Content uploaded by Ana Carolina Garriga
Author content
All content in this area was uploaded by Ana Carolina Garriga on Jun 03, 2016
Content may be subject to copyright.
Content uploaded by Ana Carolina Garriga
Author content
All content in this area was uploaded by Ana Carolina Garriga on May 20, 2016
Content may be subject to copyright.
1
Central Bank Independence in the World.
A New Dataset
Ana Carolina Garriga
Forthcoming in International Interactions
This article introduces the most comprehensive dataset on de jure central bank independence
(CBI), including yearly data from 182 countries between 1970 and 2012. The dataset
identifies statutory reforms affecting CBI, their direction, and the attributes necessary to build the
Cukierman, Webb and Neyapty index. Previous datasets focused on developed countries, and
included non-representative samples of developing countries. This dataset’s substantially broader
coverage has important implications. First, it challenges the conventional wisdom about central
bank reforms in the world, revealing CBI increases and restrictions in decades and regions
previously considered barely affected by reforms. Second, the inclusion almost 100 countries
usually overlooked in previous studies suggests that sample selection may have substantially
affected results. Simple analyses show that the associations between CBI and inflation,
unemployment or growth are very sensitive to sample selection. Finally, the dataset identifies
numerous CBI decreases (restrictions), whereas previous datasets mostly look at CBI increases.
These data’s coverage not only allows researchers to test competing explanations of the
determinants and effects of CBI in a global sample, but it also provides a useful instrument for
cross-national studies in diverse fields, such as liberalization, diffusion, political institutions,
democratization, or responses to financial crises.
Keywords: Central banks, central bank independence, datasets, measurement,
reforms
Correspondence:
Ana Carolina Garriga
Centro de Investigación y Docencia Económicas (CIDE)
División de Estudios Políticos
Carretera México-Toluca 3655
Lomas de Santa Fe, México DF 01210, MEXICO
Phone: (+52 55) 5727-9800 Ext. 2165
carolina.garriga@cide.edu
2
This article introduces the most comprehensive dataset on de jure central bank
independence (CBI) available to the date. The dataset identifies statutory reforms
affecting CBI, their direction, and the attributes necessary to build the Cukierman, Webb
and Neyapty (1992) (CWN) index in 182 countries between 1970 and 2012. The most
commonly-used datasets include fewer than 100 countries, and cover fewer years. This
dataset codes the existence of reforms in 6,764 observations, and computes the CWN
index for 5,866 observations.1 The data coverage not only allows researchers to test
competing explanations on the determinants and effects of CBI in both developed and
developing countries, but it also provides a useful instrument for cross-national studies in
diverse fields. CBI has been a variable of interest not only for studies of the determinants
and effects of monetary policy, liberalization, or diffusion, but also for the study of
political institutions, democratization, or responses to crises (Adam, Delis, and Kammas
2011, Reenock, Staton, and Radean 2013, Rosas 2006). This article shows that previous
data provide incorrect or incomplete conclusions about the dynamics of central bank
reform in the world. For example, analysis of global data refutes “the fact that during the
forty years ending in 1989 there had hardly been reforms in [central bank] legislation”
(Cukierman 2008:724) . Finally, simple regressions show that associations between CBI
and inflation, unemployment and growth are very sensitive to the sample used. Sample
selection may have affected the generalizability of previous results in a significant manner.
CBI is the central bank’s capability of controlling monetary instruments
(Bernhard 2002:21) or, inversely, CBI is the set of restrictions to the government’s
1 The largest publically available original dataset (Bodea and Hicks 2015b)
includes only 2,314 observations (34.2% of this sample). The largest compilation of
datasets, including own coding (Sadeh 2011), has 2,714 observations (40% of this
sample).
3
influence on the central bank management of monetary policy. CBI can be restricted or
increased on three dimensions: personnel, financial, and policy independence (Eijffinger
and de Haan 1996:2). Personnel independence reflects limits to the government’s
influence on the central bank board’s membership or tenure. Financial independence
restricts the government’s ability to use central bank’s loans to fund its expenditures, to
avoid monetary policy subordination to fiscal policy. Finally, policy independence
reflects the central bank’s powers to formulate and execute monetary policy. This
includes the central bank’s ability to set the goals and/or chose the instruments of
monetary policy (Debelle and Fischer 1995). Central banks’ institutional designs vary
across these dimensions, resulting in different levels of CBI. However, providing a
continuous measure for CBI with cross-sectional and temporal validity and a broad
coverage has proven to be a difficult task.
In the 1980s, CBI emerged as the recipe to avoid the pervasive inflationary
consequences of shortsighted electoral ambitions. The practical advice derived from the
“rules versus discretion” literature (Barro and Gordon 1983, Rogoff 1985) was to solve
the time-inconsistency problem (Kydland and Prescott 1977) by delegating the control of
monetary policy to independent central banks. International agencies and policy makers
embraced this advice (Bernhard, Broz, and Clark 2002:699, International Monetary Fund
1999, World Bank 1992).
The need to test the theoretical argument, and assess the consequences of
delegation to central banks, spurred the interest in measuring CBI. Many studies show
the stabilizing effects of CBI on the economy: CBI is linked to lower inflation, reduced
variation in inflation and output, increased credibility of the monetary policy, and lower
uncertainty among economic agents (Bodea and Hicks 2015a, Cukierman 1992,
Cukierman, Miller, and Neyapti 2002, Cukierman, Webb, and Neyapti 1992, Persson and
4
Tabellini 1990, Rogoff 1985). The literature also shows that CBI has important political
consequences (Bernhard and Leblang 2002, Clark, Golder, and Poast 2013).
