data frequencies, annual and monthly, covering the periods 1959–1998 and 1990–1998. All exchange rates are expressed in terms of Sri Lanka rupees per unit of foreign currency. The exchange rate data for this hypothesis is obtained from the Annual Reports and monthly bulletins of the Central Bank of Sri Lanka. Monthly spot exchange rates for the US dollar are announced in the morning of every working day, and middle rates (the average between the buying and selling rates) are given for other currencies. The speculative efficiency hypothesis is tested utilizing weekly data for 1, 3, and 6- month forward rates, together with the corresponding spot rates for the Sri Lanka/US dollar exchange rates. The observations cover the period 15 February 1996 to 31 255 Exchange Rate Efficiency and Capital Mobility December 1998. Unfortunately, forward rate data, covering a period consistent with the data used for tests of uncovered interest parity, are not available.15 The forward rate data has been obtained directly from the Central Bank of Sri Lanka. Interest Rates Country Source Description Sri Lanka Annual Reports CBSL; 3- month Treasury bill rate Monthly Bulletins CBSL U.S. IFS 3-month Treasury bill rate U.K. IFS 91-day Treasury bill rate Japan IFS 2-month private bill rate; Government bond yield India IFS Inter-bank rate Germany IFS 12-month Treasury bill rate France IFS Money market rate. (Average rates for day to day loans against private bills). CBSL = Central Bank of Sri Lanka IFS = International Financial Statistics. 15 A forward exchange market has been in operation in Sri Lanka since 1983; however, the volume of trading has been low–usually under US dollars 10 million per day. Its activities increased somewhat after 1989, albeit not to any significant size. Most of the data on forward rates was lost in a bomb attack on the Central Bank of Sri Lanka in 1996. Therefore, the data available has been used to test for forward market efficiency. It should be noted, however, that these rates are quoted rates and not transacted rates, so that actual transactions may not have taken place at these rates. 256 Exchange Rate Efficiency and Capital Mobility Due to limitations in data availability, the assets employed are not strictly comparable.16 They vary in terms of risk and maturity. The sample periods differ according to data availability. Savings, Investment, Gross Domestic product, consumption and government expenditure data are taken from the Central Bank Reports of Sri Lanka. The data cover the 1959–1998 period. As in the previous chapters, the data is first tested for non-stationarity. All the time series employed are tested for unit roots. The test results are reported in Tables 6.1 and 6.2. 16 See Chinn and Frankel (1994), Glick and Hutchison (1990), Woo and Hirayama (1996), Frankel and MacArthur (1988) for use of similar data series. 257 Exchange Rate Efficiency and Capital Mobility Table 6.1 Dickey-Fuller Test for Unit Roots for the Levels of the S eries No Trend Trend Variable ADF LM ADF LM Exchange Rates: U.S. -0.24 4.01 -2.31 8.05 U.K. -0.11 12.92 -2.20 8.54 India -1.14 2.82 -1.59 4.41 Japan -0. 72 5. 51 -1. 75 6. 79 France -0.96 5.19 -2.75 6.23 Germany -0.58 1.72 -1.61 5.73 Nominal Rates: Sri Lanka -1.45 4.72 -2.26 4.35 U.S -1.90 11.04 -1.74 11.78 U.K -2.22 10.96 -1.94 11.63 India -2.84* 8.39 -3.82** 9.78 Japan (Govt. bond yield) -0.36 8.25 -1.99 10.92 Japan (Private bond yield) -0.60 9.90 -1.64 9.48 France -1.21 7.27 -0.62 8.35 Germany Interest Rate Differential: S.L–U.S -1.37 3.73 -2.33 1.17 S . L – U. K -1. 67 5. 34 -2. 92 4. 59 S.L–India -2.79* 8.17 -4.05** 8.61 S.L–Japan -1.79 5.59 -2.16 10.68 S.L–France -1.15 2.89 -2.38 2.19 S.L–Germany -2.70* 9.03 -2.42 10.54 Real Interest Rates: Sri Lanka -3.84*** 10.14 -4.08*** 10.86 U.S -2.22 10.83 -2.39 10.03 U.K -2.38 4.90 -2.61 2.08 India -3.83*** 6.76 -5.46*** 5.68 Japan (Govt. bond yield) -2.48 5.72 -2.71 4.18 Japan (Private bond yield) -0.48 3.71 -3.36* 9.06 France -1.59 9.06 -2.06 8.68 Germany -3.11** 6.27 -3.01 6.81 Spot Rate -1.52 8.48 -1.33 5.87 Forward Rate (1 month) -1.62 8.66 -1.68 4.73 Forward Rate ( 3 months) -1.67 3.13 -0.98 4.95 F orward R ate ( 6 m onths) -1. 69 2. 73 -2. 03 7. 04 I/Y -1.43 3.74 -2.29 5.99 S/Y -1.86 6.98 -3.62** 4.48 CA/Y -3.16** 4.60 -3.30* 5.16 C -1.60 10.56 -1.16 9.54 X -0.84 2.13 -2.39 3.92 258 Exchange Rate Efficiency and Capital Mobility Table 6.1: Dickey-Fuller Test for Unit Roots for the Levels of the Series (cont) Note: T he lag length for the ADF regressions has been selected to ensure white noise residuals. A sixth order autoregressive model is used. The χ2 statistic for 6th order serial correlation in the residuals with 6 degrees of freedom is 12.59. Significance levels with trend: 1%, -4.07 : 5%, -3.46 : 10% -3.16; without trend: 1%, -3.51 : 5%, - 2.90, 10% -2.58 (Davidson and MacKinnon). *, **, *** significant at the 10%, 5% and 1% levels respectively. With the exception of the real interest rate for Sri Lanka and India, all data series appear to be non-stationary in levels. Therefore, ADF tests are carried out on the first differences of the data series that display a unit root. Table 6.2 reports the results. 259 Exchange Rate Efficiency and Capital Mobility Table 6.2 Dickey-Fuller Test for Unit Roots for First Differences of the S eries Variable ADF LM Exchange Rates: U.S -5.91*** 0.85 U.K -5.10*** 4.55 India -5.54*** 1.97 Japan -4.58*** 2.44 France -7.31*** 3.66 Germany -1.92 5.95 Nominal Rates: Sri Lanka -6.67*** 5.03 U.S -4.53*** 9.08 U.K -4.83*** 10.10 India -6.22*** 12.04 Japan (Govt. bond yield) -3.70*** 5.66 Japan (Private bond yield) -2.45 6.68 France -4.29*** 3.97 Germany -3.36*** 9.78 Interest Rate Differential: S.L –U.S -6.97*** 2.19 S.L –U.K -6.34*** 7.34 S.L –India -5.79*** 10.25 S.L–Japan -4.03** 9.94 S.L–France -7.14*** 2.38 S.L–Germany -4.07*** 10.12 Real Interest Rates: Sri Lanka -7.63*** 6.01 U.S -6.06*** 9.41 U.K -8.08*** 1.71 Japan (Govt. bond yield) -5.83*** 5.30 Japan (Private bond yield) -4.08*** 6.51 France -6.00*** 8.24 Germany -5.42*** 5.05 Spot Rate -8.34*** 2.47 Forward Rate (1 month) -12.03*** 6.09 Forward Rate ( 3 months) -15.73*** 3.44 Forward Rate ( 6 months) -10.61*** 2.63 I/Y -6.08*** 4.74 S/Y -6.87*** 10.46 CA/Y -3.57*** 11.20 C -3.26** 10.89 X -5.17*** 5.69 Note: T he lag length for the ADF regressions has been selected to ensure white noise residuals. A sixth order autoregressive model is used. The χ2 statistic for 6th order serial correlation in the residuals with 6 degrees of freedom is 12.59. Significance levels without trend : 1%, -3.51 : 5%, -2.90, 10% -2.58 (Davidson and MacKinnon). *, **, *** significant at the 10%, 5% and 1% levels respectively. 260 Exchange Rate Efficiency and Capital Mobility With the exception of the exchange rate for Germany and the private bond yield for Japan, all series appear to be stationary in the first differences. While the results for the interest rate differentials indicate that the series are I(1), the change in exchange rate (st+1 – st) rejects the presence of a unit root in support of a stationary alternative. These results suggest that UIP is very unlikely to hold. However, given the limitations of tests for unit roots, results for UIP are still presented in the following section. 6.7 RESULTS 6.7.1 Uncovered Interest Parity Given that all exchange rates are measured relative to the Sri Lankan rupee, the exchange rates are likely to be contemporaneously correlated across currencies. Such contemporaneous correlation across regressions implies that the OLS coefficients might not be efficient. Therefore, in addition to OLS, following the work of Flood and Rose (1996) and Fama (1984), Zellner’s (1962) Seemingly Unrelated Regression (SUR) procedure is employed to improve the precision of the coefficient estimates. The SUR technique involves the application of generalized least squares estimation to a system of seemingly unrelated equations. The equations are related via the non-zero covariences associated with error terms across different equations at a given point in time (contemporaneous correlation). A problem that arises when testing for UIP is that exchange rate expectations are unobservable. This problem has been circumvented by assuming rational 261 Exchange Rate Efficiency and Capital Mobility expectations. Table 6.3 reports regression results for equation (6) for the 1959–1998 period. The sample is also divided into two sub-periods, 1959–1976 and 1977–1998, to distinguish between the pre- and post-float periods respectively. OLS and SUR Estimates Table 6.3 Uncove re d Inte re st Pari ty wi th Rati onal Expectations—OLS and SUR Regressions 1959– 1998 (s t+1 – st) = α + β (i- i*)t + v Country O LS SUR α β R2 DW α β R2 DW US 0.07 0.00 0.00 1.9 0.06 0.003 0.01 1.9 (0.02) (0.003) (0.02) (0.003) UK 0.05 0.004 0.02 1.7 0.05 0.002 0.02 1.7 (0.02) (0.004) (0.02) (0.003) India 0.01 -0.00 0.00 1.8 0.01 0.0001 -0.00 1.8 (0.03) (0.01 (0.03) (0.0007) France 0.06 0.002 0.01 2.4 0.07 -0.0001 -0.00 2.3 (0.03) (0.005) (0.03) (0.005) (Standard errors reported in parenthesis) Wald Tests 1. All β equal χ2(3) = 0.70(0.874) 2. All β=0 χ2(4) = 1.19(0.75) 262 Exchange Rate Efficiency and Capital Mobility The estimated β coefficients are significantly below their hypothesized value of unity, with the coefficients on β considerably smaller under the SUR technique. For India and France, the coefficients are also incorrectly signed, suggesting perhaps of the omission of time-varying risk premia from the regression equations. A Wald test of the hypothesis that all β=0 cannot be rejected at the 1% level of significance. This is not surprising, in view of the fact that the coefficients on the interest differentials are not significantly different from zero. The intercept terms, however, are not significantly different from zero, consistent with theory. The R2 for the regression equations are in the range of –0.00 and 0.02, suggesting no explanatory power in the regression equations, while the DW statistics indicate the absence of serial correlation in the disturbance terms. Results for the pre- and post-float periods are reported in Tables 6.4 and 6.5 respectively. 263 Exchange Rate Efficiency and Capital Mobility Table 6.4 Pre Float Results 1959–1976: OLS and SUR Estimates (s t+1 – st) = α + β (i- i*)t + v Country O LS SUR α β R2 DW α β R2 DW US 0.02 -0.01 0.03 2.0 0.05 0.02 0.07 1.7 (0.02) (0.016) (0.02) (0.02) UK 0.02 0.006 0.21 1.6 0.01 0.002 0.13 1.7 (0.009) (0.003) (0.008) (0.002) India -0.007 -0.002 0.002 1.7 -0.02 -0.009 -0.02 1.8 (0.03) (0.01) (0.03) (0.01) France -0.06 -0.03 0.12 2.2 -0.08 -0.04 0.12 2.2 (0.07) (0.02) (0.04) (0.01) (Standard errors reported in parenthesis) Wald Tests 1. All β equal χ2(3) = 12.06(0.007) 2. All β=0 χ2(4) = 25.15(0.00) 264 Exchange Rate Efficiency and Capital Mobility Table 6.5 Post Float Results 1977–1998: OLS and SUR Estimates (s t+1 – st) = α + β (i- i*)t + v Country OLS SUR α β R2 DW α β R2 DW US 0.13 -0.006 0.08 1.6 0.11 -0.003 0.02 1.6 (0.04) (0.005) (0.04) (0.005) UK 0.10 -0.003 0.01 2.0 0.11 -0.005 -0.00 2.0 (0.04) (0.006) (0.04) (0.005) India 0.04 -0.005 0.008 1.9 0.04 -0.005 0.008 1.9 (0.06) (0.01) (0.06) (0.01) France 0.09 -0.001 0.001 2.4 0.10 -0.004 -0.009 2.4 (0.05) (0.008) (0.05) (0.007) Germany 0.11 -0.002 0.04 1.9 1.10 -0.0004 0.001 1.9 (0.03) (0.008) (0.02) (0.005) Japan 0.24 -0.01 0.17 2.0 0.21 -0.01 0.16 2.1 (0.08) (0.007) (0.07) (0.01) (Standard errors reported in parenthesis) Wald Tests 1977-1998 1. All β equal χ2(3) = 0.06(0.995) 2. All β=0 χ2(4) = 1.96(0.743) 1982-1998 1. All β equal χ2(5) = 6.79(0.236) 2. All β=0 χ2(6) = 7.15(0.307) 265 Exchange Rate Efficiency and Capital Mobility The estimated β coefficients for the pre-float period are significantly different from unity, with the estimated coefficients on the interest rate differential negative for the US, India and France. The coefficients on the interest rate differential for the post- float period are negative and insignificantly different from zero for all countries. T he result s ap p ear t o suggest a widening of yield rates over t ime. The negat ive estimates for β are perhaps associated with the omission of a time-varying risk premium from the uncovered interest parity equation.17 Finally, the Wald test statistics, that the slope coefficients in the regressions are equal across countries, is consistent with the data for the post-float period. This is not surprising, in view of the fact that all the coefficients are not significantly different from zero. Hence, the hypothesis that all β=0 cannot be rejected for the post-float periods. Surprisingly, the hypothesis that all β=0 is rejected for the pre-float period. A test of all β=1 for this period yields a Wald test statistic of over-unity, leading to complete rejection of the uncovered interest parity hypothesis. Hypothesis tests have also been carried out on monthly data for the period in which interest rates reflected market conditions, 1990–1998. 17 This has been termed the forward premium puzzle in the international finance literature. See Bilson(1981), Cumby and Obstfeld (1984), Flood and Rose (1996). The omission of time-varying term premia from the regression equation is found to result in a correlation between the interest rate differential and disturbance term violating the assumptions underlying equation (6). 266 Exchange Rate Efficiency and Capital Mobility Table 6.6 Uncovered Interest Parity—OLS and SUR Regressions 1990.1–1998.12 (s t+1 – st) = α + β (i- i*)t + v Country OLS SUR α β R2 DW α β R2 DW US 0.005 -0.000 0.00 1.3 0.002 0.0002 0.06 1.4 (0.002) (0.0002) (0.00) (0.0002) UK 0.01 -0.0008 0.009 1.7 0.009 -0.0006 0.007 1.7 (0.007) (0.0008) (0.006) (0.0006) India -0.005 0.0003 0.006 2.0 -0.005 0.0003 0.006 2.0 (0.004) (0.0005) (0.004) (0.0004) France 0.02 -0.001 0.01 1.8 0.01 -0.0008 0.01 1.9 (0.01) (0.001) (0.005) (0.0005) Germany 0.02 -0.002 0.03 1.8 0.01 -0.0009 0.02 1.8 (0.01) (0.001) (0.0006) (0.0005) Japan 0.03 -0.002 0.02 1.8 0.02 -0.001 0.02 1.8 (0.01) (0.001) (Standard errors reported in parenthesis) Wald Tests 1. All β equal χ2(5) = 5.74(0.16) 2. All β=0 χ2(6) = 6.38(0.12) 267 Exchange Rate Efficiency and Capital Mobility Evidence points again to the rejection of uncovered interest parity, with the coefficients on the interest rate differential negative for all countries except the U.S. and India under SUR estimation, and negative for all countries excepting India under OLS estimation. Any implication of yield convergence is totally rejected by the data. The hypothesis that all β=0 cannot be rejected at the 5% level of significance. A test of the hypothesis that all β=1 is decisively rejected by the data, corroborating the results obtained in respect of the annual data. The plots of uncovered nominal interest rate differentials (see Figures 6.1–6.6 ) confirm these conclusions, suggest ing no evidence of convergence over time. They appear, on the contrary, to have widened after the float. Interest Rate Differentials Figure 6.1 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 268 Exchange Rate Efficiency and Capital Mobility Figure 6.2 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF Figure 6.3 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 269 Exchange Rate Efficiency and Capital Mobility Figure 6.4 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF Figure 6.5 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 270 Exchange Rate Efficiency and Capital Mobility Figure 6.6 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF It follows from this that expectations are not rational or time-varying risk premia exist, or that both conditions prevail. The rejection of uncovered interest parity is also consistent with peso effects, bubbles and bandwagon effects and also with virtually all other studies. The rejection of uncovered interest parity may not, therefore, provide a clear insight on the degree of foreign exchange market efficiency. For this reason, a test of speculative efficiency is carried out in the following section. 6.7.2 Speculative Efficiency Table 6.2 indicates that forward and spot exchange rates are integrated of order one. This provides the basis for the use of cointegration. Therefore, estimates of the 271 Exchange Rate Efficiency and Capital Mobility cointegrating regressions for the speculative efficiency condition are presented in Table 6.7, with ft representing the forward rate contracted in period t for payment at the end of 1, 3 and 6 months. Table 6.7 The Dickey-Fuller Unit Root Test for the Residuals of the Cointegration Regression of the Spot Rate on the Forward Rate Cointegrating Regression ADF ______ st+1 = a+ b f t,1 + et -3.37 st+3 = a+ b f t,3 + et -2.44 st+6 = a+ b f t,6 + et -1.87 Significance levels: 1% -4.29; 5%, -3.74, 10%, -3.45 (Davidson and MacKinnon) The results point to the rejection of a long-run relationship between spot and forward exchange rates. The test statistic for the one-month forward rates is marginally below the 10% critical value while, the three-month and six-month statistics are significantly below the 10% level. The rejection of cointegraton tests, however, does not necessarily imply forward exchange market inefficiency because the failure of the tests could result from unsatisfied assumptions. More recently, Sephton and Larsen (1991) have questioned the usefulness of tests of long-run relationships based upon cointegration techniques. Therefore, hypothesis tests 272 Exchange Rate Efficiency and Capital Mobility employing the conventional inference procedures are carried out to verify the evidence implied by the cointegration tests. Table 6.8 presents OLS estimation results for tests of speculative efficiency. Table 6.8 Speculative Efficiency: OLS Estimates st+n = α + β f t,+ et Maturity α β t: α=0 t: β=1 F:α=0;β=1 R2 1 month 0.03 0.99 0.71 -0.75 3.34 0.98 (0.04) (0.01) 3 month -0.05 1.01 0.60 -0.63 1.28 0.95 (0.08) (0.02) 6 month -0.19 1.05 -1.17 1.17 0.70 0.85 (0.16) (0.04) Standard errors corrected by the Newey West procedure (1987) reported within parenthesis The β coefficients for the 1, 3, and 6-month forecast horizons are, respectively, 0.99, 1.01 and 1.05 implying that the entire variation in spot rate is explained by the forward rate data. The intercept terms are also insignificantly different from zero consistent with the theory of speculative efficiency. The null hypotheses that α=0 and β=1 cannot be rejected on the basis of a t-test for all three forecast horizons. The joint hypothesis of α=0 and β=1 also cannot be rejected at the 5% level of significance. The R2 for the regressions range from 0.85–0.98 suggesting high explanatory power of the equations. To gain further insight to this relation, Figure 6.7 plots the weekly spot and forward rates for the Sri Lanka rupee/US dollar for the 273 Exchange Rate Efficiency and Capital Mobility period under study. The figure clearly demonstrates that the spot and forward exchange rates tend to move together by approximately the same amounts. Figure 6.7 Source: Central Bank of Sri Lanka However, as pointed out by Ngama (1992), the use of classical inference procedures to non-stationary data may not be strictly correct. Hence, a number of studies of speculative efficiency employ the current spot rate as a stationary deflator (see Fama 1984, Zietz and Homaifar 1994, Cavaglia, Verschoor and Wolff 1994) so that the speculative efficiency condition is re-tested employing equation (13). 