X .. .
.y'to villus,
.r‘g-ovlslnt‘

. .00 n'f.|..b.o

I

‘ !
NM’OvIIIW-nvih‘la lil‘ld}. .
$53 .

I":

as" u a ‘un II.

. I . . .

X4315 : .

II ".“|"o v.o
. ..:>Iu.

vll\l

: 4 5f”? WWWN‘DHIJ latfowaaiwmnmwvw.
.Jufimnmmmlfihmquwfihcwwmhn

f f‘}! It

 

0U Vn)
ixflufilku.‘ . .t . pt v. ?
buzvupvilu. - $1.0} ,
LIT”. b
A : . 4'4
.rn votavuvhn '97

1th! IT I- .53
A Aur'trlu‘

I }| .
. .. - In'tcul
. A. 2.. . . . . A"-
. .. . . .. .. .1 . .. . ‘ . . p.93... ‘
.. I',.‘-I‘7>.V. 0: I v . ‘ .x v ‘ v A u
.!.-..:.-‘ux . ., . . :2 . V

yo. . . .

 

3.. ‘ [IA 0 1... .
filly»...
.. v ‘ I .

. IANIAIV} on n . n
.- .v .1'97. «:3.
0"»...0-1. Et;v

....3o:.uv...

.5714. INN 1m. x
u .‘A'

. .. . .. .. .. . .:-1..
nu... .100th , . .. 7 I. H. ‘ 1 . u 7 o..v pop .. . ‘ . .
. 7L”:;-..£.- . . . . .
39.».clcc: f . .
Ll:- o. I. .

f y : ..x-......d.€.hu.b{¥.v. ha...

.rnllpt'! l!)01|.p?

. . ..vf..6.. -i.

. u..,. .0 2.3:! nun-III.

. . \...... v ,

. 1 Bi}; I£d10h| . I f. . .
. . 3:2\ 4. .. .. -2. 2135531.”: .1. . to A . . . . ‘ . . .
..|..h.3€s!.3w-. «W. Hume]: Judi? ... u f . ..ur...«..£13-1.finlv‘i?n's. v4.1.2.2. . .. ‘ A . , - Irvvrrol fl .

. i....||'l....l\\ la. . .2...‘Nvl._pvg4 . ...lfll.. {V 300.. y .5.E)fiuu.nlu lok’.’ I _ .7. ... ... v.1 . .,.|Ir.hrl..’uv¢|u.fl'». .,I§E.flilfv
.IZ . . 1-. .. ...... ......u. 7... ‘... . .A ,.

 

 

 

 

 

:3. ..L.. 8114»

. CI, t |.,...'
‘ A f. . .. . ..‘ .1; ‘.4 I; .1: ..» r . .' _ ._ ,1 1,.. .2. . . . . ._ , .5 . . A ‘ S, ,.‘ :1
. . .(Itu ..r.l‘.\rhlr . ., . .. . . r i, L . . . h .1 . .2 . . . V. ._.t, . 6

   

'1 I All I

THESIS

 

Illlllllllllllllll

This is to certify that the

dissertation entitled

CENTRAL BANK INDEPENDENCE AND MONETARY POLICY EFFECTS
presented by

Min Chang

has been accepted towards fulfillment
of the requirements for

PhD. degree in Economics

 

    

Major pro esso

Date ////Z/? ’7‘

MS U is an Affirmative Action/Equal Opportunity Institution 0-12771

 

I LIBRARY

Michigan Qtate
Unlverstty

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINES return on or before date due.

 

MTE DUE MTE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1/98 cJCIRC/DdoOuopfiS-p.“

 

CENTRAL BANK INDEPENDENCE AND MONETARY POLICY EFFECTS
By

Min Chang

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

Department of Economics

1997

ABSTRACT
CENTRAL BANK INDEPENDENCE AND MONETARY POLICY EFFECTS
By

Min Chang

This dissertation investigates the effects of central bank independence (CBI) on
monetary policy effects, especially, liquidity effects. Chapter I ofiers a survey of previous
studies and the ideas of this thesis. Chapter II examines the variable indexes for CBI and
investigates the relationship between CBI and inflation in developing countries as well as
in developed countries. It is found that there is a significant negative relationship
between CBI and inflation in developed countries but not in developing countries. It is
argued that the levels of CBI in developing countries are too low to have an impact on the
economy.

Chapter III finds an implication for the relationship between CBI and monetary policy
effects with the basic liquidity model. Then it applies the traditional econometric
approaches to find the empirical results. It is shown that the higher CBI can result in the
stronger monetary policy effects. For more correct identification, Chapter IV constructs
both non-structural and structural vector autoregression (VAR) models to examine the
effects. The non-structural VAR models confirm the positive effects of CBI on the
monetary policy effects while the results from the structural VAR models are not

significant. Chapter V summarizes the conclusions of this study.

To my mother, parents-in-law,
and

the memory of my father
for his love and support throughout his lifetime.

iii

ACKNOWLEDGMENTS

I would like to acknowledge the important individuals who made the completion of
this document possible. My deep thanks and appreciation go to Dr. Robert H Rasche who
has been my committee chairperson. Without his constant encouragement and insightful
comments, this dissertation could never have been completed.

I gratefully acknowledge Dr. Gerhard Glomm and Dr. Christine Amsler, my other
committee members, for their help and valuable comments. My fellow graduate students
have played a major role in developing me as an economist during the last four years at
Michigan State University. I especially want to thank Ramon E Pineda, Jen-Je Su, and all
Korean economics-graduate students whom I have met here.

The greatest thanks, however, go to my family. My parents always supported me and
encouraged me to do my best and believed in me even when they did not agree with my
decision. For everything I have got today, I am deeply indebted to them. I would also
like to thank my parents-in-law for their support. It is impossible to have better in-laws.
I would be remiss if I did not thank my brother and sister for their love over the years.

I wish to acknowledge my wife, Hye Youn, for her love, support, and prayers
throughout the process of graduate studies. She has provided me with a constant source
of encouragement and inspiration even when She has struggled for her own graduate
studies. Her emotional support and patience have resulted in a debt I will never able to
repay. Finally, I would like to thank my eight-month-old son, Seung Bin, for a full of joy

he has brought me.

iv

TABLE OF CONTENTS

LIST OF TABLES ........................................................................................................ vii
LIST OF FIGURES ....................................................................................................... ix
CHAPTER I
INTRODUCTION ............................................................................................................ 1
CHAPTER 11
CENTRAL BANK INDEPENDENCE AND INFLATION ........................................ 9
1. INDEX FOR CENTRAL BANK INDEPENDENCE .......................................... 9
2. CENTRAL BANK INDEPENDENCE AND INFLATION
IN DEVELOPED COUNTRIES ......................................................................... l7
3. CENTRAL BANK INDEPENDENCE AND INFLATION
IN DEVELOPING COUNTRIES ........................................................................ 26
4. VARIABLES FOR CENTRAL BANK INDEPENDENCE
AND INFLATION ............................................................................................... 38
5. SUMMARY ........................................................................................................ 41
CHAPTER III
CENTRAL BANK INDEPENDENCE AND LIQUIDITY EFFECTS ...................... 43
1. MODEL APPROACH ......................................................................................... 44
1-1. BASIC LIQUIDITY MODEL ..................................................................... 44
1-2. GENERATING A LIQUIDITY EFFECT ................................................... 47
1-3. CBI IN BASIC LIQUIDITY MODEL ......................................................... 48
2. TRADITIONAL EMPIRICAL APPROACHES ................................................ 53
2-1. TRADITIONAL DISTRIBUTED LAG REGRESSIONS ......................... 54
2-2. DYNAMIC MULTIPLIERS METHODS .................................................. 55

2-3. DYNAMIC CORRELATIONS .................................................................. 56

2-4. ESTIMATION RESULTS ............................................................................ 58
2-5. ESTIIVIATION WITH THE ALTERNATIVE MONEY
AGGREGATES ........................................................................................... 71
3. SEEMINGLY UNRELATED REGRESSION APPROACHES ........................ 75
3-1. SURE .......................................................................................................... 75
3-2. ESTIMATION RESULTS .......................................................................... 77
4 SUMMARY ........................................................................................................ 86
CHAPTER IV
VECTOR AUTOREGRESSION APPROACHES .................................................... 87
1. NONSTRUCTURAL VAR APPROACHES ..................................................... 87
1-1. VECTOR AUTOREGRESSION ................................................................. 87
1-2. FOUR VARIABLE VAR ESTIMATION .................................................. 96
1-3. FIVE VARIABLE VAR ESTIMATION .................................................. 102
2. STRUCTURAL VAR APPROACHES ............................................................ 108
2-1. STRUCTURAL VAR ............................................................................... 108
2-2. MULTI-COUNTRY SVAR MODEL ...................................................... l 13
2-3. ESTIMATION RESULTS ........................................................................ 118
3 SUMMARY ...................................................................................................... 123
CHAPTER V
CONCLUSION ........................................................................................................... 125

vi

LIST OF TABLES

Table l - Index for Central Bank Independence ........................................................... 13
Table 2 - Rank-correlation of CBI indexes ................................................................... 15
Table 3 - CBI, Inflation rate, and M1 growth rate in developed countries .................... 21
Table 4 - Inflation rate and M1 growth rate vs. the CBI ............................................... 23
Table 5 - Variance of Inflation rate and M1 growth rate vs. the CBI ........................... 25
Table 6 - CBI, Inflation rate, and M1 growth rate in developing countries .................. 28
Table 7 - Inflation rate and M1 growth rate vs. the CBI ............................................... 31
Table 8 - Variance of Inflation rate and M1 growth rate vs. the CBI ........................... 33
Table 9 - Transformed Inflation rate and vs. Turnover rate ............................................ 35
Table 10 - Variables for legal independence index ......................................................... 39
Table 11 - Variables for CBI vs. Inflation rate(1972-1995) ........................................... 40
Table 12 - CBI and Liquidity Effects (Eq.(2. l )) ............................................................. 60
Table 13 - CBI and Liquidity Effects (Eq.(2.3)) ............................................................. 64
Table 14 - CBI and Liquidity Effects (Dynamic correlations) ........................................ 70
Table 15 - CBI and Liquidity Effects (TR) ..................................................................... 73
Table 16 - CBI and Liquidity Effects (MB) .................................................................... 74
Table 17 — Residual Correlation Matrix (1980-1989) ..................................................... 77
Table 18 - CBI and Liquidity Effects (SURE Eq.(2. l )) .................................................. 81

vii

Table 19 - CBI and Liquidity Effects (SURE Eq.(2.3)) .................................................. 85

Table 20 - CBI and Liquidity Effects (4 variable VAR) ............................................... 101
Table 21 - CBI and Liquidity Effects (5 variable VAR) ............................................... 107
Table 22 - 5 country SVAR model (1973-1989) .......................................................... 1 16
Table 23 - 6 country SVAR model (1973-1989) .......................................................... 1 17
Table 24 - CBI and Liquidity Effects (SVAR) ............................................................. 122

viii

LIST OF FIGURES

Figure 1 - Expected Liquidity Effects after Monetary Shock .......................................... 7
Figure 2 - CBI vs. Inflation rate and M1 Growth rate in Developed Countries ............ 22

Figure 3 - CBI vs. Inflation rate and M1 Growth in Developing Countries (1960-95) ..30

Figure 4 - Gap between Fitted and Actual Inflation rate ................................................ 37
Figure 5 - Response of R to M1 change (1980-1994: Eq.(2.1)) ................................... 59
Figure 6 - CBI and Liquidity Effects (M1, Eq.(2.1)) .................................................... 61
Figure 7 - Response of R to M1 Change (1980-1994, Eq.(2.3)) ................................... 63
Figure 8 - CBI and Liquidity Effects (M1, Eq.(2.3)) ..................................................... 65
Figure 9 - Dynamic Correlation b/w R(t) and M1(t-t) (1980-1994) ............................. 68
Figure 10 - CBI and Liquidity Effects (M1, 1980-1994: Dynamic Corr.) ...................... 69
Figure 11 - Response of R to M1 Change (1980-1989zSURE Eq.(2.1)) ......................... 79
Figure 12 - CBI and Liquidity Effects (1980-1989:SURE Eq.(2.1)) ............................... 80
Figure 13 - Response of R to M1 Change (1980-1989:SURE Eq.(2.3)) ......................... 83
Figure 14 - CBI and Liquidity Effects (1980-1989:SURE Eq.(2.3)) ............................... 84
Figure 15 - Impulse Responses in (R,M,Y,P) : US .......................................................... 93
Figure 16 - Impulse Responses in (Y,P,CP,R,M) : US .................................................... 94
Figure 17 - Liquidity Effects in (R,M,Y,P) ..................................................................... 95
Figure 18 - CBI and Liquidity Effects (R,M,Y,P) ........................................................... 98

ix

Figure 19 - CBI and Liquidity Effects (P,R,M,Y) ........................................................... 99

Figure 20 - CBI and Liquidity Effects (P,Y,R,M) .......................................................... 100
Figure 21 - CBI and Liquidity Effects (R,M,P,Y,CP) ................................................... 104
Figure 22 - CBI and Liquidity Effects (Y,P,CP,R,M) ................................................... 105
Figure 23 - CBI and Liquidity Effects (P,CP,R,M,Y) ................................................... 106
Figure 24 - Liquidity Effects in 5-Country SVAR (R,M,P,Y) ...................................... 120
Figure 25 - Liquidity Effects in 6-Country SVAR (R,M,P,Y) ...................................... 121

CHAPTER I

INTRODUCTION

In recent years many countries have adopted or made progress toward adopting
legislative proposals making their central banks more independent. Between 1989 and
1991, New Zealand, Chile, and Canada enacted legislation that increased the
independence of their central banks. The 1992 Treaty on European Union, Maastricht
Treaty, requires EC members to give their central banks more independence to establish
new European central bank (European Monetary Union). As a result, EC countries that
do not yet have strong independent central banks have tried to make their central banks
more independent]. Furthermore, the governments of Mexico and Brazil have announced
their intentions to introduce legislation to create more independent central banks. More
recently, in 1996, the government of Japan also announced the proposal removing the
Bank of Japan from government control. In 1997, newly-elected labor party government
in UK enhanced the independence of the Bank of England by giving more autonomy to
the bank in monetary policy decision procedure.

The changes in legislation usually give more authority to the central banks and also
direct them to focus mainly on the objective of price stability even at the cost of
disregarding other objectives such as high employment or economic growth. The success

of the highly independent Bundesbank and Swiss National Bank in maintaining

 

' To meet the level of independence described by the Maastricht Treaty, a central bank must be prohibited
from taking instructions from the government. The term for central bank governors must be set at a
minimum of five years. In addition, the central bank must be prohibited from purchasing debt instrument
directly from the government and from proving credit facilities to the government.

comparatively low rates of inflation for prolonged periods of time as well as recent
empirical and theoretical studies focused on central bank independence have contributed
this tendency.

The theoretical argument stems from the wide acceptance of the analysis of two
stylized macroeconomic facts in the papers of Kydland and Prescott (1977) and Barro and
Gordon (1983). They argue that the long-run Phillips curve is vertical, that is, inflation
has no permanent effect on real outcomes and that governments nonetheless have an
incentive to spring inflationary surprises upon the public. As a result, these papers
argued, a primary cause of inflation was govemment’s inability in the eyes of the public
to commit credibly to a low inflation policy. One could remove the time-inconsistency
problem by making government unable to renege upon a commitment to low inflation. In
Rogoff (1985), the appointment of a conservative central banker was shown to be one
means to achieve low inflation. These theoretical arguments, based upon the assumption
that central banks’ preferences are more inflation averse than those of govemment,
subsequently stimulated empirical research.

Several empirical studies including Alesina & Summers (1993), Cukierman (1992),
Cukierman et. al. (1993) and Fischer (1994) found that greater central bank independence
is associated with lower levels of inflation. Those studies conclude that the countries
which have low central bank independence have experienced high levels of inflation and
high variance of inflation rates because political and economic dependence restrict the
ability of central bank to select its policy objectives without influence by the government.
This political and economic dependence of the central bank in countries with high

inflation experiences makes agents assign low credibility to the central bank’s monetary

policy. Since these theoretical and empirical studies, legal central bank independence has
been identified with a credible commitment to the price stability. This credibility bonus
is presumed to be the source of the widely known negative correlation between central
bank independence and average inflation rates.

However, despite the theoretical and policy attention paid to central bank
independence, few studies have examined the postwar experience for evidence of
credibility effects of central bank independence. If independent central banks’ policies
are inherently more credible, not only inflation levels but also expectations of monetary
policy must differ systematically across countries with differences in central bank
independence.

Very recently Debelle and Fischer (1994), Walsh (1994), and Posen (1995) have tried
to find the evidence of credibility effects by measuring disinflationary cost which is the
ratio of output loss to the rate of disinflation. They adopted the idea that the greater
credibility attributed to independent central bank should reduce the costs of subsequent
policies to lower inflation. They estimated disinflationary costs in their sample countries
and examined the relationship between these costs and central bank independence. In
contrast to the idea, however, they found that disinflation appeared to be more costly and
no more rapid in countries with more independent central banks.

Debelle and Fischer (1994) used the disinflationary costs measured by Ball (1993) and
compared the costs of Germany with those of United States and concluded that the

disinflationary costs in Germany have been higher than those of United States.2 Walsh

 

2 Ball(l993) adopts a case study approach to estimating the sacrifice ratios associated with specific
disinflationary periods for the developed countries. He identifies disinflationary episodes from the DEC D
data anytime inflation drops for four or more straight quarters after having risen(that is, peak to trough).

(1994) measured the costs of disinfaltion in twelve EC countries for the period 1973—1986
following Ball, Mankiw, and Romer (1988).3 He also found that those EC countries with
greater central bank independence appear to also face higher costs of disinflation. Posen
(1995) also has relied upon Ball (1993) to identify disinflationary episodes of seventeen
countries from OECD data for the period 1950-1989 and also found a positive
relationship between central bank independence and disinflationary costs.4 From the
results, they argue that enhanced credibility does not appear to be the source of the
negative correlation between central bank independence and lower inflation.

These studies, however, examined the relationship with the samples of developed
countries which are considered to have more independent central banks than the other
countries. For example, Debelle and Fischer (1994) examined the relationship only with
two countries which have one of the most independent central banks in the world. They
argued that improving central bank independence level of the Federal Reserves to the
level of the Bundesbank would increase disinflationary cost and so that there is no
credibility bonus. However, is the difference in independence level in these two highly
independent central banks large enough to give us the no credibility bonus implication?

Since central bank independence is negatively correlated with average inflation rates
and cost for reducing inflation increases when the level of inflation goes down, the

positive relationship between disinflationary cost and the independence level may reflect

 

For each of these episodes Ball has computed a sacrifice ratio; total point-years of unemployment above
that at episode start, divided by total points of inflation lost.

3 Twelve EC countries are Belgium, Denmark, France, Germany, Greece, Ireland, Italy, Luxemburg,
Netherlands, Portugal, Spain, and United Kingdom. Ball, Mankiw, and Romer report a tradeoff parameter
that measures the extent to which a nominal shock affects real output. This parameter is the estimated
coefficient on the growth rate of nominal income in an equation for the level of real output. The lagged
level of output and a time trend are also included in the regression and results are reported for 43 countries.

increasing marginal costs of disinflation.5 Then, this sample bias to highly independent
central bank countries may not capture the proper credibility bonus problem in the
economies. Extending the sample to include more countries which have low central bank
independence may reveal a different relationship between central bank independence and
disinflationary costs. However, since there is no central bank independence measurement
which can be applied to both developed countries, which are considered to have highly
independent central banks, and developing countries, which are considered to have low
independent central banks and there are not enough data to measure disinflationary costs
in developing countries, it is difficult to examine the relationship between disinflationary
costs and central bank independence with the extended sample of countries. Though I
will not reexamine the relationship with extended data, this paper shows another way to
find a credibility bonus effect of central bank independence. To prove the credibility
bonus effect, we will directly examine the relationship between central bank
independence and monetary policy effects with the data of developed countries.

The purpose of this paper is to examine the possibility that the degree of monetary
policy effect could differ across countries among which central banks have different
levels of independence. One of the most pervasive real effects long-claimed for monetary
policy is its ability to affect interest rates in the short run through channels other than the
standard-expected inflation effect. The alleged short-term inverse relationship between

interest rates and monetary policy is called the “liquidity effect” of monetary policy. In

 

4 The countries are Australia, Austria, Belgium, Canada, Denmark, France, Germany, Greece, Ireland, Italy,
Japan, Netherlands, New Zealand, Spain, Switzerland, United Kingdom, and United States.

5 Walsh(1995), however, found that the tradeoff parameter from Ball, Mankiw, and Romer was significantly
and positively related to central bank independence even after controlling for average inflation in a cross
section of EC countries.

this paper, we focus on the liquidity effect to estimate the impact of monetary policy in
each country, then, concentrate on the relationship between central bank independence
and the liquidity effect in order to examine if monetary policy effects are influenced by
central bank independence (CBI).

The negative relationship between CBI and inflation rates which are found by many
empirical studies gives us the possibility of testing whether or not the monetary policy
effects are smaller in low CBI countries than in high CBI countries because of high
inflation experiences and low credibility of central bank policies in the low CBI
countries. Regarding the liquidity effect, therefore, it would also be smaller in low CBI
countries if the inflation expectation effect dominates the liquidity effect in a short period
of time after a monetary policy disturbance. So, the length of the period which the
liquidity effect dominates the inflation expectation effect after a monetary shock would be
shorter. If the responses of interest rates to a monetary shock are smaller in low CBI
countries, changes in the economic variables that respond to monetary policy through
interest rates would also be smaller in the low CBI countries. This suggests that monetary
policy effects would be smaller in low CBI countries than in high CBI countries.

This implies that to get the same effects on target variables from monetary policy, low
CBI countries may need larger changes in the monetary policy variables that they control.
Assume that a low CBI country decides to increase the money supply to stimulate the
economy by lowering nominal interest rates. Because of the low credibility of the central
bank’s policy and the relatively weak effects of monetary policy on the economy, it may
need a larger monetary expansion than a high CBI country does to lower interest rates to

the same degree or the inflation expectation effect will dominate the liquidity effect more

quickly after the monetary expansion. Further, this higher monetary expansion in the low
CBI countries would induce higher inflation rates in the future than in high CBI countries.
This negative relationship between CBI and liquidity effect, therefore, could produce a
vicious cycle which shows a strong relationship between CBI and inflation rates across
countries. This expected negative relationship between CBI and liquidity effect can be

shown as the following graph.

Ai low CBI high CBI

con? / country
~// A.

Figure 1. Expected Liquidity Effects after Monetary Shock

 

In low CBI country, the response of interest rates would be smaller and less persistent
than in high CBI country. In some period of time after monetary expansion, the inflation
expectation effect would be stronger than the liquidity effect so the changes in interest
rates would turn positive. We can expect that this turning point afier the monetary shocks
would come sooner in low CBI countries than in high CBI countries. The graph,
therefore, shows that expected liquidity effect would be weaker and be sustained for less
time in low CBI country than in high CBI country.

The organization of this paper is as follows. Chapter 11 describes the indexes for

central bank independence and sample countries used in this paper and reexamines the

relationship between CBI and inflation rates with extended data. In chapter III, we find a
theoretical implication regarding the relationship between CBI and liquidity effect
through a model economy approach. Then, with a group of the countries which show the
negative relationship between CBI and inflation rates, we estimate liquidity effects by
using traditional approaches and examine the possible relationship between CBI and
liquidity effects across countries. In chapter IV, we construct both non-structural and
structural VAR models to isolate liquidity effects and identify the monetary shock and
examine the relationship between CBI and liquidity effects. The conclusions from this

analysis are summarized in chapter V.

CHAPTER 11

CENTRAL BANK INDEPENDENCE AND INFLATION

1. Index for Central Bank Independence

It is difficult to measure the degree of legal independence a central bank has, let alone
the degree of its actual independence from government. Cukierman (1992) has pointed
out that actual, as opposed to formal, independence hinges not only on legislation, but on
many other factors such as informal arrangements with the government, the quality of
bank personnel, and the personal characteristics of the key individuals at the bank.
Because such factors are impossible to quantify, most research has focused on legal
independence and for the most part analysis is restricted to the industrial countries.

There are different measurement of CBI developed by different researchers.6 Most
frequently used indexes include those of Bade and Parkin (1988) which were extended
by Alesina (1988), Grilli, Masciandaro, and Tabellini (1991), Alesina and Summers
(1993) and Cukierman, Webb, and Neyapti (1992).

