Central Bank Behavior and Statutory Independence C. James Hueng Department of Economics

advertisement
Central Bank Behavior and Statutory Independence
C. James Hueng
Department of Economics
5412 Friedmann Hall
Western Michigan University
Kalamazoo, MI 49008-5330
Phone: 0021-269-387-5558 Fax: 0021-269-387-5637
EMail: James.Hueng@wmich.edu
Central Bank Behavior and Statutory Independence
Abstract
This paper integrates the Taylor reaction function literature with the literature on central bank
independence (CBI). The central bank’s policy reaction function describes its behaviors, which
measures the practical CBI, as opposed to the legal CBI measured by CBI indices. By analyzing
the relationship between various legal CBI indices and the central banks’ reactions to inflation for
eighteen OECD countries, we find that the difference of behaviors among central banks is
consistent with the economic measure of independence, which measures how easy it is for the
government to finance its deficits by direct access to credit from the central bank.
Key Words: Central Bank Independence (CBI), Legal CBI index, Policy Reaction Function.
JEL Classifications: E58
i
1. Introduction
The time inconsistency model by Kydland and Prescott (1977) and Barro and Gordon
(1983) shows that governments suffer from an inflationary bias. Insulating monetary policy from
the political pressure helps to reduce this inflation bias. For example, Rogoff (1985) proposes to
delegate monetary policy to a central banker with a more inflation-averse preference than the
government. With some degree of political independence, this “conservative” central banker is
able to enforce the low inflation equilibrium.
Since the establishment of this theoretical
foundation, a bulk of research has been devoted to the links between economic performance and
the degree of central bank independence (CBI).1
Extensive empirical evidence suggests that CBI helps to reduce inflation. This evidence
generally consists of a negative relationship between average inflation rates and the level of CBI
in cross-country regressions, using indices reflecting the degree of CBI. On the other hand,
studies linking CBI and output find that, especially in the industrial countries, a higher level of
CBI does not decrease output growth. As a result, it seems that CBI increases the likelihood of
achieving a lower inflation at no real cost.
The fact that many OECD countries have
implemented institutional reforms granting their central banks more independence over the last
two decades is argued to be triggered by the theoretical support as well as the empirical evidence.
However, this view has been subject to both theoretical and empirical criticism. Two
major critics are presented in the literature. First, it is argued that CBI is an endogenously
determined variable, not an exogenous variable that causes low inflation. Posen (1993), Forder
(1996), and Hayo (1998) argue that CBI is determined by social attitudes. Countries with a
commitment to price stability may have a greater propensity for independent central banks.
1
For a detailed survey of this literature, see Berger, de Haan, and Eijffinger (2001) and Hayo and
Hefeker (2002).
1
Independent central banks are successful in implementing low and stable inflation simply because
their independence reflects an anti-inflation social attitude. For example, Posen (1995) finds that
CBI does not help explain inflation if his measure of the strength of financial sector opposition
toward inflation is included as an explanatory variable. Daunfeldt and de Luna (2008) compare
the implementation dates of central bank independence reform with the long-term inflation trends
for 29 OECD-countries and find that price stability seems to have been achieved in most
countries before the central banks are granted more independence in the institutional reforms.
The second major critic on the CBI-to-inflation causality is that CBI is not necessary for
price stability. Other monetary policy designs such as fixed exchange rate, inflation targeting,
and central bank contracts may contribute to low inflation rates as well. On the other hand,
inflation, and other real measure of economic performance, can be influenced by many factors
other than CBI. For example, even if an independent central bank implements a low inflation
policy, governments still have an incentive to create unexpected inflation through an
expansionary fiscal policy. Therefore, studies such as Cukierman (1992), Posen (1995), Campillo
and Miron (1997), and Forder (1998) try to control for other political and economic variables that
are believed to be causally linked to inflation and find that the CBI-inflation relationship is not
robust with regard to various control variables.
This literature so far has been focusing on the link between CBI indices and economic
performances. Studies either investigate whether CBI measured by a CBI index leads to certain
economic performances, or evaluate various indices of CBI as good measures of practical CBI
based on their effects on economic performances.
However, it is a distant link between
institutional designs and economic performances. We argue that it is this distant link that causes
the critics mentioned above, which make the test of the relationship between CBI indices and
economic performances problematic. As Forder (1996) says, "There is no theory that says it
matters what the rules say. There is only a theory that says it matters what the behavior is."
2
Therefore, "what the rule says" would matter only if it is consistent with "what the behavior is."
The purpose of this paper is to compare what central banks actually do as opposed to what they
are legislated to do. That is, we focus on a more direct link – the link between CBI indices and
the central bank behaviors – and analyze which aspects of the legal CBI are more consistent with
the central bank’s behaviors.
A central bank's behavior is described by its reaction function, which measures how the
monetary policy instrument reacts to the information available to the central bank about the state
of the economy. Most central bank reaction functions can be reasonably described by a simple
function based on the Taylor rule. The Taylor-type reaction function specifies that the policy
interest rate responds to the inflation and output gaps.
In addition to integrating the Taylor reaction function literature with the literature on CBI,
this paper contributes to the research on the relationship between legal and practical CBI. CBI is
a theoretical concept based on the unobservable loss function of the delegated central banker,
while most of the CBI indices are legal indicators constructed by the reading of statutes.