274 Exchange Rate Efficiency and Capital Mobility Table 6.9 A Te st of Spe cul ati ve Effi ci e ncy Empl oyi ng the Curre nt S pot Rate as a Stationary Deflator: OLS Estimates st+n – st = α + β (f t – st) + et Maturity α β t: α=0 t: β=1 F:α=0;β=1 R2 1 month -0.006 0.16 -21.7 -24.8 677 0.13 (0.00) (0.03) 3 month -0.003 0.04 -18.5 -76.5 5852 0.07 (0.00) (0.01) 6 month -0.04 0.33 -50.0 -20.3 1448 0.44 (0.00) (0.03) Standard errors corrected by the Newey West procedure (1987) reported within parenthesis The results now appear to reject the speculative efficiency condition, with the coefficients on the forward-spot differential, ft – st, significantly different from unity. The results are in complete contrast to the results for the levels equation. While much of the work based upon equation (13) tend to reject the speculative efficiency condition (see Fama 1984), studies based on the levels equation tend to support the restrictions implied by this hypothesis—Hakkio and Rush (1989). The rejection of the speculative efficiency condition over the short term has been explained by way of irrationality of market participants (Froot and Frankel, 1989); risk aversion (Taylor, 1989); risk aversion and irrational expectations (MacDonald and Torrance, 1990). It is possible, therefore, to conclude that overall results point to the rejection of the speculative efficiency condition for Sri Lanka, as inferences drawn from the difference equation are likely to be more accurate, given the non-stationarity of the data series. 275 Exchange Rate Efficiency and Capital Mobility 6.7.3 Savings-Investment Correlations This section goes on to examine the extent to which deviations from interest parity may be attributable to imperfect capital mobility. The estimated regression coefficients for savings investment correlations given by equation (14) are reported below. The sample period runs from 1959–1998. The sample is divided into two sub-periods, with the 1959–1976 period representing the era of fixed exchange rates, and 1978–1998 the period of floating exchange rates. It should be noted that while Feldstein and Horioka (1980) use cross sectional data, this study uses time series data.18 However, the regressions are also carried out on the first differences of the series to make the data stationary, see Bayoumi (1990). 18 The use of time series data can be found in Frankel (1986), Bayoumi (1990) and Monadjemi (1990). 276 Exchange Rate Efficiency and Capital Mobility Table 6.10 SAVING–INVESTMENT CORRELATIONS Sample Period Regression Equation R2 DW 1959-1998 (I/Y) = 0.07 + 0.90 (S/Y) 0.36 0.68 (2.51) (4.65) 1959-1976 (I/Y) = 0.12 + 0.32 (S/Y) 0.18 1.3 (5.53) (1.85) 1977-1998 (I/Y) = 0.25 - 0.04 (S/Y) 0.00 0.65 (5.03) (-0.14) 1959-1998 ∆(I/Y) = 0.002 - 0.05 ∆ (S/Y) 0.002 2.0 (0.51) (-0.29) 1960-1976 ∆(I/Y) = -0.00 + 0.17 ∆ (S/Y) 0.06 2.3 (-0.15) (0.97) 1977-1998 ∆(I/Y) = 0.00 - 0.19 ∆ (S/Y) 0.03 1.9 (0.75) (-0.80) With a Structural Break 1959-1998 I/Y = 0.15 + 0.07 (S/Y) +0.09 D 0.69 1.0 (6.17) (0.39) (6.19) t statistics are reported in the parenthesis below the coefficients The null hypothesis that the savings retention coefficient (the coefficient on S/Y) is zero is rejected for the full sample covering the 1959–1998 period, with the estimated coefficient taking on a value of 0.90. The results for the full sample appear to suggest that 90% of the increase in the domestic investment ratio is financed by domestic savings. While the null hypothesis that the coefficient on S/Y 277 Exchange Rate Efficiency and Capital Mobility is zero is rejected at the 10% level for the pre-deregulation period, it is not rejected for the post-deregulation period. The coefficient drops from 0.32 during the period 1959–1976 to –0.04 for the period 1978–1998. Although negative, the coefficient is insignificantly indifferent from zero suggest ing a significant increase in cap it al mobility between the two periods. This is further confirmed by the sharp fall in the R2 of the regressions. The levels of the series display some evidence of serial correlation in the residuals on the basis of the DW statistics.19 The savings retention coefficient on the first differenced data is marginally higher and negative in the post-deregulation period in comparison the coefficient for the pre-deregulation period. In order to investigate if the financial deregulation has led to a weakening of the link between savings and investment, the regression of I/Y on S/Y is carried out with the inclusion of a dummy variable (see Table 6.10). This variable takes on a value of zero until 1977 and one thereafter. While the coefficient on the dummy variable is statistically significant, confirming a structural break, the restriction that the 19 Correction for serial correlation does not lead to a significant change in the savings retention coefficient. 1959–1998 (I/Y) = 0.07 + 0.95 (S/Y) - 0.37 ∆(S/Y)t-1 + 0.47 ∆(I/Y)t- 1 (2.11) (4.66)*** (-1.24) (1.83) DW=1.2 1959–1976 (I/Y) = 0.12 + 0.28 (S/Y) ) - 0.03 ∆(S/Y)t- 1 + 0.24 ∆(I/Y)t- 1 (4.42) (1.26) (-0.11) (0.96) DW =1.6 1977–1998 (I/Y) = 0.23 + 0.10 (S/Y) - 0.35 ∆(S/Y)t- 1 + 0.33 ∆(I/Y)t- 1 (4.61) (0.33) (-1.25) (1.20) DW =1.2 278 Exchange Rate Efficiency and Capital Mobility coefficient on the savings ratio is zero is not rejected, consistent with the evidence obtained for the levels regressions. Overall evidence therefore points to an increase in capital mobility. A test of capital mobility put forward by Sachs (1981), based on a regression of the current account ratio to investment ratio, is also used to confirm the above results. Sachs hypothesized that if capital was internationally mobile, investment should have a negative impact on the current account. Table 6.11 CURRENT ACCOUNT-INVES TMENT CORRELATIONS CA/Y = a + b (I/Y) Sample Period Regression Equation R2 DW 1959–1998 (CA/Y) = 0.06 - 0.52 (I/Y) 0.60 1.3 (4.43) (-7.70) 1959–1976 (CA/Y) = 0.05 - 0.47 (I/Y) 0.23 1.7 (1.44) (-2.16) 1977–1998 (CA/Y) = 0.15 - 0.86 (I/Y) 0.69 1.30 (4.77) (-6.65) 1959–1998 ∆(CA/Y) = 0.002 - 0.93 ∆ (I/Y) 0.52 2.5 (0.59) (-6.30) 1960–1976 ∆(CA/Y) = 0.001 - 0.59 ∆ (I/Y) 0.22 2.4 (0.28) (-2.05) 1977–1998 ∆(CA/Y) = 0.003 - 1.02 ∆ (I/Y) 0.61 2.6 (0.62) (-5.62) t statistics are reported in parenthesis 279 Exchange Rate Efficiency and Capital Mobility As pointed out by Penati and Dooley (1984), the inverse correlation between the current account balance and I/Y should increase over time with increasing capital mobility. The results appear to suggest increasing capital mobility, with the estimated slope coefficients rising from –0.47 to –0.86 for the levels, and from –0.59 to –1.02 for the first differences of the series between the pre- and post-deregulation periods. The results confirm increased capital mobility between the pre- and post- deregulation periods. Figure 6.8 illustrates the behaviour of the average rates of investment, savings and current account for the period under study. The graph clearly indicates a significant rise in the investment ratio over the 1977–1980 period, leading to a widening gap between the savings and investment ratios. A question arises, therefore, as to whether the absence of a correlation between savings and investment for this period was due to these few outlying observations. Hence, the regression for the post- deregulation period is re-estimated by omitting these observations. 280 Exchange Rate Efficiency and Capital Mobility Figure 6.8 Source: Central Bank of Sri Lanka Annual Reports Table 6.12 SAVINGS-INVESTMENT AND CURRENT ACCOUNT-INVESTMENT CORRELATIONS ELIMINATING OBSERVATIONS FROM 1977–1980 Sample Period Regression Equation R2 DW 1981-1998 (I/Y) = 0.23 + 0.13 (S/Y) 0.03 0.4120 (6.69) (0.68) 1981-1998 (CA/Y) = 0.09 - 0.60 (I/Y) 0.28 1.4 (1.51) (-2.51) t statistics are reported in parenthesis 20 With correction for serial correlation: 1981-1998 (I/Y) = 0.23 + 0.12 (S/Y) + 0.11∆ (I/Y)t-1 +0.06 ∆(S/Y)t-1 (6.23) (0.56) (0.50) (0.28) R2 = 0.05 DW = 0.56 281 Exchange Rate Efficiency and Capital Mobility The elimination of observations does not lead to a significant change in the estimated coefficients in the regressions of S/Y on I/Y or CA/Y on I/Y. It is possible to conclude, therefore, that financial deregulation has led to an increase in capital mobility. However, it should be kept in mind that a number of factors could bias the results in favour of the Feldstein-Horioka model. Studies by Dooley, Frankel and Mathieson (1987) and Frankel and MacArthur (1988) find a strong association between domestic savings and investment for economies with relatively open capital accounts and a weak correlation between savings and investment for developing economies that rely heavily on foreign aid to finance their current accounts. Fry (1993) shows that a rise in the debt ratio in developing countries leads to a widening of the current account ratio increasing the gap between the savings ratio and investment ratio monotonically. The results obtained above while could be attributed to increased capital mobility, it is also possible that the increase in foreign debt and widening current account imbalance in the post-deregulation period have biasd the results in favour of increased capital mobility. 6.7.4 Consumption-Income Correlations The consumption function given by equation (26) is estimated by using OLS and IV techniques. The results are reported in Table 6.13. The second and third rows report the adjusted R2 for the OLS regressions of ∆C and ∆X on the instruments. As 282 Exchange Rate Efficiency and Capital Mobility pointed out by Campbell and Mankiw (1990), et in equation (26) is an innovation and is hence orthogonal to any variable that is in the agents’ information set at time t-1. Therefore, IV estimation is also employed to eliminate the potential inconsistencies of the OLS estimates. 283 Exchange Rate Efficiency and Capital Mobility Table 6.13 Consumption-Income Correlations ∆C = υ + λ ∆X Sample Period OLS estimates IV estimates 1 2 3 1960-1998 λ 0.81 0.43 0.67 0.58 (2.3) (1.12) (1.36) (1.95) R2 for ∆C -0.03 -0.03 -0.09 R2 for ∆X 0.04 -0.02 -0.04 1960-1976 λ 1.09 1.20 0.83 1.03 (8.82) (3.02) (1.51) (5.29) R2 for ∆C -0.03 -0.16 0.07 R2 for ∆X -0.11 -0.20 0.23 1977-1998 λ 0.49 0.29 0.04 0.57 (3.13) (0.77) (-0.02) (1.99) R2 for ∆C 0.004 -0.12 0.06 R2 for ∆X 0.05 -0.15 -0.05 Instruments used are 1: Constant, ∆C t-2,…∆C t-4 2: Constant, ∆X t-2…∆X t-4 3: Constant, ∆C …∆C ,∆X …∆X ,CA t-2 t-4 t-2 t-4 21 t-2 Scaling has been carried out by dividing ∆C, ∆X and CA by Xt-1 The R2 values are the adjusted R2 from OLS regression of ∆C and ∆X on the instruments. t statistics are reported in parenthesis. 21 See Campbell and Deaton (1989). This method is employed by Campbell and Mankiw(1990), Shibata and Shintani (1998). 284 Exchange Rate Efficiency and Capital Mobility The results appear to be robust to the measures of estimation. While the coefficient on ∆X records a decline from 1.09, capital immobility, in the period 1959–1976 to approximately 0.49 in the post-deregulation period under OLS, the IV estimates record a similar trend. There is significant evidence of an increase in capital mobility between the two periods consistent with the results obtained in respect of the savings-investment correlations. The fact that consumption appears sensitive to income given a constant real interest rate, violates Michener’s (1984) proposition of a fluctuating real interest rate as the factor giving rise to the close association between the two variables. It would, however, be interesting to examine if the relaxation of the assumption of a constant real interest rate would yield similar results. 285 Exchange Rate Efficiency and Capital Mobility Table 6.14 Consumption–Income Correlations Relaxing the Assumption of a Constant Real Rate of Interest ∆C = υ + δr + λ ∆X Sample Period OLS estimates IV estimates 1 2 3 4 1960-1998 λ 0.70 -0.004 0.22 0.59 0.37 (6.57) (-0.003) (0.20) (1.95) (0.74) δ -0.006 0.01 0.01 -0.00 0.004 (-2.86) (0.50) (0.66) (-0.09) (0.40) R2 for ∆C -0.03 -0.03 -0.07 -0.07 R2 for ∆X 0.04 -0.02 -0.04 -0.06 1960-1976 λ 0.78 0.43 0.63 0.68 0.66 (5.72) (1.01) (1.87) (4.04) (3.98) δ -0.01 -0.02 -0.02 -0.12 -0.02 (-3.17) (-2.42) (-3.12) (-3.33) (-3.32) R2 for ∆C -0.03 -0.16 -0.12 -0.15 R2 for ∆X -0.11 -0.20 -0.18 -0.06 1977-1998 λ 0.29 -0.10 -2.81 0.31 0.02 (2.42) (-0.23) (-0.23) (1.32) (0.93) δ -0.008 -0.02 -0.02 -0.010 -0.008 (-4.46) (-1.72) (-0.24) (-2.44) (-1.95) R2 for ∆C 0.004 -0.12 -0.18 -0.12 R2 for ∆X 0.05 -0.15 -0.18 -0.21 Instruments used are 1: Constant, ∆C t-2 …∆C t-4 2: Constant, ∆X t-2 …∆X t-4 3: Constant, ∆C t-2 …∆C t-4,∆X t-2…∆X t-4,CA t-2 4: Constant, ∆C t-2 …∆C t-4,∆X t-2…∆X t-4, rt-2 Scaling has been carried out by dividing ∆C, ∆X and CA by Xt-1 The R2 values are the adjusted R2 from OLS regression of ∆C and ∆X on the instruments. t statistics are reported in parenthesis. 286 Exchange Rate Efficiency and Capital Mobility The coefficients on the real rate of interest are statistically significant and of the correct sign for most of the regressions. Despite the fluctuations in the real interest rate, the results are not significantly different from those of Table 6.13, confirming increased capital mobility between the pre- and post-deregulation periods. 6.7.5 Real Interest Rate Linkages The focus of this chapter has hitherto been on the nominal rate of interest. It is, however, the real rate of interest that is of importance for purposes of policy. M oreover, tests of capital mobility will only be valid in the presence of real interest rate convergence. In order to prevent savings from crowding out investment, real rates of interest would have to be equalized across countries. Therefore, in order to influence economic activity, it is necessary to influence the real rate of interest. Thus, this section attempts to examine real interest rate linkages between Sri Lanka and the US, UK, Japan, India, France and Germany. Table 6.15 reports regression results for equation (32).22 22 Note that the t statistics of the regression equations for the U.S, U.K, Japan and France may not be strictly correct given the non-stationarity of the data. Cointegration tests are not carried out, due to the inconsistency in sample periods. 