Bade and Parkin investigates the cross-country relationship between monetary policies
and the laws which establish and delimit the powers of central banks. The study is

empirical and deals with the experience of twelve industrial countries during the floating

 

6 Pollard (1993) provides a good survey paper regarding various CBI indexes.

10

exchange rate years 1972 to 1986.7 They describe the central bank laws of the twelve
countries focusing on three features: i) The relationship between central bank and
government in the formulation of monetary policy; ii) The procedures for appointing the
board of the central bank, and ; iii) The financial and budgetary relations between central
bank and the government. On the basis of features i) and ii) they classify the twelve
central banks according to their degrees of policy independence. On the basis of feature
iii) they identify the degree of financial independence from government.

Alesina uses the Bade and Parkin index of policy independence to illustrate the
relationship between the degree of politico-institutional stability and economic
performance. Alesina extends the Bade and Parkin’s sample of countries to include
Denmark, Finland, New Zealand, Norway, and Spain. The numerical values of Alesina
index of central bank independence are identical to those of the Bade and Parkin index of
policy independence, except for the case of Italy.8

Grilli, Masciandaro, and Tabellini (GMT) compare the monetary regimes of eighteen
industrial countries during post-war period (1950—1989) by focusing on political and
economic independence of central bank.9 According to GMT, political independence is
the capacity to choose the final goal of monetary policy, such as inflation or the level of
economic activity. Their index of political independence is primarily determined by the

following features: i) Relationships between central banks and government in the

 

7 Twelve countries are Australia, Belgium, Canada, France, Germany, Italy, Japan, Netherlands, Sweden,
Switzerland, United Kingdom, and United States.

8 “Bade and Parkin’s classifications disregard institutional changes in the period considered. The Italian
Central Bank obtained more economic independence in 1982. Given this change we classify Italy as 1.5
rather than 2, as in Bade-Parkin.”(Alesina, 1988, p.42)

9 Eighteen countries are Australia, Austria, Belgium, Canada, Denmark, France, Germany, Greece, Ireland,
Italy, Japan, Netherlands, New Zealand, Portugal, Spain, Switzerland, United Kingdom, and United States.

ll

formulation of monetary policy; ii) Procedures for appointing the board of the central
bank and; iii) Formal responsibilities of the central bank with the respect to monetary
policy. The first two features are the same as those of Bade and Parkin. The overall
index of policy independence is determined by a combination of these attributes. The
economic independence considers the ability of the government to determine the
conditions under which it can borrow from the central bank and the monetary instruments
under the control of the central bank. Generally, the total score for both political and
economic independence is employed as an indicator for legal independence.

Alesina and Summers do not add any new criteria for their central bank independence
index. They use the average value of Bade and Parkin (BP) and Grilli, Masciandaro and
Tabellini after constructing conversion from the GMT scale to a 1 to 4 scale comparable
with BP scale.

As a more recent index, Cukierman et. al. code the legal central bank independence
following two principles. First, they code only a few narrow but relatively precise legal
characteristics and second, they use only the written information from the charters.l0
They group the legal characteristics of the central bank as stated in its charter into four
clusters of issues: i) The appointment, dismissal, and terms of office of the chief
executive officer of the bank - usually the governor, ii) The policy formulation cluster,
which concerns the resolution of conflicts between the executive branch and the central

bank over monetary policy and the participation of the central bank in the budgetary

 

'0 Cukierman et. a1. develop four measures of central bank independence in their article. An aggregate legal
index is developed for four decades in 72 countries. In addition, they develop other three indicators of
actual independence which are the rate of turnover of central bank governors, an index based on a
questionnaire answered by specialists in 23 countries, and an aggregation of the legal index and the rate of
turnover.

12

process, iii) The objectives of the central bank, and iv) Limitations on the ability of the
central bank to lend to the public sector; such restrictions limit the volume, maturity,
interest rates, and conditions for direct advances and securitized lending from the central
bank to the public sector. Then, they aggregate the numerical coding for each different
legal variables and rank countries according to their aggregate variables for legal central
bank independence for each decade from 1950 to 1989.11

Table 1 presents these four indexes of central bank independence of eighteen countries
which we will use to estimate the effects of CBI in this paper. These countries are the
same as those in Fischer (1994). As we can expect from the fact that all these different
studies consider very similar factors for making their indexes of central bank
independence, from the table we can point out that all agree that Germany and
Switzerland have the most independent central banks and there exists a rough consensus
on CBI measures.

There are, however, a few countries which are ranked quite differently by the several
measures. As we can see, the Bank of Japan has the third lowest independence level of
all 18 countries according to Cukierman, while it has much higher level of independence
in index of Alesina and Summers. This discrepancy over the degree of independence
comes not only from differences in factors considered in measuring independence, but
also from the criteria of independence as well.

The argument for the Bank of Japan’s higher rank of independence in Alesina and
Summers come from Bade and Parkin and GMT because Alesina and Summers’ index is

the average value of those of Bade and Parkin and GMT. Bade and Parkin claimed that

 

" We examine the variables for legal central bank independence in detail at the last part of this chapter.

Table 1. Index for Central Bank Independence

 

 

 

sum of the indices for political and economic independence.

BPA GMT AS Cukierman
Australia 1 2 (8) 0.36 (8)
Austria - 9 — 0.62 (2)
Belgium 2 7 2 (8) 0.17 (16)
Canada 2 11 2.5 (4) 0.45 (6)
Denmark 2 8 2.5 (4) 0.50 (4)
Finland 2 - - 0.28 (11)
France 2 7 2 (8) 0.27 (12)
Germany 4 13 4 (l) 0.69 (1)
Italy 1.5 5 1.75 (14) 0.25 (13)
Japan 3 6 2.5 (4) 0.18 (15)
Netherlands 2 10 2.5 (4) 0.42 (7)
New Zealand 1 3 1 (16) 0.24 (14)
Norway 2 - 2 (8) 0.17 (16)
Spain 1 5 1.5 (15) 0.14(18)
Sweden 2 - 2 (8) 0.29 (10)
Switzerland 4 12 4 (1 ) 0.57 (3)
United Kingdom 2 6 2 (8) 0.32 (9)
United States 3 12 3.5 (3) 0.48 (5)
Note : The value in parentheses is the relative ranking. The GMT measure is the

14

the Bank of Japan is independent from the government in formulating and implementing
monetary policy and GMT claimed that there are no provisions for handling policy
conflicts between the Bank of Japan and the government. In contrast, Cukierman argues
that the Bank of Japan and the govemment formulate policy jointly and in the case of a
policy conflict, the government makes a final decision.

This difference is confirmed by Table 2, which shows Spearman rank correlation
coefficients of the various measures. The values above the diagonal indicate rank order
correlation in the eighteen countries and the values below the diagonal are the
correlations when we exclude Japan from the sample countries. The correlation of Bade
and Parkin with Cukierrnan’s legal independence has the lowest value while the
correlation of GMT with Alesina and Summers’ CBI index has the highest value. The
correlation of GMT to Alesina and Summers’ CBI index goes up to 0.96 when we
exclude Japan from the sample countries. This high correlation results come from the
fact that Alesina and Summers use the average values of Bade and Parkin and GMT.

Since Alesina and Summers’ measurement represents both Bade and Parkin and GMT
indexes, the most interesting statistics are the correlations of Alesina and Summers’ index
with Cukierrnan’s index. These are 0.78 in 18 countries and 0.89 when Japan is
excluded. These high correlations confirm that these different authors have very common
ideas when they set the independence level of central banks of different countries.
Therefore we can expect that the our results will be little affected by our choice of CBI

indexes for estimation.

15

Table 2. Rank-correlation of CBI indexes

 

 

BPA GMT AS Cukierman
BPA - 0.69 0.88 0.53
GMT 0.80 - 0.89 0.82
AS 0.88 0.96 - 0.78
Cukierman 0.67 0.83 0.89 -

 

All those indexes mentioned above are legal measures of central bank independence
but there are also nonlegal measures of central bank independence. Cukierman (1992)
and Cukierman et. a1. (1992) have developed a nonlegal index for CBI based on the actual
average term of office central bank governors. This indicator is based on the presumption
that a higher turnover of central bank governors indicates a lower level of independence,
at least above some threshold, and that even if the central bank law is quite explicit, it
may not be operational if a different tradition has precedence. For example, in Argentina,
the legal term of office of the central bank governor is four years but there is also an
informal tradition that the governor will resign whenever there is a change of government,
or even a new finance minister. The actual average term of oflice of the central bank
governors during 1980s were only ten months. This suggests that the turnover rate of
central bank governors can be a good indicator for the degree of CBI.

Cukierman and Webb (1995) have developed another nonlegal index of CBI. They
argue that the frequency of transfers of central bank governors reflects both the fiequency
of political change and the percentage of political changes that are followed by changes in

the govemorship of the central bank. They develop an indicator of the political

l6

vulnerability of the central bank, which is defined as the percentage of political
transitions that are followed within six months by the replacement of the central bank
governor. For the period fi'om 1950 to 1989, the average index of political vulnerability
is 0.24 for whole sample countries, 0.10 for industrial countries, and 0.34 for developing
countries.

Those existing legal and nonlegal indexes of CBI are often incomplete and noisy
indicators of actual independence. However, this does not mean that they are
uninforrnative. As Cukierman (1995) pointed out, their use should be supplemented by
judgment of the problem under consideration.

In this paper, we will use Cukierrnan’s legal CBI index which seems to be most widely
used. We, however, will use Alesina and Summers’ as well as Cukiennan’s to
reexamine the relationship between CBI and inflation rates in 18 developed countries,
since we want to see here if we can replicate existing results with the same index and
method they used. In chapter III and IV, the number of sample countries we examine is
decreased from 18 to 12 because of non-availability of the data in some countries and we
will exclude Japan from the sample countries even though we can get easily all data for
Japan since the estimated results are somewhat affected by which CBI index we choose
for J apan.12

For developing countries, Cukierman argues that the actual frequency of change of the
chief executive officer of the bank is a better proxy for central bank independence, since

the divergence between the letter of the law and actual practice seems substantially higher

 

'2 In section III and IV, when we include Japan in the sample countries,Alesina and Summers’ index for
CBI gives us more significant results in most cases than does Cukierman’s index. However, the differences
in significance level between two estimation results are within 4 percent.

17

in developing countries than in industrial countries. Therefore we will follow Cukierman
and use the turnover rate of central bank governors as CBI index for those developing

countries.

2. Central Bank Independence and Inflation in Developed Countries

Before we explore the possible relationship between CBI and liquidity effects, we
briefly reexamine the relationship between CBI and inflation rates in developing
countries as well as developed countries to get the proper sample of countries for the next
sections. Many researchers have found a negative relationship between CBI and inflation
rates in developed countries. Among them, Bade and Parkin investigate the relationship
between central bank types and monetary policy in twelve countries. In their analysis of
monetary policy they focused on two aspects; its inflation level and the variability of
inflation. They find that the two most independent central banks, those of Germany and
Switzerland, have achieved a lower inflation rate than the other central banks. Also, they
point out that the mean inflation rate of the eight government dominated central banks is
in excess of ten percent.13 On the basis of these facts Bade and Parkin conclude that there
is an association between degree of central bank policy independence and the average rate
of inflation.

Grilli, Masciandaro, and Tabellini find that economic independence was negatively

related to inflation in eighteen OECD countries over the period 1950—1989. Political

18

independence also had a negative correlation with inflation, but the relationship was not
statistically significant. Breaking the data into four decade-long subperiods, they find that
neither measure of independence has a significant effect on inflation in the first two
decades. In the 19703 both measures of independence are significant while in the 19805
only the economic independence is significant.

Alesina & Summers (1993) frnd a negative correlation between the inflation rates and
the level of CBI in 16 industrialized countries for the period of 1955-1988 and also find
that the more independent a central bank, the less variable inflation. On the other hand,
they found no correlation between the level of CBI and average economic growth or the
variability in economic growth. With these results, they argued that a higher CBI results
in lower inflation rates without any side effects on real economic variables. Fischer
(1994) found the same relationships as Alesina and Summers did, using only the GMT
index for 18 industrialized countries for the period of 1960-1992

Cukierman (1992) provides the most comprehensive analysis of central bank
independence and its relationship to inflation performance using data for 1950-1989 with
a sample of 73 countries. He concludes that the legal independence level has a
statistically significant coefficient with the predicted negative sign for the industrial
countries so that laws do make a difference.14 His estimation shows that 0.1 point

increase in legal independence index might increase the real value of money by 0.5%.

 

'3 They are Australia, Belgium, Canada, France, Italy, Netherlands, Sweden, and United Kingdom.

'4 Actually, Cukierman did not use the inflation rate. He used the real depreciation of a given amount of
money by transforming each year’s inflation rate into inflation divided by one plus the inflation rate and
then taking the geometric average for the decade. This transformed inflation is also used inCukierman et.
al. (1992)

19

However, even though these empirical studies show a negative correlation between

central bank independence and the inflation rate, this inverse relationship does not
necessarily imply a causal relation from central bank independence to inflation, since low
independent level of central bank may result from high inflation experiences or other
factors in the countries which also leads to high inflation.15 Despite of the causality
problem, the previous empirical studies have been based on the argument that central
bank independence has a significant effect on inflation rates. For example, Schaling
(1995) argues as follows to examine the relationship between central bank independence
and inflation rate.
“... the degree of central bank independence is the ultimate cause of the level of inflation.
For central bank independence is the ability and willingness to conduct an autonomous
monetary policy directed at price stability as the single policy goal. If not seriously
hampered by other elements of economic policy, such as wage increases, budget deficits,
and government debt, it will eventually lead to low sustainable inflation...”
(Schalling,1995, p.122)

In this paper we use Cukierman’s legal independence and Alesina & Summers’ index
for 18 developed countries.16 Since Cukierrnan’s index is the most comprehensive one
and Alesina & Summers’ index is the average value of those of Bade and Parkin and

GMT, these two indexes are thought to be the most representative ones. The 18

developed countries are the same as those of Fischer (1994).

 

lsCukierman et.al. (1992) did a simple Granger causality test by estimating the bivariate autoregressive
processes for inflation and turnover. They find that the coefficient of lagged turnover in the inflation
equation is highly significant, as is the coefficient of lagged inflation in the turnover equation. They
conclude that these results imply that there is a vicious circle between inflation and low levels of CBI.
'6 The eighteen countries are those in Table 1.

20

Furthermore, we examine not only the relationship between CBI and inflation rate but
also the relationship between CBI and the M1 growth rate.17 Central banks can affect the
inflation rate through controlling a money aggregate with various monetary policies.
Therefore we can assume that countries which have less independent central banks will
experience higher M1 growth rates and these higher M1 growth rates increase the
countries’ inflation rate. This means that even though the inflation rates are high in low
CBI countries, if the M1 growth rates do not have concrete relationships with the
degrees of CBI, we can not say that CBI has a direct effect on the inflation rate.

We estimate the relationship during 1960-1995 period and the post-Bretton-Woods
periods 1972-1995. During the fixed exchange rate system of Bretton-Woods countries
were fully committed to an exchange rate target and had no room to conduct autonomous
domestic monetary policy. Thus, before 1972 the empirical relation between CBI and
inflation was much less straightforward than after 1972.18

With a sample of eighteen developed countries we can replicate the previous results no
matter which index is used for CBI. Although the relationships between the M1 growth
rate and CBI are weaker than those between inflation rates and CBI, simple regression
show that CBI has negative effects on inflation rates through its effect on M1 growth in
developed countries. Figure 2 shows this estimated negative relationship for some

sample periods and Table 4 reports the regression results.

 

'7 We use Ml data because M1 is the only monetary aggregate which is available for all countries.
’8 We neglect the other exchange rate policy factor in Europe. Some European countries(EMS countries)
participates in the snake arrangement before 1982.

21

Table 3. CBI, Inflation rate, and M1 growth rate in developed countries

 

 

 

 

CBI(Legal Independence) Inflation rate Ml growth rate
‘60-’90 ‘70-’90 ‘60-’95 ‘72-’95 ‘60-’95 ‘72-’95
Germany 0.69 0.69 3.36 (1.75) 3.68 (1.94) 8.16 (3.57) 8.19 (4.06)
Austria 0.62 0.61 4.23 (1.97) 4.57 (2.25) 7.28 (3.95) 7.01 (4.77)
Switzerland 0.57 0.59 3.75 (2.20) 3.90 (2.52) 5.49 (3.53) 6.43 (6.62)
Denmark 0.50 0.50 6.33 (3.49) 6.75 (3.96) 10.84 (7.83) 11.41 (9.32)
US 0.48 0.48 4.74 (3.10) 5.72 (3.19) 6.39 (3.13) 7.63 (2.87)
Canada 0.45 0.45 5.01 (3.30) 6.20 (3.38) 8.05 (6.65) 8.90 (6.79)
Netherlands 0.42 0.42 4.36 (2.76) 4.34 (3.09) 7.95 (4.09) 7.44 (4.51)
Australia 0.36 0.36 6.30 (4.14) 8.01 (3.95) 9.29 (6.64) 11.89 (6.25)
Sweden 0.29 0.29 6.62 (3.15) 7.81 (3.06) 6.13 (8.33) 9.12 (2.86)
Finland 0.28 0.28 6.80 (4.30) 7.71 (4.71) 16.13(27.43) 20.00(32.98)
UK 0.32 0.27 7.29 (5.34) 8.81 (5.82) 9.74 (6.51) 12.77 (5.59)
Italy 0.25 0.25 8.42 (5.74) 10.70 (5.69) 14.3l(10.03) 13.45 (6.27)
New Zealand 0.24 0.24 7.83 (5.40) 9.70 (5.49) 10.33(11.67) 14.20(12.46)
France 0.27 0.24 6.00 (3.74) 6.92 (4.21) 8.63 (5.16) 7.88 (4.93)
Japan 0.18 0.18 4.99 (4.34) 4.67 (5.24) 11.96 (8.03) 8.22 (6.07)
Belgium 0.17 0.17 4.52 (3.04) 5.32 (3.34) 5.58 (3.66) 5.65 (3.77)
Norway 0.17 0.17 6.06 (3.29) 6.93 (3.31) 12.54 (7.06) 14.87 (7.33)
Spain 0.14 0.16 9.35 (5.45) 11.04 (5.63) 13.98 (6.20) 14.83 (6.68)

 

Note : Countries are ordered by the level of CBI for the period of 1970-1990. The standard deviations are
reported in parentheses.

<Inflation rate : 1960 -1995>

22

<M1 change rate : 1960 -1995>

 

    

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

10 13
' SPA . FIN
9 7 . ITA . ‘° 7
e - ' NEW ' 14 a
0 UK ’
7 — °.F|N 12 a
0
° ‘ NOR FRA 1° 7
GER
5 - 8 _.
O
4 _. AUST . 6 —
SWI ’ GER
3 *- 1 r 1 l l 4 r l l 1 l l
01 02 03 0‘ 05 06 07 08 01 02 03 04 05 0e 07 03
Legal Independence Legal Independence
inf = 9.73 - 826109“ M1 =12.94 - 939 legal
(-3.09)° (-227)-
R2 = 0.37 ' significance 0.01 level R2 = 0,24 - significance 0.04 level
<lnflation rate : 1972 -1995> <M1 Change 1319 3 1972 '1995’
14 16
SP}: .NOR
NEW
12 ~ 0 SPA 14 d . ”A UK
10 NEW ‘2 r
l 10 9
a .—
8 ..
6 —.
6 _I
4 — 4 2
. SVVI
2 1 I 1 l 2 I l I l
o 1 2 3 4 5 o 1 2 3 4 5

Independence Level (Alesina 8. Summers)

inf = 12.05 - 2.18 indep.
(4.73)‘
R2: 0.62 ' significant 0.01 level

Independence Level (Alesina & Summers)

m1 = 16.65 - 2.82 indep.

(-3.56)'

F!2 = 0.48 ' significant 0.01 level

Figure 2. CBI vs. Inflation rate and M1 Growth rate in Developed Countries

23

Table 4. Inflation rate and M1 growth rate vs. the CBI

 

Dependent Variables Explanatory Variables Alesina & Summers Cukierman

 

1960-1995 Inflation rates Intercept 9.77 8.20
CBI -1.63"'** -6.51*"‘*
(-5.02) (-3.43)
R2 0.64 0.42
M1 growth Intercept 13.45 12.94
CBI -l.74* -9.39*
(-2.25) (-2.27)
R2 0.27 0.24
1972-1995 Inflation rates Intercept 12.05 9.73
CBI -2.l8*** -8.26***
(-4.78) (-3.09)
R2 0.62 0.37
M1 grth Intercept 16.65 14.47
CBI -2.82*** -ll.57*
(-3.56) (-2.16)
R’- 0.48 0.23

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 5 percent level,
** at 3 percent level, and *** at 1 percent level.

24

We also frnd a very significant negative relationship between the variance of inflation
rates and CBI during both 1960-1995 and 1972-1995 periods no matter which index is
used for CBI. The variance of M1 growth also has a negative correlation with Alesina
and Summers’ index, but it does not have a significant relationship with Cukierman’s
index. This negative relationship seems to be a result of a correlation between the level
and variability of inflation (or M1 growth) as argued by Alesina and Summers. The
regression results are reported in Table 5.

In summary, our estimation results confirm that CBI has a significant effect on the
inflation rate in the developed countries. CBI also seems to have an effect on a central
bank’s decision how to conduct monetary policy. In developed countries, highly
independent central banks seem to be more conservative in increasing monetary aggregate
growth and they are better in reducing the inflation rate than low independent central

banks.

25

Table 5. Variance of Inflation rate and M1 growth rate vs. the CBI

 

 

Dependent Variables Explanatory Variables Alesina & Summers Cukierman
1960-1995 Variance of Intercept 6.40 5.53
Inflation CBI -1.12*** -5.16***
(-4.59) (-3.95)
R2 0.60 0.49
Variance of Intercept 10.39 10.91
MI growth CBI -1.62*"‘ -9.38
(-2.65) (-l.19)
R2 0.33 0.08
1972-1995 Variance of Intercept 6.49 5.81
Inflation CBI -l .06*** -5.31***
(-3.88) (-4.08)
R2 0.52 0.51
Variance of Intercept 8.95 9.76
M1 growth CBI -1.24* -6.54
(-1.77) (-0.65)
R2 0.18 0.03

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent level ,
** at 3 percent level, and *** at 1 percent level.

26

3. Central Bank Independence and Inflation in Developing Countries

All studies except Cukierman (1992) and Cukierman et.al. (1992) restrict the samples
to developed countries to examine the effects of CBI on the economy. Cukierman (1992)
developed four measures of central bank independence and explores their relations with
inflation. He examined these relationships in the cases of developing countries as well as
developed countries. He found that legal independence was inversely related to inflation
in industrial countries as was the turnover rate of central bank governors positively
related to inflation in developing countries.

In this paper, we use the turnover rate of central bank governors for 27 developing
countries as the index of CBI following Cukierrnan’s finding which indicates that a legal
index of independence is not usefill for studying developing countries.19 The 27
developing countries are countries for which both macroeconomic time series data and
turnover rates of central bank governors are available.

However, even though we use the same index as Cukierman did, our estimation result
does not support Cukierman’s findings (1992) which show a significant relationship
between CBI and inflation rates in developing countries since the results are strongly
dependent on which countries are included in the model. We find no relationship
between CBI and inflation rates in developing countries after excluding hyperinflation
countries. Some possible explanations for this result will be discussed after showing the

estimation results.

‘

'9 We estimate the relationship between legal CBI index and inflation rate in developing countries for
1972-1995 period. Contrary to the developed countries’ case, the relationship is positive and insignificant.

27

For the estimation, we group the developing countries as follows; Group I includes all
27 developing countries for which data are available and Group 11 includes 23 developing
countries after excluding Argentina, Brazil, Israel and Peru which experienced
hyperinflations. We exclude those 4 countries from Group Ito formulate Group H.

Since these countries have experienced very high inflation rates during 1970-80’s and
there is a strong possibility that these hyperinflation rates, which were higher than
hundreds percent or even thousands percent a year, might come as the result of abnormal
macroeconomic behavior, political instability, or regional war in these countries.
Therefore, it may be reasonable to exclude these countries from the sample developing
countries to examine more general influences of CBI on macroeconomic variables.