Regulations on the functions of the central banks cannot completely specify the limits of authority
between the central banks and the political authorities under all contingencies. Therefore, a legal
measure of independence may not reflect a central bank’s de facto level of independence. Several
studies have tried to derive the “practical CBI” as opposed to the legal CBI. For example,
Cukierman (1992), Cukierman et al. (1992) and de Haan and Kooi (2000) use the actual average
term of office of central bank governors to proxy for the practical CBI. de Haan et al. (2003) use
a latent variable approach to derive the common unobservable factor from a list of legal CBI
indices and use it as a measure of the practical CBI. We argue that central banks’ behaviors
reflect the practical CBI and therefore, the central bank reaction function measures the practical
CBI. If the legal CBI is consistent with the practical CBI, we would expect measures of CBI to
3
be positively associated with the size and the speed of the central banks’ reaction to inflation
shocks.
Using data from eighteen OECD countries during the 1980’s, we find that certain aspects
of the legislated differences in CBI generate variations in interest rate setting behaviors of the
central banks.
Specifically, limitations on the ability of the central bank to lend to the
government affect central bank’s reaction to inflation. Differences of legislation on the term of
office, appointment, and dismissal of central bank governors, on the other hand, have no
significant effect on central banks’ different reactions to inflation.
The remainder of the paper is organized as follows. Section 2 estimates the policy
reaction functions for eighteen OECD countries. Section 3 discusses various existing CBI indices
and their relationships with the estimated policy reactions. Section 4 concludes the paper.
2.
Policy Reaction Functions
The monetary policy reaction function, which measures how the monetary policy
instruments react to the information available to the central bank about the state of the economy,
adequately represents the central bank’s behaviors. Taylor (1993) postulates a simple interestrate feedback rule, now known as the Taylor rule, to be followed by central banks:
it = r * + π + λ (π t − π T ) + α y yt ,
(1)
where r* is the equilibrium real interest rate, π t is the annual inflation rate (percentage change of
the price level from a year ago to present), π T is the target level of π t , and yt is the percentage
deviation of real output from its potential level. The Taylor rule states that policy responds only
to the current values of two target variables, inflation and output gap. Taylor argues that equal
weights of 0.5 on reactions to inflation and output ( λ = α y = 0.5 ), 2% equilibrium real interest
rate, and 2% inflation target describe U.S. monetary policy very well in the late 1980s and the
4
early 1990s. The Taylor-type reaction function can be specified as the Taylor rule with general
coefficients plus a policy shock:
it = α + απ π t + α y yt + ν t ,
(2)
where α = r* + (1 − απ )π T , απ = 1 + λ , and ν t is the policy shock.
Studies have shown that this Taylor-type rule and its various approximate forms are
optimal for a central bank that has a quadratic loss function over inflation and output [e.g.,
Rudebusch and Svensson (1999)]. The coefficients in the rule are linked to the underlying
structure of the economy and the weight the central bank places on inflation versus output
deviations.
Taylor (1999) shows that this monetary policy rule is an implication of many
different monetary systems. Even for open economies, Taylor (2001) argues that an interest-rate
reaction to the exchange rate in the rule is not necessary because the rule implies an indirect
interest-rate reaction to the exchange rate through the effect of the exchange rate on inflation and
output. Therefore, this type of simple feedback rule has been used widely as an account of actual
policy in other countries as well.
It is argued that for financial market stability, for fear of loss of credibility from sudden
large policy reversals, and for the need for consensus building to support a policy change, central
banks usually avoid large changes in interest rates and adjust interest rates in a gradual fashion.
Studies such as Judd and Rudebusch (1998) and Woodford (2001) incorporate interest-rate inertia
and estimate the policy reaction function in the context of an error correction model. In particular,
the actual interest rate is replaced by its target level set by the central bank:
itT = α + απ π t + α y yt ,
(3)
T
where it is the policy target rate that is achieved through gradual adjustment. Clarida et al.
T
(1998) specify the dynamics of adjustment of the actual interest rate to it as:
it = γ itT + (1 − γ )it −1 + ν t .
5
(4)
That is, the actual interest rate is a weighted average of the target and the past interest rate. If
γ = 1 , (3) becomes the original Taylor reaction function. If γ = 0 , the interest rate is a random
walk process and does not respond to changes in inflation and output gap.
Judd and Rudebusch (1998) use a richer dynamic and suggest that the change in the
interest rate partially corrects the error between last period’s setting and the target level, as well
as maintaining some of the momentum from the changes in the last period:
∆it = γ (itT − it −1 ) + ρ∆it −1 + ν t .
(5)
It is clear that (4) is a special case of (5) when ρ = 0. That is, Clarida et al. (1998) assume that
the interest rate does not depend on its level two periods before.
Substituting (5) into (3) yields the equation to be estimated:
∆it = α + γαπ π t + γα y yt − γ it −1 + ρ∆it −1 + ν t ,
(6)
where α = γ r* + γ (1 − απ )π T . If ρ = 0 and γ = 1 , then there is no interest-rate inertia and the
original Taylor reaction function (2) applies. Even though the equilibrium real interest rate r * and
the inflation target π T cannot be separately identified, the parameters usually of particular interest,
i.e., the weights on inflation ( απ ) and output ( α y ), can be estimated by nonlinear least squares.
Policy reaction functions like (2) and (6) are “backward looking” and can be theoretically
derived from an economic model where the objective function is quadratic and the expectation is
adaptive [e.g., Rudebusch and Svensson (1999)]. In accordance with considerations of rational
expectations in economic behaviors, studies such as Clarida et al. (1998, 2000) consider a
forward looking version of the backward looking reaction (6):
∆it = α + γαπ E t [π t + j ] + γα y E t [ yt + k ] − γ it −1 + ρ∆it −1 + ν t ,
(7)
where Et is the expectation based on the information set at time t. That is, the central bank’s
target rate depends on j-period ahead expected inflation and k-period ahead expected output.