287 Exchange Rate Efficiency and Capital Mobility Table 6.15 OLS Estimates for Real Interest Rate Linkages r = a + br* Country Instrument Period b R2 DW US 3m TB Rate 1959-1998 0.66 0.08 1.18 (1.76) 1959-1976 0.88 0.23 1.15 (2.18) 1977-1998 0.37 0.01 1.23 (0.59) Japan Govt.Bond Yield 1969-1998 0.48 0.10 1.20 (1.79) 1969-1976 0.47 0.47 1.24 (2.29) 1977-1998 -0.23 0.04 1.26 (-0.30) Private Bond Rate 1981-1998 -0.47 0.03 1.57 (-0.71) India Call Money Rate 1960-1997 0.38 0.16 1.29 (2.62) 1960-1976 0.24 0.27 0.80 (2.35) 1977-1997 0.66 0.14 1.38 (1.78) Germany 12m TB Rate 1981-1998 -0.60 0.02 1.32 (-0.58) France Govt.Bond Yield 1959-1998 0.95 0.27 1.49 (3.73) 1959-1976 0.85 0.23 0.93 (2.20) 1977-1998 1.23 0.25 1.65 (2.59) UK 3m TB Rate 1959-1998 0.34 0.08 1.33 (1.81) 1959-1976 0.17 0.06 1.08 (0.99) 1977-1998 0.41 0.05 1.37 (1.06) t statistics reported within parenthesis 288 Exchange Rate Efficiency and Capital Mobility T he result s in general ap pear t o suggest part ial linkage, wit h t he b coefficient s for t he full sample lying in the region of 0.95 and 0.34. The estimated coefficient for France is 0.95, which is not significantly different from unity. An examination of the results for the sub-samples appear to suggest a weakening of the link between the Sri Lanka-U.S and Sri Lanka-Japan real rates, while a strengthening of the link is observed between Sri Lanka and India, France and the U.K. The coefficient on the real rate of interest for France rises from 0.85 in the pre-deregulation period to 1.23 in the post-deregulation period, implying full linkage. The partial support for real interest rate equality could be attributed to the increased capital mobility observed in the previous section. The b coefficients for Japan and Germany however, are significantly negative in this period, suggesting the absence of a link. While there appears to be no consistency in the results for the post deregulation period, the results for the US, UK, Japan and France should be interpreted with caution, given the non-stationarity of the data. The lack of consistency in results is not surprising in view of the fact that there are capital controls in Sri Lanka. The results for India probably are the most reliable, given the stationarity of this data series. Hence, these results are examined below in greater detail, with the real rate in Sri Lanka regressed on both the real rate in India and a lagged real rate in Sri Lanka. In the event of some form of adjustment in the pricing mechanism, the coefficient on the lagged real rate would be non-zero. Real interest rate equality between India and Sri Lanka, on the other hand, would imply that the coefficient on the lagged real rate be zero. 289 Exchange Rate Efficiency and Capital Mobility Table 6.16 A Partial Adjustment Model for Real Interest Rate Linkages between India and Sri Lanka r = a + λrt-1 + γr* Country Instrument Period λ γ β = γ/1−λ R2 LM ______ India Call Money Rate 1960-1997 0.39 0.30 0.49 0.28 0.05 (0.15) (0.14) 1960-1976 0.55 0.25 0.56 0.55 2.61 (0.19) (0.08) 1977-1997 0.26 0.47 0.64 0.20 0.25 (0.23) (0.40) standard errors reported in parenthesis The coefficients on the real rate for India are not significantly different from the regression, excluding the lagged real rate. However, the current real interest rate in Sri Lanka appears to be related to the lagged real interest rate. The results indicate some form of rigidity in the foreign exchange market, with evidence of a decrease between the pre- and post-deregulation periods. Section 6.4.5 above showed that real interest rate equality held only in the absence of deviations from PPP and UIP. The lack of total support for real interest rate equality could be attributed in part to the rejection of UIP. It is important to note, however, that while evidence totally rejected UIP, some support was found for real interest rate linkage. A likely explanation for the partial support for real interest rate equality 290 Exchange Rate Efficiency and Capital Mobility could perhaps be associated with the smaller deviations in PPP. Caution, however, should be exercised in interpreting these results due to the non-stationarity of some of the data series. Figures 6.9 to 6.14 illustrate the real interest rate linkages between Sri Lanka and the countries to which the rupee in linked. Real Interest Rate Linkages Figure 6.9 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 291 Exchange Rate Efficiency and Capital Mobility Figure 6.10 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF Figure 6.11 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 292 Exchange Rate Efficiency and Capital Mobility Figure 6.12 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF Figure 6.13 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 293 Exchange Rate Efficiency and Capital Mobility Figure 6.14 Source: Computed from Central Bank of Sri Lanka Annual Reports and International Financial Statistics, IMF 6.8 IMPLICATIONS This chapter tests a number of alternative theories for foreign exchange market efficiency. While the evidence points to the rejection of UIP and speculative efficiency, the results are in general supportive of the hypothesis of increased capital mobility consistent with the findings of Haque and Montiel (1991), and Dooley and Mathieson (1994). However, the lack of support for UIP is hardly surprising, given the rejection of the theory for a number of developed countries.23 23 See Froot and Thaler (1990) for a survey; also see McCallum (1994). 294 Exchange Rate Efficiency and Capital Mobility Although the capital controls in place give rise to problems in interpreting results of UIP, the unanimous rejection of the theory of uncovered interest parity and the incorrectly signed coefficients on the interest rate differentials is perhaps a reflection of time-varying risk premia. From 1990–1993, the Sri Lankan Government pursued a policy of sterilization in order to absorb the excess liquidity that had resulted from large capital inflows. Such a policy would undoubtedly create risk premia for which foreign investors would have to be adequately compensated in order to invest in Sri Lanka. To the extent that the Sri Lankan authorities are able to enforce policies to control capital flows, the country would retain a degree of independence in pursuing domestic stabilization policies. Moreover, the domestic economy would be insulated from external shocks to some degree. Another source of the inconsistency with the theory of uncovered interest parity could be attributed to Central Bank intervention in the foreign exchange market. Official aid and long-term capital inflows in Sri Lanka are often governed by political considerations. Furthermore, the domination of the foreign exchange market by the two state-owned banks, whose speculative activities are closely monitered by the Government, does not contribute towards narrowing the interest rate differential. It is also likely that the war and its perceived consequences have discouraged some capital inflows. Short-term capital flows are affected to some degree by the uncertainity generated by political instability and the lack of complete information. The greater influence, however, is likely to be from the capital controls employed. Non-nationals, for instance, cannot invest in government securities. 295 Exchange Rate Efficiency and Capital Mobility Further, most foreign investors tend to look at Sri Lanka as part of South Asia, rather than as an important country for investment in its own right. An impact on the Colombo Stock Exchange was observed when the East Asian crisis erupted and also when India and Pakistan carried out their nuclear tests. However, it is difficult to identify the exact cause for the rejection of uncovered interest parity, given the capital controls. While interest rates in Sri Lanka were observed to have generally risen through the 1980s and 1990s, the same rate of increase was not observed in the economies of the West, which perhaps is a likely explanation for the widening nominal interest rate differentials in the post-deregulation period. During much of the 1990s, these countries experienced decreases in inflation and interest rates. The impact of these developments is possibly the phenomenon that has been captured by this study. These results are consistent with the evidence obtained for speculative efficiency. This hypothesis is conditional upon the assumptions of risk-neutrality and rational expectations. The overall evidence pointing to the rejection of the speculative efficiency condition is perhaps a reflection of the violation of these assumptions and also the thinness of the forward exchange market. It is shown that real interest rate equality only holds in the absence of deviations from PPP and UIP. While Frankel (1986) pointed out that the main source of the rejection of real rate equality for the developed countries was the failure of PPP, 296 Exchange Rate Efficiency and Capital Mobility Cumby and Obstfeld (1984) showed that UIP was equally hard to achieve in the case of developed countries. The lack of total support for real interest rate linkage in this chapter, while perhaps is partly a reflection of the failure of UIP, the partial support is probably explained by the smaller deviations in PPP. It is important to note, however, that the results need to be interpreted with some caution, given the non- stationarity of some of the data series. It was further pointed out in the previous chapter that the shortcoming in the CCPI tended to understate the true rate of inflation. Therefore, to the degree that the rate of inflation is understated, the real rate is overstated, resulting in a bias in favour of the equality of real interest rates and PPP. 6.9 CONCLUSION This chapter has attempted to evaluate the applicability of foreign exchange market efficiency for Sri Lanka. Consistent with previous findings, tests of capital mobility demonstrated that capital flows have become increasingly more mobile. The theory of uncovered interest parity was, however, decisively rejected by the data. This could perhaps be a reflection of the capital controls in place. First, capital controls, if effective, raise transaction costs of portfolio adjustments and discourage investors from arbitraging away yield differentials. Second, capital controls may be associated with increased political risk and uncertainty about future capital controls, increasing the variability of expected yields and reducing arbitrage by risk-averse investors. 297 Exchange Rate Efficiency and Capital Mobility The evidence that emerges from these results is clearly mixed with indications of increased capital mobility but inefficiency in respect of the forward exchange market and rising disparities between uncovered nominal interest rate differentials. Overall evidence, however, points to an increase in capital mobility, suggest ing an enhanced role of the exchange rate in the monetary transmission process. It is possible to conclude, therefore, that Sri Lanka is on its way to achieving greater efficiency in respect of the foreign exchange market. 298 Conclusion and Policy Implications CHAPTER 7 CONCLUSION AND POLICY IMPLICATIONS 7.1 OVERALL FINDINGS Drawing on the experience of Sri Lanka, this study attempts to evaluate the implications for efficiency in the light of the financial reform programme. To this end, the study employs a number of standard tests for market efficiency, which include the expectations hypothesis of the term structure, the Fisher hypothesis, uncovered interest parity, speculative efficiency, real interest rate equalization and tests of capital mobility. An investigation of the financial structure in Sri Lanka suggested that Sri Lanka had made appreciable progress in the last two decades in expanding its financial structure and deepening its financial markets. The empirical results presented in this study however, suggest that the financial sector still falls short of achieving market efficiency. The expectations hypothesis of the term structure according to which the forward interest rate is an unbiased predictor of the future spot rate is rejected for Sri Lanka. Similarly, the predictive content of the spread in forecasting the future direction of short rates is limited. However, the spread predicts in the correct direction of future short rates. While the structural weaknesses of the Sri Lankan financial system are cited as a potential cause of the rejection of the expectations hypothesis, the findings need not necessarily be interpreted as evidence against market 299 Conclusion and Policy Implications efficiency. It is possible that the rejection of the hypothesis stems from the use of the wrong economic model. A test of the Fisher effect employing rational and adaptive approaches for modeling inflationary expectations indicates that greater support is for the adaptive expectations model. The results suggest a lag in the reaction of interest rates to changes in expected inflation. The evidence in support of an inverted Fisher Effect in the post-deregulation period suggests that inflationary expectations could have an indirect effect on the rate of interest through its impact on the real rate. The accuracy of these results depend on the measure of inflationary expectations used as pointed out by Stammer and Valentine (1982) and therefore need to be interpreted with caution. Updating the Colombo Consumer Price Index (CCPI) would perhaps increase the precision of the results to some degree. Another likely cause for the slow reaction of interest rates to changes in inflation is perhaps the lack of an inflation hedge to investors in Sri Lanka.1 In the absence of such alternative assets, there is no market pressure for inflation to push up the rate of interest. The introduction of treasury bonds in 1997 should contribute to the promotion of a meaningful inflation hedge and strengthen secondary markets for government securities. A time lag in the adjustment of interest rates to inflationary expectations could also 1 The existence of an alternative asset of which the yield increases with inflation. See Stammer and Valentine (1982). 300 Conclusion and Policy Implications arise if Sri Lankan interest rates are largely determined by overseas interest rates rather than the rate of inflation. This however, is not supported by the results for UIP which indicate a widening of the interest rate differential between Sri Lanka and other countries in the post deregulation period. While the results for speculative efficiency are not clear-cut suggesting the existence of time varying risk premia or irrational expectations, there is some support for real interest rate equalization. Consistent with previous findings, the tests of capital mobility based on the Feldstein-Horioka (1980) and Shibata-Shintani (1998) models appear to suggest an increase in capital mobility. However, as pointed out by Tesar (1991) and Shibata and Shintani (1998) among others, the observed correlation between savings and investment, and consumption and income need not necessarily arise from an increase /decrease in capital mobility but changes in a number of other factors including technological progress, population growth and government policy. The evidence of an increase in capital mobility nevertheless has important policy implications for Sri Lanka in that it suggests an enhanced role of the market mechanism in the monetary transmission process. Thus, the experience of Sri Lanka suggests that financial deregulation implemented in isolation will not automatically promote market efficiency unless accompanied by positive policy action to reinforce the impact of these reforms. There are a number of obstacles inherent to developing countries such as Sri Lanka, as will be seen in the next section, that slow down the process of growth of financial markets. It is important, therefore, that the monetary authorities address the question of what 301 Conclusion and Policy Implications action should be taken to foster an effective market structure and promote competitive market behaviour. In conclusion, this study, therefore, makes an attempt to provide several recommendations which could help to reinforce the impact of financial deregulation on market efficiency. 7.2 POLICY IMPLICATIONS AND IMPEDIMENTS TO REFORM A policy lesson to be learnt from this analysis is that interest and exchange liberalization alone will not guarantee market efficiency unless accompanied by fiscal consolidation and other developments in the financial system. Despite the fact that Sri Lanka has made appreciable progress in respect of the financial sector, the failure to correct certain fundamental structural weaknesses in the system has constrained the impact of the financial reforms. Notwithstanding the establishment of new banks over the past decade, the market share of the two state-owned commercial banks has remained relatively constant at 59% of the total assets of the banking system. Due to the high volume of non-performing loans from lending to priority sectors, the solvency of these banks remains dependent on credit from the Central Bank. In 1993, these banks were recapitalized at a cost of 24 billion rupees, and in 1996 at a cost of 19.4 billion rupees.2 Moreover, the high intermediation costs of these banks have led to large spreads between lending and deposit rates, rendering monetary policy instruments, in particular, the interest rate policy, ineffective. 2 See Central Bank of Sri Lanka (1998),“Economic Progress of Independent Sri Lanka.” 302 Conclusion and Policy Implications Table 7.1 Bank Performance State Commercial Banks Private Banks (5 largest) 1995 1996 % of total 1995 1996 % of total In Rupees Million Capital and Reserves 12,358 14,604 - 7,922 9,431 - Assets 203,890 232,462 59.6 95,296 118,684 35.4 Deposits 149,838 164,845 62.9 74,987 93,807 41.1 Advances 119,240 119,264 58.9 56,239 66,425 35.5 Return on Assets 1.2 1.4 - 2.1 2.0 - (% pre tax) Return on Equity 4.8 18.5 - 15.0 16.1 - (% pre tax) Credit-Deposit Ratio 0.8 0.7 - 0.8 0.7 - Risk Weighted Capital Adequacy Ratio (range) 6.7-14.7 6.3-13.6 - 8.8-14.6 7.6-17.8 - Liquid Asset Ratio 21.6-26.0 18.4-25.7 - 20.9-32.4 19.6-28.0 - (range) Non-performing Loans/ Total Advances (%) 16.2 18.9 - 9.0 11.8 - Non-performing Loans/ Total Assets (%) 9.5 9.7 - 5.3 6.6 - Source: IMF Staff Country Report 1997 303 Conclusion and Policy Implications In a system such as this, monetary and fiscal policies are conducted through the banking system. Therefore, to the degree that the Government is induced to run debt-financed fiscal deficits, the main function of the money market would be to engage in the primary sale of government securities for financing the servicing requirements of the government debt. This has constrained the use of open market operations as an effective means of policy and distorted the level and structure of interest rates. Hence, while the inefficiency of the market could be attributed to government intervention, this inefficiency would again provide the incentive for the Government to influence the availability of securities and thereby the yield. Given the lack of depth in securities markets, the main source of corporate financing in Sri Lanka is also the commercial banking system. While well- developed equity markets can absorb risk without government intervention in the financial system through the provision of risk-absorbing equity contracts, this is not possible in underdeveloped stock markets. Therefore, the underdeveloped nature of security markets again would provide the rationale for government intervention in the banking system in order to absorb the risk of the corporate sector.3 The main task of monetary policy in the context of an open economy is to promote the process of financial intermediation by ensuring positive real interest rates. The control of inflation is important for ensuring positive real rates. The success of monetary policy in achieving this objective has so far been limited due to the 3 See Cho (1984). 