For example, every year Argentina experienced inflation rates which were higher than
100% a year during 1975-1991 period - particularly for 1989-1990 annual inflation rates
were almost 3000% and changed the central bank governor more than once per year.
Such frequent changes in the central bank governor might induce high inflation rates, or
the government might dismiss the governor frequently because of his responsibility for
high inflation. Whichever is true, there is a great possibility that these very high
correlations between CBI and turnover rate dominated in Cukierman’s result. We
examine the relationship between M1 growth and the turnover rate as well as between the
inflation rate and the turnover rate in each group. Figure 3 depicts the relationship for

the period 1960-1995.

28

Table 6. CBI, Inflation rate, and M1 growth rate in developing countries

 

 

 

 

CBI(tumover rates) Inflation rate M1 growth rate
‘60-’90 ‘72-’90 ‘60-’95 ‘72-’95 ‘60-’95 ‘72-’95
Malaysia 0.09 0.10 3.42 (3.56) 4.68 (3.63) 10.65 (8.62) 13.53 (7.56)
South Afiica 0.12 0.10 9.40 (5.26) 12.65 (2.86) 11.73 (9.35) 14.80 (10.12)
Thailand 0.14 0.18 5.27 (5.51) 7.04 (5.84) 10.88 (5.88) 12.77 (6.22)
Israel 0.14 0.17 48.75 (81.61) 70.21 (93.22) 50.93 (69.59) 69.59 (79.96)
Philippines 0.15 0.10 11.48 (9.75) 13.85 (10.60) 13.75 (6.39) 15.97 (5.61)
Morocco 0.15 0.10 5.99 (4.12) 7.79 (3.73) 12.20 (5.51) 13.78 (4.54)
Honduras 0.16 0.24 7.36 (7.34) 10.16 (7.62) 12.00 (7.97) 14.07 (8.06)
Mexico 0.17 0.22 27.25 (33.85) 39.31 (35.85) 32.07 (27.36) 43.07 (27.95)
Nigeria 0.17 0.18 17.18 (17.44) 23.34 (18.31) 20.98 (22.13) 26.41 (23.76)
Ethiopia 0.20 0.30 7.37 (9.53) 8.78 (10.08) 10.87 (9.19) 13.83 (7.55)
Portugal 0.21 0.28 12.15 (8.43) 16.05 (7.66) 13.84 (7.26) 17.09 (6.66)
Greece 0.22 0.29 11.88 (8.29) 16.77 (5.39) 16.93 (4.30) 18.48 (3.41)
Nepal 0.23 0.18 8.98 (5.96) 10.04 (5.18) 17.50 (7.23) 17.13 (5.42)
Egypt 0.25 0.22 10.01 (7.29) 13.69 (5.41) 13.16 (7.45) 16.52 (5.97)
Panama 0.25 0.10 3.06 (3.73) 4.01 (4.28) 9.03 (11.18) 9.80 (12.90)
Ghana 0.26 0.23 34.43 (33.47) 42.33 (34.14) 29.03 (18.88) 39.00 (14.38)
Pakistan 0.29 0.28 8.11 (5.80) 10.22 (5.82) 13.38 (6.03) 15.54 (5.43)
Singapore 0.30 0.30 3.26 (4.93) 4.21 (5.69) 11.83 (6.15) 12.36 (6.17)
Venezuela 0.33 0.38 15.93 (20.03) 23.15 (21.11) 18.05 (19.83) 24.66 (21.06)
Peru 0.34 0.34 367.33(1345) 546.45(1629) 248.77(858.8) 368.l7(1047)
Zambia 0.38 0.38 34.92 (51.29) 49.48 (57.06) 32.34 (31.28) 35.16 (34.27)
India 0.38 0.40 7.94 (5.97) 9.11 (6.34) 12.62 (5.43) 14.32 (5.83)
Korea 0.43 0.32 10.69 (7.34) 10.27 (7.98) 23.38 (11.52) 21.96 (1 1.19)
Turkey 0.43 0.39 34.53 (31.31) 49.40 (28.12) 34.43 (20.56) 45.19 (16.96)
Brazil 0.56 0.59 368.04(738.3) 531.17(863.7) 310.55(790.5) 454.35(959.3)
Costa Rica 0.70 0.64 14.01 (16.47) 19.88 (17.43) 18.20 (14.27) 22.92 (14.86)
Argentina 0.99 0.94 278.29(633.5) 411.45(752.4) 314.59(619.4) 371 .94(628.7)

 

Note : Countries are ordered by the level of turnover rate for the period of 1960-1990. The standard

deviations are reported in parentheses.

29

In Group I, we find a very significant relationship between turnover rates and either
inflation rates or M1 grth rates as did Cukierman. The higher the turnover rates of the
central bank governor, the higher the inflation rate. This means that more independent
central banks are associated with lower levels of inflation (or M1 growth) as we expected.
In Group I, however, the hyperinflation countries seem to have a very strong influence on
the results of the regression. Since these hyper-inflation countries have relatively
dependent central banks, their strong relationship between inflation rates and CBI
dominates the relationship in all other countries and leads us to the same conclusion as
we reached with the developed countries, no matter what relationship prevails in the other
developing countries. Therefore, it is unreasonable to conclude that there exists a
positive relationship between inflation rates and turnover rates in developing countries.

In Group H, we do not find any significant relationship between inflation rates and
turnover rates, even though the significance level is as high as 0.01 in Group I. This
confirms that the relationships in the four hyperinflation countries dominates in Group I,
and that Cukierman’s result may be biased by the hyperinflation countries in his sample.
The relationships between M1 growth rates and turnover rates show the same patterns
as those between inflation rates and turnover rates in the both groups. Therefore, these
results do not support the previous view that there exists a negative relationship between

inflation rates and CBI in developing countries.

30

1. Group 1 countries

 

 

 

 

 

 

 

 

 

 

<lnflation m1e> <M1 growth>
‘00 r
l . 0
35° ‘ PER BRA
°°° I
250 «
2m 1
150 1 ‘ i
100 < . ‘
50 i 1
o 4
L - . . . ,
0.0 o 2 04 06 0.0 1 o
Turnover Rate Turnover Rate
inf=-35.9+290.511m m1 = -37 7+291.3 turn
(3.26)‘ 2 (4.29)‘
R2 = 030 - significan1001 level R = 0.42 ' elgnrficant 0.01 level

2. Group 11 countries

 

 

 

 

 

 

 

 

<lnllation rete> <M1 growth>
40 f 40 I --~
35 ‘ e ' 0 35 i e
e 0 i
30 30 . .
I
25 1 25 .
20 1 20 J
15 15 i
10 1 1° 7
1.
5 1 S 1
O . O
o T T V Y Y f T o Y T V Y Y Y 1
0.0 0.1 0.2 0.3 04 05 06 07 0.8 00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 00
Turnover Rate Turnover Rate
1n1=752+2191um m1=1119+207tum
(1.44) (1.80)
Rzgoiog R2=0.13

Figure 3. CBI vs. Inflation rate and M1 Growth in Developing Countries (1960-1995)

31

Table 7. Inflation rate and M1 growth rate vs. the CBI

 

 

Dependent Variables Explanatory Variables Group I Group II
1960-1995 Inflation rates Intercept -35.88 7.52
CBI 290.51” 21.91
(3.26) (1.44)
R2 0.30 0.09
M1 grth Intercept -37.71 11.92
CBI 291.28" 20.73*
(4.29) (1.80)
R2 0.42 0.13
1972-1995 Inflation rates Intercept -66.57 8.16
CB1 473.33" 36.96
(3.52) (1.66)
R2 0.33 0.12
M1 grth Intercept -54.74 14.13
CBI 405.92“ 25.95
(4.03) (1.59)
R2 0.39 0.11

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent
level ,** at 1 percent level.

32

We also find a positive relationship between variance of inflation rates and turnover
rates in Group I during both 1960-1995 and 1972-1995 periods. The relationship is
significant at the 3 percent level during both periods. This seems to be a result of a
correlation between the level and variability of inflation as argued by Alesina & Summers
(1993). In Group II, however, the relationship is not significant for the 1960-1995 period
while it is significant at 10 percent level for 1972-1995 period. This result implies that
CBI is neither correlated with the variance of inflation rates nor with the inflation rate in
the developing countries of the Group II. Therefore, it may not be quite accurate to argue
that more independence of central banks in developing countries has resulted in a lower
variance of and a lower level of inflation rates. This is confirmed by the relationships
between the variance of the M1 grth rates and turnover rates. In Group I, the
relationship is significant at 1 percent level but it is not significant at all in Group H.

This argument, however, does not mean that high independence of central bank is not
useful for achieving a low inflation rate in developing countries. That is, it suggests that
there could be no relationship between CBI and inflation rate only across developing
countries. However, if the level of CBI is increased in a country while other conditions
remain constant, inflation rate could decrease as argued by previous analyses. The recent

experiment of New Zealand gives us a good example.20

‘

2° see F. Holmes (1994).

33

Table 8. Variance of Inflation rate and M1 growth rate vs. the CBI

 

 

Dependent Variables Explanatory Variables Group I Group II
1960-1995 Variance of Intercept -83.46 5.95
Inflation CBI 666.46“ 28.24
(2.40) (1.48)
R2 0.19 0.09
Variance of Intercept -96.15 8.03
M1 growth CBI 647.78*** 14.81
(3.14) (1.25)
R2 0.28 0.07
1972-1995 Variance of Intercept -121.75 3.86
Inflation CBI 872.41 *"‘ 37 .44*
(2.51) (1.75)
R’- 0.20 0.13
Variance of Intercept -120.25 7.12
M1 growth CBI 783.34**" 17.27
(3.05) (1.29)
R2 0.27 0.07

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent
level ,** at 3 percent level, *** at 1 percent level.

34

However, the one thing we need to note is that Cukierman (1992) and Cukierman et.

al. (1992) did not use inflation rate itself. Actually, they use the transformed inflation rate
in order to reduce heteroskedasticity of the error and thus improve the efficiency of the
estimate. Cukierman et. a1. argue the reason for using transformed inflation rate as
follows:
“ .. Most countries had average inflation rates of 20 percent or less, but a few had three-
digit inflation rates in some decades. Using the straight inflation rate would give undue
weight to these outlier observations. So we transformed each year’s inflation rate into
inflation divided by one plus the inflation rate and then took the geometric average for the
decade. This variable represents the annual real appreciation of a given amount of
money; we call it D:

D = 112/ (1 + 1t)
where rt is the inflation rate and D (hence, the transformed inflation rate) takes a value
from 0 to 1. When inflation is 100 percent a year, D is 0.5. ..” (Cukierman et.al., p370)

In spite of their argument for using the transformed inflation rates, even if we use the
transformed inflation rate, we find that the estimation result is still biased by these
hyperinflation countries. In Group I, we find very significant relationship between
transformed inflation and turnover rate during both periods. They are significant even at
0.1 percent level. However, in group II, this significant relationship disappears for the
1960-1995 period and the relationship is significant only at 10 percent level for the post
Bretton Woods period.

The results of estimations for developing countries without the hyperinflation
countries contradict the previous argument that CBI has important impacts on economic
behavior in developing countries. I think there can be many explanations of this

contradiction. First, the turnover rate may not be a proper index for CBI in developing

countries. In some developing countries, a low turnover rate of central bank governors

35

Table 9. Transformed Inflation rate vs. Turnover rate

 

 

Explanatory Variables Group I Group II
1 960- 1 995 Intercept 0.029 0.064
CBI 0.360 ** 0.127
(4.04) (1.52)
R2 0.40 0.10
1972-1995 Intercept 0.026 0.071
CBI 0.509 ** 0.213*
(4.39) (1.79)
R2 0.44 0.13

 

Note : The t-statistics are reported in parentheses. * indicates significance at the
10 percent level , ** at 0.1 percent level,

does not mean that central bank has a high level of independence because a relatively
subservient governor may stay in office for a long time, or because a long-term governor
is sometimes an intimate person or even a relative of the president or the prime minister
in some despotic states.21

Second, the utility function of the central bank in developing countries may be
different from that of central banks in developed countries. In developing countries, a
central bank may place more weight on other economic objects such as economic growth
than on price stability regardless of its independence levels. Then we could find that
some central banks have tried to reduce inflation rates while others have tried to pursue
other objects. In this case, CBI will not have any concrete relationship with other

economic variables across countries.

 

2' Cukierman (1992) argues that this may be true for countries with exceptionally low turnover rates, such
as Denmark, Iceland, and the United Kingdom.

36

As a third explanation, there exits no relationship between CBI and economic
performance in developing countries not because CBI is unimportant in these countries,
but because level of CBI is too low to have an impact on the economy. Conversely,
within the countries which have little CBI, the difference of CBI is unrelated to the
difference in economic behavior. On the other hand, within developed countries, which
have high independent central banks, many analyses show that the difference of CBI is
strongly correlated with the difference in economic behavior. Therefore, we may suspect
that CBI should be higher than some critical independence level in order to have some
impacts on economic behavior. If CBI is below this level, the difference of CBI is
unlikely to be one of the factors which characterize the economy.

We draw the following graph, Figure 4, by using the CBI data of Cukierman et. al.
(1992) to see the significance of CBI in economy. The Y-axis represents the difference
between fitted and actual inflation rates and the X-axis represents the overall
independence level in their paper. The higher value on X-axis means the lower overall
independence of central bank. The graph shows that the gap between fitted and actual
inflation rates increases sharply when the overall independence of central bank decreases.

The large gap means that there exists no relationship between CBI and inflation rates
because the error of estimation is too big for CBI to have a significance in the economy.
Conversely, if CBI has a significant relationship with inflation we can expect small gap
between the fitted and the actual inflation rates because the fitted value comes from a
linear function of the independence level of the central bank. Hence, as the gap
decreases, the relationship between CBI and inflation rates becomes stronger. The dotted

line indicates the median of CBI (0.17) of 73 countries which includes the developing

37

countries as well as the developed countries. The overall indexes of all industrial
countries are below the median - actually, all of those countries are below 0.11; and most
of the developing countries are above the median.

Therefore, within the developing countries it is more difficult to find a relationship
between CBI and inflation rates since the gap between fitted and actual inflation is much
bigger than in the developed countries. If CBI increases to some levels above the median
independence level the gap will decrease and we can expect much more stable
relationship between CBI and economic performance in the developing countries. This
gives an implication for countries which are going to reform their central banks by giving
more independence to get some benefit from high CBI. It says that, if a country has the
central bank which independence level is much below the median level, it is very difficult
to get the effects they want with slight improvement of CBI, so that they may need much

greater reform in central bank system in order to firlfill the purpose.

#1
O
I

u
o
1

—L
o
1

 

 

 

Gap b/w Fitted & Actual lnf.(%)
N
O

 

O

0.1 0.2 0.3 0.4
<-higher-- Overall Independence Level --lower->

Figure 4. Gap between Fitted and Actual Inflation rate

38

4. Variables for Central Bank Independence and Inflation

Based on many studies which showed the inverse relationship between CBI and
inflation rates and the success of the Bundesbank system in Germany, recently some
countries are trying to reform the central bank system by giving more independence to the
bank. However, it is uncertain which factors are really important in enhancing CBI level
since there are many factors which affect the independence level of central bank. Among
these factors, there can be some key elements which really contribute to improvement of
CBI, while others are of little help for improving CBI. Therefore, although a country
reforms the central bank system by raising overall CBI index, if there is not much
improvements in these key factors which play important roles in the actual practice of
central bank, it would be hard to get real effects of central bank reform.

Regarding this issue, it is very useful to examine the variables which are components
of Cukierrnan’s legal index. Cukierrnan’s legal independence index consists of 4
categories which are subdivided into 16 different legal variables coded on a scale of 0
(lowest level of independence) to 1 (highest level of independence). These coded values
are added up for each of the four categories, and finally the weighted average of the
values for the four categories becomes the legal index. Table 10 summarizes these

categories.

39

Table 10. Variables for legal independence index.

 

 

Variables weight Variables weight
1. CEO variables 0.20 4. Limitation on lending to the government 0.50
a. Terms of office (0.05) 3. Limitation on nonsecuritized lending (0.15)
b. Who appoints CEO? (0.05) b. Securitized lending (0.10)
c. Dismissal procedure (0.05) c. Terms of lending (0.10)
d. Possibility of holding other office (0.05) (1. Potential borrowers from the bank (0.05)
2. Policy formulation 0.15 e. Limits on central bank lending (0.025)
a. Who formulate monetary policy (0.05) f. Maturity of loans (0.025)
b. Who has frnal word in resolution of conflict (0.05) g. Interest rates on loans (0.025)
c. Role in govemment’s budgetary process (0.05) h. Primary market prohibition (0.025)
3. Central bank objectives 0.15

 

Using data of 18 developed countries, first we regress each of 16 variables on inflation
rates. All coefficients but one have negative signs, which confirm the inverse relationship
between legal independence and inflation rates. Among these 16 coefficients, the
coefficients of the final authority variable in policy formulation category and some
variables in limitation on lending category are significant. The final authority variable
and the securitized lending variable have inverse relationships with inflation rates at
significance level 0.03. Although the relationship between inflation rate and each sixteen
legal variables give us an idea which ones are more important factors for CBI, the
relationship between four categories and inflation rate has more general implication since

each sixteen legal variables has a different weight to make four categories and this

40

combination of each legal variable can draw a different result regarding the relationship
with inflation rate.

Therefore, next we regress the four categories on inflation rates. The results are
reported in Table 11. From these results, it is obvious that the most important variables
for CBI are both the variables concerning the policy initiatives of the central bank in the
decision making process and the variables concerning the conditions attached to central
bank credit to the government. The relationship either between policy variable and
inflation rates or between combinations of policy and credit variable are a little stronger
than the relationship between overall legal independence and inflation rates. The
variables regarding CEO or bank objectives have no relationship with inflation rates at
all. If we make a new CBI index using only policy and credit variables, it shows a more
significant inverse relationship with inflation rates than does the overall legal

independence level.

Table 11. Variables for CBI vs. inflation rates (1972-1995)

 

 

Intercept 7.40 8.27 6.99 8.79 8.96 9.00 9.73 8.54
CEO -5.99 2.21
(-0.38) (0.18)
Policy -51.91** -36.53* -37.62*
(-3.25) (-1.95) (-1.90)
Objective -0.49 1.21
(-0.28) (0.81)
Lending -12.30** -7.00 -8.17
(-2.86) (-1.46) (-1.56)
Real CBI -11.60**
(-3.32)
Overall -8.26**
legal index (-3.09)
R2 0.01 0.40 0.01 0.34 0.47 0.41 0.37 0.50

 

Note : The t-statistics are reported in parentheses. Real CBI is the weighted sum of policy and lending
variable!“ indicates significance at the 10 percent level, ** at 1 percent level.

41

This implies that even though a country raises the degree of central bank independence
through reforming the variables like the term of office for the central bank governor or
the objectives of central bank, it might not contribute to real CBI enhancement. Since the
result suggests that real improvement of CBI comes from reforming variables which are
directly related with actual monetary policy, not from improving formal variables like
literal objectives of the central bank, the central bank might not have real independence as
long as the bank does not have certain degree of independence in executing monetary

policy.

5. Summary

In this chapter, we examined the various CBI measurements and found that there is a
rough consensus on level of independence level of each central bank in developed
countries. Then, with these different CBI indexes, we reexamined the relationship
between CBI and the level and variance of inflation rates (or M1 growth rates). With the
sample of developed countries, the estimation results confirmed the previous studies
which argued that CBI has a negative effect on inflation rates. However, we found that
this inverse relationship between CBI and inflation rates disappeared in developing
countries after we exclude four hyper-inflation countries from the sample of developing
countries. This implies that the previous studies might be biased by these hyper-inflation

countries since these countries have very dependent central bank systems.

42

As the most important factor we did not find the inverse relationship in developing
countries, we argue that, in these countries, the level of CBI is too low to have an impact
on the economy. This is supported by the existence of much bigger gaps between fitted
and actual inflation rates in these countries than in developed countries. This gives an
important implication to the countries which are going to reform their central bank
system. It says that, if a country has a highly dependent central bank, slight improvement
of CBI would not be helpful in achieving low inflation and so they may need much
greater reform in the central bank system.

The similar implication comes from the estimation of the relationship between the
variables for CBI and inflation. The estimation results imply that real improvement of
CBI comes from reforming policy variables which are directly related with actual
monetary policy, not from reforming formal variables like literal objectives of the central
bank. In summary, this chapter suggests the countries which are going to make their
central bank more independent need to make more rapid and greater reform in the central
bank system, especially, in policy variables among many factors regarding CBI in order to

get the effects they want from the reform.

CHAPTER III

CENTRAL BANK INDEPENDENCE AND LIQUIDITY EFFECTS

In this chapter, we examine the possibility that the degree of monetary policy effect on
economy could differ across countries among which central banks have different levels of
independence. In order to measure the monetary policy effect, we estimate the liquidity
effect which is a short-run negative response of interest rates to an increase in the money
supply.

To examine the relationship between CBI and liquidity effects, first, we set a model
economy which produces liquidity effects and investigate the effect of CBI on the
liquidity effects in the model. Then, we apply empirical approaches to estimate the
relationship between CBI and liquidity effects by adopting three different traditional
approaches and seemingly unrelated regression (SURE) method. We model only
developed countries to explore the possible relationship between CBI and liquidity effects
since the results of previous chapter suggest the possibility that CBI has little effect on
economic variables in developing countries.

To estimate the liquidity effects, we concentrate on the relationship between a money
aggregate, M1, and the money market interest rate in 12 OECD countries.22 For CBI
index, we use the legal independence index suggested by Cukierman (1992). We use
money market interest rates which are consistent with those used in most previous

studies. These interest rates for 12 countries come from IFS series number 60B in each

 

22 We estimate the relationship with alternative money aggregate, total reserves and monetary base, in the
end of the first section.

43

44

country section. This is the federal funds rate in US. and the call rate in other countries.
These rates are extremely short-term interest rates which allow us to separate liquidity
effects from expected inflation effects without imposing a theory of the term structure and
expected inflation. For the series on M1, we use IFS series number 34 which is the sum
of currency outside banks plus demand deposits held with the monetary system by the rest
of domestic economy other than central government. We use the monthly series for each
variable.23

We estimate the liquidity effect over 1970’s with a sample of countries for which data
are available and over 1980 tol994 period with all 12 countries as well as over the full
1971 to 1994 period. The sample countries are Australia (’69.7), Canada (’75.1),
Denmark (’74.1), Finland (’77.1), France (’64.1), Germany (’64.1), Italy (’71.l),
Netherlands (’60.1), Norway (’71.8), Sweden (’66.l), UK (’72.1) and US (’57.1). The
month in parenthesis indicates the first month when data are available in each country.
For Germany, we will consider only West Germany before unification in 1990 to avoid

effects of the large economic shocks on some economic variables after unification.

1. Model Approach

l-l. Basic Liquidity Model

 

23 Geweke and Runkle(1995) suggest that time aggregation from a biweekly interval to a quarterly interval
is not a problem when identifying monetary policy and that time aggregation does not seem to be a problem
when evaluating the dynamic effects of typical changes in variables.

45

We consider a basic liquidity model in Christiano and Eichenbaum (1992b) which is a
simplified version of the model in Christiano (1991). The following argument for the
model mainly comes from Christiano and Eichenbaum (1992b). In basic liquidity model,
the only source of uncertainty in the agents’ environment pertains to monetary policy.
The model has three types of agents: households, goods-producing firms, and financial
intermediaries. At the start of period t, the representative household possesses the
economy’s entire beginning-of-period money stock M,. The household allocates Q,
dollars to purchases of the consumption good, C,, and lends the rest, M,-Q,, to financial
intermediaries. Consumption purchases must be fully financed with cash that comes
from two sources: Q, and current-period wage earnings. The household chooses Q,, C,,

and the fraction of period t devoted to work (L,) to maximize the expected value of the

criterion E ,B'U (C, ,1 — L,). Here U(C,, I-L,) denotes the household’s utility function,
1:0

given by (1 -}71n(C,) + fln(I-L,). Also, ,6 and yare scalars between 0 and 1.

In the basic liquidity model, the household’s contingency plan for Q, is not a function
of the period-t realization of monetary policy. The maximization occurs subject to the
cash constraint that nominal consumption expenditures, P,C,, can not exceed Q, plus W,L,.
Here P, and W, denote the period-t dollar price of goods and labor, respectively. In
addition, the household must obey its budget constraint,

M+1=R1(M'QJ +D,+F,+ (Q:+ le‘PlCt)
where R, is the gross interest rate in period t and F, and D, denote period-t dividends

received from firms and financial intermediaries respectively.