6
Since there is no consensus on whether the central banks are backward looking or forward
looking in the 1980's, we estimate both the backward looking and the forward looking reaction
functions to show the robustness of our results.
The empirical work involves time series regressions for each country under consideration
and a cross-country comparison of the policy reaction functions and CBI. Both parts need
sufficient sample data to show convincible results. Therefore, this paper focuses on OECD
countries because of the data availability, as well as their wide coverage in the CBI and the
monetary policy literatures. Monthly data are used in the time series regressions.2
The sample covers the period from January 1980 to December 1989, a total of 120
observations for each country. We only focus on this period because first, we wish to see
whether the monetary policy reforms during the 1990s triggered by proponents of CBI can be
justified. Therefore, we cover the period that is prior to the reforms and the transition to the
reforms. Second, in theory, under a fixed exchange rate system monetary policy cannot be set
exogenously. Therefore, we avoid the Bretton-Woods era. Extensions to earlier periods would
also reduce the data availability. Furthermore, inflation targeting and incentive contract for
central bankers, which are other alternative monetary policy designs, have not yet been adopted
during this period. Finally, there is a general agreement in the literature that CBI is fixed during
this period for most of the countries.
2
We are aware of the literature on the policy reaction function using real time data [e.g.,
Orphanides (2003)]. However, the availability of real time data is limited to a few countries.
Without the data we can only assume that the biases caused by the revised data are similar across
these OECD countries. This is also a reason why we rely on the monthly data because the CPI
and industrial production data are not revised as often as the GDP data.
7
If it is available, the interbank loan rate is used as the policy rate. If this rate is not
available, the central bank discount rate is used.3 The inflation rate is the percentage change in
CPI from the same month a year ago. The industrial production index is used to measure output
and the Hodrick-Prescott filter is used to derive the potential output.4 It turns out that eighteen of
the OECD countries have complete data to be used in our analysis.
Panel (A) of Table 1 shows the results for the backward looking model (6) from the
nonlinear least squares estimations. Panel (B) shows the results for the forward looking model
(7), using a 12-month forecasting horizon for both inflation and output targets. We use the
Instrumental Variable estimations with lags of the right-hand-side variables as the instruments for
model (7).5 First of all, the estimates of γ are all well below one, indicating that all these central
banks adjust their interest rates gradually, with a high degree of inertia. However, the estimates
of γ are all significantly different from zero at the 5% level, indicating that an interest rate target
exists and it reacts to variables other than itself. The dynamics of the adjustment vary across
countries: ρ̂ is statistically significant in some countries and insignificant in the others.
Our focus is on the policy reaction to inflation, which is positive and significantly
different from zero at the 5% level in all countries except for Norway in Panel (A) and Greece in
Panel (B). In general, policy reaction to inflation is stronger in the forward looking model than in
3
The discount rate is used for Austria, Finland, Greece, Ireland, Italy, and Portugal.
4
We also use a detrended log industrial production with a quadratic trend to measure the output
gap. The estimation results are slightly different, but the general conclusions do not change. The
results are available from the author upon request.
5
We also run regressions with the change of log real effective exchange rate added. Consistent
with Taylor’s (2001) argument, the interest rate reaction to the change in real exchange rate is
only statistically significant at the 5% level in Denmark.
8
the backward looking model. Only a few countries follow the Taylor principle ( αˆπ > 1 ) in the
1980’s in the forward looking model.
On the other hand, the policy reactions to output gap are mostly statistically insignificant,
consistent with the findings in the previous studies. The only exceptions are Canada and the
United States, both of which also have a relatively high reaction to inflation.
3.
CBI Indices and Policy Reactions
3.1 Legal CBI indices
In the most comprehensive study of central bank institutional arrangements, Cukierman
(1992) and Cukierman, Webb, and Neyapti (1992) create an index of CBI based on differences in
central bank laws for a large sample of countries. Only the written information from the charters
in various legal dimensions is used, with relatively few subjective judgments. Their weightedaverage index (CWN) is aggregated from sixteen legal characteristics of central bank charters,
which are grouped into four clusters of issues [the followings are directly quoted from Table 1 in
Cukierman et al. (1992)]:
(1) The term of office, appointment, and dismissal of the chief executive officer (CEO) of the
central bank.
(2) The central bank’s authority to formulate and finalize monetary policy.
(3) The legal mandate of the central bank to pursue the objective of price stability.
(4) Limitations on the ability of the central bank to lend to the government. This cluster
encompasses five more detailed variables: (4.1) limitation on nonsecuritized lending;
(4.2) limitation on securitized lending; (4.3) who determines the terms of lending
(maturity, interest, and amount); (4.4) potential borrowers from the CB; and (4.5) type of
limits (maturity, interest, amount, and whether the CB is prohibited from the primary
market of the government securities).
9
Eight legal variables are constructed from these four clusters of characteristics (the first three
clusters plus five variables in the fourth cluster of characteristics) and then aggregated into the
weighted-average index CWN.
The first column of the top panel in Table 2 lists this index for those eighteen OECD
countries. The bottom panel shows the Spearman's rank correlation coefficient (Spearman's ρ )
between the CBI index and the estimated policy reaction to inflation ( αˆπ ), along with the P-value.