304 Conclusion and Policy Implications Government's policy of deficit budgeting. The large and persistent budget deficits have led to inflation failing to compensate savers fully for price increases crowding out private sector investment. Therefore, market-determined interest rates alone will not lead to higher financial savings unless accompanied by adequate fiscal adjustment (see Table 7.2). An examination of the national savings rate of Sri Lanka with some of the high-performing East Asian economies indicates that Sri Lanka’s national savings rate is only 18% of the GDP in comparison to 30%–35% in the comparator countries for the 1991–1996 period. 305 Conclusion and Policy Implications Table 7.2 Determinants of Savings Rates: A Comparison of Trends 1981-1985 1986-1990 1991-1996 Average Average Average Nation al S avin g (percentage of GDP) Indonesia 21.3 25.7 32.0 Korea 26.5 36.2 35.1 Malaysia 30.4 29.4 30.8 Thailand 20.4 29.5 34.9 Sri Lanka 16.6 15.1 18.1 Real Per Capital Income (US Dollars) Indonesia 1,343 795 847 Korea 3,098 4,834 6,663 Malaysia 1,518 1,505 2,127 Thailand 917 1,122 1,757 Sri Lanka 274 233 188 Real Per Capital GDP Growth (percentage) Indonesia 3.2 4.3 2.7 Korea 6.6 9.0 6.4 Malaysia 2.5 4.1 6.1 Thailand 3.2 8.6 7.1 Sri Lanka 3.7 1.9 3.7 Age Dependency Ratio (% of Working Age Population to Total) Indonesia 0.73 0.68 0.64 Korea 0.55 0.47 0.43 Malaysia 0.71 0.72 0.72 Thailand 0.67 0.57 0.53 Sri Lanka 0.65 0.62 0.57 Inflation Rate (percentage) Indonesia 9.7 7.5 8.9 Korea 7.3 5.5 6.0 Malaysia 4.7 1.8 3.8 Thailand 5.0 3.9 5.0 Sri Lanka 12.2 12.6 10.8 306 Conclusion and Policy Implications Table 7.2: Determinants of Savings Rates: A Comparison of Trends (cont) 1981-1985 1986-1990 1991-1996 Average Average Average Variance of Inflation Indonesia 8.9 1.9 0.6 Korea 65.3 7.4 3.1 Malaysia 11.8 1.4 0.5 Thailand 21.3 3.3 1.2 Sri Lanka 43.2 31.8 4.9 Black Market Ex change Rate Premium (percentage) Indonesia 2.8 9.5 8.0 Korea - 7.7 1.7 Malaysia 0.8 1.2 1.2 Thailand - 2.1 2.1 Sri Lanka 21.5 17.3 7.5 Broad Money (percentage of GDP) Indonesia 17.5 28.6 41.2 Korea 35.1 43.3 59.2 Malaysia 90.6 117.7 115.5 Thailand 47.7 62.5 75.0 Sri Lanka 34.8 35.3 36.0 Public Saving (percentage of GDP) Indonesia 9.3 6.0 6.9 Korea 6.0 7.3 8.4 Malaysia 12.5 8.6 14.5 Thailand 3.9 8.1 10.7 Sri Lanka 1.5 -0.2 -0.2 Private Saving (percentage of GDP) Indonesia 12.0 19.8 25.1 Korea 20.5 28.9 26.6 Malaysia 17.8 20.8 16.2 Thailand 16.5 21.5 24.3 Sri Lanka 15.1 15.3 20.1 Source: IMF Staff Country Report 1997 307 Conclusion and Policy Implications While Sri Lanka records public dissaving, private savings is 20% of the GDP, which is comparable to that of the other economies. Of the determinants of savings rates, the gap between Sri Lanka and the comparator countries are largest in per capital income levels, inflation rates, financial sector indicators and public savings. Increasing public savings is perhaps the most effective means of raising national savings in the short run. It is observed, however, from Table 7.3 that the size of the public sector in Sri Lanka is significantly higher than that of the high-performing East Asian economies. Table 7.3 Size of the Public Sector in Selected Countries Government Exp endit ure as a % of GDP Country 1980 1996 Indonesia 22.1 14.6 Korea 17.2 18.6 Phillippines 13.4 18.5 Sri Lanka 41.4 27.7 Thailand 18.8 16.5 Source: World Development Indicators, World Bank (1999) Adequate fiscal adjustment would help raise public savings, contain inflation and enhance the effectiveness of monetary policy. A sustained increase in long-term savings would require the Government to reduce its resort to captive financing, thereby giving the public greater flexibility in their portfolio allocation. Therefore, it is important that the Government clearly distinguishes between its allocative and monetary policy roles. 308 Conclusion and Policy Implications A second implication of this investigation relates to the time lag in inflationary expectations formation. To the degree that the CCPI understates the true rate of inflation, monetary policy formulated on the basis of these prices will prove inefficient. In such circumstances, the Government has an incentive to run debt- financed fiscal deficits, which perhaps is a partial explanation for the high level of public debt of the Sri Lankan Government. Today ,Sri Lanka's public debt stands at 91% of GDP. A third implication of this analysis relates to the impact of increased capital mobility on the conduct of monetary policy. Chapter 6 suggests that, with the removal of barriers to capital flows, Sri Lanka experienced a steady rise in the volume of capital inflows in the post-deregulation period. This suggests that the exchange rate should also be taken into consideration when formulating monetary policy, as the role of the exchange rate policy in the monetary transmission process would be enhanced. Despite the increase in capital flows, an analysis of the capital flows to less developed countries indicates that Sri Lanka has attracted a disproportionately small share of the total flows (see Figure 7.1). About 90% of private capital flows to less developed countries have been concentrated in East Asia, Latin America and China. 309 Conclusion and Policy Implications Figure 7.1 Source: IMF World Economic Outlook 1998 and Central Bank of Sri Lanka Annual Reports While the hypothesized benefits from capital flows include increased access to capital and faster productivity growth, and risk diversification, it also requires stricter discipline in economic management, as it leaves less room for policy errors. Greater openness leads to greater vulnerability of an economy to changes in investor sentiment. The East Asian crisis serves to highlight this. However, sound macroeconomic policies could help to mitigate the potential adverse effects from capital flows. The potential benefits from capital flows also depend on the uses to which they are channelled. Figure 7.2 demonstrates that, while inflows into East Asia and Latin America have led to increases in investment, in Sri Lanka they have fuelled increases in consumption. 310 Conclusion and Policy Implications Figure 7.2 Source : World Bank, World Development Indicators 1999 Sri Lanka is the only country in which government consumption exceeds gross domestic investment. Therefore, in order to sustain the inflow of capital and ensure that the benefits from capital flows are fully realized, greater coordination of fiscal, monetary and exchange rate policies are required. To benefit from greater openness, the country also needs a well-developed corporate sector. The corporate sector in Sri Lanka unfortunately has adopted a cautious approach to participation in investment activity. As pointed out by Reisen (1989), when the budget deficit exceeds the current account deficit, the public sector becomes a net user of household and corporate savings, which are then unavailable for private investment, depressing private investment in many developing countries, as in the case of Sri Lanka. 311 Conclusion and Policy Implications 7.3 RECOMMENDATIONS Building credibility in the face of macroeconomic uncertainty is a task facing monetary policy-makers in Sri Lanka. To be successful, financial policies need to be underpinned by other reforms, including the privatization of the state-owned commercial banks, efficient capital markets, effective legal, accounting, management and supervisory structures, and fiscal consolidation. The continued dependence of the state-owned commercial banks on the Government has not only inhibited their efficient functioning, but also constrained the competition and efficiency of the entire banking system. Political pressures to finance unviable projects and limited incentives for screening and monitoring projects has led to a rising volume of non-performing loans by these banks. Therefore, the privatization of these two banks is crucial for the accelerated growth of the banking sector. Consequently, it would be possible for the imposition of commercial standards within priority sectors, as in the banks of Taiwan, Korea and Singapore. Strengthening the management and risk evaluation capabilities of bank managers should also form an important part of the restructuring procedure. Legislation put forward to permit the two banks to function autonomously was met with opposition in Parliament in 1996 and consequently withdrawn. Drawing lessons from the East Asian crisis, however, the Executive directors of the Central Bank subsequently observed that financial sector weaknesses need to be addressed immediately. While privatizing the two banks was seen as imperative, improving 312 Conclusion and Policy Implications the standard of prudential regulations for the financial system was also seen as important. The traditional development finance role of development banks should be extended to cover the mobilization of private sector resources. By mainstreaming and co- financing operations with the private sector, and creating favourable conditions through the restructuring of their development finance operations, development banks could promote private sector participation in economic activity. A sustained increase in long-term savings would require the Government to reduce its reliance on captive financing, thereby giving market participants greater flexibility in their portfolio decisions. While the development of an active market for securities of over one year remains a priority, it has to be supported by an institutional framework that includes efficient clearing and settlement arrangements, credit rating agencies and insurance. The introduction of new instruments and incentives through the establishment of a new policy framework would greatly assist in the development of the bond market. This would promote greater participation by the private sector in investment activity. In order to promote the further development of the capital market, the Government could set up a parallel stock exchange to facilitate the listing of smaller companies that do not adequately satisfy the criteria of the Colombo Stock Exchange.4 Less 4 Indonesia introduced such a system in 1987. See Binhadi (1994). 313 Conclusion and Policy Implications stringent capital adequacy requirements by the new stock exchange would give smaller companies the opportunity of participating actively in the capital market with access to lower-cost credit. A sustainable reduction in fiscal deficits would enable the hastening of financial sector reforms, as the need to depend on the financial sector for finance would be progressively lower. The budget deficit as a percentage of GDP has averaged – 10.5 in the 986–1996 period in comparison to -0.1, -0.5, -2.4, and 2.1 for South Korea, Indonesia, Malaysia and Thailand, respectively.5 The items that impose a strain on the budget are defence expenditure, interest on public debt, the civil service wage bill, pension bill, and social welfare expenditure. The civil service wage bill and the pension bill are areas in which cuts could be made. Therefore, greater efforts need to be made in curtailing current expenditure and diverting resources to their most productive uses. Reducing its borrowings to a permissible level would permit the Government to borrow directly from the financial markets in competition with the private sector. Further, the Government should ensure that the investments financed by these resources have equivalent rates of return to private sector investments. Sri Lanka’s revenue has averaged 20.2 % of GDP over the 1986–1996 period, with tax revenues accounting for 17.7% of GDP, and current expenditure 21.4% in the same period. From 1987 onward, Sri Lanka’s recurrent expenditure has exceeded 5 World Economic Outlook 1998 and Central Bank of Sri Lanka Annual Reports 314 Conclusion and Policy Implications its revenue. Therefore, widening the tax base is imperative, given the need to produce a surplus on the current account. Of concern in this respect is the position of industries set up under the Board of Investment (BOI) which enjoy a number of tax concessions. The preferential treatment given to these industries as against the non-BOI investment has caused a major distortion in the tax system discriminating against local investors. Moreover, these incentives are granted through bureaucratic administrative procedures that foster corruption. Therefore, a tax system without discretionary incentives and equitable rates would be better suited to both local and foreign investors. Further, indicators that assist in the formulation of monetary policy need to be developed. Given the oligopolistic structure of the banking system, interest rate liberalization alone will not guarantee optimal rates. Interest rates will have to be managed so as to reduce the spread between lending and borrowing rates, while concurrently ensuring positive real interest rates. External and internal shocks also affect the conditions under which the economy is functioning. Agricultural economies such as Sri Lanka are likely to be particularly vulnerable to such shocks, as climatic factors play an important role is these economies. Given that foreign interest rates are also subject to wide fluctuations, it is desirable that interest rate management policies be kept flexible. Over the longer term, a policy directed at price stability could overcome the uncertainty of the market and the public, and also limit the adverse impact on 315 Conclusion and Policy Implications interest rates and output. A policy framework based on inflation-targeting is a potentially promising approach that may be worth considering.6 Inflation-targeting includes:7 establishing targets for inflation (together with conditions under which the authorities should accept deviations from these targets); forecasting inflation based upon unchanged policies; and formulating and implementing changes in policy in response to deviations from targets. This would help build credibility through greater transparency in monetary policy, provided it is combined with increased accountability of the Central Bank for attaining price stability and is accepted by the Government and the public. If monetary policy is constrained by fiscal considerations, large fiscal deficits would give rise to inflationary pressures that would undermine the effectiveness of monetary policy in achieving a target rate of inflation. This, however, also requires the support of greater transparency in accounting standards and proper legal systems. Sustainability of economic policies can reduce private sector concerns about policy reversals and foster private sector confidence. Programmes could be based on medium-term fiscal frameworks that set medium-term targets for money creation, inflation and government debt. 6 This approach has been adopted in Canada, Finland, New Zealand, Australia, Sweden the U.K, Spain and Isreal. 7 See Lane, Griffiths and Prati (1995). 316 Conclusion and Policy Implications The coordination of fiscal, monetary and exchange rate policies are important in establishing policy credibility. A related issue concerns the price index to be employed in the inflation target. All inflation-targeting countries have employed some variant of the CPI. In the case of Sri Lanka, prior to inflation-targeting, improving the efficiency of the Colombo Consumer Price Index (CCPI) remains a crucial issue. As the CCPI is used to estimate certain components of the GDP deflator, the deflator does not provide a suitable alternative. The IM F (1995) has suggested several possible approaches to improving the efficiency of the CCPI. First, to remove the effect of duty waivers and administered price deductions from the index. Second, to estimate the general price level by examining the miscellaneous component comprising, among other items, household goods, alcohol, tobacco products, school fees, medicine, and transport, that are not affected by duty waivers and administered price changes. Third, to relate underlying inflation to monetary expansion; and fourth to examine the rate of increase of factors such as the rental price of capital and unit labour costs. The objective of transparency, however, would be best supported by the use of the CCPI. Therefore, the simplest approach might be to revise the overall CCPI by updating its base year, incorporating important present-day consumption items and making it more representative of price changes of the entire economy while concurrently eliminating the effect of duty waivers and administered price deductions from the index. 317 Conclusion and Policy Implications In view of the enhanced role of the exchange rate, given the increased capital inflows, the adoption of a monetary conditions index as an operational target of policy is an approach worth considering.8 The Bank of Canada has employed the concept of monetary conditions, which is a combination of the movement in interest rates and the exchange rate, in a similar manner to which interest rates were used in the past.9 As defined by Freeman (1994), this index is a combination of the changes in short-term interest rates (the 90-day commercial paper rate) and the multilateral exchange rate (the G-10 rate) from some arbitrary base period. Taking into consideration forecasts of factors such as the movements in foreign variables, domestic exogenous variables and the momentum of the economy, the Bank of Canada recommends a path for monetary conditions that would be consistent with a targeted outcome for the rate of inflation. If a shock hits the system, the path for the operational target is revised in order to counteract the impact of the shock on the ultimate target. As the mix of monetary conditions is primarily the result of market expectations, some mixes of monetary conditions may not be as favorable as others for the economy over time. A relative merit of the monetary conditions index is that it ensures the incorporation of variations in exchange rate in policy formulation. While the monetary conditions index is an useful measure of monetary conditions, its limitations as a policy measure should also be recognized. One, the precision of 8 This has already been pointed out by the Central Bank of Sri Lanka (1998). 