46

The financial intermediary has two sources of funds: M, - Q, and lump-sum injections
X, of cash by the monetary authority. These funds are lent over the period in perfectly
competitive markets to firms at the gross interest rate R,. The financial interrnediary’s net
cash position at the end of the period is distributed, in the form of dividends, to the
financial interrnediary’s owner, the household, alter the consumption-good market has
closed.

The period-t technology for producing new goods is given by
f(K,, z,L,) = K,“(2,L,)"" + (1-6)K, for 0 < a < 1 and 0 < 5< 1.
Here K, is the beginning-of-period-t stock of capital, 6 is the rate of depreciation on
capital, and the function f(-,-) denotes new period-t output plus the undepreciated part of
capital. Also, 2, is the state of technology at period t, which grows at the constant
geometric rate ,u > 0. Firms must borrow working capital W,L, from financial
intermediaries to cover their labor costs. Loans must be repaid to the financial
intermediaries at the end of period t. Consequently, the total period-t cost associated with
hiring labor equals R,W,L,.

Firms own the stock of capital, which evolves according to I, = K,. 1 - (1 - (5) K, where
1, denotes period-t gross investment. Unlike labor, capital is assumed to be a credit good,
so that the firm need not borrow funds from the financial intermediary to finance
investment activities. At the end of the period, after consurnption-good market closes, the
firm’s net cash position is distributed to its owner, the household. The perfectly
competitive firm maximizes the expected present discounted value of dividends by choice
of contingency plans which satisfy 1, and L, as functions of model variables dated period t

and earlier.

47

The shock in the model economy is disturbance to the rate of grth of money, x,.
Here, x, a X, / M,, where X,, again, is cash injections from the monetary authority to the
financial intermediaries and M,., = M, + X,. We assume that the shock enters this way :
x, = (1 -p,,) x + pxx,-1+ a”

The shock to money growth, 5,), is mutually uncorrelated at all leads and lags and are
uncorrelated with x,.], j > 0. It is the part of x, that can not be predicted based on past
values of the variables in the model. For this reason, 8,, is referred to as the unexpected
components of x,. The parameter p, in equation controls the autocorrelation properties of
x,. In particular, the correlation between x, and x,., is just ,0," for j > 0. Finally, x is the

unconditional mean of x,.

1-2. Generating a Liquidity Effect

The key feature of the basic liquidity model which lets it generate a substantial
liquidity effect is that the assumed rigidity in Q, prevents an increase in the money supply
from being distributed proportionally among all agents. If Q, does not respond to X,, a
positive money shock increases the total percentage of the money supply available to
financial intermediaries to lend to firms. However, this requires that firms absorb a
disproportionately large share of new cash injections. For firms to do so voluntarily,
interest rates must fall. Of course, if the growth rate of money displays positive
persistence, then the expected inflation effects of a change in the growth rate of money
exert countervailing pressure on interest rates. Under these circumstances, whether

interest rates fall or rise depends on which effect is stronger.

48

1-3. CBI in Basic Liquidity Model

The theoretical argument for CBI is based on the view that political policy-makers are
subject to an inflationary bias. Monetary policy enables them quickly, but temporarily to
achieve various real objectives such as high employment or low interest rates. In the
process, high-powered money is increased fueling inflation and inflationary expectations
and creating an inflationary bias that persists long after the desirable effects of monetary
expansion have disappeared. The inflationary policy bias can be eliminated by
precommitting policy prior to contracting time to price stability or to a low rate of
inflation. One way of implementing this commitment in practice is to give sufficient
independence to the central bank and to direct it by law and/or other means to focus on
price stability even if that implies a relative neglect of other objectives.

Therefore, the central banks would show different characteristics in their monetary
policies depending on the levels of their own independence. This implies that the time
path of money growth rate also might be affected by CBI. We assumed the money
growth rule as follows in the model : x, = (I - [9,) x + p, x,-, + 8,).

Now, consider how this growth path would be changed by diflerence in levels of CBI.
First, x is the unconditional mean of x,. From the previous chapter, we know that low
CBI countries have experienced higher average growth rate of money than high CBI
countries, that is, x would be greater in low CBI countries than in high CBI countries.
The country which has higher x value should have experienced higher inflation rate.
Therefore, higher x value leads to stronger expected inflation effect in the country and so,

as Christiano and Eichenbaum (1992) mentioned, this strong expected inflation effect

49

would make interest rates fall less given monetary expansion in the model. However, this
does not mean that high CBI has a positive effect on monetary policy in this model since
p, also could be affected by CBI.

In monetary policy rule above, p, controls the autocorrelation properties of x, and
determines the dynamic characteristics of time path of money growth rate since it
represents the persistency of money shock in the model economy. High value of p,
means that money shock is much more persistent in the economy. Here, we need to
examine why inflation is persistent.

Cukierman (1992) accounts for the persistence by linking the central bank’s current
actions to the public’s expectations about the future. Current inflation provides
information to the public, either about the central banker’s type or about the central
bank’s private information about serially correlated aggregate supply shocks. Therefore,
high inflation today raises expected inflation next period as the public infers that the
central bank is soft on money grth in the first case or because the public infers that a
persistent aggregate supply shock has raised the marginal benefit of inflation in the
second case. The rise in expected inflation in the following period also raises the actual
rate of inflation for that period. Thus, persistence arises because of learning
considerations, not because of wage and price rigidities in the economy. Cukierman
(1992)’s models imply that inflation’s persistence is due fimdamentally to the persistence
properties of the shock, not the policy response.24 If all shocks are transitory, so is
inflation, or if the public shares the same information as the central bank, inflation’s

persistence is exactly that of the shock. Cukierman (1992) and Ball (1990) also argue

50

that central bank preference has an important role in explaining the persistence of
inflation.

Since the only shock in our model economy is a monetary shock and inflation history
of a country has been affected by the level of CBI of the country, Cukierman’s argument
gives us an empirical implication that persistence of money shock in the model could be
different if the country has the different levels of CBI. Low CBI countries have shown
more persistent inflation history and their inflation rates have been increasing for last
couple of decades. In a simple model in which only shock in the economy is monetary
shock, inflation persistence is due to the persistence properties of the monetary shock.
Therefore, we infer that persistence of monetary shock is stronger in low CBI countries
even though there are other factors which affects inflation persistence.

To examine the relationship between CBI and persistence of money shock in real
economy, we follow the estimation method in Christiano (1991), and Christiano and
Eichenbaum (1992). We make first-order autoregressive models for each country in our

sample and estimate the coefficient p, of each country for 1970-1994. Then we regress

CBI on the value of p,,. Our estimated result is as follows;

p, = 0.427 - 0.337 CBI R2 = 0.65
(-2.36)

and the coefficient of CBI is significant at 10 per cent level.
This estimation confirms the preceding argument that low CBI countries have had more
persistent money growth shock during last couple of decades. This might be explained by

the facts that in low CBI countries, money growth rate has been increasing and the

 

2" See Chapters 9 ~16 in Cukierman (1992).

51

government in these countries might have strongly affected the conduct of monetary
policy by affecting institutional structure or preference of the central bank.

Now consider how persistence in money shock affects the liquidity effect in the
model. A money shock increases not just x,, but also x,., for j>0. And the change in x,+,
would increase when p, approaches to 1. The unexpected upward revision in the forecast
of x,. 1 exerts upward pressure on P,+1/P,, that is, a persistent jump in money growth raises
anticipated inflation. The more persistence in money growth shock would result in the
higher anticipated inflation. Therefore, this higher anticipated inflation exerts more
countervailing pressure on interest rates. Under these circumstances, interest rates might
fall less or even might rise by stronger expected inflation effects. Christiano (1991)
shows the quantitative results that high value of p, decreases the liquidity effect in the
model economy.25

The last element to examine is 5,, in money growth rule. In previous chapter, we did
not find any relationship between Cukierrnan’s index for CBI and the variance of money
growth rate while we found a relationship between Alesina and Summers’ index and the
variance of money growth rate. The variance of money growth rate is greater in low CBI
countries than in high CBI countries when we use Alesina and Summers’ index for CBI.

However, this does not mean that the variance in money growth rule in our model, 0'“. ,
also has a relationship with CBI because we control the autocorrelation properties of
money growth rate in the model. 63,, is the unexpected component of x, that can not be

predicted based on past values of variables in the model. The size of 5,, depends on

monetary policy authorities’ decision or economic circumstances at each time and so does

52

the variance, a“. Policy authorities decide the money grth rate after they consider all

available economic and political information. Therefore, the size of 5,, is determined
when they set the money grth rate at each time and is independent of degree of
persistence in money growth rate rule. If large variance of money growth rate results
mainly from strong persistence of shock in money grth rate, the variance of money

growth rate would have little relationship with the variance in our model, 0'”. If the

variance of money growth rate comes mainly from other sources, the variance in our
model could have a relationship with the variance of money growth rate and also could
have a relationship with CBI, particularly, Alesina and Summers’ index. However, in
developed countries in our sample for 1970-1994, we do not find any relationship either
between the variance of money grth rate and the variance in our model or between

CBI and the variance in our model,0' no matter which CBI index is used. These

£,x ’

results suggest that 0'” can be considered as a parameter which is independent of CBI.

In summary, level of CBI affects the money growth rule in the basic liquidity model
we examined. The lower CBI results in higher average money growth rate (high x) and
more persistence (high p,) of money growth shock. High average money grth rate
and/or more persistent money grth shock raises anticipated inflation. This stronger
expected inflation effect might dominate liquidity effect or at least lessen the liquidity
effect of a given expansionary money shock. Therefore, in the basic liquidity model, low
CBI results in weak monetary policy effect, that is, weak liquidity effect as we assumed at

the beginning of the paper.

 

2’ See the Table 2 in page 19 in Christiano (1991).

53

2. Traditional Empirical Approaches

In this section we use three different traditional methods to estimate liquidity effects.
The traditional analysis of the liquidity effects is firmly rooted in the comparative statics
of money demand: An increase in the rate of growth of the money supply, holding output
and prices constant, causes the nominal interest rate to fall. After an increase in money
growth there will be a period over which the interest rate is depressed. Eventually,
inflation will adjust to the new money growth rate and the long-run correlation between
the interest rate and money grth is positive. The long-run tendency for changes in
money growth to be reflected in expected inflation and thus, nominal interest rates, is
referred to as the expected inflation effect. The negative interest elasticity of money
demand produces the liquidity effect, but ultimately the expected inflation effect
dominates the liquidity effect.

There are two traditional empirical approaches to estimate the liquidity effect. One is
the reduced-form approach which calculates reduced-form correlations between interest
rate and money aggregate and the other is the aggregate money demand approach which
estimates a money demand equation. The reduced-form approach regresses interest rate
against current and past monetary aggregate. The aggregate money demand approach
estimates the interest elasticity that underlies the liquidity effect by conditioning on a
broader set of variables and imposing more restrictive assumptions on dynamics than
does the reduced-form approach. In this paper, we replicate the reduced-form approach to

estimate the liquidity effect and examine the relationship between CBI and liquidity

54

effect. In addition to the reduced-form approach, we also use dynamic correlation method
following Christiano and Eichenbaum (1992).

However, although the traditional approaches give us clear methodologies for seeking
liquidity effects, these approaches assume that the money supply is exogenous. When the
variation in money supply is not independent of money demand, these empirical results
do not distinguish how much of the money-interest correlation is due to interest elasticity
of money demand and how much of the correlation arises from the dependence of money
supply and interest rates on other variables. Furthermore, as Leeper and Gordon (1992)
argued, few economists would disagree that the traditional approaches correctly
characterizes the economy’s short-run response to an exogenous monetary shock and
there is considerably less agreement on how to identify such shocks empirically.
Therefore, in order to compensate these deficiencies, in next chapter, we will try to

identify money supply shock with vector autoregression (VAR) approaches.

2-1. Traditional Distributed Lag Regressions.

The traditional empirical approach to measuring the liquidity effect, which is
associated with Cagan and Gandolfi (1969), Melvin (1983) and Cochrane (1989),

regresses interest rates against current and past growth of a monetary aggregate.

SS

r1=a +fl(1')p,+81

" 1
Where fl (L)=1§0’BJ'L (2.1)
r , = level of interest rate
,0 t = grth rate of money

If it is found that the cumulative coefficients from this regression are significantly
negative over some relatively short horizon, it is interpreted as evidence that the liquidity
effect dominates the relationship between money growth and interest rates in the short
run. Since these regressions condition only on money growth rates, to interpret the
coefficients as reflecting the effects of monetary policy on interest rates we must assume
that money grth and interest rates do not have a strong tendency to respond jointly to
other variables. To investigate the relationship between money growth and the interest
rate, we will use the following lag lengths for each period : l97l.1~1979.l2 (18 lags),
1980.1~1994.12 (24 lags), l97l.1~l994.12 (36 lags). In choosing lag lengths for the
regressions, usually there exists a trade-off between including enough lags to exhaust the
information in the data and overfitting the data. Here, we use the lag lengths which are

consistent with those used in previous studies.26

2-2. Dynamic Multipliers Methods

These distributed lag regressions have been largely supplemented by estimates of
dynamic multipliers associated with unanticipated changes in money growth. To obtain

the dynamic multipliers implied by traditional regressions, we append to the interest rate

56

equation describing the evolution of money growth, which allows the data to characterize
money innovations. A money innovation is defined as a one-tirne change in the residual
of the money equation and the two equations are used to trace out the paths of money
growth and interest rates associated with a typical money innovation. Here, we maintain
the assumption that money growth is exogenous and estimate a univariate autoregressive
process for it.

pt : §O+§ (UPI + ”I

n i (22)
where 6 (L) = £161.14

Combining (2.1) and (2.2) yields the multipliers y(L) in
rz=a/+7(L)771+81
where a/ a a +fl (1) [1—5 (1)]‘15O (2.3)
y (L) a ,6 (1.)[1—5(L)]_1
To estimate liquidity effect, we examine the path of the interest rates for 36 months

following a one-percentage point innovation to the growth rate of the money. (that is, we

set 11, = I, and n,+k= 0 fork¢ 0)
2-3. Dynamic Correlations

In addition to the previous reduced-form approaches, we estimate the liquidity effect

by investigating dynamic correlations between the variables. Following Christiano and

 

2" Leeper and Gordon (1992) summarize the previous studies which used traditional distributed lag
regressions to get the liquidity effects.

57

Eichenbaum (1992), we get liquidity effects through dynamic correlations between
interest rate at time t and money growth rate at time tit. Here, we assume that money is
positively correlated over time and consider a benchmark scenario in which the only
shocks are those of the money supply. Christiano and Eichenbaum (1992) describes the
scheme for dynamic correlation as follows:

“... Consider first the correlation between FF, (Federal funds rate) and future values of M,
(money). Suppose that, at time t, there was an unanticipated increase in the money
supply. Given a liquidity effect, this would be associated with a decline in F F,. With M,
positively correlated over time, high values of M, would be associated with high values of
M,+,, for 1 >0. Other things being equal, we would expect FF, to be negatively correlated
with future values of M, with the exact magnitude of the correlation depending on the size
of the liquidity effect and the degree of serial correlation in M,.

Next consider the correlation between FF, and past values of M,. Suppose that at time
t-r, for r > 0, there was an unanticipated increase in money supply. This would exert
negative pressure on FF,-,. Suppose that M, is sufficiently autocorrelated that the initial
increase in M,., is associated with higher growth rates in M,.,,,- for j 2 1. This, we expect,
would generate an increase in the anticipated rate of inflation from time t-T+j, for j 2 1. If
the liquidity effect lasted only one period, then the inflation effect would dominate after
one period, so that FF,-,+,-, for j 2 1 would rise. Consequently, FF,-,+,-, for j 2 1, would be
positively correlated with M ,.,, that is, p(FF,, M,.,) > 0 for r 2 l, where p( , ) denotes the
correlation operator. In fact, there is no reason to believe that the liquidity effect lasts for
only one period. Suppose instead that liquidity effect dominated the expected inflation
effect for k periods. Then p(FF,, M,-,) would be negative for r s k, but positive for 1.‘ >
k. In this sense, k can be thought as measuring the persistence of the liquidity effect.

While useful for pedagogical purposes, the logic of the previous scenario holds only if
the sole source of aggregate uncertainty is shocks to the money supply. With other
shocks to the system, the dynamic correlation between FF and the stock of money
depends, at least in part, on the way the FOMC reacts to the other shocks. ...” (p.343)

Here, we will get the dynamic correlation, p(i,, M,-,), for -18 s r .<_ 18 (1971.1~1979.12),
-24 s r s 24 (1980.1~1994.12) and -36 s r s 36 (1971.1~1994.12). The choice of value
of t, the lengths of leads and lags, is consistent with the choice of the lag length in the

previous reduced-form approaches. Furthermore, we will calculate the value of k to

measure the persistency of the liquidity effect for each period.

I
t

‘tg

Sig

d1

16'

58

2-4. Estimation Results

With the equation (2.1), we get a response path of interest rates given monetary shock.
Figure 5 depicts the cumulative lag coefficients, the sum of [31's in equation (2.1), on the
growth rate of M1. In high CBI countries, the cumulative coefficients are consistently
significantly negative for over more than two years. This implies there exists persistent
and strong liquidity effects in high CBI countries. Liquidity effects are strong enough to
dominate the expected inflation effects for the two year time span in the high CBI
countries. Liquidity effects, however, seem to be dominated by expected inflation effects
in the low CBI countries. The responses of interest rates in the low CBI countries show
little negative relationship with the monetary shock and in a short period of time after
monetary shock, interest rates begin to increase because of the expected inflation effects.

To examine the relationship between CBI and liquidity effects, we pick the value of
contemporary liquidity effect ([30) and minimum value of liquidity effect (value of the 213,
which reaches the minimum point in the given time span) for each country. In addition,
we calculate average liquidity effect which represents response of interest rates on
average during the periods of time we examine. This number would be positive if, on
average, expected inflation effects are stronger than liquidity effects in the given time
span. A negative sign would imply that an expansionary monetary shock, on average,
decreases interest rates through a liquidity effect rather than increasing interest rates
through expected inflation effects during those periods. After computing these values, we

regress each of these values on the CBI index. The results are reported in Table 12.

Responeeolfitem Charge

mambmcw

Response of R, to M1 Change

59

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.6
— U.S.(0.46)
------ NET(0.42)
— - GERtO.69)
1.2 v V V v v v v
0 4 6 12 16 20 24
0.6
0.4 .1 — NOR(0.17)
...... smwgg)
0.2 4 g — — U.K.(0.27)
0.0
-0.2 -1
-0.4 -
-0.6 -
-0.6 -
-1.0 ~
-1.2 . . T . v .
0 4 6 12 16 20 24
0.6
— us.
0" _( ..... NET
/ — - GER
..._ — NOR
3.. A I \J SM
00 1&- 1‘; -. \-./ . -/ — U'K
'/ '~/ ~/ ' \ " 4:;R /
-0.2 T '
-0.4 a
-0.B-1 \/ """"" \-—-\ ,’/\\
-0.6 -1 \ a
1.0 - .........
'1.2 V l T T r I V“ f T V '
0 4 6 12 16 20 24
Time (Month)

Figure 5. Response of R to M1 Change (1980~1994: Eq.(2.1))

 

60

Table 12. CBI and Liquidity Effects (Eq.2.1)

 

 

Dependent Explanatory Variables 1970-1979 1980-1994 1970-1994
Variables
Contemporary Intercept 0.029 0.494 0. l 22
Liquidity Effect CBI -l.074* -l.597**** -0.807"‘***
(-1.80) (-4.39) (-3.18)
R2 0.32 0.66 0.50
Minimum Value Intercept 0.029 0.420 0.009
of Liquidity CBI -1.409*** -1.914*** -0.896**
Effect
(-2.96) (-2.94) (-2.40)
R2 0.56 0.46 0.37
Average Intercept 0.143 0.481 0.117
Liquidity Effect CBI -0.700 -1 .597 1"" -0.691*
(-1.20) (-3.36) (-2.18)
R2 0.17 0.53 0.32

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent
level, ** at 5 percent level, *** at 3 percent level, and ***"‘ at 1 percent level.

The results confirm our hypothesis that CBI level influences the liquidity effect. The
greater CBI, the stronger liquidity effects. Contemporary responses of the interest rate to
monetary shocks are larger in the high CBI countries than in the low CBI countries in
every sample period. In particular, the inverse relationship between CBI and the
contemporary liquidity effect is significant at 1 percent level for both 1980-1994 and
1970-1994 periods. The minimum value of the liquidity effect also seems to have an
inverse relationship with CBI. For the periods of 1970-1979 and 1980-1994, this inverse
relationship is significant at the 3 percent level, and at the 5 percent level for 1970-1994
period. This implies that the response of interest rates to monetary disturbances is
stronger in high CBI countries than in low CBI countries. Figure 6 clearly shows these

relationships for some sample periods.

l.

1980 - 1994

< Contemporary Liquidity Effects >

61

 

0.6

0,4 -

0.2 <

 

 

 

 

2

T Y 1 Y

1 0.2 0.3 0.4 0.5 00 0 7
Central Bank Independence

L10 8 0.494 - 1.597 C31
(4.39) ‘

Rz=0.66 ' significance 0.001 level

.1970 -1994

< Contemporary Liquidity Effects >
2

0.0

 

 

 

 

T Y T I 7

0.1 0.2 0 3 0.4 0.5 0.0 07
Central Bank Independence

L10 = 0.122 - 0.807 CBI
l-3.181 '

n’-o.5o ° significance 0.01 level

0.8

< Minimum Value of Liquidity effects>

 

 

 

 

 

 

 

 

 

 

 

04
0.2 - '
00 '
-0.2 4
-0.4 4
00 -
00
10 «
O
1.2 T . . . . .
0.1 0.2 0.3 0.4 05 as 0.7 0.0
Central Bank Independence
L10 -= 0.420 - 1.914 081
l-2.94l°
l=12 =o.46 ' significance 0.02 level
<Minimum Value of Liquidity effects>
'08 I T I T T T
0.1 0.2 0.3 0.4 0.5 0.0 0.7 00
Central Bank Independence
LIQ = 0.009 - 0.696 CBI
I-2.401 '

lit2 =O.37 ' significance 0.04 level

Figure 6. CBI and Liquidity Effects (M1, Eq.(2.1))

62

The average liquidity effect is also significantly affected by CBI. For all three sample
periods, we find an inverse relationship between CBI and the average liquidity effect
although the relationship is not significant for 1970—1979 period. These regression
results, therefore, suggest that the overall response of interest rates to monetary shock
would be deeper and greater in high CBI countries than in low CBI countries and
confirms our expectation at the beginning of the paper about the response path of interest
rates to a monetary shock.

We can draw the same conclusion from other traditional methods. Allowing money
growth to evolve according to its own past does not appreciably alter the characterization
drawn by the cumulative coefficients in the previous approach. In high CBI countries,
monetary innovations have a negative contemporaneous correlation with the interest rate
for all periods. Figures 7 shows response paths of interest rates given monetary shock at
time 0 for some countries during 1980-1994 period. In the figure, the point estimate
shows that the interest rate declines at impact and stays below its initial level for about
two years in high CBI countries like United States or Netherlands. However, in low CBI
countries, monetary innovations seem to have even a positive contemporaneous
correlation with the interest rate and interest rate maintains almost the initial level despite
of a monetary shock given at the initial time. This confirms that high CBI countries have
a stronger and more persistent liquidity effect than low CBI countries. To investigate the
relationship between CBI and the liquidity effect, we pick the values of contemporary and
minimum liquidity effect for each country from the values of correlation between
monetary innovation and the interest rate and estimate the relationship between these

variables and CBI. We also calculate average value of responses in interest rate. The

Responudfitomchange

Response at R, to M1 Change

Responuolmomcnmoe

63

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

— U.S.
..... “R
— - NET
— NOR
-— SM
-- UK

 

 

 

 

‘06 Y vvvvvvvvvvv v y . r V , ff
0 5 10 15 20 25 30 35
0.4
" / /\ / .' VF /\/‘
0.0 thvAh \/I\ 4/ \_ 1M
_-' .I - \J’
-0.2-4
04-
-0.6 a . e , f
0 5 10 15 20 25 30 35
0.6
0.4-1
0.2-1 ‘.
.’\_ '. / .3
/‘ "-/‘1 . - .C’S
. A. v, \. . \ X,./7 V ,
0'0“ V '. ' -. ' . - I \V/ .. ' "
\ 3"‘\ ' t \f"/"‘/
/ x/ J
-0.2- /
' /
. \
044 .' \4/
‘0-6 l l r r 1 Y r r W—r ' r 1 2 fit V 1
O 5 10 15 20 25 30 35
Time(Month)

Figure 7. Response of R to M1 Change (1980~1994, Eq.(2.3))

 

 

64

regression results confirm our previous result which shows inverse relationship between
liquidity effects and CBI. The estimation results are reported in table 13 and Figure 8
depicts these relationships for some sample periods.