The Spearman's rank correlation coefficient is a non-parametric measure of the statistical
dependence between the ranks of two variables. If the rank of CBI is the same as the rank of
αˆπ across countries, the Spearman's ρ is equal to one. Only the rank of the variables matters and
the magnitudes of the differences between observations do not affect the Spearman's ρ . For
comparison, we also list the simple correlation coefficients, along with the P-values, under the
Spearman's statistics. Since the conclusions drawn from these two measures of correlations are
basically the same, we only discuss the Spearman's statistics in the following analysis.
The Spearman's ρ between απ and CWN is 0.229 for the backward looking model and
0.186 for the forward looking model. Therefore, the most popular CBI index constructed by
Cukierman (1992) and Cukierman et al. (1992) has a positive relationship with the central bank
behavior. This relationship, however, is highly insignificant (P-value is 0.381 for the backward
looking model and 0.478 for the forward looking model).
Are there certain characteristics of the central banks' legal status that are more consistent
with their behaviors than the others? To answer this question, we list the scores of those four
clusters of characteristics that are used to construct CWN in the second to the fifth columns of
Table 2 (denoted as CWN1, CWN2, CWN3, and CWN4), and calculate their correlations with the
policy reaction to inflation. Clearly, the characteristics in the first and the third clusters are not
consistent with the central bank behaviors. Their correlations with the policy reaction to inflation
10
are negative and highly insignificant. That is, the term, appointment, and dismissal of the central
bankers, and the monetary stability objective in law do not have significant influence on the
different behaviors.
On the other hand, the second and the fourth clusters of characteristics are more
consistent with the behaviors of the central banks. Their Spearman's correlations with the policy
reaction to inflation are positive and statistically significant at the 5.5% and 13.9% levels,
respectively, for the backward looking model and at the 13.6% and 9.8% levels, respectively, for
the forward looking model. That is, relatively, the central bank's authority to formulate and
finalize monetary policy and the limitations on central bank lending to the government are more
consistent with the differences in central bank behaviors.
The above result, however, is only marginally significant. To check the robustness of
this finding, we consider another measure of CBI developed by Grilli, Masciandaro and Tabellini
(1991). Their index (GMT) contains some characteristics similar to those in CWN. Therefore, it
provides comparable categories to CWN, but from different authors’ subjective readings of law
and with a different weighting scheme.6
GMT distinguishes the political independence from the economic independence. The
political measure of independence (GMT_P), with a range from zero to eight, sums up the scores
from the following eight categories [directly quoted from Grilli, Masciandaro and Tabellini’s
(1991) Table 12]: (1) Governor not appointed by government; (2) Governor appointed for more
than five years; (3) All the Board not appointed by government; (4) Board appointed for more
than five years; (5) No mandatory participation of government representative in the Board; (6) No
government approval of monetary policy formulation is required; (7) Statutory requirements that
central bank pursues monetary stability amongst its goals; (8) Legal provisions that strengthen the
6
Segolato et al. (2007) provide a detailed comparison between CWN and GMT.
11
central bank's position in conflicts with the government are present. The central bank scores one
if the answer to the category is yes and zero if the answer is no.
It is clear that categories (1), (2), and (5) are essentially CWN’s first cluster of
characteristics [(3) and (4) are about the terms and the appointment of the board, which is not
considered in CWN]. Therefore we integrate these three categories and call it GMT_P1, which is
comparable to CWN1.
Categories (6) and (8) are similar to CWN’s second cluster of
characteristics. Therefore we sum up these two scores and call it GMT_P2, which is comparable
to CWN2. Category (7) alone (GMT_P3) is comparable to CWN’s third cluster of characteristics
(CWN3).
The economic measure of independence in Grilli et al. (1991) is defined as the ability of
the central bank to use monetary policy instruments without restrictions. This index (GMT_E),
with a range from zero to seven, sums up the scores from the following seven categories [directly
quoted from Grilli, Masciandaro and Tabellini’s (1991) Table 13]: (1) Direct credit facility: not
automatic; (2) Direct credit facility: market interest rate; (3) Direct credit facility: temporary; (4)
Direct credit facility: limited amount; (5) Central bank does not participate in primary market for
public debt; (6) Discount rate set by central bank; (7) Banking supervision not entrusted to the
central bank or not entrusted to the central bank alone.7 Clearly, categories (1) – (5) essentially
measure how easy it is for the government to finance its deficits by direct access to credit from
the central bank, equivalent to the fourth cluster of characteristics in CWN. Therefore, we sum
up the scores of these five categories and call it GMT_E4, which is comparable to CWN4.
7
Grilli et al. (1991) combine the political and economic independence indices to serve as an
indicator for legal independence.
12
These decomposed GMT indices are listed in the sixth to the eleventh columns of Table
2.8 The economic independence index (GMT_E) has a positive and highly significant Spearman's
correlation coefficient with the central bank's reaction to inflation (0.693 for the backward
looking model and 0.612 for the forward looking model, with both P-values less than 1%).
Therefore, the ability of the central bank to use monetary policy instruments without restrictions
shapes the central bank's behaviors.
In particular, the part of GMT_E that specifies the
limitations on the ability of the central bank to lend to the government, GMT_E4, is also positive
and highly significant (0.503 for the backward looking model and 0.491 for the forward looking
model, with both P-values less than 5%). This is consistent with our earlier finding with the
CWN4 index.
On the other hand, the political independence (GMT_P) is relatively unimportant. The
decompositions of the political independence (GNT_P1, GMT_P2, and GMT_P3) show little
variation across the countries and therefore, the correlation results may not be very reliable.