9 See Freedman (1994). 318 Conclusion and Policy Implications the weights used in constructing the monetary conditions index is approximate and could vary over time. Although the monetary authorities is able to control the level of monetary conditions, it is not able to sustainably influence the mix of monetary conditions. Two, the monetary conditions index incorporates only two channels of policy influence on activity and inflation pressures, namely the interest and the exchange rate. Persistent current account deficits could lead to continued changes in the mix of monetary conditions. On the external front, the East Asian crisis serves to highlight the need for greater transparency and macroeconomic stability prior to capital account liberalization. Therefore, the Central Bank should proceed cautiously with respect to capital account liberalization. A sound macroeconomic environment conducive to the growth of private sector enterprise should precede external liberalization. Further, the structure of capital markets should be reinforced through appropriate policies in the areas of information and accounting mechanisms, regulation and supervision, property rights and taxation regimes. These will remain the main elements in determining the transfer of risk. Moreover, the authorities should be vigilant of capital flows generated by “herd instincts”, or flows that finance a consumption boom. Imposing temporary margins on credit growth could help to avert the risks associated with external liberalization. A conventional attraction of less developed countries for their cheap labour is progressively becoming less important in investment decisions. High-quality 319 Conclusion and Policy Implications productive labour and managerial discipline is also required to reinforce their comparative advantages. Therefore, the Government could assist in upgrading technology and labour skills. Thus, the achievements notwithstanding, a number of impediments to reform remain. Liberalization without substantial development of the financial sector will not be effective or sustainable. Success in reforming the financial sector will depend on the Government’s ability to attain and sustain macroeconomic stability, curtail large and persistent budget deficits and maintain its commitment to restructuring and privatizing the two state-owned commercial banks. It is important that monetary policy be insulated from the pressures created by the Government to finance its budget deficit. Increasing Central Bank autonomy will greatly assist in achieving this. An extensive programme to develop public debt management and securities markets is necessary to permit the Government to meet its financing requirements though the market, and pay market rates of interest on debt. The Central Bank could assist in the development of the market infrastructure, including payment and settlement systems, legal and regulatory framework of the markets, and introduction of longer-term market instruments. Hence, it is possible to conclude that financial deregulation will generate market efficiency only if the reform programme is designed to promote agents to acquire information that will be reflected in market prices. Continued intervention by the Government will limit the incentive faced by agents to acquire information. In 320 Conclusion and Policy Implications particular, as the Government is in a position to influence prices, information is likely to be less valuable to market participants when it is influenced than when the market process is left undisturbed. Thus, unless a concerted effort is made by the Government to reduce its dependence on the financial sector as a means of achieving its allocative goals, it could become a major impediment to further liberalization of the financial sector. 321 Bibliography BIBLIOGRAPHY Aghevli B B and Khan M S (1977) “Inflationary Finance and the Dynamics of Inflation: Indonesia, 1951-72,” American Economic Review 67(3), 390–403 Aigbokhan B E (1991) “The Naira Exchange Rate Depreciation and Domestic Inflation,” Indian Journal of Economics 71(283), 507–517 Akaike H (1973) “ Information Theory and an Extension of the Maximum Likelihood Principal,” in Petrov B N and Csaki F eds., Proceeding of the Second International Symposium on Information Theory, Akademiai Kiado, Budapest Akhtar M A (1975) "The Inflation Problem in Developing Economies: India and the Philippines," Indian Economic Journal 23(2), 144–165 Alexander S (1961) “Price Movements in Speculative Markets: Trends or Random Walks?” in Cootner (1964) ed., The Random Character of Stock Market Prices, MIT Press, MA Aliber R (1973) “The Interest Rate Parity Theorem: A Reinterpretation,” Journal of Political Economy 81(6), 1451–1459 Almon S (1965) “The Distributed Lag between Capital Appropriations and Expenditures,” Econometrica 33(1), 178–196 Amsler C E (1986) “The Fisher Effect: Sometimes Inverted, Sometimes Not?” Southern Economic Journal 52(3), 832–835 Apergis N and Eleftheriou S (1997) “The Efficient Hypothesis and Deregulation: The Greek Case,” Applied Economics 29(1), 111–117 322 Bibliography Athukorale P and Rajapathirana S (1993) "Domestic Financial Market and the Trade Liberalizaion Outcome: Evidence from Sri Lanka," World Development 21(7), 1191–1203 Atkins F J (1989) “Cointegration, Error Correction and the Fisher Effect,” Applied Economics 21(12), 1611–1620 Attfield, Demeray and Duck (1985) Rational Expectations in Macroeconomics: An Introduction to Theory and Evidence, Basil Blackwell Ltd., Oxford Ayadi O Felix and Pyun C S (1994) "An Application of Varience Ratio Test to the Korean Securities Market," Journal of Banking and Finance 18 (4), 643–658 Bachelier L (1900) “Theory of Speculation,” in Cootner (1964) ed., The Random Character of Stock Market Prices, MIT Press, MA Baillie R T, Lippens R and McMahon P C (1983) “Testing Rational Expectations and Efficiency on the Foreign Exchange M arket,” Econometrica 51(3), 553–563 Ball Ray (1988) "What do we know about Stock Market Efficiency" in Guimaraes R, Kingsman B and Taylor S eds., A Reappraisal of the Efficiency of Financial Markets NATO ASI Series, Springer- Verlag, Berlin Barnea A, Dotan A and Lakonishok J (1979) "The Effect of Price Level Uncertainty on the Determination of Nominal Interest Rates: Some Empirical Evidence" Southern Economic Journal 46(2), 609–614 Barnes Paul (1986) "Thin Trading and Stock Market Efficiency: The Case of the Kuala Lumpur Stock Exchange" Journal of Business Finance and Accounting 13(4), 609–617 323 Bibliography Barro R J and Gordon D B (1983) “Rules, Discretion and Reputation in a Model of M onetary Policy” Journal of Monetary Economics 12 (1), 101–121 Barro R J (1977a) “Unanticipated Money Growth and Unemployment in the United States,” American Economic Review 67(2), 101–115 Barsky R B (1987) "The Fisher Hypothesis and the Forecastability and Persistence of Inflation," Journal of Monetary Economics 19 (1), 3–24 Bayoumi T (1990) “Saving-Investment Correlations” IMF Staff Papers 37(2), 360– 387 Beaver W H (1981) “ Market Efficiency,” Accounting Review 56(1), 23–37 Begg D K H (1982) The Rational Expectations Revolution in Macroeconomics: Theories and Evidence, Phillip Allan, Oxford Beja A (1979) “State Preference and the Riskless Interest Rate – A Markov Model of Capital M arkets,” Review of Economic Studies 46(3), 435–446 Bekaert G and Hodrick R J (1993) “On Biases in the Measurement of Foreign Exchange Risk Premiums,” Journal of International Money and Finance 12(2), 115–138 Bekaert G, Hodrick R J and Marshall D A (1997) "On Biases in Tests of the Expectations Hypothesis of the Term Structure of Interest Rates," Journal of Financial Economics 44(3), 309–348 Bernanke B S and Mishkin F S (1997) “Inflation Targeting: A New Framework for M onetary Policy?” Journal of Economic Perspectives 11(2), 97–116 Bhatti R and Moosa I (1995) “An Alternative Approach to Testing Uncovered Interest Parity,” Applied Economics Letters 2(12), 478–481 324 Bibliography Bilson J (1981) “The Speculative Efficiency Hypothesis,” Journal of Business 54(3), 435–451 Binhadi (1994) “Financial Deregulation and Bank Supervision: The Case of Indonesia,” in Faruqi S ed; Financial Sector Reforms, Economic Growth and Stability: Experiences in Selected Asian and Latin American Countries, World Bank, Washington D.C Blanchard O J (1981) “Output, the Stock Market and Interest Rates,” American Economic Review 71(1), 132–143 Blejer M I and Khan M S (1983) “The Foreign Exchange Market in a Highly Open Developing Economy: The Case of Singapore,” Journal of Development Economics 12(1/2), 237–249 Bomberger W A and Makinen G E (1979) "Some Further Tests of the Harberger Inflation Model Using Quarterly Data," Economic Development and Cultural Change 27(4), 629–644 Bonham C S (1991) "Correct Cointegration Test of the Long Run Relationship Between Nominal Interest and Inflation," Applied Economics 23(9), 1487–1492 Brealy R and Schaefer S (1977) "Term Structure With Uncertain Inflation," Journal of Finance 32(2), 277–289 Brennan M J and Schwartz E S (1979) “Continuous Time Approach to the Pricing of Bonds,” Journal of Banking and Finance 3 (2), 133–155 Brenner R J and Kroner K F (1995) "Arbitrage, Cointegration, and Testing the Unbiasedness Hypothesis in Financial Markets," Journal of Financial and Quantitative Analysis 30(1), 23–42 325 Bibliography Brown R L, Durbin J and Evans J M (1975) “Techniques for Testing the Constancy of Regression Relat ions Over T ime,” Journal of the Royal Statistical Society Series B 37, 149–163 Cacizoneri M B (1985) “Monetary Policy Games and the Role of Private Information,” American Economic Review 75(5), 1056–1070 Cagan P (1956) “The Monetary Dynamics of Hyper-Inflation,” in Friedman M ed, Studies in the Quantity Theory of Money, University of Chicago Press, Chicago Campbell J and Mankiw N (1990) “Permanent Income, Current Income, and Consumption,” Journal of Business and Economic Statistics 8(3), 265–279 Campbell J and Shiller (1991) “Yield Spreads and Interest Rate Movements : A Bird’s Eye View,” Review of Economic Studies 58(3), 495–514 Campbell J Y and Deaton A S (1989) “Why is Consumption So Smooth,” Review of Economic Studies 56(3), 357–374 Campbell J Y and Mankiw N G (1989) “Consumption, Income, and Interest Rates: Reinterpreting the Time Series Evidence,” in Blanchard O and Fischer S eds., NBER Macroeconomics Annual 1989, MIT Press, MA Campbell J Y and Mankiw N G (1990) “Permanent Income, Current Income and Consumption,” Journal of Business and Economic Statistics 8(3), 265–279 Campbell J Y and Mankiw N G (1991) “The Response of Consumption to Income: A Cross country Investigation,” European Economic Review 35(4), 723–767 Campbell J Y and Shiller R J (1987) "Cointegration Tests of Present Value Models," Journal of Political Economy 95(5), 1062–1088 Campbell J Y, Lo A W and MacKinlay A C (1997) The Econometrics of Financial 326 Bibliography Markets, Princeton University Press, NJ Campbell J Y (1986) "A Defense of Traditional Hypotheses About the Term Structure of Interest Rates," Journal of Finance 41(1), 183–193 Cargill T F (1977) “Direct Evidence of the Darby Hypothesis for the United States,” Economic Inquiry 15(1), 132–134 Carlson J A (1977) "Short Term Interest Rates as Predictors of Inflation: Comment," American Economic Review 67(3), 469–475 Carmichael J and Stebbing P W (1983) “Fisher’s Paradox and the Theory of Interest,” American Economic Review 73(4), 619–630 Carr J, Pesando J E and Smith L B (1976) “Tax Effects, Price Expectations and the Nominal Rate of interest,” Economic Inquiry 14(1), 259–269 Cavaglia S M FG, Verschoor W F C and Wolff C C P (1994) “On the Biasedness of Forward Foreign Exchange Rates: Irrationality or Risk Premia?” Journal of Business 67(3), 321–343 Central Bank of Sri Lanka Annual Report: Various Issues, Central Bank of Sri Lanka, Colombo Central Bank of Sri Lanka (1990) Fortieth Anniversary Commemorative Volume of the Central Bank of Sri Lanka 1950–1990, Central Bank of Sri Lanka, Colombo Central Bank of Sri Lanka (1991) Growth and Social Progress in Sri Lanka, Statistics Department Central Bank of Sri Lanka, Colombo Central Bank of Sri Lanka (1998) Sri Lanka Socio-Economic Data 1999, Statistics Department Central Bank of Sri Lanka, Colombo Central Bank of Sri Lanka (1998), Economic Progress of Independent Sri Lanka, 327 Bibliography Central Bank of Sri Lanka, Colombo Central Bank of Sri Lanka Annual Review of the Economy: Various Issues, Central Bank of Sri Lanka, Colombo Chance D M (1985) " A Semi-Strong Form Test of the Efficiency of the Treasury Bond Futures Market," Journal of Futures Markets 5(3), 385–405 Chandavarkar A G (1971) "Some Aspects of Interest Rate Policies in Less Developed Economies: The Experience of Selected Asian Countries," IMF Staff Papers 18(1), 48–112 Charemza W W and Deadman D F (1992) New Directions in Econometric Practice, Edward Elgar, London Chinn M D and Frankel J A (1994) “Financial Links Around the Pacific Rim : 1982– 1992,” in Glick R and Hutchison M M eds., Exchange Rate Policy and Interdependence: Perspectives from the Pacific Basin Cambridge University Press, Cambridge Cho Y J (1984) “On the Liberalization of the Financial System and Efficiency of Capital Allocation Under Uncertainty,” Ph.D Dissertation Stanford University Cho Y J (1986) "Inefficiencies from Financial Liberalization in the Absence of Well- Functioning Equity Markets," Journal of Money, Credit and Banking 18(2), 191– 199 Choe C and Moosa I A (1999) “Financial System and Economic Growth: the Korean Experience” World Development 27(6), 1069–1082 Choi I (1994) “Residual Based Test for the Null of Stationarity with Application to US 328 Bibliography Macroeconomic Time Series,” Econometric Theory 10(3/4), 720–746 Choudhry A (1997) “Cointegration Analysis of the Inverted Fisher Effect: Evidence from Belgium, France and Germany,” Applied Economics Letters 4(4), 257–260 Coats W L Jr. And Khatkhate D R (1984) "Monetary Policy in Less Developed Countries: Main Issues," Developing Economies 22(4), 329–348 Cochrane John H (1991) "Volatility Tests and Efficient M arkets "Journal of Monetary Economics 27 (3), 463 –485 Colombage S S (1993) Monetary Policy in an Open Economy: The Experience of Sri Lanka, Sri Lanka Economic Association, Colombo Cooper Kerry S and Fraser D R (1982) The Financial Market Place, Addison–Wesley Publishing Co., MA Cootner P H (1964) “The Random Character of Stock Market Prices,” MIT Press, MA Cox J C, Ingersoll J E Jr. and Ross S A (1981) "A Reexamination of Traditional Hypothesis About the Term Structure of Interest Rates," Journal of Finance 36(4), 769–799 Cukierman A and Meltzer A H (1986) “A Theory of Ambiguity, Credibility and Inflation Under Discretion and Asymmetric Information,” Econometrica 54(5), 1099–1128 Cukierman A and Wachtel P (1979) "Differential Inflationary Expectations and the Variability of the Rate of Inflation : Theory and Evidence," American Economic Review 69(4), 595–609 Culbertson J M (1957) "The Term Structure of Interest Rates," Quarterly Journal of 329 Bibliography Economics 71(4), 485–517 Cumby R and Mishkin F (1986) “The International Linkage of Real Interest Rates : The European-US Connection,” Journal of International Money and Finance 5(1), 5–23 Cumby R E (1988) “Is it Risk? Explaining Deviations from Uncovered Interest Parity” Journal of Monetary Economics 22 (2), 279–299 Cumby R E and Obstfeld M (1981) “Exchange Rate Expectations and Nominal Interest Rates: A Test of the Fisher Hypothesis,” Journal of Finance 36(3), 697– 707 Cumby R E and Obstfeld M (1984) “International Interest -Rate and Price-Level Linkages Under Flexible Exchange Rates: A Review of Recent Evidence,” in Bilson J F and Marston eds., Exchange Rates: Theory and Practice University of Chicago Press, Chicago Daniels J P, Nourzad F and Toutkoushian R K (1996) "Testing the Fisher Effect as a Long Run Equilibrium Relation," Applied Financial Economics 6(2), 115–120 Darby M R (1975) “The Financial and Tax Effects of Monetary Policy on Interest Rates,” Economic Inquiry 13(1), 266–276 Davidson R and MacKinnon J G (1993) Estimation and Inference in Econometrics, Oxford University Press, Oxford Davis K (1980) “Bank Deregulation: An Appraisal,” Australian Financial System Inquiry Technical Paper No. 12 Bank Deregulation Seminar , Sydney Dawson S M (1984) "The Trend Toward Efficiency For Less Developed Stock Exchanges: Hong Kong," Journal of Business Finance and Accounting 11(2), 151– 330 Bibliography 161 Department of Census and Statistics (1998) Statistical Profile of Sri Lanka, Department of Census and Statistics, Colombo Department of Census and Statistics (1998) Labour Force Survey Fourth Quarter 1998, Department of Census and Statistics, Colombo Dickey D A and Fuller W A (1979) "Autoregressive Time Series With a Unit Root," Journal of the American Statistical Association 74(366), 427–431 Dickey D A and Fuller W A (1981) “ Likelihood Ratio Statistics for Autoregressive Time Series with a Unit Root,” Econometrica 49(4 ), 1057–1072 Dillen H (1997) "A Model of the Term Structure of Interest Rates in an Open Economy with Regime Shifts," Journal of International Money and Finance 16(5), 795–819 Diz A C (1970), "Money and Prices in Argentina 1935–1962," in Meiselman D, ed; Varieties of Monetary Experience, University of Chicago Press, Chicago Dooely M P, Frankel J and M athieson D (1987), “ International Capital M obility: What Do Saving-Investment Correlations Tell