The relationships between CBI and liquidity effects which are derived by monetary
innovation equation (2.3) satisfy our expectations. All but one sign of coefficients are the
same as we expected and the results coincide with the previous one in which we used
equation (2.1) to derive liquidity effects. Both the contemporary liquidity effect and the
minimum value of liquidity effect are significantly affected by CBI for all 3 different

sample periods, but the relationship between CBI and the average liquidity effect is

significant only for 1970-1994 period.

Table 13. CBI and Liquidity Effects (Eq.2.3)

 

 

Dependent Explanatory Variables 1970-1979 1980-1994 1970-1994
Variables
Contemporary Intercept 0.049 0.282 0.281
Liquidity Effect CBI -1.479** 0846"“ -1.174****
(-2.61) (-3.03) (-4.48)
R2 0.49 0.48 0.67
Minimum Value Intercept -0.176 0.063 0.050
of Liquidity Effect CBI -1.132* -0.712*** -0.894***
(-1.82) (-2.69) (-2.84)
R2 0.32 0.42 0.45
Average Intercept -0.066 0.198 0.044
Liquidity Effect CBI 0.512 -0.467 -0.262***
(0.56) (-l.68) (-2.51)
R2 0.04 0.22 0.39

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent

level, ** at 5 percent level, 1'" at 3 percent level, and **** at 1 percent level.

1. 1980 -1994

< Contemporary Liquidity Effects >

65

< Minimum Value of Liquidity effects >

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 3 01
e
0.2 1 . op .
0'1 I -o.1
0.0 1 I
-0.2
01
.03 4
-0 2
'04 1
0.3
e
04 '°‘5
'0-5 Y Y 1 r v r ‘06 T r v r r f
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0 7 o e
CBI CBI
L10 '3 0.293 - 1.004 CBI L10 8 0.067 - 0.711 CBI
(-3.52)° (.201)’
R2 I 0.55 ' significance 0.01 level R2 8 0.44 ' signifiemee 0.02 level
< Contemporary Liquidity Effects > < Minimum Value 61 Liquidity effects >
0.2 0.0
O
01
0.0 -
-0.2
-0.2 <
-o 3
-0.4 04
.05
00 .
-0.0
e
05 r r 1 1 v T .o_7 r r y ' r r
0" °~2 0-3 0-4 05 ° 6 0 7 0-8 0.1 0.2 0.3 0.4 0.5 0.0 0.7 0.0
C81 C81

L10 3 0.208 - 1.008 CBI
(-2.95)'
2

R 8 0.47 ’ Significance 0.02 level

Figure 8.

L10 = 0.076 - 0941 CBI
(-3.16)'
2

R = 0.50 ' significance 0.01 level

CBI and Liquidity Effects (M1, Eq.(2.3))

ii

iii
gil

C11

1101

We

iii;

and

66

Finally, Using the dynamic correlation method to estimate the liquidity effect still does
not change our conclusion about the relationship between CBI and the liquidity effect.
From Figure 9, we can find easily that liquidity effects persist up to 3 years in Germany
and more than 2 years in other high CBI countries. In low CBI countries, however, the
figure shows that the persistence of liquidity effects is very weak.

To examine the effect of CBI on the liquidity effect, we use the same kind of values as
we did in the previous traditional methods. We pick up the contemporary and minimum
values of dynamic correlation and also calculate the average value of the liquidity effect
given time span. In addition, we test the persistency of liquidity effects by using k in
Christiano and Eichenbaum’s argument which is the number of months which show
liquidity effects, in dynamic correlation method. We regress CBI on these values and
estimate the relationship. Figure 10 depicts the relationship for the sample period of
1980-1994 and Table 14 reports the regression result.

When we use the dynamic correlation method to estimate the liquidity effect, we do
not find any significant result for 1970-1979 period, though all coefficients have the signs
we expected. On the other hand, the estimated results for 1980-1994 period are all
significant at 1 percent level. CBI has especially strong positive effect on the persistence
of the liquidity effect at 1 percent significance level during the sample period. The result
suggests improving CBI by 0.1 would increase the liquidity effects by 7 months.

In summary, the estimation results support our hypothesis that there exists an inverse
relationship between CBI and the liquidity effect no matter which traditional method we
use. In high CBI countries, responses of interest rates given monetary shocks are stronger

and deeper than those of low CBI countries. Interest rates are more sensitive to monetary

67

innovation and the persistence of liquidity effects is stronger in high CBI countries than in
low CBI countries. These results, therefore, imply monetary policy effects on economy

may be greater in high CBI countries than low CBI countries.

Correlation

Correlation

Correlation

68

 

 

—- US (0.48)
------ GER(O.69)
-- - NET(O.42)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. ’/
\\\ ’4'—/—/
.1.0 V t V v V v W v—r V v r v VVVVVVVVVV f r v ‘7
-24 -20 -16 -12 -6 0 4 6 12 16 20 24
0.6
0.4 _ -—- NOR(0.17)
...... SM(°.29)
02 - —— U.K(0.27)
0.0
02-
-0.4 d
-0.6 -<
-0.6-i
-1° vvvvvv 'Vv 'vvr**YVVVVv vvvvv 'VVV'VT' VVVVVVY—Vfi—"V
-24 -20 -16 -12 -6 -4 0 4 8 12 16 20 24
0.6
,_ — us
0.4 —l or \‘- """ GER
f—I. \. __ NET
.-/" ’\. / __ T'OR
0.2 "‘ A ' ./' ._. _ SM
' — U.K

 

 

 

 

 

 

 

 

va‘Y‘IVVVTIYTYT—rYTV‘f'vIVI'1

-24 -20 -16 -12 -8 -4 0 4 8 12 16 20 24

Time (Month)

Figure 9. Dynamic Correlations b/w R, and M1,., (1980~l994)

< Contemporary Correlation >

69

< Minimum value of Correlation >

 

 

 

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 e 0 e T
l
0.4 « . 0.4 l
I, a
0.2 < 02 ‘
0.0 1 00
02 - -02 4
-04 1 -O,4 1
-0.0 < 00 1
0.0 « -0.e -
1.0 Y fir V 7* V I '10 Y r v 1 1 7
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.6 0 1 0,2 o_3 0.4 0,5 0,5 0,7 on
081 CBI
LIQ = 0.504 — 2.022 car “0 2 0135‘ -1592 93'
(3.13). 2 (4.09)
R2 = 0.79 e (affirm 0001 Incl R 3 063 ' 6190M 0.002 level
< Average Liquidity effect > < Number of Months which shows quuidrty effect >
0.6 1
0 3
0.2 1
0,1 4
0.0 -
m l
-0 2
-0.3 -
0‘ r r r 1 T f o 1 J‘: , y y r fi
01 0 2 0.3 0.4 0.5 0.6 0.7 0.8 0.1 0.2 0.3 0.4 0.5 0.0 0.7 o e
CBI CBI
LIO - 0.401 - 1-096 08' DLIQ - 41.111 + 78.029 081
2 (4.05)‘ (3.64)‘
R = 0-38 ' “Mm 090‘ 10V“ R2 = 0.57 ‘ signitlmnce 0.005 level

Figure 10. CBI and Liquidity Effects (Ml, l980~1994: Dynamic Corr.)

Table 14. CBI and Liquidity Effects (Dynamic correlation)

 

 

Dependent Explanatory Variables 1970-1979 1980-1994 1970-1994
Variables
Contemporary Intercept 0. 1 54 0.594 0.33 5
Liquidity Effect CBI -0.992 -2.022** -1.358**
(-1.71) (-6.13) (-3.73)
R2 0.30 0.79 0.58
Minimum Value Intercept -0.229 0.354 0.036
of Liquidity Effect CBI -0.469 -1.592** -0.846*
(-1.13) (-4.09) (-2.18)
1?.2 0.15 0.63 0.32
Average Intercept 0.01 8 0.401 0. 1 29
Liquidity Effect CBI -0.111 -1.096** -0.309
(-0.63) (-4.65) (-1.72)
R2 0.05 0.68 0.23
Persistency of Intercept 12.960 -6.111 18.519
Liquidity Effect CBI 14.310 78.029" 37.447
(1.09) (3.64) (0.98)
R2 0.15 0.57 0.09

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 5 percent
level, ** at 1 percent level.

71

2-5. Estimation with the Alternative Money Aggregates

In previous sections, to estimate the liquidity effects, we investigated the relationship
between a money aggregate, M1, and the market interest rates. Although the estimation
results show the inverse relationship between these liquidity effects and CBI, since there
are alternative money aggregates which can be used to estimate the liquidity effects, there
is a possibility that the liquidity effects from these alternative money aggregates have a
different relationship with the CBI.

To investigate this possibility, we estimate the liquidity effects with other money
aggregate, total reserves(TR) and monetary base(MB), by applying three different
traditional approaches and then examine the relationship between these liquidity effects
and CBI. We use IFS data series number 14 for TR and number 20 for MB.27 We
estimate the liquidity effects only for the periods of 1980-1994 because of data
availability ad the number of sample countries are 10 for TR liquidity effect and 11 for
MB liquidity effect.28

Using TR to estimate liquidity effect does not change our previous results. We do not
find any significant difference in the relationship between CBI and TR liquidity effect
from the relationship between CBI and M1 liquidity effect. Although the significance

level is a little weaker, the TR liquidity effect has a negative correlation with CBI. Every

 

27 IFS data series number 14 is the reserve money of monetary authority and number 20 comprises domestic
currency holdings and deposits with the monetary authorities. These are somewhat different from exact
amount of TR and MB in each country. We use them because they are calculated under the same criteria as
well as they are available.

2’ We exclude Norway and U.K. for TR liquidity effect and U.K. for MB liquidity effect because data are
non-available in these countries.

72

relationship meets our expectation no matter which methods are used to estimate the
liquidity effect. Table 15 reports the estimation results with TR.

While the liquidity effect with TR has some significant relationships with CBI, the
liquidity effect with MB does not have a significant relationship with CBI though all of
the estimated signs of the coefficients are as we expected. However, we should not
interpret this result as the evidence that makes our previous results less important. Since
we use IFS data to get MB for each country because of data availability and this data is
not the correct MB for every country because of difference in calculating MB, this
estimation result might be less meaningful than others.

Although the significance of the inverse relationship between these liquidity effects
and CBI are a little weaker than those when we use M1 for the estimation of liquidity
effect, all estimated signs are exactly the same as we expected. Therefore, since generally
it seems that the estimation results with alternative money aggregates support our
previous results and M1 is the only money aggregate data which are available for whole
our sample periods, we will concentrate only on M1 for the estimation of liquidity effect

in next sections.

73

Table 15. CBI and Liquidity Effects (TR)

 

 

Dependent Explanatory Variables EQ. 2.1 EQ. 2.3 Dynamic
Variables Correlation
Contemporary Intercept 0.053 -0.01 0 0.4 1 8
Liquidity Effect CBI -0.141 -0.014 -1.026***
(-1.13) (-0.10) (-2.46)
R2 0.14 0.00 0.43
Minimum Value Intercept 0.095 -0.027 0.118
of Liquidity Effect CBI -0.466* -0.014 -0.681
(-1.89) (-0.11) (-1.72)
R2 0.31 0.00 0.27
Average Intercept 0.197 0.052 0.31 8
Liquidity Effect CBI -0.526*" -0.089 -0.709**
(-2. l 8) (-0.98) (-2.22)
R2 0.37 0.11 0.38
Persistency of Intercept - - -6.329
Liquidity Effect CBI - - 79.87 ***
- - (2.60)
R2 - - 0.46

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent
level, 1”" at 6 percent level, *** at 4 percent level.

74

Table 16. CBI and Liquidity Effects (MB)

 

 

Dependent Explanatory Variables EQ. 2.1 EQ. 2.3 Dynamic
Variables Correlation
Contemporary Intercept 0.073 -0.009 0.332
Liquidity Effect CBI -0.272 -0.081 -1.017
(-0.91) (-0.30) (-1.64)
R2 0.09 0.01 0.23
Minimum Value Intercept 0.159 -0.054 0.114
of Liquidity Effect CBI -0.907 -0.091 -0.804
(-1.35) (-0.51) (-1.73)
R2 0.17 0.03 0.25
Average Intercept 0.362 0.085 0.238
Liquidity Effect CBI -0.975 -0.106 -0.620
(-1.41) (-0.55) (-1.32)
R2 0.13 0.03 0.16
Persistency of Intercept - - 5.618
Liquidity Effect CBI - - 51.38
- - (1.70)
R2 - - 0.24

 

Note : The t-statistics are reported in parentheses.

75

3. Seemingly Unrelated Regression Approaches

3-1. SURE(Seemingly Unrelated Regression Estimation)

The previous section derived the liquidity effect by applying the same equation
country-by-country, since we assume that the economic variables for each country are
independent of those in the other countries. These equations, however, might be
connected not because they interact, but because their error terms are related. For
example, a shock affecting the interest rate for one country may spill over and affect the
interest rate for other countries. In this case, estimating these equations as a set, using a
single (large) regression, should improve efficiency and better reflect the real world.
Actually, in the real world, there are many factors which affect the economies of all
countries simultaneously. As a simple example, a change in world interest rates would
affect the level of domestic interest rates in all countries, or supply side factors like an oil
shock would spill over and change the price level of many countries at the same time.
We, therefore, apply the SURE (Seemingly Unrelated Regression Estimation) to the
previous equations and reestimate liquidity effects.

SURE consists of writing a set of individual equations as one giant equation. Suppose
there are N equations Y, = X, B, + a, , where the subscript i refers to the ith equation.

These equations are written as

q l— —I- -1 l— q

1'

Y1
Y2 X2 162 52

 

 

 

 

 

 

 

 

76

or Y* = X* [3* + 3*.

If we allow contemporaneous correlation between the error terms across equations, so
that, for example, the tth error term in the ith equation is correlated with the tth error term
in the jth equation, the variance-covariance matrix of 8* will not be diagonal. Estimating
these error correlations and the diagonal elements(by using the residuals from each
equation estimated separately) should allow estimation of the variance-covariance matrix
of 8* and generation of GLS estimated of 0*. Correlation of residuals are reported in
Table 17. The values above the diagonal are correlations of the residuals from SURE and
those below diagonal are the correlations of the residuals from OLS estimation for the
period of 1980-1989.

From the table, for example, we observe that the residuals in equations of US and
Germany or those of US and Canada are highly correlated and move together to the same
direction. These high correlations raise the possibility that the variance-covariance
matrix of 8* is not diagonal. Breusch and Pagan (1980) suggest the Lagrange multiplier
statistic to test whether the variance-covariance matrix of 8* is diagonal.29 Based on the
OLS results, the Lagrange multiplier statistic is 145.79, with 66 degrees of freedom. The
1 percent critical value is about 100, so the hypothesis that the variance-covariance matrix
of 8* is diagonal can be rejected. Therefore, the estimation by SUR will give us more

efficiency than the OLS estimation of the previous section.

 

29 Following Breusch and Pagan (1980), we estimate the correlation coefficient between the i"I and j‘h
residuals from OLS. The sample size times the sum of all these squared estimated correlations is distributed
as a chi-square with degrees of freedom equal to the number of correlations.

77

Table 17. Residual Correlation Matrix (1980-1989)

 

 

U. S GER AUS CAN DEN FRA F IN ITA NET NOR SWE U.K

US 1 0.608 -0.586 0.690 0.230 0.630 -0.269 0.448 0.367 -0. l 66 0.363 0.018
GER 0.374 I -0.633 0.333 0.132 0.666 -0.558 0.258 0.331 -0.104 0.002 0.166
AUS -0.479 -0.491 1 -0.107 -0.181 -0.655 0.481 -0.145 -0.350 0.435 0.019 0.002
CAN 0.542 0.079 0.047 1 0.218 0.338 -0.21 1 0.244 0.248 -0.012 0.404 0.138
DEN 0.088 0.043 -0.110 0.136 1 0.447 -0.l6l 0.217 0.054 -0.309 0.119 0.163
FRA 0.413 0.545 -0.520 0.149 0.361 1 -0.527 0.425 0.129 -0.205 0.162 0.039
FIN -O.300 -0.490 0.321 -0.227 -0. 102 -0.479 1 0.315 0.184 0.035 0.362 -0.498
ITA 0.209 0.147 -0. 127 0.026 0.161 0.308 0.320 1 0.113 0.026 0.454 -0.395
NET 0.161 0.177 -O.298 0.118 0.019 0.011 0.204 0.020 1 -O.383 0.258 -0.182
NOR -0.047 -0.029 0.337 0.065 -0.264 -0.120 0.028 0.068 -0.292 I -0. 156 0.034
SWE 0.160 -0.l71 0.064 0.313 0.098 0.055 0.350 0.350 0.172 -0.094 1 ~0.272
U.K 0.136 0.154 0.049 0.200 0.178 0.155 -O.43l -0.262 -0. 194 0.032 -0.207 1

 

3-2. Estimation Results

We estimate the liquidity effect for 4 sample periods which are 1973-89, 1973-94,
1980-89, and 1980-94. Since SURE methods restricts the sample period to the common
available period for all sample countries, our sample period for all countries can be
extended only to December, 1989 since we are considering German data up to
reunification. However, we estimate the equations for the period until 1994 with only the
countries for which data are available by excluding Sweden (for which data can be

extended only to 1990) and Germany. And we also exclude Canada and Finland for the

78

estimation for the periods which begin in 1973, since available data for these countries
only goes back to the mid 1970s.

Figure 11 depicts the response of interest rates to the change of M1 for the period of
1980-1989 applying SURE to equation (2.1). In high CBI countries, liquidity effects are
quite strong and persistent while in low CBI countries liquidity effects are much weaker
than those of high CBI countries. Therefore, applying SURE for equation (2.1) does not
change our previous results in any of the sample periods. Table 18 reports the regression
results and Figure 12 depicts the estimation results clearly for the period of 1980-1989.

In all sample periods, the estimated results show that there is a strong inverse
relationship, not only between CBI and the contemporary liquidity effect, but also
between CBI and the rrrinirnum value of the liquidity effect. Also, it verifies the inverse
relationship between CBI and the average liquidity effect. Finally, we count the number
of months which show liquidity effects for a given time span and use this number as the
value which represents duration of liquidity effects for the comparison with the results
from using k values in dynamic correlation method that suggest a positive relationship
between CBI and the duration of the liquidity effects. The estimation results confirm that
CBI has strong positive effects on the duration of the liquidity effect. It seems that the
liquidity effect would last about 5 months longer if we increased the CBI measure by 0.1

point.

79

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.6
0.4 -1 — GER(0.69)
------ DEN(0.50)
0.2 1 — — NET(0.42)
0.0
02 4 I
\ ’- \
-0.3 T \ .... ’ .— / \ ‘‘‘‘
\
1.0 fii Y V\
0 4 8 12 16 20 24
0.6
— SWE(0,29)
0.4 4 ------ U.K(0.27)
— —- NOR(0.17)
0.2 -I
00’*\ “““““““““ \4”’ ~~~~~~ ->”“§;-:;L,
-0.2 —
~04 -
-0.6 1
-0.8 ~
.10 e . - 3 . - -
0 4 6 12 16 20 24
0.6
— GER
..... DEN
04 -r — - NET
— SM
— UK
—— NOR

 

 

 

 

 

 

 

 

Time (Month)

Figure 11. Response of R to M1 Change (1980~1989 : SURE Eq.(2.1))

< Contemporary Liquidity Effect >

80

< Minimum Value of Liquidity Effect >

 

 

 

 

 

 

 

 

 

 

0.5 02 f
° I e

04 < 0.0 '*

02 * -02 <

0° ‘ -04 «

02 ‘05 1

‘0‘ " _oa 4

.06 Y Y Y Y Y Y 10 v Y 1 . F r

0.1 012 0 3 0.4 05 0‘6 0‘7 03 0'1 0'2 03 0‘ 05 06 07 0,5

CBI

LIO I 0.317 -1.173CB|
(~3.57)‘
2

R - 0.56 ' significance 0.01 level

CBI

LIQ = 0.247 - 1.57OCBI
(-2.96)'

R2 I 0.47 ' eignlficatce 0.02 level

 

 

 

 

 

 

 

 

 

 

< Average Liquidity Effect > < Number of Months which shows Liquidity Effect >
are 35
0e 0 so «
0‘ ‘ 5 ‘1
0.2 <
20 4
oo «
15 ~
43 2 <
10 «
44
5 4
as
O ' .
'0‘ a V f f r v o L r A. Y T v r r _ ' ”1
01 02 03 04 0.5 06 07 oe 0‘ 03 °-3 °-‘ °5 0° °7 0‘
CBI CBI

LIQ = 0.493 - 1.517CBI
(-2.79)’
2

R I 0.44 ' significance 0.02 level

LIQ = -2332 + 48.40408l
(390).
R2 = 0.47 - significance 0.02 level

Figure 12. CBI and Liquidity Effects (1980~1989 : SURE Eq.(2.1))

81

Table 18. CBI and Liquidity Effects (SURE Eq.(2. 1))

 

 

Dependent Explanatory ‘73-‘89 ‘73-‘94 ‘80-‘89 ‘80-‘94
Variables Variables
Contemporary Intercept 0. l 57 0.362 0.377 0.365
Liquidity Effect CBI -O.862**** 4.636" -l.173**** -1.276****
(-4.58) (-2.61) (-3.57) (-4.95)
R2 0.72 0.53 0.56 0.75
Minimum Value Intercept 0.107 0.266 0.247 0.317
of Liquidity Eff. CBI -1.041*** -1.602*"‘* -1.570*** -1.636*
(-3.23) (-2.91) (-2.96) (-1.92)
R2 0.57 0.59 0.47 0.32
Average Intercept 0.156 0.155 0.493 0.347
Liquidity Effect CBI -O.727* -O.752* -1.517*** -1.250*
(-1.86) (-2.06) (-2.79) (-1.95)
R2 0.30 0.41 0.44 0.32
Duration of Intercept 1.449 -2.482 -2.832 -7.949
Liquidity Effect CBI 37.196“ 54.595*** 48.404*** 64.470"*
(2.05) (2.88) (3.00) (2.88)
R2 0.34 0.58 0.47 0.51

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent
level, *"‘ at 5 percent level, *" at 3 percent level, and **** at 1 percent level.

82

With equation (2.3), SURE also confirms strong relationship between CBI and the
liquidity effect. Figure 13 depicts the response of interest rate to the monetary shock by
applying SURE to the equation (2.3). In Figure 13, we can find the strong and persistent
liquidity effects in high CBI countries and much weaker liquidity effects in low CBI
countries. To investigate the relationship between liquidity effect and central bank
independence, we use the same kind of values as we did in the previous sections. Then,
we regress the measure of central bank independence on these values. The regression
results are reported in Table 19 and Figure 14 depicts the estimation results for some
sample countries for the period of 1980-1989.

From the regression, we find that there is a significant inverse relationship between
CBI and the contemporary liquidity effect for all 4 sample periods. For the periods of
1973-1989 and 1980-1989, CBI is also significantly inversely related to the minimum
value of the liquidity effect, while for the periods of 1973-1994 and 1980-1994 there exist
inverse relationships that are significant level at 11 percent and 13 percent respectively.
CBI also has significant inverse relationship with the average liquidity effect and positive
relationship with the duration of liquidity effect for the period of 1980-1989. For the
period of 1980-1994 the estimation also shows a significant relationship between CBI and
the duration of the liquidity effect. For the other sample periods, the estimated results
have the signs we expected in the relationship between CBI and average liquidity effect
and the relationship between CBI and duration of liquidity effect. However, these

expected results are not significant.