GMT_P1, which measures the terms of office and the appointment of the central banker and is
similar to CWN’s first cluster of characteristics, has relatively more variation. Consistent with
the CWN1 result, it does not affect the central banker’s behaviors.
3.2 Other measures of practical CBI
This subsection compares our measure of practical CBI – the central bank’s interest rate
reaction to inflation – to two other measures of practical CBI in the literature. The first one is the
Turnover index in Cukierman (1992), Cukierman et al. (1992) and de Haan and Kooi (2000).
This index uses the actual average term of office of central bank governors to proxy for the
practical CBI. It is based on the presumption that more rapid turnover, i.e., a shorter term of
8
Data for Finland, Norway, and Sweden, which are not available in GMT, are constructed by
Segolato et al. (2007).
13
office, of central bank governors indicates a lower level of independence. Therefore, it is related
to CWN’s first cluster of characteristics and GMT_P1 (the term of office, appointment, and
dismissal of the governors). We report this index in Table 2, denoted as “Turnover”, and show its
correlation with the policy reaction to inflation. Apparently, this index does not reflect the central
banks’ different reactions to inflation. This result implies that not only the term of office,
appointment, and dismissal of the central bank governors specified in laws, but also the actual
turnover rate of the governors, is not important in shaping the central bank’s behaviors.
Another attempt to measure the practical CBI is the Factor Score constructed by de Haan
et al. (2003). They argue that the practical CBI is an unobservable latent variable embedded in
measures of legal CBI. They use factor analysis techniques to retrieve the common unobservable
factor from a list of legal CBI indices.9 Optimal weighting of these indices is derived and used to
obtain the best approximation of unobservable CBI. This index is denoted as “Latent” and shown
in the last column of Table 2. Interestingly, the Spearman's correlation coefficient between this
latent indicator and the interest reaction to inflation is above 0.5 and is significantly different
from zero at the 3% level in both models. Therefore, this latent indicator does a good job in
picking up the central bank's anti-inflation behavior.
9
de Haan et al. (2003) include CUK (the equally-weighted measure of CWN), CWN, GMT, the
index by Alesina (1988), and the index by Eijffinger and Schaling (1992). Before deriving the
common unobservable factor, they adjust the indicators CUK, CWN and GMT such that the
issues related to the objective of monetary policy is removed. They do this because they are only
interested in independence proper and want to remove the conservativeness of central banks
embodied in law. However, as shown earlier in this paper, the price stability objective is not
important in affecting CB’s behaviors. Therefore, we do not expect that the result would be
different if the objective of monetary policy is not removed.
14
3.3 Robustness check: adjusting the estimation errors
In the above analysis, we run a time-series regression of the Taylor reaction function to
estimate the interest rate response to inflation for each of the countries considered. Then these
estimated coefficients are compared to the CBI indices in a cross-sectional analysis. One may
argue that the estimation errors from the time series regression may affect the results in the crosssectional analysis. To deal with this concern, we use an Estimated Dependent Variable (EDV)
model to address the estimation error problem. An asymptotically efficient feasible generalized
least squares estimator proposed by Hanushek (1974) is used to access the relationship between
central banks’ reaction to inflation and CBI.
Specifically, let β be the vector of the interest rate responses to inflation from those T
countries (T = 18 in our case): β = (απ 1 απ 2 ... απ T )′ ; and βˆ = (αˆπ 1 αˆπ 2 ... αˆπ T )′ be the
estimated value from the time series regressions in the previous section. The model of interest is
β = Xγ + v , where X is a Tx1 vector of CBI index. However, β is not observable. Rather we
have β
β̂ , where βˆ = β + u , u is a vector of sampling error ui (i = 1, 2, . . ., T) in the regression
estimates from the time series regression, and E(ui) = 0. Since the time series estimations use
data from different countries, it is not unreasonable to assume that ui's are heteroscedastic [V(ui) =
ωι2 ] and mutually uncorrelated [E(uiuj) = 0 for i≠j].
The feasible regression model
becomes βˆ = X γ + ε , where ε = u + v , and u and v are independent. Assume that E( vv′ ) = σ2I.
Then
 σ 2 + ω12
0 

⋱
V( ε ) = 
.
 0
2
2
σ + ωT 

Since we have estimates of ωι2 's from the estimated coefficients in the time series regressions, we
only need an estimate of σ2 in order to apply feasible generalized least squares to obtain
15
asymptotically efficient estimate of γ .
σˆ =
2
∑ εˆ −∑ ωˆ
2
i i
2
i
i
Hanushek (1974) shows that σ2 can be estimated by
−1
+ tr ( X ′X ) X ′GX 

 , where G is the estimated V(u).
T −K
Table 3 shows the estimated coefficient of αˆπ on each CBI index. The results confirm
the conclusions in Section 3.2: significant relationships exist between central bank's behaviors
and GMT's economic independence index (GMT_E), GMT's measure the government's ability to
finance its deficits through the central bank (GMT_E4), and the latent indicator proposed by de
Haan et al. (2003).
4.
Conclusion and Discussion
Do central banks react to economic shocks according to what they are legislated to do?
We try to measure the central bank’s response to economic situations as a means of providing
information about practical central bank independence.
A central bank is thought to be
“independent” if it reacts to an unexpected increase in inflation by increasing nominal interest
rates. Using data from eighteen OECD countries during the 1980’s, we find that certain aspects
of the legislated differences in CBI generate variations in interest rate setting behavior of the
central banks.