Us?,” IMF Staff Papers 34(3), 503– 530 Dooley M P (1996) “A Survey of Literature on Controls over International Capital Transactions,” IMF Staff Papers 43(4), 639–687 Dooley M P and Isard P (1980) “Capital Controls, Political Risk, and Deviations from Interest Parity,” Journal of Political Economy 88(2), 370–384 Dooley M P and Mathieson D J (1994) “Exchange Rate Policy, International Capital M obility, and M onetary Policy Instruments,” in Glick R and Hutchison M M eds., 331 Bibliography Exchange Rate Policy and Interdependence: Perspectives from the Pacific Basin Cambridge University Press, Cambridge Drake P J (1977) "Securities Markets in Less-Developed Countries," Journal of Development Economics 13(2), 73–91 Driffill J, Psaradakis Z and Sola M (1997) “A Reconciliation of Some Paradoxical Empirical Results on the Expectations Model of the Term Structure,” Oxford Bulletin of Economics and Statistics, 59(1), 29–42 Dunham D and Kelegama (1997) “Does Leadership Matter in the Economic Reform Process? Liberalization and Governance in Sri Lanka, 1989–1993,” World Development 25(2), 179–190 Dunham D and Kelegama S (1994) Stabilization and Liberalizaation: A Closer Look at the Sri Lankan Experience 1977–1993, Research Studies Macroeconomic Series No.8, Institute of Policy Studies, Colombo Dutt S D and Ghosh D (1995) "The Fisher Hypothesis: Examining the Canadian Experience," Applied Economics 27(11), 1025–1030 Dyckman T R and Morse D (1986) Efficient Capital Markets and Accounting: A Critical Analysis , Second Edition , Englewood Cliffs, NJ Eaton J and Turnovsky S (1983), “Covered Interest Parity, Uncovered Interest Parity and Exchange Rate Dynamics,” Economic Journal 93(371), 555–575 Ebrill L P and Fries S M (1991) "Broad Money Growth and Inflation in the United States," IMF Staff Papers 38(4), 736–750 Edey M and Hviding K (1995) “An Assessment of Financial Reform in OECD Countries,” Working Paper OECD No.154, Paris 332 Bibliography Edwards S (1983) “Floating Exchange Rates, Expectations and New Information,” Journal of Monetary Economics 11(3), 321–336 Edwards S and Khan M (1985) "Interest Rate Determination in Developing Countries: A Conceptual Framework ," IMF Staff Papers 32(3), 377–403 Edwards S (1988) "Financial Deregulation and Segmented Capital Markets : The Case of Korea," World Development 16(1), 185–194 El-Erian M A and Kumar M (1995) “Emerging Equity Markets in Middle Eastern Countries,” IMF Staff Papers 42(2), 313–343 Elyasiani E, Perera P and Puri T (1996), “Market Efficiency and Calendar Anomalies in Emerging Capital M arkets: Evidence from the Colombo Stock Exchange,” Journal of International Financial Markets, Institutions and Money 6(4), 59–77 Enders W (1995) Applied Econometric Time Series, John Wiley & Sons Inc., New York Engle R F and Granger C W C (1987) "Co-integration and Error Correction: Representation, Estimation and Testing," Econometrica 55(2), 251–276 Engle R F, Lilien D M and Robins R P (1987) “Estimating Time Varying Risk Premia in the Term Structure: The ARCH-M Model” Econometrica 55(2), 391–407 Engsted T (1995) "Does the Long Term Interest Rate Predict Future Inflation? A Multi-Country Analysis," Review of Economics and Statistics 77(1), 42–54 Engsted T and Tanggaard C (1994) "Cointegration and the US Term Structure," Journal of Banking and Finance 18 (1), 167–181 Errunza V R (1979) "Efficiency and the Programs to Develop Capital Markets The Brazilian Experience," Journal of Banking and Finance 3 (4), 355–382 333 Bibliography Estrella A and Hardouvelis G A (1991) " The Term Structure as a Predictor of Real Economic Activity," Journal of Finance 46(2), 555–576 Fahimuddin and Rao B B (1997) "Unit Roots in Selected Indian Macroeconomic Time Series," Paper Presented at Staff Seminar Giri Institute of Development Studies, Lucknow Fama E F (1965) “The Behaviour of Stock Market Prices,” Journal of Business 38(1), 34–105 Fama E F (1975) "Short Term Interest Rates as Predictors of Inflation," American Economic Review 65(3), 269–282 Fama E F (1976) "Inflation Uncertainty and Expected Returns on Treasury Bills," Journal of Political Economy 84(3), 427–448 Fama E F (1977) " Interest Rates and Inflation: The Message in the Entrails," American Economic Review 67(3), 487–496 Fama E F (1984) “Forward and Spot Exchange Rates,” Journal of Monetary Economics 14(3), 319–338 Fama E F and Gibbons M R (1984) "A Comparison of Inflation Forecasts," Journal of Monetary Economics 13(3), 327–348 Fama E F (1984) "The Information in the Term Structure," Journal of Financial Economics 13(4), 509–528 Fama E F (1990) "Term Structure forecasts of Interest Rates, Inflation and Real Returns," Journal of Monetary Economics 25(1), 59–76 Fama E F and Bliss R R (1987) “The Information in Long Maturity Forward Rates, American Economic Review 77(4), 680–692 334 Bibliography Fama E F, Fisher L, Jensen M, Roll R (1969) "The Adjustment of Stock Prices to New Information " International Economic Review 10(1), 1– 21 Fama E F (1970) "Efficient Capital Markets : A Review of Theory and Empirical Work " Journal of Finance 25(2), 383– 417 Fama E F (1976) Foundations of Finance , Basic Books, New York Fama E F (1991) "Efficient Capital Markets: II," Journal of Finance 46(5), 1575– 1617 Faruqee H (1992), “Dynamic Capital Mobility in Pacific Basin Developing Countries: Estimation and Policy Implications,” IMF Staff Papers 39 (3 ), 706–717 Feldman D (1989) "The Term Structure of Interest Rates in a Partially Observable Economy," Journal of Finance 44(3), 789–812 Feldstein M (1980a) “Tax Rules and the Mismanagement of M onetary Policy,” American Economic Review 70(2) Supplement, 182–209 Feldstein M (1983), “Domestic Saving and International Capital Movements in the Long Run and the Short Run,” European Economic Review 21(1/2), 129–151 Feldstein M and Horioka C (1980) “ Domestic Saving and International Capital Flows,” Economic Journal 90(358), 314–329 Feldstein M S (1976) “Inflation, Income Taxes and the Rate of Interest: A Theoretical Analysis,” American Economic Review 66(5), 809–820 Fernando G (1994) "A Review of Financial Reforms as a Continuing Process and Sri Lankan Experience in Selected Areas," Bankers Journal Sri Lanka 13(1), 11–24 335 Bibliography Fieleke N (1982) “National Saving and International Investment,” in Saving and Government Policy,” Conference Series No.25 Boston Federal Reserve Bank, Boston Finnerty J E (1976) “ Insiders and Market Efficiency,” Journal of Finance 31(4), 1141–1148 Fisher I (1930) The Theory of Interest, M acmillan, New York Fisher W H and Turnovsky S J (1992) "Fiscal Policy and the Term Structure of Interest Rates: An Intertemporal Optimizing Analysis," Journal of Money, Credit, and Banking 24(1), 1–26 Flood R and Rose A (1996) “Fixes: of the Forward Discount Puzzle,” Review of Economics and Statistics 78(4), 748–752 Frankel J A (1986) “International Capital Mobility and Crowding Out in the U S Economy: Imperfect Integration of Financial Markets or of Goods Markets?” NBER Working Paper No.1773 in Hafer R ed., How Open is the U S Economy? Lexington, MA Frankel J A (1986) “The Implications of Mean-Variance Optimization for Four Questions in International Macroeconomics,” Journal of International Money and Finance 5 Supplement, 53–75 Frankel J A and MacArthur A T (1988) “Political vs. Currency Premia in International Real Interest Differentials: A Study of Forward Rates for 24 Countries,” European Economic Review 32(5), 1083–1121 Freedman C (1994) “The Use of Indicators and of the Monetary Conditions Index in 336 Bibliography Canada,” in Balino T J T and Cottarelli C eds., Frameworks for Monetary Stability: Policy Issues and Country Experiences, IMF, Washington D.C Frenkel J A and Levich R M (1975) “Covered Interest Arbitrage: Unexploited Profits?” Journal of Political Economy 83(2), 325–338 Frenkel J A and Levich R M (1977) “Transaction Costs and Interest Arbitrage: Tranquil versus Turbulent Periods,” Journal of Political Economy 85(6), 1209– 1226 Friedman M and Schwartz A (1982) Monetary Trends in the U S and the U K: Their Relation to Income, Prices and Interest Rates 1867–1975, University of Chicago Press, Chicago Froot K A and Frankel J A (1989) “Forward Discount Bias: Is it an Exchange Risk Premium?” Quarterly Journal of Economics 104 (1 ), 139–161 Froot K A and Thaler R H (1990) “Anomalies: Foreign Exchange,” Journal of Economic Perspectives 4(3), 179–192 Froot K A (1989) "New Hope for the Expectations Hypothesis of the Term Structure of Interest Rates," Journal of Finance 44(2), 283–305 Fry M J (1978) “Money and Capital or Financial Deepening in Economic Development? ” Journal of Money, Credit and Banking 10(4 ), 464–475 Fry M J (1978) “ Money and Capital or Financial Deepening in Economic Development? ” Journal of Money, Credit and Banking 10(4), 464–475 Fry M J (1980) “Saving, Investment, Growth and the Cost of Financial Repression,” World Development 8(4), 317–327 Fry M J (1993) “ Direct Investment in Southeast Asia: Differential Impacts” ASEAN 337 Bibliography Economic Research Unit, Institute of Southeast Asian Studies Fry M J (1995 ) Money, Interest and Banking in Economic Development, Second Edition, John Hopkins University Press, Baltimore Fry M J (1995) " Financial Development in Asia: Some Analytical Issues," Asian Pacific Economic Literature 19(1), 40–57 Fuller W A (1976) Introduction to Statistical Time Series, Wiley, New York Fung Hung-Gay and Isberg S C (1992) "The International Transmission of Eurodollar and US Interest Rates: A Cointegration Analysis," Journal of Banking and Finance 16 (4), 757–769 Gaab W, Granziol M and Horner M (1986), “On Some International Parity Conditions: An Empirical Investigation,” European Economic Review 30(3), 683–713 Galbis V (1977) “ Financial Intermediation and Economic Growth in Less Developed Countries: A Theoretical Approach,” Journal of Development Studies 13(2), 58–72 Gandhi D K, Saunders A and Woodward R S (1980) "Thin Capital Markets: A Case Study of the Kuwaiti Stock Market," Applied Economics 12(3), 341–349 Gandolfi A E (1982) “Inflation, Taxation and Interest Rates,” Journal of Finance 37(3), 797–807 Garcia M G P (1993) " The Fisher Effect in a Signal Extraction Framework : The Recent Brazilian Experience," Journal of Development Economics 41(1), 71–93 Geweke J and Feige (1979) “Some Joint Tests of the Efficiency of Markets for Forward Foreign Exchange,” Review of Economics and Statistics, 61(3), 334–341 Ghosh A (1995) “International Capital Mobility Amongst the Major Industrialised Countries : Too Little or Too Much,” Economic Journal 105(428), 107–128 338 Bibliography Gibson W E (1970) "Price-Expectations Effects on Interest Rates," Journal of Finance 25(3), 19–34 Gibson W E (1972) "Interest Rates and Inflationary Expectations: New Evidence," American Economic Review 62(5), 854–865 Glick R and Hutchison M (1990) “Financial Liberalization in the Pacific Basin: Implications for Real Interest Rate Linkages,” Journal of the Japanese and International Economies 4(1), 36–48 Glick R and Hutchison M M (1994) Exchange Rate Policy and Interdependence: Perspectives from the Pacific Basin, Cambridge University Press, Cambridge Gochoco M S (1991) "Financial Liberalization and Interest Rate Determination : The Case of the Philippines 1981–1985," Journal of Macroeconomics 13(2), 335–350 Godfrey L G (1978a) “Testing Against General Autoregressive and Moving Average Error M odels when the Regressors Include Lagged Dependent Variables,” Econometrica 46(6), 1293–1301 Godfrey L G (1978b) “Testing for Higher Order Serial Correlation in Regression Equations when the Regressors Include Lagged Dependent Variables,” Econometrica 46(6), 1303–1310 Golub S (1990) “International Capital Mobility : Net Versus Gross Stocks and Flows,” Journal of International Money and Finance 9(4) , 424–439 Gonzales J, Spencer R and Walz D (1999) “ The Information in the Mexican Term Structure of Interest Rates: Capital Market Implications,” Journal of International Financial Markets, Institutions and Money 9(2), 149–161 339 Bibliography Graham F C (1988) “The Fisher Hypothesis: A Critique of Recent Results and Some New Evidence,” Southern Economic Journal 54(4), 961–968 Granger C W C (1986) "Developments in the Study of Cointegrated Economic Variables," Oxford Bulletin of Economics and Statistics 48(3), 213–227 Granger C W J and Newbold P (1974) “Spurious Regressions in Econometrics,” Journal of Econometrics 2( 2), 111–120 Groenewold N and Kang K C (1993) "The Semi-Strong efficiency of the Australian Share M arket," Economic Record 69(207), 405–410 Grossman J (1981) "The 'Rationality' of Money Supply Expectations and the Short Run Response of Interest Rates to Monetary Surprises," Journal of Money, Credit and Banking 13(4), 409–424 Grossman S (1976) " On the Efficiency of Competitive Stock Markets where Traders have Diverse Information " Journal of Finance 31(2) 573–585 Grossman S (1995) "Dynamic Asset Allocation and the Informational Efficiency of Markets ", Journal of Finance 50(3), 773– 787 Grossman S and Stiglitz J (1980) "On the Impossibility of Informationally Efficient Markets ", American Economic Review 70(3), 393– 408 Guimaraes R M C, Kingsman B G and Taylor S J (1988) eds., A Reappraisal of the Efficiency of Financial Markets, NATO ASI Series, Springer–Verlag, Berlin Gujarati D (1995) Basic Econometrics, Third Edition, Mc-Graw Hill Inc., NY Gunatilake L K (1994) "Financial Innovation – Global Trends and the Experience of Sri Lanka," Bank ers Journal Sri Lanka 13(1), 45–53 Gunatilleke G (1993) Development and Liberalization in Sri Lanka Trends and 340 Bibliography Prospects, Marga Institute, Colombo Gweke J and Feige E (1979) “Some Joint Tests of the Efficiency of Markets for Forward Foreign Exchange,” Review of Economics and Statistics 61(3), 334–341 Hakkio C S (1981) “ Expectations and the Forward Exchange Rate,” International Economic Review 22(3), 663–678 Hakkio C S and Rush M (1989) “ Market Efficiency and Cointegration: An Application to the Sterling and the Deutschemark Exchange Markets,” Journal of International Money and Finance 8 (1), 75–88 Hall A D, Anderson M H and Granger C W J (1992), "A Cointegration Analysis of Treasury Bill Yields," Review of Economics and Statistics 74(1), 116–126 Ham J and Choi W (1991) “The Analysis of the Determination of Interest Rate: The Case of Korea,” Monthly Bulletin of the Bank of Korea, 3–49 Hamori S (1997) "A Simple Method to Test the Fisher Effect," Applied Economics Letters 4(8), 477–479 Hansen L and Hodrick R (1980) "Forward Exchange Rates as Optimal Predictors of Future Spot Rates : An Econometric Analysis," Journal of Political Economy 88(5), 829–853 Haque N and Montiel P (1991) “Capital Mobility in Developing Countries: Some Empirical Tests,” World Development 19(10), 1391–1398 Haque N and Montiel P (1991), “How Mobile is Capital in Developing Countries?,” Finance and Development 28(3), 38–39 Harberger A C (1963) "The Dynamics of Inflation in Chile," in: Christ C et al eds., Measurement in Economics: Studies in Mathematical Economics and Econometrics 341 Bibliography in Memory of Yehuda Grunfeld, Stanford University Press, CA Hardouvelis G A (1988) "The Predictive Power of the Term Structure During Recent M onet ary Regimes," Journal of Finance 43(2) , 339–356 Hardouvelis G A (1994) " The Term Structure Spread and Future Changes in Long and Short Rates in the G7 Countries: Is there a Puzzle?," Journal of Monetary Economics 33(2), 255–283 Hardouvelis G A (1987) "Reserves Announcements and Interest Rates : Does Monetary Policy Matter?," Journal of Finance 42(2), 407–422 Hardy D C (1997) “Informational Efficiency, Interest Rate Variability, and Central Bank Operations,” IMF Working Paper WP/97 /26, IMF, Washington D.C Hargis K and Maloney W (1997) "Emerging Equity Markets: Are They For Real?," Journal of Financial Research 20(2), 243–262 Harris L (1981) Monetary Theory, McGraw-Hill, New York Harvey C R (1988) “The Real Term Structure and Consumption Growth,” Journal of Financial Economics 22(2), 305–333 Hawtrey K M (1997) "The Fisher Effect and Australian Interest Rates," Applied Financial Economics 7(4), 337–346 Hess P J and Bicksler J L (1975) “Capital Asset Prices Versus Time Series Models as Predictors of Inflation: The Expected Real Rate of Interest and Market Efficiency,” Journal of Financial Economics 2(4), 341–360 Hettiarachchi W (1988) “Globalization – Liberalizing the Capital Account in Sri Lanka,” Macroeconomic Series No.11, Institute of Policy Studies, Colombo Hettiarachchi W and Herat U (1995) “Coping with Capital Inflows: The Case of Sri 342 Bibliography Lanka,” Paper presented at the Eleventh Pacific Basin Central Bank Seminar on Capital Flows, Hong Kong Hettiarrachchi W (1986) "Growth and Expansion of Banking," in Rasaputra W, Tilakaratna W M, Fernando S T G and Fernando L E N eds., Facets of Development in Independent Sri Lanka: Ronnie de Mel Felicitation Volume, Ministry of Finance and Planning Sri Lanka Hickman W B (1942) “The Term Structure of Interest Rates: An Exploratory Analysis,” in Kessel R A (1965) The Cyclical Behaviour of the Term Structure of Interest Rates,” NBER Occasional Paper 91, NBER, NY Hicks J R (1939) Value and Capital: An Inquiry into Some Fundamental Principals of Economic Theory, Clarendon Press, Oxford Holmes M J and Wu Y (1997) “Capital controls and Covered Interest Parity in the EU: Evidence from a Panel-Data Unit Root Test,” Weltwirtschaftliches Archiv 133(1), 76–89 Hu Z (1993) “The Yield Curve and Real Activity,” IMF Staff Papers 40(4), 781–806 Huizinga J and Mishkin F S (1984) “Inflation and Real Interest Rates on Assets with Different Risk Characteristics,” Journal of Finance 39(3), 699–712 IMF (1995) Staff Country Report No.95/60, IMF, Washington D.C IMF (1996) Staff Country Report No.96/100, IMF, Washington D.C IMF (1997) Staff Country Report No.97/95 , IMF, Washington D.C IMF (1998) World