83

 

 

— GER

02 —1 f/\ ...... U-S
— — NET
. /\-/\ v A ,

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

06-1
-o.a 4 - - - . fl . - 4+ ......
o 4 e 12 16 20 24 20 32 36
0.4
—SWE
“‘ /\ /\ / 1'33):
\ ’\ /.. _ '
00 \\r/ \ (mm?f\/\:\ / X J-LA - "
J \ \VJ'\~’ w \ / \1
412- V
04—
00-
4.8 VVVVVVVVVVVVVVVVVVV ViT V r1 ' r ' ' v """""
o 4 e 12 1e 20 24 20 32 30
0.4
-—GER
.'/\ ----- us
'- ——NET
0.2-1 / \\ /\ _ SM
/\ .' ‘ —-—U.K
A A“ ‘M A .d. /‘ _ /Q — NOR
°'°' "’ - 1 ' ‘ \/\
./ .. / J \- /
\-.d. C" "I \ .. ..U
/ -. _- .
.02 4 \f _. . \./ \ \th .'
\ ‘ ' e I .0
-0.4-1
-0.e-'
.03 ....,.f-.r-...,-..4,++.4,....,-fi.-,....,....
o 4 a 12 1e 20 24 20 32 30

Time (Month)

Figure 13. Response of R to M1 Change (1980~1989 : SURE Eq.(2.3))

< Contemporary Liquidity Effect >

 

0.4

0.2 <

 

 

 

 

.1 0.2 0.3 0.4 0.5 0.6 0.7 08

C81

LIQ 8 0.292 - 0.986 CBI

(- .
R’ - 0.24 'sioniflcance 0.10 level

< Average Liquidity Effect >

 

0.0

0.6 1

0.4 .1

0.2 1

001

 

 

 

 

CBI

LIQ = 0.703 - 0.001 CBI
(-1 051'
R’ - 0.2a 'eignif'loance 0.08 level

84

< Minimum Value of Liquidity Effect >

 

 

 

 

 

 

 

 

 

 

0.2
0.0 <
O
0.2
.04
-0.6 .
i
l
’08 L Y V V T Y Y
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
CBI
LlQ . 0.060 - 0.516 CBI
(-1.93)°
R’ - 0.27 ‘eignificance 0.08 level
< Number of Months which shows Liquidity Effect >
40
35
30 <
25 4
20
15 1
I
10
O
5 . e
0 I Y 1 Y Y 1 Y
0.1 o 2 0.3 o 4 0.5 0.0 0.7 0.8

CBI

LIQ = 2.311 + 49.153 CBI
(2.23)‘
R’ -- 0.33 'eigniflcance 0.05 level

Figure 14. CBI and Liquidity Effects (1980~1989 : SURE Eq.(2.3))

Table 19. CBI and Liquidity Effects (SURE Eq.2.3)

85

 

 

Dependent Explanatory ‘73-‘89 ‘73-‘94 ‘80-‘89 ‘80-‘94
Variables Variables
Contemporary Intercept 0.143 0.388 0.292 0.472
Liquidity Effect CBI -O.899**** -1.839** -0.986* -1.707****
(-3.84) (-2.50) (-1 .80) (-3.34)
R2 0.65 0.51 0.24 0.58
Minimum Value Intercept 0.059 0.173 0.066 0.097
of Liquidity Eff. CBI -0.931*** -1.449* -O.516* -0.998
(.327) (-1.88) (-1.93) (-1.70)
R2 0.57 0.37 0.27 0.27
Average Intercept 0.060 0.024 0.363 0.1 14
Liquidity Effect CBI -0.186 -0.157 -O.881* -0.357
(-0.84) (-0.79) (-1.95) (-1.53)
R2 0.08 0.09 0.28 0.23
Duration of Intercept 8.983 10.029 -2.311 8.198
Liquidity Effect CBI 24.024 24.449 49.153M 41 .820*
(1.19) (0.97) (2.23) (1.85)
R2 0.15 0.14 0.33 0.30

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent

level, *“‘ at 5 percent level, "* at 3 percent level, and **** at 1 percent level.

86

4. Summary

In this chapter, we estimated the liquidity effect by using three different traditional
approaches and then we regress these liquidity effects on CBI measurement. The
estimation results confirmed that the liquidity effects are inversely related with CBI. We
found that the liquidity effects are much stronger and more persistent in high CBI
countries than in low CBI countries. These results did not change when we applied
SURE to the traditional approaches. These estimation results imply that CBI has
significant effects not only on inflation level itself but also on monetary policy effects, so
that credibility bonus effect exists.

However, as we pointed out at the beginning of this chapter, the traditional approaches
do not correctly identify the liquidity effects. Therefore, in next chapter, we seek to
identify the liquidity effect more clearly and correctly by using vector autoregression
approaches and then we will investigate the relationship between these liquidity effects

and CBI again.

CHAPTER IV

VECTOR AUTOREGRESSION APPROACHES

1. Nonstructural VAR Approaches

1-1. Vector Autoregression (VAR)

The traditional approaches suggest that there are inverse relationships between CBI
and liquidity effects. The liquidity effects estimated by the traditional methods, however,
can not be taken as evidence that unanticipated expansionary monetary policy
disturbances drive interest rates down, because the traditional empirical approaches
implicitly assume that no other variables induce interest rates and money aggregates to
move together to generate the correlations estimated in these regressions. Therefore, the
estimated relationship between CBI and liquidity effect may not be the real relationship
because of problems in estimating liquidity effects by using the traditional approaches.
At a minimum, providing such evidence requires identifying assumptions that are
sufficiently strong to isolate a measure of the monetary policy disturbance. As it turns out
from many studies, inference regarding the effects of monetary policy on interest rates
hinges critically on two factors; the identifying assumptions used to obtain measures of
unanticipated shocks to monetary policy, and the measure of money used in the analysis.

There are many different ways to solve the identification problem. These include

event analysis (Romer and Romer, 1989, 1990), nonstructural VARs which are not

87

88

designed to be invariant to policy regime changes (Strongin, 1995), structural VARs
which are designed to be invariant to policy regime changes (Leeper & Gordon, 1994),
traditional general equilibrium models with detailed financial sectors (Gilles, Coleman, &
Labadie, 1993), and real business cycle models with an appended monetary sector
(Christiano, 1991)

With regard to the measure of money, some studies use a monetary aggregate as the
measure of exogenous policy disturbances (M-rule). Among the monetary aggregates
used we find non-borrowed reserves (NBR), MO, and M1. When monetary aggregates
are used to measure exogenous policy disturbances, a liquidity puzzle problem arises
consistently which is that irmovations in the monetary aggregates seem to be associated
with rising, rather than falling interest rates. This puzzle has been surveyed by
Reichenstein (1987) and was recently redocumented by Leeper and Gordon (1992).
Leeper and Gordon (1992) show that the relationship between innovations in the
monetary aggregate and interest rates is highly uncertain, varies across time, and usually
has the opposite sign to the theoretical prediction. Christiano and Eichenbaum (1992)
also found no liquidity effect with M0 or M1, but with NBR they found a persistent
liquidity effect. However, we need to note that all these studies have been done only on
US data.

The liquidity puzzle with a money based measure for monetary policy led Sims (1992)
and Bemanke and Blinder (1992), among others, to identify monetary policy directly with
innovations in interest rates (R-rule). This identification scheme is relatively successful
in producing results consistent with a priori expectations about the effects of monetary

policy. However, according to Strongin (1995), this approach also has a problem. He

89

argues that, without any demonstrated empirical linkage between central bank actions and
interest rate movements, it is unclear how innovation in interest rates can reasonably be
attributed to monetary policy.

In this paper, we will apply R-rule to each country since with M-rule we could find a
liquidity effect only in a few European countries among all eleven sample countries.30
This is consistent with other studies which found no liquidity effects with M-rule (M1) in
US. Therefore, in the next section, we estimate the relationship between CBI and the
liquidity effect only with R-rule. We use M1 for the monetary aggregate because it is
available for all the countries we examine here.

To clarify the nature of the identifying assumptions that have been used in this paper,
suppose that the economy evolves according to Ayt = B(L)yt + (it. Here, yt denotes the
time t values of the variables summarizing the state of the economic system. yt includes
the time t values of the observable endogenous nonpolicy variables (ylt) as well as the
time t values of the policy instruments (yzt). The fimdamental sources of uncertainty in
this economy are summarized by the i.i.d. random variable 8:, which has the property that

Estet’= I where I denotes the identity matrix. The vector 8: is partitioned as 8i = [8n 82t]’.

With this notation, 821 represents the fundamental disturbances to policy variable (ya).
The constant matrix A summarizes the manner in which the contemporaneous values of
y, are related to each other, while B(L) is a matrix polynomial in positive powers of the

lag operator L.

 

30 In VAR approaches, we exclude Australia from the previous 12 countries because of non-availability in
data

90

Now suppose we are interested in examining the historical effects of policy
disturbances, that is, we want to characterize the dynamic effects of past variations in yzi ,
arising from different values of 821, on y“. Given values for A and B(L), these responses

can be calculated from the moving average representation of the system

yt = C(L) e, = E; C,c,_, where C(L) = A"[ I - B(L)]‘l.

Under our assumptions, the (k, j) element of Cs gives the response of the kth element of
y”, to a unit disturbance in the jth element of 8!,

Suppose we are modeling a four variable VAR system. For example, if we define that
the first element of y! as the price level , the second element as the income level, the third

element as the interest rate level and the fourth element as the level of the money

aggregate. A pth-order vector autoregression, denoted VAR(p), can be written as

yt =c+d>1yt-1 +032 yt-2 +... +<I>p yt.p +et. (4.1)

where yt =[Pt Yt Rt Mt ]’
31:181t €2t 83t 84t 1’

Here 0 denotes an (4 x 1) vector of constant and (I) an (4 x 4) matrix of autoregressive
coefficients for j = l, 2, ..., p. The (4 x 1) vector 8 is a vector generalization of white
noise:

B(St) = 0 (4.2)

E(8t 31’) = Q for t = r

0 otherwise,

with Q an (n x n) symmetric positive definite matrix.

91

Using lag operator, (4.1) can be written in the form
[1.0),L-<1>2L2-...-<1>,L"]y.=c+et (4.3)
or <D(L)y, = C + 8t,

where <D(L) indicates an (4 x 4) polynomial matrix in the lag operator L.

Now, VAR(p) can be written in vector MA(oo) form as

y = k + at + ‘1’] 3H + ‘1’; at; + (4.4)

or yt = k + ‘1’(L)et

where ‘P(L) = [<D(L)]" and k = [<1>(1)]"c

Thus, the matrix ‘P has the interpretation

6y... / as; = ‘1’, ; (4.5)
that is, the row i, column j element of ‘1’, identifies the consequences of a one-unit
increase in the jth variable’s innovation as date t (8,1) for the value of 1th variable at time
t+s (yius), holding all other innovations at all dates constant.

If we were told that the first element of 81(8):) changed by 5;: at the same time that the

second element changed by 6y, the third element changed by ER, and the fourth element

by 5M, then the combined effect of these changes on the value of vector yr+s would be

given by
Ayers = (6yr+s/681()8p+(6yt+s/082t)5y+(6yr+J6831)53+(awn/5840514 = 11156, (4.6)
where 8 = [ 8p, 5y, 5R, 5M]’.

Now set yH = yt_2 = . . . = yt_p = 0, an = l, and all other elements of at to zero, and

simulate the system (4.1) for dates t, t +1, t +2, . . . , with c and 8m , 8r+2 , . . . all zero.

92

The value of the vector yt+s at date t + s of this simulation corresponds to the jth column of
1115 . By doing a separate simulation for impulses to each of the innovations (i = 1,2,3,4),
all of the columns of 111, can be calculated.

A plot of the row i, column j element of 111, ,

6 yms / 6 Sjt, (4.7)
as a function of s is called the impulse response function. It describes the response of yim
to a one-time impulse in yjt with all other variables dated t or earlier held constant.

Here, we derive the impulse response fimction of the variables to the shock of each
variable to get the liquidity effects in each country. For example, Figure 15 and Figure 16
depict these impulse response functions for a 4 variable VAR system of US in (R,M,Y,P)
order and for a 5 variable VAR system in (Y,P,CP,R,M) order respectively for the period
of 1980-1994.“ The dotted lines represent the estimated confidence bands constructed
from Monte Carlo integration method. In the Figure 15, for example, the first graph in
the second column represents the liquidity effects since the impulse response of Mt to the
shock of K depicts the liquidity effect in the R-rule. By applying VAR to each country,
we get the impulse responses which indicate the liquidity effects from all sample
countries. Figure 17 depicts the impulse response functions of all countries in the sample

which represent the liquidity effects in a 4 variable VAR (R,M,Y,P) order.

 

3 ' In 5 variable VAR system, we include commodity price index(CP).

93

mm a E on w. 0. w 0 mm on 0N ON 0. 0. m 0 00 8 K On w. 0. w 0 0n 8 0N a w. 0. w 0

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

wed.
8.0
r000
080.0 9.0.0 0.0.0 8.0
mm 8 8 8 m. 0. w 0
IDDIDDIIFFFIF’F’FPPD PF PPPFD g?
\I f-
I
I ‘-
1') I
130
‘ \
\l I II 0 I \ I \
v88.0
0000.0 30.0 0.0.0 8.0
acumen—0‘00 8 £830.00 090090.83800 88880.0.m0
hhhhhh bFDEPPLFhI-ubbh-D-bthbbbbhhh goon hhhhhbbhh-hnInhbpb-b>-th-thhbhhh- KS9 nF-bhhpbD-nhhhhbhhhbhbhhhhhbhbh-hnun §e°l hhPhbbhhhbththPthbPht-hhhhubnbbh 9.0!
0\ cl I \. 1: \ J
l l r t t l
I. 1 i \t 886 \ > >/ \. I 8.0
J \I I \ I \ r\ f 1‘ \1
\ \ \ l \ f 3 \
r 88.0 r .306 r 80.0 r 0‘0
30 30.0 20.0 80
8894.80.ng “swagggwo
’PVDIIDD’P PF’ DDPDDD Eb? 88? P lPPthDDFhDDID 88?
I I
‘ I
, I‘ r I! II II
I}; 101.1 / \I \o .\ I
i 30 ll LY» 950
1 \ I. .r \ \ l ll<il 34
ll ' I \ I I! I \ 0| 0‘ f
I.
188.0 tnSo
0080 0.0.0 0.0.0 8.0

 

 

 

 

 

 

 

 

 

 

 

 

Figure 15. Impulse Responses in (R,M,Y,P) : U.S.

94

gonna—«0.0.0 0

 

 

. 008.0.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

3.33:3:u_...m_::u:. .o

 

‘

   

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

«5.0

20.9

 

 

 

 

 

 

 

 

 

 

 

 

 

 

10.8.0.

13.0

r280

0000.0

1 0.8.0.

an v.90

 

 

flog?

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 16. Impulse Responses in (Y,P,CP,R,M) : U.S.

0000
0001
0002

-O.(X13 ‘

01W

0“ ..

0000
0002
OW

0m'

0005

0035
0030
0025
0020
0015
0010

00010
00005
00000

«OCXHO -

00015
00020

95

 

 

 

 

00025-

U S
J
o 1 a 7 12 1e 20 iii 21 2'. W 32 36
Canada 7‘ _ ___ , ,_

/\/\ f \/\/\\ I)

V \ \/
\er
o 4 o 12 1e 20 24 23 32 36
Denmark
0 4 a 12 111 I 20 21 21) 3.2 “—— Se
Finland
\v
(1 a 0 1277716 20 21 23 32 36
France
/\ / . .
V\ \
o 4 a 12 1e 20 21 211 32 36
Time (Month)

Figure 17.

2D 24

/\
’\ A A _ /\/ \‘i a;
A y \1 1. j “\
\/ W V ‘v 1
4 I 12 20

16

0002
0001
0000
0W1
.0002
OW
0

 

Netherlands
0 003
o 002
o 001
o 000 — —— —
-o 001 1
0002 1

0W4:
0M1-

 

 

 

00010
~0CX)15 - /
00020 .

0.0025 1
00030 A 7 s . ,
o a a 12 16 20 24

Time (Month)

20 32 )6

Liquidity Effects in (R,M,Y,P)

96

1-2. Four Variables VAR Estimation.

We will measure the shock to monetary policy as an orthogonalized component of the
innovation to the short interest rates with the ordering as (R,M,Y,P) suggested by
Christiano and Eichenbaum (1992), which corresponds to the identification scheme
imposed by Bemanke and Blinder (1992) as well as Sims (1986). We examine the
relationship with the alternative ordering (P,Y,R,M) in which the unanticipated change in
monetary policy is measured at the point of the innovation in interest rate that is
orthogonal to the innovations in both Pt and Yt, but not M. We also analyze the
relationship with the alternative ordering (P,R,M,Y), which the unanticipated change in
monetary policy is measured at the point of the innovation in interest rate that is
orthogonal to the innovation in Pt, but not Mt and Yt, This means that in setting K the
monetary authority looks at P, but not Y, and Mt.

With these 3 different orderings, we estimate impulse responses of each variable to the
monetary shock. The impulse response of Mt to the shock of K depicts the liquidity
effect in the R-rule. In the Figure 15, for example, the first graph in the second column
represents the liquidity effects. From this impulse response function, we get the
contemporary liquidity effect and the minimum value of liquidity effect given the time
span. We count the number of months which show liquidity effect given the time span
and consider this number as the duration of the liquidity effect. We also sum the
response of M1 after the R—shock and use this aggregate response as the average liquidity

effect, as we did in the previous sections.

97

The estimated results for the relationship between CBI and liquidity effect seem to be
little affected by choice of orderings for constructing liquidity effect.32 The Figures 18,
19 and 20 shows these relationships. CBI is very significantly related to the minimum
value of liquidity effect in all 3 orderings. The duration of the liquidity effect is
positively related with CBI at the 8% to 10% significance level for all 3 orderings. The
results show that 0.1 point increase in CBI will make the liquidity effect last about 4
months longer. We can get the same relationship as we expected in both the
relationship between CBI and contemporary liquidity effect and the relationship between

CBI and average liquidity effect. But these are significant only at the 11% to 18% level.

 

32 Here, we report the estimation results only for the period of 1980-1994. Actually, there was little
difference in the estimation results between for the period of 1980-1994 and 1970-1994. For the period of
1970-1979, we found weaker relationships between CBI and liquidity effects than those for other two
sample periods.

Contemporary Liquidity Effects

98

 

 

 

Minimum Value of Liquidity Effects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

°-°‘° 0.002
e
0005 .
oooo «
o 006 <
0 004 a .0 002 e
0002 <
41004 4
0000 «
.0002 < -0006 .
.0004 <
-o 008 <
o 006 4
e
”008 ‘E ' ' . ' ' “_*‘—‘ .0010 . a . . . .
°‘ °2 °3 °-‘ °5 0° °»7 03 01 02 03 04 0.5 06 07 08
HQ = 0.004 - 0.012 CBI LIQ = 0_oo1_ 0.014 CBI
2 (.145) (4.68)'
R = 0-19 R2 = 0.60 ' significance 0.01 level
Average Value of Liquidity Effects Duration of Liquidity Effects
0.020 so
0
0015 “ ‘0 -1
o
0.010 -‘ 3° .
0 005 a 20 .
0000 1 10 1
-0005 1 o 4 o .
.0 0‘0 7 V V 1 Y Y Y ‘I V V V V
0.1 02 03 04 05 06 07 on 01 0.2 0.3 04 as as 0.7 on
CBI CBI

LIQ = o 009 - o 021 CBI
(-174)
R2= 0.25

Figure 18.

LIQ = 5.50 + 47.40 CBI
(1.94)‘

R2 = 0.29 ‘ significance 0.10 level

CBI and Liquidity Effects : (R,M,Y,P)

< Contemporary Liquidity Effect >

99

< Minimum Value of Liquidity Effect >

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.009 0W
0
0,006 < . 0002 .
0
CW 1
0.002 <
0000 <
00% 1
«0.004 .
an 1
0” Y 1 Y Y Y T 0.010 Y Y I Y Y Y
0.1 0.2 0.3 0.4 0.5 0.6 o 7 0.5 0‘ 03 0-3 0‘ °-5 0-5 °-7 0-5
CBI CBI
“Q = o 003 - o 008 CBI LIQ = 0.002 - 0.016 CBI
' ' -3.18)'
.149 (
R2 = o 20 ( ) R2 = 0.53 ' significane 0.01 level
< Average Liquidity Effect > < Duration of Liquidity Effect >
0020 40
O
0.015 4
3o .
O
0010 -
20 -t
0.005 4
10 <
01130 4
0.005 < ° ‘ ' °
0.010 1 Y Y Y I T T 1' Y 1 I Y
O 1 O 2 0.3 O 4 O 5 0,6 0.7 0.6 0,1 0.2 0,3 0.4 0.5 0.6 0.7 0 8
CBI CBI

LIQ = 0.009 - 0.021 CB:
(-155)
R2 = 0.23

LIQ = 8.565 + 42.52 CBI
(1.75)
R2 = 0.25

Figure 19. CBI and Liquidity Effects : (P,R,M,Y)

Contemporary Liquidity Effects

100

Minimum Value of Liquidity Effects

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0000 0.004
. .
0W 1 cm -
O
0.002 1
0.000 .
0002 1
0004 <
0.006 -
0.000 . , , , , Y 0.010 . . , . 1 ,
0.1 0,2 0.3 0,4 0.5 0.6 047 0.5 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
CBI CBI
LIQ = 0.002 - 0.008 CBI LIQ = 0.003 - 0.016 CBI
(4.52) (-3.56)'
R2 = 0.20 R’ = 0.59 ' simificanoe 0.01 level
Average Value of Liquidity Effects Duration 0‘, Liquidity Effects
0020 ——1 40
O
0.015 «
3O .4
O
0010 <
20 -
0.005 «
10 .
0.000 «
0015 1 0 4 0 0
0.010 , I w r 1 1 , , Y 1 ,
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.0 0 1 0'2 0.3 0.4 0.5 0.6 0.7 0 6
CBI CBI

LIQ = 0.010 - 0.023 081
(.177)'
R’ = 0.26 1 significance 0.10 level

LIQ = 6.45 + 45.57 CBI
(1 .87)'
R2 = 0.28 ' significance 0.10 level

Figure 20. CBI and Liquidity Effects : (P,Y,R,M)

101

Table 20. CBI and Liquidity Effects (4 variable VAR)

 

 

Dependent Explanatory (R,M,Y,P) (P,R,M,Y) (P,Y,R,M)
Variables Variables
Contemporary Intercept 0.004 0.003 0.002
Liquidity Effect CBI —0.012 -0.008 -0.008
(-1.45) (-l .49) (-1.52)
R2 0.19 0.20 0.20
Minimum Value Intercept 0.001 0.002 0.003
of Liquidity Eff. CBI -0.014** -0.016** -0.016**
(-3.68) (-3. l 8) (-3.56)
R2 0.60 0.53 0.59
Average Intercept 0.009 0.009 0.010
Liquidity Effect CBI -0.021 -0.021 -0.023"‘
(-1.74) (-1.65) (~1.77)
R2 0.25 0.23 0.26
Duration of Intercept 5.501 8.565 6.45
Liquidity Effect CBI 47.401 * 42.520 45.57“
(1.94) (1 .75) (1.87)
R2 0.29 0.25 0.28

 

Note : The t-statistics are reported in parentheses. * indicates significance at the

10 percent level, 1'" at 1 percent level.

102

1-3. Five Variables VAR Estimation.

In this section, we include a measure of commodity prices (CP) to avoid the “price
puzzle” which is the result that positive orthogonalized innovations to interest rates are
associated with a prolonged rise in the price level.33 Sims (1992) conjectured that this
response reflects the fact that the Fed has an indicator of inflation in its reaction function
that is missing fiom the VAR underlying the policy shocks measure. After including
commodity prices in VAR system many authors find no price puzzle.34 We will examine
this possibility with a multicountry sample. Here, we add a commodity price index
CP(IFS..00176AXD) to a four variable VAR system.

Following Sims (1992), we will use R-rule in which the identification schemes used to
interpret the data will rely mainly on postulating that innovations .in short interest rates
represent monetary policy disturbances and estimate impulse responses for the five-
variable VAR estimates for 11 countries. The responses are for orthogonalized
innovations with the ordering as (R,CP,M,P,Y). Next, we will assume that monetary
authority sets a monetary policy after watching the macroeconomic aggregates, and that
monetary policy can not affect these aggregates in the contemporaneous period. This
assumption corresponds to the ordering as (Y ,P,CP,R,M). Finally, we will assume that
the monetary authority can not consider the level of Y, when they set the policy. Since
there is a time lag in obtaining Y: data, it is impossible to consider Yt of the period during

which they set the policy. But they can get contemporaneous price variables before

 

33 Actually, we found no price puzzle with US data for the sample periods that extend to 1994. The price
level decreases after a monetary contraction. This response of price level is shown in the fourth column of
the first graph in the Figure 13.