Specifically, limitations on the ability of the central bank to lend to the
government affect central bank’s reaction to inflation. That is, if it is easy for the government to
finance its deficits by direct access to credit from the central bank, the deficits would be
inflationary and the central bank’s ability to fight inflation is retarded.
On the other hand, differences of legislation on the term of office, appointment, and
dismissal of central bank governors have no significant effect on central banks’ different
reactions to inflation. However, one has to be cautious in generalizing this argument to other
countries.
It is possible that this is only true in these developed countries because the
government’s control on the central bankers is monitored by the democratic process, even if the
16
term of office and the government’s power of appointing and dismissing the central banker are
specified in law. Due to the data limitation, we leave this possibility for future research.
Another possible drawback of this paper is on the estimation of the policy reaction
function. We only use two representative models to estimate the central banks' reaction functions.
In a cross-sectional analysis, it is up for debate whether all the central banks have the same policy
reaction function and the same degree of forward looking. In addition, policy reaction functions
are known to be temporally unstable. However, again, the limitation on data availability does not
allow us to consider the dynamic of the reaction.
Nonetheless, this paper makes the first attempt to integrate the literature on Taylor
reaction function and the literature on CBI. We provide a new direction for future studies on
evaluating the central bank independence. Future studies should try to conquer the limitations
mentioned above.
17
Reference
Alesina, A., 1988. Macroeconomics and politics. NBER Macroeconomic Annual. MIT Press,
Cambridge, MA, pp. 13 –52.
Barro, R.J., Gordon, D., 1983. Rules, discretion, and reputation in a positive model of monetary
policy. Journal of Monetary Economics 12, 101 –121.
Berger, H., de Haan, J., Eijffinger, S.C.W., 2001. Central bank independence: an update of
theory and evidence. Journal of Economic Surveys 15, 3 –40.
Campillo, M., Miron, J.A., 1997. Why does inflation differ across countries? in: Romer, C.D.,
Romer, D.H. (Eds.), Reducing Inflation: Motivation and Strategy. University of Chicago
Press, Chicago, pp. 335 – 357.
Clarida, R., Gali, J., and Gertler, M., 1998.
Monetary Policy Rules in Practice: Some
International Evidence. European Economic Review, 42, pp. 1033-1067.
Clarida, R., Gali, J., and Gertler, M., 2000. Monetary Policy Rules and Macroeconomic Stability:
Evidence and Some Theory. Quarterly Journal of Economics, 115(1), pp.147-180.
Cukierman, A., 1992.
Central Bank Strategy, Credibility, and Independence.
MIT Press,
Cambridge, MA.
Cukierman, A., Webb, S. B. and Neyapti, B., 1992. Measuring the independence of central banks
and its effects on policy outcomes. The World Bank Economic Review, 6, 353-398.
Daunfeldt, S., and de Luna, X., 2008. Central bank independence and price stability: evidence
from OECD-countries. Oxford Economic Papers, 60 (3), 410-422
de Haan, J. and Kooi. W., 2000. Does central bank independence really matter?: New evidence
for developing countries using a new indicator. Journal of Banking and Finance, 24 (4),
643-664.
18
de Haan, J., Leertouwer, E., Meijer, E. and Wansbeek, T., 2003.
Measuring Central Bank
Independence: A Latent Variables Approach. Scottish Journal of Political Economy, 50 (3),
326-340.
Eijffinger, S. C. W. and Schaling, E., 1992. Central Bank Independence: Criteria and Indices,
Research Memorandum No. 548. Department of Economics, Tilburg University, processed.
Forder, J. 1996. On the measurement and assessment of ‘institutional’ remedies. Oxford
Economic Papers 48, 39–51.
Forder, J. 1998. The case for an independent central bank: a reassessment of evidence and
sources. European Journal of Political Economy, 14, 53-72.
Grilli, V., Masciandaro, D., Tabellini, G., 1991. Institutions and policies. Economic Policy 6,
341 - 392.
Hanushek, Eric A., 1974. Efficient Estimators for Regressing Regression Coefficients. The
American Statistician, 28(2), 66-67.
Hayo, B., 1998. Inflation culture, central bank independence and price stability.
European
Journal of Political Economy, 14, 241 –263.
Hayo, B. and C. Hefeker, 2002. Reconsidering central bank independence. European Journal of
Political Economy, 18, 653–676.
Judd, J. P. and G. D. Rudebusch, 1998. Taylor's rule and the Fed, 1970-1997. Economic Review,
Federal Reserve Bank of San Francisco, 3-16.
Kydland, F.W., Prescott, E.C., 1977. Rules rather than discretion: the inconsistency of the
optimal plans. Journal of Political Economy 85, 473 –491.
Orphanides, A., 2003. Historical Monetary Policy Analysis and the Taylor Rule. Journal of
Monetary Economics, 50 (5), pp. 983–1022.
19
Posen, A.S., 1993. Why central bank independence does not cause low inflation. In: O’Brian, R.
(Ed.), Finance and the International Economy, vol. 7. The Amex Bank Review Prize
Essays. Oxford Univ. Press, Oxford, pp. 40 –65.
Posen, A., 1995.
Declarations are not enough: financial sector sources of central bank
independence. NBER Macroeconomic Annual 1995, Cambridge, MA.
Rogoff, K., 1985. The optimal degree of commitment to an intermediate monetary target.
Quarterly Journal of Economics 100, 1169 –1190.