Economic Outlook, IMF, Washington D.C Inder B and Slvapulle P (1993) “Does the Fisher Effect Apply in Australia?” Applied Economics 25(6), 839–843 343 Bibliography Institute of Policy Studies (1995) Sri Lanka: State of the Economy, Institute of Policy Studies, Colombo International Monetary Fund International Financial Statistics, IMF, Washington D.C Ippolito R A (1989) “Efficiency with Costly Information: A Study of Mutual fund Performance, 1965–1984,” Quarterly Journal of Economics 104(1), 1–23 Ismail D and Khatkhate D (1995) " Regulated Deregulation of the Financial System," World Bank Discussion Paper No. 292, World Bank, Washington D.C Jacques K (1995) "Unit Roots, Interest Rate Spreads and Inflation Forecasts," Applied Economics 27(7), 605–608 Jarque C M and Bera A K (1980) “ Efficient Tests for Normality, Homoscedasticity and Serial Independence of Regression Residuals,” Economics Letters 6(3), 255– 259 Jayamaha R (1986) "Changes in Monetary Policy," in Rasaputra W, Tilakaratna W M, Fernando S T G and Fernando L E N eds., Facets of Development in Independent Sri Lanka: Ronnie de Mel Felicitation Volume, Ministry of Finance and Planning Sri Lanka Jennergren L P and Korsvold P E (1975) "The Non-Random character of Norwegian and Swedish Stock Market Prices," in Elton E J and Gruber M J eds., Internal Capital Markets, North Holland Publishing Co., Amsterdam Jensen M C (1972) "Capital Markets: Theory and Evidence," Bell Journal of Economics and Management Science, 3(2), 357–398 Jensen M C (1978) "Some Anomalous Evidence Regarding M arket Efficiency " Journal of Financial Economics 6(2/3), 95–101 344 Bibliography Johansen S (1988) "Statistical Analysis of Cointegration Vectors," Journal of Economic Dynamics and Control 12(2/3), 231–254 Johansen S and Juselius K (1990) “Maximum Likelihood Estimation and Inference on Cointegration – with Applications to the Demand for Money" Oxford Bulletin of Economics and Statistics 52(2), 169–210 Johnson D R (1993) "International Interest Rate Linkages in the Term Structure," Journal of Money, Credit, and Banking 25(4), 755–770 Johnson O E G (1984) "On Growth and Inflation in Developing Countries," IMF Staff Papers 31(4), 636–660 Joines D (1977) “Short Term Interest Rates as Predictors of Inflation: Comment,” American Economic Review 67(3), 469–475 Jones D S and Roley V V (1983) "Rational Expectations and the Expectations Model of the Term Structure: A Test Using Weekly Data," Journal of Monetary Economics 12(3), 453–465 Jordan J (1983) "On the Efficient Market Hypothesis," Econometrica 51 (5), 1325– 1343 Judge G, Hill R, Griffiths W, Lutkepohl H and Lee T (1988) Introduction to the Theory and Practice of Econometrics, Second Edition, John Wiley, NY Juttner D J and Hawtrey K M (1997) Financial Markets, Money and Risk, Fourth Edition, Longman, Melbourne Juttner J (1980) “Entry to Banking,” Australian Financial System Inquiry Technical Paper No. 12 Bank Deregulation Seminar, Sydney Karfakis C I and Parikh A (1994) “Uncovered Interest Parity Hypothesis for Major 345 Bibliography Currencies,” Manchester School 62(2), 184–198 Karunasena A G (1996) "Changing Financial Systems in Open Economies: Sri Lanka's Experience," Bank for International Settlements Policy Papers No. 1, 114 –125 Karunatileke H N S (1973) Central Banking and Monetary Policy in Sri Lanka, Lake House, Colombo Karunatileke H N S (1987 ) The Economy of Sri Lanka, Centre for Demographic and Socio-Economic Studies, Colombo Keane S M (1983) Stock Market Efficiency: Theory, Evidence and Implications, Phillip Allan, Oxford Kendall M G (1953) “The Analysis of Economic Time Series, Part 1: Prices,” in Cootner (1964) ed., The Random Character of Stock Market Prices, MIT Press, MA Kessel R A (1965) “The Cyclical Behaviour of the Term Structure of Interest Rates,” NBER Occasional Paper 91, NY Khatkhate D (1982) "Anatomy of Financial Retardation in a Less Developed Country : The Case of Sri Lanka, 1951–1976," World Development 10(9), 829–840 Khatkhate D (1982) "National and International Aspects of Financial Policies in LDCs: A Prologue” World Development 10(9), 689–694 Khatkhate D (1988) "Assessing the Impact of Interest Rates in Less Developed Countries," World Development 16(5), 577–588 Khor H and Rojas-Suarez L (1991) “Interest Rates Interest in Mexico: The Role of Exchange Rate Expectations and International Creditworthiness,” IMF Staff Papers 38(4), 850–871 Kiem D B and Stambaugh R F (1986) "Predicting Returns in the Stock and Bond 346 Bibliography Markets," Journal of Financial Economics 17(2), 357–390 Kihlstrom R and Mirman L (1975) "Information and Market Equilibrium" Bell Journal of Economics 6(1), 357– 376 Kim S B (1993), “Do Capital Controls Affect the Response of Investment to Saving? Evidence from the Pacific Basin,” Economic Review Federal Reserve Bank of San Francisco(1), 23–39 Kim S M (1989) “Monetary Aggregate and Market Rate of Interest as the Intermediate Target of Monetary Policy,” Monthly Bulletin of the Bank of Korea, 17–30 Kindleberger Charles P (1987) "Financial Deregulation and Economic Performance : An Attempt to Relate European Financial History to Current LDC Issues," Journal of Development Economics 27(1/2) Special Issue, 339–353 Koyck L M (1954) Distributed Lags and Investment Analysis, North Holland Publishing Co., Amsterdam Krol R (1996), “ International Capital Mobility: Evidence from Panel Data,” Journal of International Money and Finance, 15(3), 467–474 Kuen Tse Yiu and Song Tan Kim (1996) “Interest Parity and Dynamic Capital M obility: The Experience of Singapore,” in Ito T and Krueger A O eds., Financial Deregulation and Integration in East Asia, NBER- East Asia Seminar on Economics Volume 5 , Chicago University Press, Chicago Kuo S W Y (1990) "Liberalization of the Financial Market in Taiwan in the 1980s," in Rhee S G and Chang R P eds., Pacific-Basin Capital Markets Research, North- Holland, Amsterdam 347 Bibliography Lahiri K (1976) "Inflationary Expectations: Their Formation and Interest Rate Effects," American Economic Review 66(1), 124–131 Lahiri K and Lee J (1979) “Tests of Rational Expectations and Fisher Effect,” Southern Economic Journal 46(2), 413–424 Lai K S (1997) "Is the Real Interest Rate Unstable? Some New Evidence," Applied Economics 29(3), 359–364 Lai K S and Lai M (1991) “A Cointegration Test for Market Efficiency,” Journal of Futures Markets 11(5), 567–575 Lane T D, Griffiths M and Prati A (1995) “Can Inflation Targets Help Make Monetary Policy Credible?” Finance and Development 32(4), 20–23 Lanyi A and Saracoglu R (1983) “Interest Rate Policies in Developing Countries,” IMF Occasional Paper 22, IMF, Washington D.C Latham Mark (1986) "Informational Efficiency and Information Subsets" Journal of Finance 41(1), 39 –52 Lee Sang-Sub (1995) "Macroeconomic Sources of Time-Varying Risk Premia in the Term Structure of Interest Rates," Journal of Money, Credit, and Banking 27(2), 549–569 Leland H E and Toft K B (1996) "Optimal Capital Structure, Endogenous Bankruptcy, and the Term Structure of Credit Spreads," Journal of Finance 51(3), 987–1019 LeRoy S F (1976) " Efficient Capital Markets: Comment " Journal of Finance 31(1), 139–141 LeRoy S F (1989) “Efficient Capital Markets and Martingales,” Journal of Economic Literature 27(4), 1583–1621 348 Bibliography LeRoy S F and Porter R D (1981) “The Present Value Relation: Tests Based on Implied Variance Bounds,” Econometrica 49(3) , 555–574 Letite S P (1982) "Interest Rate Policies in West Africa," IMF Staff Papers 29(1), 48– 76 Letite S P and Sundararajan V (1990) "Issues in Interest Rate Management and Liberalization," IMF Staff Papers 37(4), 735–752 Levi E and Nobay R (1986) “The Speculative Efficiency Hypothesis: A Bivariate Analysis,” Economic Journal 96 Conference Supplement, 109–121 Levi M D and Makin J H (1978) "Anticipated Inflation and Interest Rates: Further Interpretation of Findings and the Fisher Equation," American Economic Review 68(5), 801–812 Levi M . D and Makin J H (1979) " Fisher, Phillips, Friedman and the M easured Impact of Inflation on Interest," Journal of Finance 34(1), 35–52 Levine R (1991) "Stock Markets, Growth, and Tax Policy," Journal of Finance 46(4), 1445–1465 Lin B (1999) “ Fitting the Term Structure of Interest Rates for Taiwanese Government Bonds,” Journal of Multinational Financial Management 9(3/4), 331–352 Lizondo J S (1983) “Interest Differential and Covered Arbitrage,” in Armella, Dornbusch and Obstfeld eds., Financial Policies and the World Capital Marke: The Problem of Latin American Countries, University of Chicago Press, Chicago Longstaff F A (1990) "Time Varying Term Premia and Traditional Hypothesis About the Term Structure," Journal of Finance 45(4), 1307–1314 Longstaff F A (1992) "Multiple Equilibria and Term Structure Models," Journal of 349 Bibliography Financial Economics 32(3), 333–344 Loopesko (1984) “Relationships Among Exchange Rates, Intervention and Interest Rates: An Empirical Investigation,” Journal of International Money and Finance 3(3), 257–277 Lowinger T C (1978) "Domestic Inflation and Exchange Rate Changes: The Less- Developed Countries' Case," Weltwirtschaftliches Archiv 114(1), 85–99 Lutz F A (1940) "The Structure of Interest Rates," Quarterly Journal of Economics 55, 36–63 Lynch D (1996) “Asia Pacific Bond Markets,” CMBF Paper No.16, Centre for Studies in Money, Banking and Finance, Maquarie University, Sydney Macaulay F R (1938) “The Movements of Interest Rates, Bond Yields and Stock Prices in the United States since 1856,” NBER Working Paper No.33, NBER, NY M acDonald R (1988) “Floating Exchange Rates: Theories and Evidence,” Unwin Hyman Ltd, London MacDonald R and Murphy P D (1989) "Testing for the Long Run Relationship Between Nominal Interest Rates and Inflation Using Cointegration Techniques," Applied Economics 21(4), 439–447 MacDonald R and Speight A E (1991) “The Term Structure of Interest Rates Under Rational Expectations: Some International Evidence,” Applied Financial Economics 1(4), 211–221 MacDonald R and Taylor M P (1989) “Interest Rate Parity: Some New Evidence,” Bulletin of Economic Research 41(3), 255–274 MacDonald R and Taylor M P (1993) "The Monetary Approach to the Exchange Rate: 350 Bibliography Rational Expectations, Long Run equilibrium, and Forecasting," IMF Staff Papers 40(1), 89–107 MacDonald R and Torrance T S (1990) “Expectations Formation and Risk in Four Foreign Exchange Markets,” Oxford Economic Papers 42(3 ), 544–561 MacKinnon J G (1990) "Critical Values for Cointegration Tests," in Engle R F and Granger C W J eds., Long Run Economic Relationships: Readings in Cointegration,” Oxford University Press, Oxford Maddala G S (1992) Introduction to Econometrics, Second Edition, MacMillan Publishing Co., NY Malkiel B (1970) The Term Structure of Interest Rates: Theory, Empirical Evidence, and Applications, McCaleb Seiler Publishing Co., NY M andelbrot B (1966) “Forecasts of Future Prices, Unbiased Markets and ‘M artingale’ Models” Journal of Business 39(1), 242–255 Mankiw N G (1986) "The Term Structure of Interest Rates Revisited," Brookings Papers on Economic Activity 1, 61–96 Mankiw N G and Miron J A (1986) "The Changing Behaviour of the Term Structure of Interest Rates," Quarterly Journal of Economics 101(2), 211–228 Mankiw N G and Summers L H (1984) “Do Long Term Interest Rates Overreact to Short term Interest Rates?” Brookings Papers on Economic Activity 1, 223–242 M argaritis D (1994) “Time Varying Risk Premia in the Term Structure of Interest Rates in New Zealand ,” Applied Financial Economics 4(2), 111–120 Mark N C (1985) “On Time Varying Risk Premia in the Foreign Exchange Market: An Econometric Analysis” Journal of Monetary Economics 16(1), 3–18 351 Bibliography Mathieson D J (1980) “ Financial Reform and Stabilization Policy in a Developing Economy,” Journal of Development Economics 7(3), 359–395 Mathieson D J (1982) " Inflation, Interest Rates and the Balance of Payments During a Financial Reform: The Case of Argentina," World Development 10(9), 813–827 McCallum B T (1976) “Rational Expectations and the Estimation of Econometric M odels: An Alternative Procedure,” International Economic Review 17(2), 484– 490 McCallum B T (1994) “A Reconsideration of the Uncovered Interest Parity Relationship,” NBER Working Paper No. 4113, NBER, MA McCallum B T (1996) International Monetary Economics, Oxford University Press Inc., Oxford McKinnon R I (1973) Money and Capital in Economic Development, Brookings Institution, Washington D.C McKinnon R I (1991) The Order of Economic Liberalization, John Hopkins Press, Baltimore M cNowan R and Wallace M S (1994) "Cointegration Tests of the Monetary Exchange Rate Model for Three High-Inflation Economies," Journal of Money, Credit, and Banking 26(3), 396–411 Meiselman D (1962) The Term Structure of Interest Rates, Prentice-Hall Inc., N.J Melino Angelo (1988) "The Term Structure of Interest Rates: Evidence and Theory," Journal of Economic Surveys 2(4), 335–366 Melvin M and Schlagenhauf D (1985) “A Country Risk Index: Econometric Formulation and an Application to Mexico,” Economic Inquiry 23(4), 601–619 352 Bibliography Mendoza E G (1992) "Fisherian Transmission and Efficient Arbitrage Under Partial Financial Indexation: The Case of Chile," IMF Staff Papers 39(1), 121–147 M ichener R (1984) “Permanent Income in General Equilibrium,” Journal of Monetary Economics 13(3), 297–306 Minford P and Peel D (1983) Rational Expectations and the New Macroeconomy, Martin Robertson, Oxford Mishkin F (1982) “Monetary Policy and Short-Term Interest Rates: An Efficient Markets – Rational Expectations Approach,” Journal of Finance 37(1), 63–72 Mishkin F (1984) “Are Real Interest Rates Equal Across countries? An Empirical Investigation of International Parity Conditions,” Journal of Finance 39(5), 1345– 1357 Mishkin F S (1978) “Efficient – Markets Theory: Implications for Monetary Policy,” Brookings Papers on Economic Activity 3, 707–752 Mishkin F S (1984) "The Real Interest Rate: A Multi-country Empirical Study," Canadian Journal of Economics 17(2), 283–311 Mishkin F S (1990a) “What Does the Term Structure Tell Us About Future Inflation?” Journal of Monetary Economics 25(1), 77–95 M ishkin F S (1992) "Is the Fisher Effect for Real? A Reexamination of the Relationship between Inflation and Interest Rates", Journal of Monetary Economics 30(2), 195–215 Mishkin F S and Simon J (1995) “An Empirical Examination of the Fisher Effect in Australia,” NBER Working Paper No.5080, NBER, MA Mishkin Frederic S (1988) "The Information in the Term Structure Some Further 353 Bibliography Results," Journal of Applied Econometrics 3(4), 307–314 Moazzami B (1991) "The Fisher Equation Controversy Re-examined," Applied Financial Economics 1(3), 129–133 Modigliani F R and Sutch R (1966) “Innovations in Interest Rate Policy,” American Economic Review 56(2), 178–197 Monadjemi M (1990) “Testing the Degree of International Capital Mobility,” Australian Economic Papers 29(54), 30-39 Moosa I A and Bhatti R H (1997) International Parity Conditions: Theory, Econometric Testing and Empirical Evidence, Macmillan Press, London Moreno R and Yin N (1992), “ Exchange Rate Policy and Shocks to asset Markets: The Case of Taiwan in the 1980s,” Economic Review Federal Reserve Bank of San Francisco (1), 14–34 Moser G G (1995) "The Main Determinants of Inflation in Nigeria," IMF Staff Papers 42(2), 270–289 Mundell R (1963) "Inflation and Real Interest," Journal of Political Economy 71(3), 280–283 Murphy R G (1986) “Productivity Shocks, Non Traded Goods and Optimal Capital Accumulation,” European Economic Review 30(5), 1081–1095 Muth J F (1961) “Rational Expectations and the Theory of Price Movements,” Econometrica 29(3), 315–335 Nam Joo-Ha (1993) " The Liquidity, Income and Fisher Effects of Money on Interest : The Case of Developing Country," Seoul Journal of Economics 6(3), 223–239 Nelson C (1972) “Estimation of Term Premiums from Average Yield Differentials in 354 Bibliography the Term Structure of Interest Rates,” Econometrica 40(2), 277–287 Nelson C (1972) The Term Structure of Interest Rates, Basic Books, Inc. NY Nelson C and Plosser C I (1982) "Trends and Random Walks in Macroeconomic Time Series: Some Evidence and Implications," Journal of Monetary Economics 10(2), 139–162 Nelson C and Schwert G W (1977) "Short-Term Interest Rates as Predictors of Inflation : On Testing the Hypothesis that the Real Rate of Interest is Constant," American Economic Review 67(3), 478–486 Newey W and West K (1987) “A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica 55(3), 703–708 Ngama Y L (1994) “A Re-Examination of the Forward Exchange Rate Unbiasedness Hypothesis,” Weltwirtschaftliches Archiv 130(3), 447–460 Nicholas H V B (1990) “Inflation in Sri Lanka 1971–87: A Preliminary Study,” Macroeconomic Series No.3, Institute of Policy Studies, Colombo Nicholas H V B and Yatawara R A (1991) "Inflation in Sri Lanka in 1990" Institute of Policy Studies Working Paper No.2, Institute of Policy Studies, Colombo Nielsen N C (1981) “Inflation and Taxation: Nominal and Real Rates of Return,” Journal of Monetary Economics 7(2), 261–270 Nugent J B and Glezakos C (1979) "A Model of Inflation and Expectations in Latin America," Journal of Development Economics 6(3), 431–446 Obstfeld M (1986) “Capital Mobility in the World Economy: Theory and M eas urement ,” Carnegie Rochester Conference Series on Public Policy 24, 55–104 Olekalns N (1996) "Further Evidence on the Fisher Effect," Applied Economics 28(7), 355 Bibliography 851–856 Otani I (1975) "Inflation in an Open Economy: A Case Study of the Philippines," IMF Staff Papers 22(3), 750–774 Otani I and Tiwari S (1981 ) “Capital Controls and Interest Rate Parity: The Japanese Experience 1978–1981,” IMF Staff Papers 28(4) 793–815 Pagan A (1984) “Econometric Issues in the Analysis of Regressions with Generated Regressors,” International Economic Review 