103

setting the policy. This assumption corresponds to the ordering as (P,CP,R,M,Y). From
the impulse responses of the 5 variable VAR system, we get the liquidity effects with
which to estimate the relationship with CBI. For example, in the Figure 16, the fourth
graph in the last column depicts the liquidity effects in the economy and we find no price
puzzle which is represented by the fourth graph in the second column.

The estimated results for the relationship between CBI and liquidity effect meet our
expectations which are represented by the Figures 21, 22 and 23.35 Table 21 summarized
estimation results. The contemporaneous liquidity effect is significantly inversely related
with CBI in the (Y ,P,CP,R,M) and the (P,CP,R,M,Y) orderings. The minimum value of
the liquidity effect is also inversely related with CBI at 10% significance level in
(Y,P,CP,R,M) ordering, but the relationship is not significant in the (R,M,P,Y,CP) and
the (P,CP,R,M,Y) orderings, even though the signs of coefficients are negative as we
expected. We also find the inverse relationship with CBI and the average liquidity effect
in all orderings. Finally, duration of liquidity effect shows the strongest relationship with
CBI in these 5 variable VAR systems. CBI has significant positive effects on the duration
of the liquidity effects in all 3 orderings. We can expect 5 more months of the liquidity

effect after a monetary shock if we improve CBI measurement by 0.1 point.

 

3‘ See Eichenbaum (1992), Sims (1992), Christiano et a1. (1996))

35 We report the estimation results only for the period of 1980-1994. For the period of 1970-1994, the
estimation results have little difference from those we report here. But the relationships are weaker for the
period of 1970-1979 as they were in the 4 VAR system.

< Contemporary Liquidity Effects >

104

< Minimum Value of Liquidity Effects >

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.000 0.000 a
O
0.004 4 0002 < . °
0 O .
0.002 « -0004 1 3\
. O
0000 < 0000 < e .
0002 « 0000 «
0 . -0.010 e
'0” fi a Y 1 ‘0012 v r Y T 1 T
0.1 o 2 o 3 0.4 O 5 06 0 7 0-5 0.1 0.2 0.3 0.4 0.5 0.0 0.7 0.0
CBI CBI
l = . - 0.004 BI
LIQ = 0.002 - 0.008 CBI L Q '0 003 (0 60) C
(-1.55) R2 = 0 04 I
R2 = 0.21 '
< Average Liquidity Effects > < Duration of Liquidity Effect >
0.012 40
0°10 ‘ . 35 4
0.000 «
3° ..
0005 i
25 ..
0W 1 .
2° .
0.002 «
15
0.000 4
.0002 « 1° ‘
0.004 « s . o e
‘0.” v Y Y t 7 V o T 1 V 1 1 v
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0 1 0.2 0.3 0.4 0.5 0.0 0 7 0.0

CBI

LIQ = 0.005 - 0.013 CBI
(.174)
R2 = 0.25

CBI

me = 7.045 + 44.10 CBI
(2.15)-

R’ = 0.34 1 significance 0.05 level

Figure 21. CBI and Liquidity Effects : (R,M,P,Y,CP)

105

Contemporary Liquidity Effects

 

GUM

oar-1

OUD1

-OGM1

 

-OGM

Minimum Value of Liquidity Effects

 

 

 

 

-0G3
0

0014

0012

0010

OGN1

Odfl1

GUM“

QGH1

OUD<

-oun~

1 Q2 03 04 OS 00 O] 00

L10 = 0.0017 - 0.0063 C31
(-1.00)1
R2 = 0.28 ' significance 0.10 level

Average Value of Liquidity Effects

 

.4

q

 

 

 

 

Q1 02 03 O! 05 06 07 00

CBI
1.10 = 0.0057 - 0.0143 CBI

(4.86)
R’ = 0.27 1 significance 0.10 level

Figure 22. CBI and Liquidity Effects : (Y,P,CP,R,M)

 

 

 

 

 

 

 

 

° 0
'O.m7 Y Y 1’ Y
01 02 03 04 05 06 07 00
LIQ = 00012 - 0.0070 CBI
(-1.86)‘
R2 = 0.20 1 significance 0.10 level
Number of Months with Liquidity Effects
50
‘0 a
30.
20 .4
101
. 0
o Y T Y Y Y V
01 02 03 04 05 00 07 08

CBI

LIQ = 347+ 53.17 CBI
(2.72)
R2 = 0.45 1 significance 0.03 level

106

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

< Contemporary Liquidity Effects > < Minimum Value of Liquidity Effects >
0004 oooo
. o 001
O
0.002 4 c
O 0002 3 o O
0000 4
cone 4
0002 ‘ oooi «
-ooos «
coo: « e
coca <
. 0
0m V Y Y fir Y 7 0W7 1 V7 r Y T v
0'1 0.2 03 0.4 0'5 0-6 0.7 0.8 0.1 0.2 0.3 0.4 0.5 0.6 07 on
CBI CBI
L|Q = 0.002 - 0.006 CBI LIQ = -0.001 - 0.006 CBI
(11-93)' 2 {-1.72)
R2 = 0.29 1 significance 0.1 level R = 0-25
< Average Liquidity Effects > < Duration of Liquidity Effects >
0014 so
0.012 1 .
0010 < so «
o.ooe «
oooe - 30 J
OW 4 O
can 4 20 i
0G!) 1 ‘
oooe « lo « i
.0004 < e e
‘0” V Y Y Y f f O V T T 1 T V 1
0.1 0.2 0.3 0.4 0.5 0.6 0.7 o a 0.1 0.2 0.3 0.4 0.5 0.6 o 7 0.3
CBI CBI
L|Q = 0.005 - 0.014 CBI LIQ = 4.627 + 49.28 CBI
(4.73) (2-75)'
R2 = 025 R’ = 0.46 1 significance 0.03 level

Figure 23. CBI and Liquidity Effects : (P,CP,R,M,Y)

107

Table 21. CBI and Liquidity Effects (5 variable VAR)

 

 

Dependent Explanatory (R,M,P,Y,CP) (Y,P,CP,R,M) (P,CP,R,M,Y)
Variables Variables
Contemporary Intercept 0.002 0.002 0.002
Liquidity Effect CBI -0.008 -0.006* -0.006*
(-1.55) (-1.88) (—1.93)
R2 0.21 0.28 0.29
Minimum Value Intercept -0.003 -0.001 -0.001
of Liquidity Eff. CBI -0.004 -0.007"' -0.006
(-0.60) (-l .86) (-l .72)
R2 0.04 0.28 0.25
Average Intercept 0.005 0.006 0.005
Liquidity Effect CBI -0.013 -0.014* -0.014
(-1.74) (-1.86) (-1.73)
R2 0.25 0.27 0.25
Duration of Intercept 7.045 3.471 4.627
Liquidity Effect CBI 44.184" 53.170*** 49.284***
(2.15) (2.72) (2.75)
R2 0.34 0.45 0.46

 

Note : The t-statistics are reported in parentheses. * indicates significance at the 10 percent
level, *"' at 5 percent level, *** at 3 percent level.

108

2. Structural VAR Approaches.

2-1. Structural VAR.

In a VAR system we estimated liquidity effect by the (kj) element of Cs which gives
the response of kth element of yes to a unit disturbances in the jth element of at, that is,
monetary innovation from the system yt = C(L)8 t = 2 CS ens, where C(L) = A"[I - B(L)]".
The problem with this procedure is that we can not directly observe or estimate 82¢ which
represents the vector of policy disturbances. The VAR of yt is given by y! = G(L) yH + pt
where G(L) = A"B(L), g, = A42, and Else; = A" (A")’ = D.

Without additional restrictions on the system, we can estimate D and G(L), but A and
B(L) are not identified. We can calculate the moving average representation as yt =Z(L)ut
where Z(L) =[I-G(L)]". But without very special assumptions regarding matrix A, ut’s
are not equal to 81’s. This means that dynamic response of nonpolicy variables in y, to
shocks in St will not coincide with the dynamic response of those variables to shocks in
pt.

In order to resolve this problem, sufficiently strong restrictions must be imposed to
identify the matrix A. While various procedures have been adopted by empirical studies,
the type of restrictions most relevant for the existing liquidity effect is restrictions on the
contemporaneous nature of feedback between the elements of y, that is, restrictions on

the matrix A. Most researchers in the area have proceeded by adopting a particular Wold

109

causal interpretation of the data. The general idea is to assume that the matrix A is
triangular when the variables yt are ordered according to their causal priority. Under this

assumption, there is a unique A which satisfies A‘I (A")’ = D for a given covariance

matrix D.
For example, we can write 4 variable VAR system in vector form as

Boy! = k + B(yH + Bzyg + ...... + Bth-p + [it (4.8)

where yi = (Rt, Ml, Pb Yt)’

Pi = (PM, P21, PM #1" Y

i 1 -fl?. W. 43?."
Bo = —fl(2)l 1 W333 - (2)4
-fl(3)l -1622 1 ‘76::
__:Bfil 'flgz -1623 1

 

 

k =(k19 1(2) k3, k4),
and Bs is a (4 x 4) matrix whose row i, column j element is given by 430-5
for s = 1,2,...,p.

If each side of (4.8) is premultiplied by Bo'l, the result is

y. = c + (hiya) + (Dzygz + . . . + (DPyH, + at. (4.9)
where c = B()'1 k (4.10)
cps: 130'1 B, for s= 1,2,...,p (4.11)
e. = B0“ is. (4.12)

Assuming that (4.8) is parameterized sufficiently richly that in is vector white noise, then
at will also be vector white noise and (4.9) will be recognized as the vector autoregressive

representation for the dynamic structural system(4.8). Thus, a VAR can be viewed as the

110

reduced form of a general dynamic structural model. In previous sections, we calculated
the impulse response function as 6y”, we), in (4.7) which describes the effect of an
innovation in the jth variable on future values of each of the variables. According to
(4. 12), the VAR innovation Ejt is a linear combination of the structural disturbances pt.

For example, it might turn out that

8“ = 0.1 u“ - 0.2 (12, - 0.3 (13. - 0.4 LL41.

so that if an turns out to be negative, it might reflect the effects of either 1.12; , p3. , p4. or
any combination of these disturbances. Therefore, the impulse response function using
VAR might not be interesting since St represents any combination of the different
influences that matter for the variables in the economy. This gives us the reason why we
need to identify pit from St

In VAR system, for Q = E(etet’), we found a lower triangular matrix A and a diagonal
matrix D such that Q = ADA’. We then constructed the vector Ale. and calculated the
consequences of changes in each element of this vector for future values of yL Recall
from (4.12) that the structural disturbances pt are related to the VAR innovations at by [it
= B0 8.. Suppose that it happens to be the case that the matrix of structural parameters B0
was exactly equal to the matrix A". Then the orthogonalized innovations would coincide
with the true structural disturbances it. = Boat = A'lat. Since A is lower triangular, this
requires B0 to be lower triangular.

For example, one might argue that prices respond to other economic variables only
with a lag and that money and interest rates also influence income only with a lag. One

might argue further that the interest rates affect desire money holding at the point of

111

time. These assumptions suggest ordering the variables as y = (Pt, Yt, Rb Mt)’ for which

 

 

 

 

 

 

 

 

 

 

 

 

the structural model would be
'P.‘ ”k.' '0 0 0 0”P. " l. 5:. fl}. 1." P.-.
Y. k. #3. 0 o 0 Y. +55. 13;. .61. ‘2. Y.-. (413)
R. 1‘: .33. .332 0 0 R. flil .532 '33 We. Rr-l
_M._ Jail flil El: .533 0__M._ - ii .312 is fliidtMnn
' f. r. l; r. 8-.“ yr
+...+ fl; flgz 1653 1654 Yt-p + .11:
:1 [Big fl3p3 :4 Rt-p i“:
_flfl flfz :64le flfu _M"P_ _luin_l

 

 

 

 

 

 

Suppose there exists such an ordering of the variables for which B0 is lower triangular.

Write the dynamic structural model (4.8) as

BOYt= ’rxt + “t (4-14)
where -1" [nx(np+1)] E [k B1B2...Bp]
_ 1 _
yl-l
Xi [(np+1)x1] E y.-2
_yi_pd

 

 

Suppose that the disturbances in the structural equations are serially uncorrelated and

uncorrelated with each other :

E(u().l..’) = D for t= 1: (4.15)
0 otherwise,
where D is a diagonal matrix.

The VAR is the reduced form of the dynamic structural model (4.9) and can be written as

112

yt=IT’ Xt+et (4.16)
where H’ = -B()'1 F
St 1' Bo.l ill.
Letting Q denote the variance-covariance matrix of et’ we get

0 = B(eiei’) = B." B(nlul’x Bo")’ = Bo" D (B.." )’ (4.17)
If the only restrictions on the dynamic structural model are that B0 is lower triangular with
unit coefficients along the principal diagonal and that D is diagonal, then the structural
model is just identified. To see this, note that these restrictions imply that Bo'I must also
be lower triangular with unit coefficients along the principal diagonal. Given any positive
definite symmetric matrix Q, there exists a unique lower triangular matrix A with IS
along the principal diagonal and a diagonal matrix D with positive entries along the
principal diagonal such that Q = ADA’. Thus, unique values 80'1 and D of the required
form can always be found such that satisfy (4.17). Moreover, any Bo matrix of this form
is nonsingular, so that F can be calculated uniquely from Bo and II as F = - Bo H’. Thus
given any allowable values for the reduced form parameters (1'1 and (2), there exist unique
values for the structural parameters (Bo, F and D) of the specified form, establishing that

the structural model is just identified.
Since the model is just identified, full-information likelihood (F IML) estimates of (B0,
F and D) can be obtained by first maximizing the likelihood function with respect to the
reduced form parameters (IT and Q) and then using the unique mapping from reduced
form parameters to find the structural parameters. The maximum likelihood estimates of

IT are found from OLS regressions of the elements of y. on X, , and the MLE of Q is

113

obtained from the variance-covariance matrix of the residuals from these residuals. The

estimates Bo'1 and D are then found from the triangular factorization of Q.

2-2. Multi-Country SVAR Model

For estimation of liquidity effects by structural VAR system (SVAR), we use the three
different models in Giannini (1992) which are the K—model, the C-model, and the AB-
model. However, since our model failed to converge in the K-model and the C-model,
here we report the results only with the AB-model. According to Giannini, AB-model is
the most generalized one among those 3 models and it is defined as follow:

“ A, B are (n x n) invertible matrices36

AA(L) Yr: Aer

A31: Bet

where E(e,) = [ 0 ] and B(etet’) = In .
The A matrix induces a transformation on the at disturbances vector, generating a new
vector(Aet) that can be conceived as being generated by a linear combination(through B
matrix) of n independent(orthogonal) disturbances, which we will refer to as et_
From A8. = Bet, A(et a’)A’= B(et et’)B’ , AZA’ = BB’.
For 2 known, this equation always imposes a set of n(n+1)/2 non-linear restrictions on
the parameters of the A and B matrices, leaving overall 2n2 - n(n+1)/2 free element.

..”(p.5-6)

With the AB-model in Giannini, we construct the multi-country SVAR model. In
previous section, we applied VAR model country by country and then estimated the
relationship between the coefficients which represented the liquidity effect and CBI.
However, since these coefficients came from the separate systems of each country, not

from one system, it might be less meaningful to compare the coefficients of each country

 

36 For the discussion on the size of matrix B, see Giarmini, p.4~5.

114

in the view of econometrics. In SVAR system, since we can impose the restrictions on
the system and also include more than one country in one model, it is possible to consider
the interactions of the economies and so, the estimation results would be more useful to
compare the coefficients of each country as long as the restrictions are reasonable.

Before we construct SVAR model, we need to restrict the number of countries in the
model because our time series data are not long enough to include all 11 countries in one
SVAR model. With 9 months lag in each variable and time series data for about twenty
years, six country is the maximum number of countries we can include in modeling a 4
variable SVAR system. And since the SVAR does not impose any restrictions on the
interaction between lag variables cross the countries or on the effects of past variables to
present variables, we also need to choose sarnple countries as reasonable as possible.

Under these conditions, first we construct a 5-country SVAR model for the period of
1973-1989. These countries are US, Germany, Japan, France, and U.K. Because these
countries are the five largest countries in the world economy, it is reasonable to assume
that the economic variables and economic policies are not free from those of other
countries with a lag, that is, for example, the income level of one country is affected by
the economic variables like income or price level of other countries. Second, we
construct a 6-European country SVAR model for the same period. These are Germany,
France, UK, Italy, Denmark, and Norway. We choose these 6 countries by the order in
the size of trade. The amount of international trade of these countries are lager than any

other European countries. Since these 6 countries are the biggest ones in economic power

115

in Europe and they are located nearby, we also assume that these countries have
interactions in the economic policies and the economic variables among each other.37

For the estimation, we use quarterly data as well as monthly data. Quarterly data are
used here with the expectation that that the model with quarterly data can reflect
interactions in economic policies and variables across countries better than monthly data,
since usually it takes time for economic variables of one country to affect other’s
economic policy or variables. We will report the results mostly with quarterly data
because actually we found that there are little difference in results using different time
frequency data.

We will use systems as (P,R,M,Y), (R,M,P,Y), and (Y,P,R,M).38 In (P,R,M,Y)
system, for example, price levels of each country are determined only by past values of
economic variables of the country and those of other countries. Interest rates, however,
are determined not only by past values of economic variables of all countries in the model
but by present value of price level of the country which is predetermined. Money
aggregates are affected by present price level and the interest rate of the country with all
past economic history of all countries in the model. Finally, the income level of the
country is determined by the current value of other variables in the country and the past
history of the economic variables of the country as well as those of other countries.
Table 22 shows our over-identified structural VAR system in a 5 country model and

Table 23 in a 6 country model.

 

37 We estimated the relationship with 7 countries which include US into this 6 country model. We used 6-
months lags because of the limitation of the data length. The results in 7 country model were very similar to
the results we report.

116

Table 22. 5 country SVAR model (1973-1989)

 

 

(P,R,M,Y) (R,M,P,Y) (Y,P,R,M)
us. R =156.13P M =0.003R P =0.056Y
(2.57) (1.325-4) (37453)
M =1 .125P - 1.105-4R P = 0.003R + 0.156M R =43.791Y + 82.342P
(0.038) (1 .645-4) (4.775-5) (0.01) (0.67) (2.19)
Y =-0.569P +0.004R +1.224M Y =0.004R + 1.224M - 0.569P M =07le - 0.224P - 0.014R
(5.85E-4)(2.33E-6)(2.18E-4) (2.33E-6) (2.18E-4)(5.85E-4) (0.22) (0.06) (3.1 15-4)
GER R =211.55P M =2.38E-4R P =0.033Y
(3.73) (2.105-4) (1 .955-3)
M =-0.996P + 0.002R P =0.002R - 0.029M R =-3.033Y + 30.146P
(0.09) (2.755-4) (3.555-5) (2.605-3) (0.44) (2.72)
Y =3.039P + 0.006R + 0.277M Y =0.006R + 0.277M + 3.039P M =0.251Y - 2.465P - 0.021 R
(0.13) (4.0054) (0.02) (4.0354) (0.02) (0.13) (0.03) (0.16) (7.3754)
JAP R =42.782P M =-0.010R P =-0.037Y
(1.16) (2.3354) (3.84E-3)
M =-0.748P - 0.006R P =0.002R - 0.327M R =8.449Y + 38.582P
(0.02) (2.325-4) (1 .625-4) (0.01) (0.33) (1.07)
Y =0.404P + 0.003R - 0.005M Y =0.003R - 0.005M + 0.404P M =0.079Y - 0.096P - 0.005R
(0.02) (1.9354) (0.01) (1 .935-4) (0.01) (0.02) (0.02) (0.05) (5445-4)
FRA R =213.66P M =1 .425-4R P =-0.206Y
(1.89) (2.915-4) (0.01)
M =5.92P - 0.02R P =0.004R + 0.068M R =24.590Y + 11.323P
(0.11) (4.525-4) (2415-5) (1.305-3) (1.92) (3.02)
Y =0.310P + 0.009R - 0.100M Y =0.009P - 0.100M + 0.310P M =-1.074Y - 0.131P - 0.003R
(1 .70E-3)(6.67E-6)(l .355-4) (6.67E-6)(1 .85E-4)(1.7OE-3) (0.08) (0.13) (5395-4)
U.K R =-12.614P M =0.057R P =0.008Y
(2.71) (5.5754) (4.305-3)
M =-1.085P + 0.056R P =0.001R - 0.027M R =19.563Y + 90.778P
(0.10) (5.505-4) (1.6OE-4) (2405-3) (0.68) (1.95)

Y = -0.528P -0.011R +0.018M
(0.01) (9.295-5)(1.405-3)

Y = -0.011R +0.018M -O.523P
(9295-5) (1.4053) (0.01)

Note : The standard errors are reported in parentheses

M =-O.l40Y -l.l49P + 0.014R
(0.04) (0.12) (4.28E-5)

 

3” We also construct 5-variable SVAR models which include commodity price in 5 country, 6 country. or 7
country 4-variable SVAR model but we do not report them here since the regression results regarding C Bl
do not have any significance.

117

Table 23. 6 country SVAR model (1973-1989)

 

 

(P,R,M,Y) (R,M,P,Y) (Y,P,R,M)
GER R = 11.739? M = -0.004R P = 0.046Y
(3.50) (3.85E-4) (37453)
M = -0.292P - 0.004R P = 1.82E-4R - 0.008M R = 8.366Y + 5.609P
(0.09) (3.85E-4) (6.63E-5) (2605-3) (0.886) (3.527)

FRA

U.K

ITA

DEN

NOR

Y = 0.744? + 0.003R + 0.134M
(0.06) (2.52E-4) (9.8OE-3)

R = -31.469?
(2.76)

M = -3.1 12? + 0.008R
(0.08) (4.27E-4)

Y = 1.006? - 0.003R + 0.056M
(0.05) (3.00E-4) (8.90E-3)

R = 34.543?
(1.99)

M =1.849P - 0.014R
(0.10) (7.535-4)

Y = 0.269P - 0.004R + 0.042M
(0.03) (1 .89E-4) (3.705-3)

R = 72.261 P
(1.89)

M = 3.548P - 0.013R
(0.07) (5.1254)

Y = -0.817P +0.007R + 0.242M
(0.06) (3.375-4)(9.305-3)

R = 47.279?
(3.09)

M = 0.110P - 0.0113R
(0.07) (331134)

Y = -0.911P +0.003R + 0.067M
(0.03) (1 .67E-4) (6.8OE-3)

R = 11.496P
(3.53)

M = 0.990P - 0.007R
(0.14) (5.7754)

Y = 2.067P - 0.004R + 0.004M
(0.05) (2.08E-4) (5.405-3)

Y = 0.003R + 0.134M + 0.744?
(2.52E-4) (9.8OE-3) (0.06)

M = 0.011R
(49054)
P = -2.005-5R - 0.085M
(7305-5) (2105-3)
=-0.003R + 0.056M + 1.006P
(2.60E-4) (0.01) (0.05)

M = -0.01 1R
(75554)
P = 0.002R + 0.038M
(1 .06E-4) (2.105-3)
Y=-0.004R + 0.042M + 0.269P
(1.89E-4) (3.705-3) (0.03)

M = -3.205-4R
(5.48E-4)

P = 0.004R + 0.098M
(7.405-5) (2005-3)

Y = 0.007R + 0.242M - 0.817P
(3.3754) (0.01) (0.06)

M = -0.01 1R
(3225-4)

P = 0.001 R + 0.005M
(7.995-5) (3.305-3)

Y = 0.003R + 0.067M - 0.911P
(1 .67E-4) (6.8OE-3) (0.03)

M = -0.007R
(57954)

P = 3.005-4R + 0.012M
(6.54E-5) (1 .705-3)

Y=-0.004R + 0.004M + 2.067P
(2085-4) (5405-3) (0.05)

Note : The standard errors are reported in parentheses

M = 0.308Y - 0.508? - 0.005R
(0.023) (0.089) (3.81E-4)

P = 0.089Y
(4.56E-3)

R = -8.061Y - 24.190P
(0.894) (2.848)

M =0.161Y - 3.246P + 0.008R
(0.026) (0.081) (42954)

P = 0.064Y
(0.009)

R = -27.151Y + 39.661P
(1.091) (1.874)

M = 0.702Y + 1.605P - 0.011R
(0.061) (0.103) (79354)

P = 0.043Y
(0.005)

R = 8.221Y + 69.399P
(0.654) (1.867)

M = 0.552Y + 3.526P - 0.015R
(0.021) (0.069) (4.85E-4)

P = -O.159Y
(0.006)

R = 24.764Y + 66.625P
(1.436) (3.185)

M = 0.329Y + 0.407P - 0.012R
(0.033) (0.074) (3.38E-4)

P = 0.135Y
(0.003)

R = -21.222Y + 54.409?
(1.022) (3.953)

M = 0.030Y + 0.927P - 0.007R
(0.043) (0.161) (6.0454)

118

From the estimation results, we regress interest elasticity which is the coefficient of R
in M equation on CBI. Regression results show no strong relationship between this
elasticity and CBI although the coefficients of CBI for the liquidity effect have negative
signs in all 3 cases. In the (Y,P,R,M) order, the relationship is significant at 1 percent
level while in other two orders, it is not significant. But here, since we have only five
countries in the model, it is not sufficient to interpret these results as the general
relationship. In 6 country SVAR model which is reported in Table 18, we also find that
the inverse relationship between CBI and the elasticity of interest rate but the relationship
is not significant. However, since the elasticity of interest rates does not correspond to
the liquidity effect, we need to calculate impulse response functions of the variables to the
shock of each variable from this structural VAR model and examine the relationship

between CBI and the impulse response functions which represent liquidity effects.