Rudebusch, G. D. and Svensson, L.E.O. 1999. Policy Rule for Inflation Targeting, in John
Taylor (ed.), Monetary Policy Rules, Chicago, IL: The University of Chicago Press.
Segolato, J.F., M. Sommer, M. Arnone and B. Laurens 2007. Central bank independence: lessons
from global trends. IMF working paper 07/88 (2007).
Taylor, J., 1993. Discretion versus policy rules in practice. Carnegie_Rochester conference
series on public policy 39: 195-214.
Taylor, J. B., 1999. An Historical Analysis of Monetary Policy Rules, in John Taylor (ed.),
Monetary Policy Rules, Chicago, IL: The University of Chicago Press.
Taylor, J. B. 2001. The Role of the Exchange Rate in Monetary Policy Rules. American
Economic Review, 91 (2), 263-267..
Woodford, M. 2001. The Taylor Rule and Optimal Monetary Policy. American Economic
Association Papers and Proceedings, 91 (2), pp. 232-237
20
Table 1: Estimation of the Policy Reaction Functions
(A) Backward Looking Model
(B) Forward Looking Model
Country
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Greece
Ireland
Italy
Japan
Netherlands
Norway
Portugal
Spain
Sweden
UK
US
απ
αy
γ
ρ
απ
αy
γ
ρ
0.586
0.734
0.066
0.048
0.927
0.833
0.064
0.026
(0.004)
(0.090)
(0.034)
(0.578)
(0.006)
(0.069)
(0.037)
(0.769)
0.773
0.269
0.357
-0.020
0.931
0.277
0.454
0.009
(0.000)
(0.152)
(0.000)
(0.819)
(0.000)
(0.059)
(0.000)
(0.919)
0.911
0.649
0.409
0.047
0.909
0.794
0.254
-0.034
(0.000)
(0.003)
(0.000)
(0.591)
(0.000)
(0.037)
(0.000)
(0.700)
0.990
0.031
0.410
0.100
1.126
0.035
0.302
0.081
(0.000)
(0.733)
(0.000)
(0.251)
(0.000)
(0.787)
(0.000)
(0.379)
0.324
-0.038
0.082
0.011
0.581
-0.047
0.050
-0.013
(0.001)
(0.734)
(0.005)
(0.898)
(0.034)
(0.793)
(0.027)
(0.881)
0.641
0.490
0.147
0.375
0.738
0.926
0.098
0.336
(0.000)
(0.219)
(0.000)
(0.000)
(0.000)
(0.130)
(0.000)
(0.000)
0.966
-0.023
0.076
0.334
1.322
0.081
0.082
0.293
(0.000)
(0.943)
(0.041)
(0.000)
(0.000)
(0.787)
(0.013)
(0.001)
0.188
-0.070
0.092
-0.012
0.316
-0.136
0.049
-0.025
(0.003)
(0.394)
(0.003)
(0.892)
(0.204)
(0.417)
(0.073)
(0.787)
0.413
0.193
0.148
0.181
0.476
0.336
0.109
0.168
(0.000)
(0.242)
(0.000)
(0.043)
(0.000)
(0.148)
(0.001)
(0.062)
0.505
-0.233
0.068
0.035
0.650
-0.364
0.061
0.018
(0.000)
(0.444)
(0.004)
(0.631)
(0.000)
(0.284)
(0.004)
(0.810)
0.799
0.457
0.116
0.371
1.488
0.992
0.060
0.303
(0.000)
(0.129)
(0.001)
(0.000)
(0.003)
(0.091)
(0.003)
(0.000)
0.435
0.055
0.147
-0.198
0.655
0.016
0.166
-0.210
(0.014)
(0.721)
(0.001)
(0.008)
(0.002)
(0.904)
(0.000)
(0.004)
0.242
0.115
0.110
0.014
0.376
0.091
0.139
0.023
(0.168)
(0.331)
(0.012)
(0.879)
(0.012)
(0.312)
(0.003)
(0.798)
0.741
0.315
0.076
-0.045
1.428
0.784
0.040
-0.063
(0.000)
(0.380)
(0.002)
(0.445)
(0.004)
(0.256)
(0.008)
(0.263)
0.433
0.251
0.246
0.060
0.544
0.225
0.264
0.068
(0.036)
(0.576)
(0.000)
(0.502)
(0.020)
(0.590)
(0.000)
(0.449)
0.412
0.051
0.126
0.040
0.780
0.183
0.079
0.026
(0.000)
(0.638)
(0.001)
(0.655)
(0.018)
(0.324)
(0.010)
(0.775)
0.393
0.284
0.312
-0.065
0.738
0.614
0.254
-0.102
(0.000)
(0.294)
(0.000)
(0.477)
(0.004)
(0.092)
(0.000)
(0.259)
0.756
0.846
0.220
0.389
1.045
1.384
0.128
0.332
(0.000)
(0.002)
(0.000)
(0.000)
(0.002)
(0.005)
(0.000)
(0.000)
The backward looking model is (6): ∆it = α + γαπ π t + γα y yt − γ it −1 + ρ∆it −1 + ν t . The forward
looking model is (7): ∆it = α + γαπ E t [π t + j ] + γα y E t [ yt + k ] − γ it −1 + ρ∆it −1 + ν t , with j=k=12.
The numbers in parentheses are P-values. A P-value of 0.000 indicates a significance level lower
than 0.05%.