25(1), 221–247 Papaioannou G (1984) “Informational Efficiency Tests in the Athens Stock Exchange,” in Hawawini G A and M ichel P A eds., European Equity Markets: Risk, Return and Efficienc, Garland Publishing Inc., NY Parikh A (1994) “Tests of Real Interest Parity in International Currency Markets,” Journal of Economics 59(2), 167–191 Park S B (1999) “A Test of the Market Efficiency Hypothesis with an Application to Canadian Treasury Bill Yields,” CEP 99-03 Carelton University, Department of Economics, Ottawa Payne J E and Ewing B T (1997) “Evidence from Lesser Developed Countries on the Fisher Hypothesis: A Cointegration Analysis,” Applied Economics Letters 4(11), 683–687 Peek J (1982) “Interest rates, Income Taxes and Anticipated Inflation,” American Economic Review 72(5), 980–991 Peek J and Wilcox J A (1983) "The Postwar Stability of the Fisher Effect," Journal of Finance 38(4), 1111–1124 Pelaez R F (1995) "The Fisher Effect: Reprise," Journal of Macroeconomics 17(2), 356 Bibliography 333–346 Penati A and Dooley M (1984), “ Current Account Imbalances and Capital Formation in Industrial Countries, 1949–81,” IMF Staff Papers 31(1), 1–24 Peng W (1995) "The Fisher Hypothesis and Inflation Persistence Evidence from Five Major Industrial Countries," IMF Working Paper /95/118, IMF, Washington D.C Perez – Rodriguez J V, Saez M and Murillo C (1997) “ Expectativas y volatilities condicionada. Los tipos de interes en el mercado interbancario (Expectations and Conditional Volatility: Interest Rates in the Inter-bank Market. With English Summary ) Revista de Economia Aplicada 5(13), 83–107 Pesando J (1981) “On Forecasting Interest Rates : An Efficient M arkets Perspective,” Journal of Monetary Economics, 8(3), 305–318 Pesando J E (1979) "On the Random Walk characteristics of Short and Long Term Interest Rates In an Efficient Market," Journal of Money, Credit, and Banking 11(4), 457–466 Pesaran M H (1987a) The Limits to Rational Expectations, Basil Blackwell, Oxford Pesaran M H and Pesaran B (1997) Working with Microfit 4.0: Interactive Econometric Analysis, Oxford University Press, Oxford Pesaran M H and Smith R J (1994) “A Generalized R2 Criterion for Regression Models Estimated by the Instrumental Variables Method,” Econometrica 62(3), 705-710 Phillips L and Pippenger J (1976) “Preferred Habitat vs. efficient Market: A Test of Alternative Hypothesis,” Federal Reserve Bank of St Louis Review 58(5), 11–19 Phillips L and Pippenger J (1979) “The Term Structure of Interest Rates In the MIT- OENN-SSRC M odel,” Journal of Money, Credit and Banking, 11(2), 151–164 357 Bibliography Phillips P C B and Oularis S (1990) “ Asymptotic Properties of Residual Based Tests for Cointegration,” Econometrica 58(1), 165–193 Phylaktis K and Blake D (1993) "The Fisher Hypothesis: Evidence From Three High Inflation Economies," Weltwirtschaftliches Archiv 129(3), 591–599 Poole W (1976) “Rational expectations in the Macro Model,” Brookings Papers on Economic Activity 2, 463–505 Pope P F and Peel D A (1989) " Analyst Expectations and Risk Premia in the Foreign Exchange Market: An Empirical Examination," in Guimaraes R, Kingsman B and Taylor S eds., A Reappraisal of the Efficiency of Financial Markets NATO ASI Series, Springer-Verlag, Berlin Ramanathan R (1989) Introductory Econometrics with Applications, Harcourt Brace Jovanovich Inc., Florida Ramsey J B (1969) “ Test for Specification Errors in Classical Linear Least Squares Regression Analysis,” Journal of the Royal Statistical Society B 31(2), 350–371 Rana P B (1998) “Surges and Volatility of Private Capital Flows to Asian Developing Countries: Implications for M ultilateral Development Banks,” Asian Development Bank Occasional Papers No.19, Manila Rao B B, Fahimuddin and Bajpai D K (1997) " The Relationship Between the Price Level, Money and Output in India," Unpublished Manuscript Reisen H (1989) “Public Debt, External Competitiveness and Fiscal Discipline in Developing Countries,” Princeton Studies in International Finance No.66, NJ Reisen H and Yeches H (1993) “Time Varying Estimates on the Openness of the Capital Account in Korea and Taiwan,” Journal of Development Economics 41(2), 358 Bibliography 285–305 Report of the Presidential Commission on Finance and Banking (1993), Sri Lanka Reserve Bank of New Zealand (1997) Monetary Policy Statement, December 1997 Roberts H V (1959) “Stock Market ‘Patterns’ and Financial Analysis: Methodological Suggestions,” in Cootner (1964) ed., The Random Character of Stock Market Prices, MIT Press, MA Roe A (1982) "High Interest Rates: A New Conventional Wisdom for Development Policy? Some Conclusions From Sri Lankan Experience," World Development 10(3), 211–222 Roley V V and Sellon G H Jr. (1995) "Monetary Policy Actions and Long Term Interest Rates," Economic Review Federal Reserve Bank of Kansas City 80(4), 73–89 Roll R (1970) The Behaviour of Interest Rates : An Application of the Efficient Market Model to U.S Treasury Bills, Basic Books Inc., NY Roll R (1972) "Interest Rates on Monetary Assets and Commodity Price Index Changes," Journal of Finance 27(1), 251–277 Romer D (1996) Advanced Macroeconomics, Mc-Graw Hill Companies Inc., NY Rose A K (1988) " Is the Real Interest Rate Stable?," Journal of Finance 43(5), 1095– 1112 Rubinstein M (1975) " Securities M arket Efficiency in an Arrow-Debreu Economy " American Economic Review 65(5), 812– 824 Rupananda W (1994) “Econometric Model for Inflation in Sri Lanka,” Economic Review Sri Lanka 20(7), 25–28, Colombo, Sri Lanka 359 Bibliography Sachs J (1981) “The Current Account and Macroeconomic Adjustment in the 1970s,” Brookings Papers on Economic Activity 1, 201–268 Sahu A P, Jha R and Meyer L H (1990) "The Fisher Controversy: A Reconciliation of Cont radict ory Result s," Southern Economic Journal 57(1), 106–113 Saini K G (1982) "The Monetarist Explanation of Inflation: The Experience of six Asian Countries," World Development 10(10), 871–884 Saini K G (1984) "Can Monetary Growth Explain Inflation in India," Indian Economic Journal 32(1), 61–71 Samarakoon L P (1996) “Stock Market Returns and Inflation: Sri Lankan Evidence,” Sri Lankan Journal of Management 1(4), 293–311 Samarakoon L P (1996) “Predictability of Short -Horizon Returns in the Sri Lankan Stock Market,” Sri Lankan Journal of Management 1(3), 207–224 Samuelson P A “Proof that Properly Anticipated Prices Fluctuate Randomly,” Industrial Management Review 6, 41–49 Sarel M (1996) "Nonlinear Effects of Inflation on Economic Growth," IMF Staff Papers 43(1), 199–215 Sargent T (1979) “A Note on Maximum Likelihood Estimation of the Rational Expectations M odel of the Term Structure,” Journal of Monetary Economics 5(1), 133–143 Sargent T J (1969) "Commodity Price Expectations and the Interest Rate," in: Gibson W E and Kaufman G G eds., Monetary Economics : Readings on Current Issues, M c Graw Hill Book Co., NY Sargent T J (1972) "Anticipated Inflation and the Nominal Rate of Interest," 360 Bibliography Quarterly Journal of Economics 86(2), 212–225 Sargent T J (1972) "Rational Expectations and the Term Structure of Interest Rates," Journal of Money, Credit, and Banking 4(1), 74–97 Sargent T J (1973) "Interest Rates and Prices in the Long Run: A Study of the Gibson Paradox,” Journal of Money, Credit and Banking 5(1), 385–447 Savundranayagam T (1986) “Balance of Payments,” in Rasaputra W, Tilakaratna W M, Fernando S T G and Fernando L E N eds., Facets of Development in Independent Sri Lanka: Ronnie de Mel Felicitation Volume, Ministry of Finance and Planning Sri Lanka Schwert G W (1989) "Tests for Unit Roots: A Monte Carlo Investigation," Journal of Business and Economic Statistics 7(2), 147–159 Sephton P S and Larsen H K (1991) “ Tests of Exchange Market Efficiency: Fragile Evidence from Cointegration Tests,” Journal of International Money and Finance 10(4 ), 561–570 Seyhun H N (1986) “Insiders’ Profits, Costs of Trading, and Market Efficiency,” Journal of Financial Economics 16 (2), 189–212 Sharma J L and Kennedy R E (1977) "A Comparative Analysis of Stock Price Behaviour on the Bombay, London and New York Stock Exchanges," Journal of Financial and Quantitative Analysis 12(3), 391–413 Shaw E S (1973) Financial Deepening in Economic Development, Oxford University Press, Oxford Sheehan R G (1996) "Adaptive Versus Rational Expectations and the Fisher Effect," Applied Economics 28 (1), 1–11 361 Bibliography Sheehey E J (1979) "On the Measurement of Imported Inflation in Developing Countries, " Weltwirtschaftliches Archiv 115(1), 68–79 Shibata A and Shintani M (1998), “Capital Mobility in the World Economy : An Alternative Test,” Journal of International Money and Finance 17(5), 741–756 Shiller R J (1979) "The Volatility of Long-Term Interest Rates and Expectations Models of the Term Structure," Journal of Political Economy 87(6), 1190–1219 Shiller R J (1981) "The Use of Volatility M easures in Assessing M arket Efficiency ," Journal of Finance 36(2), 291–311 Shiller R J (1989) "Price-Conditional Vector Autoregressions and Theories of Stock Price Determination," in Guimaraes R, Kingsman B and Taylor S eds., A Reappraisal of the Efficiency of Financial Markets NATO ASI Series, Springer-Verlag, Berlin Shiller R J (1990) “The Term Structure of Interest Rates,” in Friedman B M and Hahn F H eds., Handbook of Monetary Economics, Volume 1, North Holland, Amsterdam Shiller R J and Perron P (1985) “ Testing the Random Walk Hypothesis: Power vs. Frequency of Observations” NBER Technical Paper 45, NBER, MA Shiller R J, Campbell J Y and Schoenholtz K L (1983) "Forward Rates and Future Policy: Interpreting the Term Structure of Interest Rates," Brookings Papers on Economic Activity 1, 173–217 Shome D, Smith S D and Pinkerton J M (1988) "The Purchasing Power of Money and Nominal Interest Rates : A Re-Examination," Journal of Finance 43(5), 1113– 1125 Silber W L (1975) "Thinness in Capital Markets: The Case of the Tel Aviv Stock 362 Bibliography Exchange," Journal of Financial and Quantitative Analysis 10(1), 129–142 Slayer K D (1990) "The Term Structure and Time Series Properties of Nominal Interest Rates: Implications from Theory," Journal of Money, Credit and Banking 22(4), 478–490 Smyth D J (1994) "Inflation and Growth," Journal of Macroeconomics 16(2), 261– 270 Snowden P N (1987) "Financial Market Liberalisation in LDCs: The Incidence of Risk Allocation Effects of Interest Rate Increases," Journal of Development Studies 24(1), 83–93 Solnik B H (1973) “Note on the Validity of the Random Walk for European Stock Prices,” Journal of Finance 28 (5), 1151–1159 Spiegel M (1990) “Capital Controls and Deviations from Proposed Interest Parity: Mexico 1982,” Economic Inquiry 28(2), 239–248 Stammer D W and Valentine T J (1982) "Inflation and the Financial System," Working Paper 8203A, Centre for Studies in Money, Banking and Finance, Macquarie University , Sydney Stiglitz J E and Uy M (1996) “Financial Markets, Public Policy, and the East Asian M iracle,” World Bank Research Observer 11(2), 151–177 Stiglitz J R (1981) "The Allocation Role of the Stock Market Pareto Optimality and Competition," Journal of Finance 36(2), 235–251 Summers L H (1978) “Tax Policy in a Life Cycle Model,” NBER Working Paper No.302, MA Summers L H (1983) "The Nonadjustment of Nominal Interest Rates: A Study of the 363 Bibliography Fisher Effect," in: J Tobin, ed., Macroeconomic Prices and Quantities: Essays in Memory of Arthur Okun, Brookings Institution, Washington D.C Summers L H (1988) “Tax Policy and International Competitiveness,” in Frenkel J ed., International Aspects of Fiscal Policies, University of Chicago Press, Chicago Swamy P A V B and Tavlas G S (1989) "Financial Deregulation, the Demand for Money, and Monetary Policy in Australia," IMF Staff Papers 36(1), 63–101 Tanzi V (1980) “Inflationary Expectations, Economic Activity, Taxes and Interest Rates,” American Economic Review 70(1), 12–21 Taylor M (1987) “Risk Premia and Foreign Exchange: A Multiple Time Series Approach to Testing Uncovered Interest Rate Parity,” Weltwirtschaftliches Archiv 123(4 ), 579–591 Taylor M (1989) “Vector Autoregressive Tests of Uncovered Interest Parity with Allowance for Conditional Heteroscedasticity,” Scottish Journal of Political Economy 36(3), 238–252 Taylor M P (1991) "The Hyperinflation Model of Money Demand Revisited," Journal of Money, Credit, and Banking 23(3), 327–349 Tease W J (1988) “Speculative Efficiency and the Exchange Rate: Some Evidence since the Float,” Economic Record 64(184), 2–13 Tease W J (1988) "The Expectations Theory of the Term Structure of Interest Rates in Australia," Economic Record 64(185), 120–127 Telser L G (1967) "A Critique of Some Recent Empirical Research on the Explanation of the Term Structure of Interest Rates," Journal of Political Economy 75(46) Supplement, 546–561 364 Bibliography Tesar L (1991) “Savings, Investment and International Capital Flows,” Journal of International Economics 31(3), 55–78 Thornton J (1996) "The Adjustment of Nominal Interest Rates in Mexico : A Study of the Fisher Effect," Applied Economics Letters 3(4), 255–257 Tobin J (1965) “Money and Economic Growth,” Econometrica 33(4)2, 671–684 Tseng W and Corker R (1991) “Financial Liberalization, Money Demand, and M onetary Policy in Asian Countries,” IMF Occasional Paper No.84, IMF, Washington D.C Turnovsky S J (1989) "The Term Structure of Interest Rates and the Effects of M acroeconomic Policy," Journal of Money, Credit, and Banking 21(3), 321–347 Turnovsky S J and M iller M (1984) “The Effects of Government Expenditure on the Term Structure of Interest Rates,” Journal of Money, Credit and Banking 16(1), 16–33 Urich T and Wachtel P (1984) "The Effects of Inflation and Money Supply Announcements on Interest Rates," Journal of Finance 39(4), 1177–1188 Urich T and Watchel P (1981) "Market Response to the Weekly Money Supply Announcements in the 1970s," Journal of Finance 36(5), 1063–1072 Vasiceck O (1977) “An Equilibrium Characterization of the Term Structure,” Journal of Financial Economics 5(2), 177–188 Vidanapathirana U (1986) " Pattern of Industrialisation: Strategies and Responses," in Rasaputra W, Tilakaratna W M, Fernando S T G and Fernando L E N eds., Facets of Development in Independent Sri Lanka: Ronnie de Mel Felicitation Volume, Ministry of Finance and Planning Sri Lanka 365 Bibliography Visco I (1975) "Anticipated Inflation and the Nominal Rate of Interest: Further Results," Quarterly Journal of Economics 89(2), 303–310 Vogel R (1974) "The Dynamics of Inflation in Latin America 1950–1969," American Economic Review 64(1), 102–114 Vredin A and Warne A (1991) “Current Account and Macroeconomic Fluctuations,” Scandinavian Journal of Economics 93(4), 511–530 Wai U Tun and Patrick Hugh T (1973) "Stock and Bond Issues and Capital Markets in Less Developed Countries," IMF Staff Papers 20(2), 253–317 Wallace M S and Warner J T (1993) "The Fisher Effect and the Term Structure of Interest Rates: Tests of Cointegration," Review of Economics and Statistics 75(2), 320–324 Weerakoon D and Salih R (1998) “Capital Account Liberalization and Financial Crises in East Asia: Lessons for Sri Lanka,” Research Studies: Macroeconomic Policy and Planning Series No.12, Institute of Policy Studies, Colombo Weeresekera (1992) “Inflation in Sri Lanka: A Causality Analysis,” South East Asian Central Bank s Centre Staff Paper No.46, South East Asian Central Banks Research and Training Centre, Kuala Lumpur White H (1980) “ A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity,” Econometrica 48(4), 817–838 White K (1993) SHAZAM: The Econometrics Computer Program User's Reference Manual Version 7.0, Second Printing, McGraw-Hill Book Co., NY Wickramanayake J (1993) “Financial Policy and Economic Development in Developing Countries: The Case Study of Sri Lanka," Ph.D. Dissertation La 366 Bibliography Trobe University, Bundoora Winbergen S (1983) "Credit Policy, Inflation and Growth in a Financially Repressed Economy," Journal of Development Economics 13(1/2), 45–65 Wong C Y P (1997) “Black M arket Exchange Rates and Capital M obility in Asian Economies,” Contemporary Economic Policy 15(2), 21–36 Wong D Y (1990) “What do Saving-Investment Relationships Tell Us About Capital M obility?” Journal of International Money and Finance 9(1), 60–74 Woo W and Hiramya K (1996) “Monetary Autonomy in the Presence of Capital Flows : and Never the Twain Shall Meet, Except in East Asia? ” in Ito T and Krueger A O eds., Financial Deregulation and Integration in East Asia, NBER- East Asia Seminar on Economics Volume 5 , Chicago University Press, Chicago Woodward G T (1992) “Evidence of the Fisher Effect from UK Indexed Bonds,” Review of Economics and Statistics 74(2), 315–320 World Bank (1989) World Development Report 1989, Oxford University Press, New York World Bank (1999) World Development Indicators, World Bank, Washington D.C World Bank (1999) World Development Report 1998/99, Oxford University Press, New York Yohe W P and Karnosky D S (1969) "Interest Rates and Price Level Changes," in Gibson W E and Kaufman G G (1971)eds., Monetary Economics: Readings on Current Issues, M cGraw Hill Book Co., NY Yuhn K (1996) “Is the Fisher Effect Robust? Further Evidence,” Applied Economics Letters 3(1), 41–44 367 Bibliography Zaken D (1998) “ The Effect of Changes in the Term Structure of Interest Rates on the Exposure to Interest Rate Risk of Financial Institutions in Israel,” Bank of Israel Review 0(6), 50–72 Zellner A (1962) “An Efficient Method of Estimating Seemingly Unrelated Regressions, and Tests for Aggregation Bias,” Journal of the American Statistical Association 57(298), 348–368 Zietz J and Homaifar G (1994) “Exchange Rate Uncertainty and the Efficiency of the Forward Market for Foreign Exchange,” Weltwirtschaftliches Archiv 130(3), 461– 475 Zilberfarb B (1989) "Interest Rate Determination in a High Inflation Economy," Journal of Macroeconomics 11(4), 533–549 368