2-3. Estimation Results

Now, from these SVAR models, we can identify the impulse response function of the
variables to the shock of each variable in the model which is 0yt+s / 611,139 Figure 24 and
25 depict the impulse response fimctions in 5-country and 6-country SVAR model
respectively, which represent liquidity effects with (R,M,P,Y) order. From the impulse
response functions, we obtain the same kinds of values as we used in previous sections

and then regress these values on CBI measurement. The regression results are reported in

119

Table 24. In most cases, we get the expected relationships, which are inverse relationship
between CBI and the contemporary, minimum or average liquidity effect and positive
relationship between CBI and duration of liquidity effect. However, no relationship is
significant and the R2 values are very small in most cases, so that it is very difficult to get
implications from these regressions. 40

In SVAR model, we assume that economic variables or economic policy in one
country are affected not only by past history of economic variables of the country but
also by past history of economic variables or economic policies of the all other countries
in the model. This assumption may not be reasonable since the economic authority
would not consider past histories of all countries even though they might consider recent
economic variables, or of the economic policies of some countries which have
considerable effects on their economy. For example, Germany would not consider the
past movements of economic variables or economic policies of small countries like
Norway though small countries might consider the past movements of economic variables
or economic policies of big countries. This unrealistic assumption in SVAR model could
result in illogical impulse response functions of the variables and so the liquidity effect
derived from these impulse response functions might not reflect the real relationship
between money aggregate and interest rates. It, therefore, is very hard to find useful

implications in the relationship between CBI and liquidity effect in SVAR model.

 

39 We use VMA procedure for RATS in Giannini(p.122~127) to get the impulse responses in SVAR model.
4° We also found no significant relationship between CBI and liquidity efiects in 5-variable SVAR models.

<U.S>

< Germany >

< Japan >

<U.K>

< France >

120

 

0.20 -i
0.15 —i
0.10 =
005 =
000 —
006 =
~040

 

 

 

-OJ§

 

0.00
0.06
0.04 -1
00241
0.00 -«
-0.02 =
-004 =
-006 —
006 -
010 ~

 

 

 

«0.12

 

0.08 -l
0,04 -
002 =
0,00 —
0.02 ~
.004 -

 

 

 

 

 

 

 

 

0.0
0.4 d
0.2 -‘
0,0 A
-0.2 ~
-0.4 =
-0.0 -1

 

 

 

-0.0

12 10 20 24

Time (Month)

Figure 24. Liquidity Effects in S-Country SVAR (R,M,P,Y)

< Germany >

< France >

<U.K>

< Italy >

< Denmark >

< Norway >

121

 

0.2 ‘1—

0.1

0.0 I
-0.1

-0.2 *

 

20 24

 

0.8
0.6 4
0.4 1
0.2 4
0.0 -*
-0.2 -i

 

 

 

12 16 20 24

 

 

 

 

20 24

 

0.6
0.4 1
0.2

 

20 24

 

 

 

 

20 24

 

 

 

 

Time (Month)

Figure 25. Liquidity Effects in 6-Country SVAR (R,M,P,Y)

122

Table 24. CBI and Liquidity Effects (SVAR)

 

 

5-9mm“ 5mg -
QR.M,P.Y) (P.R.M.Y) (R-M-P-Y) (P,R,M,Y)
Contemporary Intercept 0. 1 74 0.245 0.266 O. l 26
Liquidity Effect CBI -0.061 -0.078 -0.747 -O.785
(-1.19) (-0.92) (-1.13) (-1.06)
R2 0.32 0.22 0.24 0.22
Minimum Value of Intercept -0.128 -0.978 -0.678 -0.l62
Liquidity Effect CBI -0.003 - 0.248 0.771 -0.378
(-0.04) (0.96) (2.06) (-042)
R2 0.00 0.24 0.52 0.04
Average Intercept 0. 122 0.023 0.077 0. 123
Liquidity Effect CBI -0.043 -0.016 0.129 -0.189
(-1.42) (-O.84) (0.12) (-1.83)
R2 0.40 0.19 0.00 0.46
Duration of Intercept 0.182 2.939 4.685 1.127
Liquidity Effect CBI 1.364 0.879 -2.882 4.357
(1.53) (0.64) (-O.69) (1.26)
R2 0.44 0.12 0.11 0.28

 

Note : The t-statistics are reported in parentheses

123

3. Summary

To complement the defects resulted fi'om the traditional approaches, we adopted VAR
system to identify the liquidity effects. First, we applied four variable nonstructural VAR
approach to country by country to estimate the liquidity effect and then we examined the
relationship between these liquidity effects and CBI. Applying VAR in our model did not
change our previous results which came from the traditional empirical approaches.
Liquidity effects are much stronger and more persistent in high CBI countries than in low
CBI countries. It means that monetary policy effects are significantly affected by the
degree of independence of the central bank. The difference in ordering of the variables
also did not change the results.

Next, we added one more variable, commodity price, to the four variable VAR system
to avoid price puzzle. The estimation results from this five variable VAR confirmed our
previous results no matter which ordering was adopted. Finally, we constructed structural
VAR model to get the more accurate impulse response functions and so the more accurate
liquidity effects. And we included all countries considered in one SVAR model while we
applied nonstructural VAR system country by country. The estimation results of this
multi-country SVAR model revealed the inverse relationship between the liquidity effect
and CBI as we expected. However, we did not find any significance in these estimated
relationship. Even though the coeflicients from the multi-country SVAR model are more
meaningful for the comparison in the View of econometrics, this model has an illogical
assumptions which are not suitable to the real economy. For example, a big country

would not consider past histories of economy of other small countries as this model

124

assumed. Since the impulse response function from this multi-country SVAR model
might be unreasonable in real world, our estimation results regarding the liquidity effect

in this model have little policy implication.

CHAPTER V

CONCLUSION

Recent tendency toward reforming the central bank system to give more independence
in many countries is primarily based on the success of the highly independent
Bundesbank and Swiss National Bank in maintaining comparatively low rates of inflation
for prolonged periods of time as well as the theoretical and empirical literatures which
show the significant effect of CBI on inflation rates. After time consistency theory, many
empirical researches have confirmed the theoretical arguments that greater central bank
independence is associated with lower levels of inflation. Since political and economic
dependence of central bank restrict the ability of the central bank to select its policy
objectives without influences by the government and these dependence with high
inflation experiences makes agents assign low credibility to the central bank’s monetary
policy, legal central bank independence has been identified with a credible commitment
to the price stability. This credibility bonus is presumed to be the source of the negative
correlation between central bank independence and inflation rates.

This study started from the idea that if independent central banks’ policies are
inherently more credible, not only inflation levels but also expectations of monetary
policy must differ systematically across countries with differences in central bank
independence. No studies have examined the postwar experiences for evidence of
credibility effect of central bank independence until very recently, some studies tried to

find the evidence. All of these studies regarding credibility effects, which are Debelle

125

126

and Fischer (1994), Walsh (1994), and Posen (1995), adopted the idea that the greater
credibility attributed to independent central bank should reduce the costs of subsequent
policies to lower inflation and so examined the relationship between disinflationary costs
and CBI. However, they found that disinflation appeared to be more costly and no more
rapid in countries with more independent central banks.

Even though they have the same results, we should note some limitations in these
studies before we accept their conclusion that there is no credibility effect of central bank
independence. Because of data availability and CBI measurement problem for
developing countries, they restrict the sample countries only to developed countries like
most literature in this field. Since these developed countries usually have high
independent central bank systems and have maintained relatively low inflation rates, the
difference of CBI might not be sufficient to show the difference of disinflationary costs
clearly. In addition, since central bank independence is negatively correlated with
average inflation rates in these developed countries, the positive relationship between
disinflationary cost and the independence level of central bank may reflects increasing
marginal costs of disinflation. Therefore, their conclusion of no credibility effect of CBI
should be interpreted very carefully and these limitations give us the necessity of further
investigations for the credibility efi‘ect.

In this paper, we tried to find the credibility effect by directly examining the
relationship between monetary policy effect and CBI. Our study is based on the
assumption that the credibility effect can be shown in monetary policy effect on the
economy and that higher central bank independence can make monetary policy effects

stronger. This means that the degree of monetary policy effects can differ across

127

countries if their central banks have different level of independence. In order to measure
monetary policy effects, we calculated liquidity effects in each country and then examined
the relationship between central independence level and the size of liquidity effects.

When we applied the various traditional approaches to find liquidity effects in each
country, the regression results confirmed that there was a strong relationship between CBI
and liquidity efiects. We could see CBI enhanced the degree of monetary policy effects
across countries. Applying SURE to the traditional approaches to find liquidity effects in
each country did not change our original results in traditional methods. The regression
results showed the significant relationship between CBI and liquidity effects.
Furthermore, the liquidity effect estimated by non-structural VAR also shows a
significant negative correlation with CBI. However, on the contrary, we could not find
any significant relationships between these two variables in either 4-variable or 5-variable
SVAR model even though the signs of most coefficients are as we expected.

How should we interpret these contradictory results in SVAR model? Is this result
strong enough to deny the preceding results which show a significant negative correlation
between CBI and liquidity effects? In SVAR system, we assumed that economic policies
or economic variables of each country are affected by past history of economic policies
and economic variables of all countries in the model. Since we use quarterly data, this
assumption means that economic policy authorities in each country get all information of
past economic fluctuation and policy changes, including the previous quarter, of all other
countries and consider all these information when they set their own economic policies.

However, in practice, there are much more considerable time lags to get the exact

information of changes in economic variables of other countries and the economic policy

128

authorities might have not enough time to wait to get all information regarding economic
changes in other countries for setting their own policy. For example, a country would
need more than one time lag, which is 3 months in our model, to get the information
about other countries’ previous period income level. We do not know how many time
lags should be needed for policy authority to get a specific information about the
economic fluctuations in other countries. This implies that the estimated liquidity effects
in SVAR model should be changed if we change our SVAR structure and that our
estimated relationship between CBI and liquidity effect also might be changed. Simple
change in lag structure in SVAR model should result in changes in estimated liquidity
effect in the model. Practically, it would need huge matrix structures and more precise
computational program to consider the case that a country needs different time lags to get
the information of each economic variables of other countries. The problem in SVAR
model is that we do not know the right structure of multi-country model even though
SVAR model gives us an advantage in which we get the all countries’ liquidity effects
from one model structure, not from country-by-country models. Therefore, we need more
future studies in constructing multi-country SVAR model.

Although the SVAR model did not support our previous results, this paper suggests
that there is a significant relationship between CBI and liquidity effects and also that there
is a vicious circle which shows us the inverse relationship between CBI and inflation
rates if a low CBI country needs larger changes in monetary aggregates to affect interest

rates which is suggested from the relationships between CBI and liquidity effects.

LIST OF REFERENCES

LIST OF REFERENCES

Alesina, Alberto, “Macroeconomics and Politics,” NBER Macroeconomics Annual 1988,
pp.13-52.

Alesina, Alberto and Lawrence H. Summers, “Central Bank Independence and
Macroeconomic Performance : Some Comparative Evidence,” Journal of Money, Credit
and Banking (May 1993), pp. 151-62

Bade, Robert and Michael Parkin, “Central Bank Laws and Monetary Policy,”
unpublished manuscript, University of western Ontario, 1985.

Ball, Laurence, “Time-consistent Policy and Persistent Changes in Inflation,” NBER
Working Paper no.3529, Dec. 1990.

, “What Causes Inflation?” Business Review, FRB of Philadelphia (March/
April 1993), pp.3-12.

 

 

, “What Determines the Sacrifice Ratio?” In Monetary Policy ed. by N.G.
Mankiw, Univ. Of Chicago Press, 1994.

 

, “Credible Disinflation with Staggered Price-Setting,” The American Economic
Review 84 (March 1994), pp.282-289.

 

, N.G. Mankiw, and D. Romer, “The New Keynesian Economics and the
Output-Inflation Tradeoff,” Brookings papers on Economic Activity, 1988 Spring, pp.1-
65.

Barro, Robert, and David Gordon, “Rules, Discretion, and Reputation in a Model of
Monetary Policy,” Journal of Monetary Economics 12 (July 1983), 101-22.

129

130

Bemanke, Ben, and Alan Blinder, “The Federal Funds Rats and the Channels of
Monetary Transmission,”American Economic Review 82 (1992), pp.901-921.

Bemanke, Ben, and Ilian Mihov, “Measuring Monetary Policy,” FRB of San Francisco
Working Paper 95-09 (Mar., 1995).

Bemanke, Ben, and Mark Gertler, “Inside the Black Box : The Credit Channel of
Monetary Policy Transmission,” Journal of Economic Perspectives 9 (Fall, 1995), pp.27—
48.

Breusch, T., and A. Pagan, “The LM Test and Its Applications to Model Specification in
Econometrics,” Review of Economic Studies 47 (1980), pp.239-254.

Bruno, M., and J. Sachs, The Economics of Worldwide Stag'lation, Cambridge, Harvard
Univ. Press, 1985.

Cagan, Phillip, and A.Gandolfi, “The Lag in Monetary Policy as Implied by the Time
Pattern of Monetary Effects on Interest Rates,” American Economic Review 59 (1969),
pp. 277-284.

Christiano, Lawrence, “Modeling the Liquidity Effects of a Money Shock,” FRB of
Minneapolis, Quarterly Review (Winter, 1991), pp.3-34.

 

, “Commentary on ‘ Resolving the Liquidity Effects’,” FRB of St. Louis, Review
77 (1995), pp.55-61.

 

, “Identification and the Liquidity Effect : A Case Study,” FRB of Chicago,
Economic Perspectives (1996), pp.2-l3.

Christiano, Lawrence, and Martin Eichenbaum, “Identification and the Liquidity Effect of
a Monetary Policy Shock” in Political Economy, Growth and Business Cycles ed. by
Cukierman et.al, MIT Press (19923), pp.335-370.

131

 

, “Liquidity Effects and the Monetary Transmission Mechanism,” American
Economic Review 82 (1992b), pp.346-353.

 

, “Liquidity Effects, Monetary Policy, and the Business Cycle,” Journal of
Money, Credit, and Banking 27 (1995), p.1113-1136.

Christiano, Lawrence, Martin Eichenbaum, and Charles Evans, “Identification and the
Effects of Monetary Policy Shocks” in Financial Factors in Economic Stabilization and
Growth ed. by Mario Blejer et.al., Cambridge Univ. Press (1996), pp.36-74.

 

, “The Effects of Monetary Policy Shocks : Evidence from the Flow of Funds,”
The Review of Economics and Statistics (Feb.l996), pp.16-34.

Cochrane, John, “The Return of the Liquidity Effect : A Study of the Short-Run
Relationship between Money Growth and Interest Rates,” Journal of Business and
Economic Statistics 7 (1989), pp. 75-83.

Croushore, Dean, “What are the Costs of Disinflation?” Business Review (May/June
1992), F RB of Philadelphia, pp.3-16.

Cukierman, Alex, Central Bank Strategy, Credibility, and Independence: Theory and
Evidence, MIT Press, 1992.

 

, Steven B. Webb, and Bilin Neyapti, “Measuring the independence of Central
Banks and Its Effect on Policy Outcomes,” The World Bank Economic Review
(September 1992) pp.353-98

 

,Pantelis Kalaitzidakis, Lawrence H.Summers, and Steven B.Webb, “Central
Bank Independence, Growth, Investment, and Real Rates,” Carnegie-Rochester
Conference Series on Public Policy 39 (autumn 1993), pp. 95-140.

 

, “Commitment through Delegation, Political Influence and Central Bank
Independence” in A Framework for Monetary Stability edited by Wijnholds et al. ,Kluwer
Academic Publishers, 1994.

132

Debelle, G., and S. Fischer, How Independent Should a Central Bank Be, Paper
presented at the Center for Economic Policy Research-FRB of San Francisco Conference,
March 1994.

De Long, j. Bradford, and Lawrence H. Summers, “Macroeconomic Policy and Long-Run
Growth,” F RB of Kansas City, Economic Review (fourth quarter, 1992), pp. 5-30.

Edelberg, Wendy, and David Marshall, “Monetary Policy Shocks and Long-Term Interest
Rates,” F RB of Chicago, Economic Perspectives (1995), pp.2-17.

Eichenbaum, Martin, “Comments on ‘Interpreting the Macroeconomic Time Series Facts
: The Effects of Monetary Policy, by Christopher Sims,” European Economic Review 36
(June, 1992), pp.1001-1011.

Epstein, Gerald, “Monetary Policy in the 19903: Overcoming the Barriers to Equity and
Growth.” In Transforming the US. Financial System edited by G. Epstein (1993) pp. 65-
99.

 

, “A Political Economy Model of Comparative Central Banking.” in New
Perspectives in Monetary Macroeconomics, Univ. of Michigan press, 1993.

Fischer, Stanley, “The role of macroeconomic factors in growth,” Journal of Monetary
Economics 32 (1993), pp.485-512.

 

, “Modern Central Banking” in The Future of Central Banking edited by Capie
et a1., Cambridge Univ., 1994, pp. 262-305.

Friedman, Benjamin and K. Kuttner, “Money, Income, Prices and Interest Rates,”
American Economic Review, Vol.82, No.3, (June, 1992), pp. 472-492.

Fuhrer, Jeffrey, and George Moore, “Monetary Policy Trade-offs and the Correlation
between Nominal Interest Rates and Real Output,” American Economic Review
(Mar.,1995), pp.219-239.

133

Gali, Jordi, “How Well Does the IS-LM Model Fit Postwar US Data?” Quarterly
Journal of Economics (May, 1992), pp.709-738.

Garfinkel, Michelle, and Daniel Thorton, “The Information Content of the Federal Funds
Rats : Is It Unique?” Journal of Money, Credit, and Banking 27 (Aug.,1995), pp.838-847.

Geweke, John, and David Runkle. “A Fine Time for Monetary Policy,” FRB of
Minneapolis, Quarterly Review (Winter, 1995), pp 18-31.

Giannini, Carlo, Topics in Structural VAR Econometrics, Springer-Verlag, 1992.

Goodhart, C.A.E, The Cetral Bank and The Financial System, the MIT Press, 1995.

Gordon, David, and Eric Leeper, “The Dynamic Impact of Monetary Policy : An Exercise
in Tentative Identification,” Journal of Political Economy 102 (1994), pp. 1228-1247.

Gordon, Robert, and S. King, “The Output Cost of Disinflation in Traditional and Vector
Autoregressive Models,” Brookings Papers on Economic Activity, 1982. pp. 205-244.

Grier, Kevin, and G. Tullock, “An Empirical Analysis of Cross-National Economic
Growth, 1951-80,” Journal of Monetary Economics 24 (1989), pp.259-276.

Grilli, Vittorio, Donato Masciandaro, and Guido Tabellini. “Political and Monetary
Institutions and Public Financial Policies in the Industrial Countries,” Economic Policy
13 (October 1991), pp. 341-92.

Hakkio, Craig, and H. Byron, “Costs and Benefits of Reducing Inflation,” Economic
Review (Jan. 1985), FRB of Kansas City, pp.3-15.

Holmes, Frank, A New Approach to Central Banking : The New Zealand Experiment and
Comparison with Australia, the Institute of Policy Studies, 1994.

134

Kydland, Finn, and Edward Prescott. “Rules Rather than Discretion: The Inconsistency of
Optimal Plans,” Journal of Political Economy 85 (June 1977), pp. 473-90.

Leeper, Eric, “Reducing Our Ignorance about Monetary Policy effects,” FRB of Atlanta,
Economic Review (1995), pp.1-38.

Leeper, Eric, and David Gordon, “In Search of the Liquidity Effect,” Journal of Monetary
Economics 29 (1992), pp.341-369.

Levine, Ross, and D. Renelt, “A Sensitivity Analysis of Cross-Country Growth
Regressions,” ,” The American Economic Review82 (Sep, 1992), pp.942-963.

Lucas, Robert, “Some International Evidence on Output-Inflation Tradeoffs,” The
American Economic Review 63 (June 1973), pp.326-334.

McDonough, William, “An Independent Central Bank in a Democratic Country: The
Federal Reserve Experience,” F RB of New York, Quarterly Review (Spring 1994), pp. 1-
6.

Melvin, Michael, “The Vanishing Liquidity Effect of Money on Interest : Analysis and
Implications for Policy,” Economic Inquiry 21 (1983), pp.188-202.

Meyer, Laurence, and R. H. Rasche, “On the Costs and Benefits of Anti-Inflation
Policies,” Review, FRB of St. Louis ( Feb. 1980), pp.3-l4.

Motley, Brian, Growth and Inflation: A Cross Country Study, Center for Economic
Policy Research Publication No.395, 1994.

Pagan, Adrian, and John Robertson, “Resolving the Liquidity Effects,” FRB of St. Louis,
Review (1995), pp.33-54

Pollard, Patricia, “Central Bank Independence and Economic Performance,” Federal
Reserve Bank of St. Louis, Review 75 (July-Aug. 1993), pp.21-36.

135

Posen, Adam, “Central Bank Independence and Disinflationary Credibility: A Missing
Link?,” FRB of New York, Stafl Reports number 1 (May 1995).

Reichenstein, “The Impact of Money on Short-Term Interest Rates,” Economic Inquiry 25
(1987), pp.67-82.

Rogoff, Kenneth, “The Optimal Degree of Commitment to an Intermediate Monetary
Target,” Quarterly Journal of Economics (November 1985), pp. 1169-90.

Romer, Christina, and David Romer, “Does Monetary Policy Matter? : A New Test in the
Spirit of Friedman and Schwartz,” NBER Macroeconomics Annual 1989, pp. 12 1-170.

 

, “ New Evidence on the Monetary Transmission Mechanism,” Brookings
Papers on Economic Activity 1990, pp.149-214.

Sims, Christopher, “Are Forcasting Models Usable for Policy Analysis?,” FRB of
Minneapolis, Quarterly Review (Winter, 1986), pp.2-16.

 

, “Interpreting the Macroeconomic Time Series Facts : The Effects of Monetary
Policy,” European Economic Review 36 (June, 1992), pp.975-1000.

Strongin, Steven, “The Identification of Monetary Policy Disturbances : Explaining the
Liquidity Puzzle,” Journal of Monetary Economics 35 (1995), pp.463-497.

Taylor, John, “The Monetary Transmission Mechanism : An Empirical Framework,”
Journal of Economic Perspectives 9 (Fall, 1995), pp.11-26.

Walsh, Carl, “ Central Bank Independence and the Costs of Disinflation in the EC,”
Working Paper 94-04, FRB of San Francisco, 1994.

 

, “Is There a Cost to Central Bank Independence?” FRB of San Francisco
Weekly Letter, 94-05, Feb. 4, 1994.