21
Table 2: CBI Indices and Their Correlation with Policy Reaction to Inflation
Country\Index
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Greece
Ireland
Italy
Japan
Netherlands
Norway
Portugal
Spain
Sweden
UK
US
Backward Spearman's
P-value
looking
Spearman's
Forward
looking
P-value
Backward
looking
Forward
looking
ρ
ρ
Corr. with αˆπ
P-value
Corr. with αˆπ
CWN
0.61
0.17
0.45
0.50
0.28
0.24
0.69
0.55
0.44
0.25
0.18
0.42
0.17
0.41
0.23
0.29
0.27
0.48
0.229
0.381
0.186
0.478
0.307
0.216
0.218
CWN1 CWN2
0.58
0.55
0.25
0.05
0.83
0.18
0.08
0.67
0.33
0.00
0.44
0.47
0.69
0.67
0.54
0.38
0.77
0.00
0.60
0.00
0.52
0.17
0.48
0.18
0.52
0.18
0.31
0.38
0.31
0.08
0.50
0.00
0.58
0.00
0.44
0.13
-0.134 0.470
0.611 0.055
-0.199 0.377
0.447 0.136
-0.118 0.499
0.641
-0.164
0.035
0.418
CWN3 CWN4
0.60
0.65
0.00
0.20
0.20
0.45
0.60
0.58
0.80
0.17
0.00
0.17
1.00
0.61
0.80
0.54
0.80
0.25
0.20
0.20
0.00
0.09
0.80
0.35
0.00
0.08
0.00
0.58
0.60
0.13
0.20
0.33
0.20
0.24
0.40
0.62
-0.122 0.374
0.645 0.139
-0.307 0.413
0.234 0.098
-0.134 0.425
0.597
-0.232
0.079
0.391
GMT_P
3
1
4
3
2.4
2
6
2
3
4
1
6
2
1
2
1.5
1
5
0.274
0.291
-0.016
0.954
0.260
0.298
0.036
GMT_E GMT_P1 GMT_P2 GMT_P3
6
0
2
1
6
0
0
0
7
2
1
1
5
1
1
1
0
1.2
0
1.2
5
1
0
0
7
2
2
1
2
0
1
0
4
1
1
1
1
3
0
0
5
0
0
1
4
2
2
1
1
2
0
0
2
1
0
0
3
1
0
0
2
1.5
0
0
5
1
0
0
7
1
2
1
0.693
-0.076
0.352
0.405
0.001
0.774
0.166
0.107
0.612
-0.180
0.236
0.335
0.008
0.495
0.366
0.189
0.716
-0.025
0.277
0.393
0.001
0.505
0.922
-0.173
0.265
0.144
0.107
0.312
GMT_E4
3
3
4
2
0
2
5
1
3
1
3
3
1
1
2
2
4
5
0.503
0.038
0.491
0.043
0.534
Turnover
0.10
0.20
0.10
0.00
0.20
0.20
0.10
0.20
0.20
0.00
0.30
0.10
0.10
0.30
0.10
0.10
0.10
0.10
-0.047
0.860
0.110
0.678
-0.045
Latent
0.730
-0.226
0.081
0.178
-0.317
-0.213
2.199
-0.207
-0.322
-0.869
0.214
0.240
-0.566
-0.477
-1.221
-0.359
-0.263
1.138
0.534
0.025
0.521
0.030
0.551
0.023
0.411
0.858
0.252
0.018
0.525
P-value
0.385
0.514 0.084
0.354 0.108
0.888
0.033
0.492
0.569
0.208
0.090
0.313
0.025
CWN: Cukierman, Webb and Neyapti (1992) weighted-average index.
CWN1 - CWN4: Four clusters of characteristics used to construct CWN.
GMT_E: Grilli, Masciandaro and Tabellini (1991) economic independence index. GMT_P: GMT political independence index.
GMT_P1 (GMT_P2, GMT_P3): Component of GMT_P that is similar to CWN1 (CWN2, CWN3). GMT_E4: Component of GMT_E that is similar to CWN4.
Turnover: Cukierman (1992) average term of office.
Latent: de Haan et al. (2003) factor score.
22
Table 3: Estimated Coefficient of Policy Reaction to Inflation on CBI index, Adjusted for Estimation Errors
Index
Backward
looking
Forward
looking
CWN
CWN1 CWN2
CWN3 CWN4 GMT_P
GMT_E GMT_P1 GMT_P2 GMT_P3
GMT_E4
Turnover
Latent
Coefficient
0.317
-0.267
0.481
-0.187
0.442
0.035
0.078
-0.006
0.070
0.167
0.085
-0.233
0.193
P-value
0.441
0.404
0.052
0.287
0.139
0.371
0.001
0.940
0.363
0.128
0.028
0.744
0.024
Coefficient
0.336
-0.392
0.560
-0.299
0.597
-0.001
0.073
-0.072
0.043
0.184
0.093
0.939
0.240
P-value
0.543
0.377
0.106
0.207
0.139
0.991
0.029
0.475
0.680
0.234
0.099
0.331
0.026
CWN: Cukierman, Webb and Neyapti (1992) weighted-average index.
CWN1 - CWN4: Four clusters of characteristics used to construct CWN.
GMT_E: Grilli, Masciandaro and Tabellini (1991) economic independence index. GMT_P: GMT political independence index.
GMT_P1 (GMT_P2, GMT_P3): Component of GMT_P that is similar to CWN1 (CWN2, CWN3). GMT_E4: Component of GMT_E that is
similar to CWN4.
Turnover: Cukierman (1992) average term of office.
Latent: de Haan et al. (2003) factor score.
23
Download