Beyond the consequences of CBI, researchers in the fields of international and
comparative political economy (Bernhard 2002, Broz 2002, Clark 2002, Hallerberg 2002),
and those interested in the politics of delegation (Bendor, Glazer, and Hammond 2001),
reforms (Acemoglu, Johnson, Querubin, and Robinson 2008), and diffusion (Polillo and
Guillén 2005) have paid particular attention to the determinants of CBI. Although
economic reasons would justify the establishment of independent central banks, the
variance CBI across countries is not explained just by economic fundamentals. However,
the few studies on the determinants of CBI have limitations. Most of these works show
the determinants of CBI in developed countries (Bernhard 2002, Broz 2002, Clark 2002,
Pistoresi, Salsano, and Ferrari 2011). Although there is agreement regarding the
possibility that the determinants of CBI are different in developed and developing
countries, data seldom allow us to test competing explanations on both sets of countries.
Studies including developed and developing countries are either cross-sectional analyses
(Crowe and Meade 2007) or are based on not-necessarily representative samples
(Berggren, Daunfeldt, and Hellström 2014, Bodea and Hicks 2015b).
Limited data availability suggests limits to our knowledge on these matters.2
Because the countries included in previous datasets are not representative world or
regional samples, it is possible that results on global samples are biased. And additional
2 Siklos (2008:803) suggests that data problems may even affect the definition of
CBI because empirical studies usually define CBI “sufficiently loosely […] to fit the
particular needs of the group of countries under investigation.” He attributes this to “the
inevitable constraints imposed by the availability of limited data as well as variations in
the quality of the data across countries.”
5
problem is that most previous data coding efforts focused on legislation that was in force
in certain years, providing valuable cross-sectional information, but little insight on
variation within countries.
To solve these issues, one needs to collect legislation from most countries in the
world, including legislation that has been revoked, and partial reforms affecting CBI. The
sources are not centralized,3 and include primary and secondary legislation, and central
banks’ internal rules. Furthermore, lack of translations for legislation in countries where
English is not the official language poses additional challenges.
Measuring CBI
Most empirical studies using CBI as dependent or independent variable base their
measures of CBI on central banks statutes (de jure CBI) (Alesina, Mirrlees, and Neumann
1989, Cukierman 1992, Grilli, Masciandaro, and Tabellini 1991). Some scholars have
used measures of de facto CBI, based on questionnaires (Blinder 2000, Cukierman, et al.
1992, Fry, Goodhart, and Almeida 1996) or in the turnover rate (TOR) of central bankers
(Cukierman and Webb 1995, Cukierman, et al. 1992, de Haan and Siermann 1996).
However, questionnaires may not be the most reliable measure of CBI, particularly
because of their narrow coverage, their problematic cross-sectional comparability, and
their little within-country variation. Furthermore, although Cukierman and others found
the TOR predicts inflation in developing countries, Dreher, Sturm, and de Haan (2008)
show that endogeneity explains this finding: central bankers unable to control inflation
are replaced more often.
3 The IMF’s Central Bank Legislation Database (CBLD) is restricted to central
banks and IMF personnel. Access to this data may help completing or contrasting the
sources used for this codification.
6
Measures based on statutes have been criticized because laws do not contemplate
all contingencies that might affect the relations between the central bank and the
government. Furthermore, deviations from the law are not infrequent. Even
independent central banks can be influenced by the government’s appointments and
threats to the bank’s independence (Lohmann 1998). Siklos (2008:804), on the other
hand, regrets that the literature on CBI “has downplayed to an excessive degree the
importance of the design of central bank legislation.” In spite of criticisms, reliance on a
legal-based measure is useful for several reasons. First, a measure of statutory CBI allows
collecting comparable cross-sectional data across time. These data allows looking for
systematic differences across observations. Second, and more importantly, the utility of
the measure depends on the research question for which it is used: statutory measures of
CBI are useful to assess governments’ institutional choices, that is, when and to what
extent governments give independence to their central banks – or limit it.
Although there are differences among different scores of de jure CBI (Alesina, et
al. 1989, Cukierman 1992, Grilli, et al. 1991), their correlation with inflation variables is
comparable. I use Cukierman, Webb, and Neyapti’s (CWN) criteria instead of other
available measures of CBI (Alesina, et al. 1989, Grilli, et al. 1991) for several reasons:
First, CWN’s criteria for coding are clear and easily replicable. Second, CWN’s
component variables are exhaustive and allow further recodifications for other purposes,4
and allow the study of particular components of the index (Banaian, Burdekin, and
Willett 1998. Furthermore, it has been widely used – “the current state of the art of
measurement of de jure CBI” (Acemoglu, et al. 2008:20) – , and its larger cross-sectional
and historical coverage allowed me to check the reliability of my own coding.
4 Arnone et al. (2007:39-40) include a table to convert CWN scores to the Grilli,
Masciandaro, and Tabellini (1991) scale.
7
Other scholars have extended CWN’s original coding of up to 72 countries
between 1950 and 1989 (on a country-decade basis), and up to 26 post-communist
countries for 1991-1998. For example, Polillo and Guillén (2005) extend this index to the
period 1990-2000 for 71 countries, Crowe and Meade coded 76 countries in 2003, Sadeh
combined extant sources and his own coding to cover 93 countries between 1968 and
2005, and Bodea and Hicks (2015b) coded CBI for 81 countries between 1972 and 2008.
Although the literature reports results based on other datasets, they are not all publicly
available (Daunfeldt, Hellström, and Landström 2013, Wessels 2006). For a list of
countries and observations included in publicly available datasets, see online appendix.
The Dataset
Coding Process and Descriptive Information
The dataset codes central bank legislation in 182 countries.5 I coded over 840
documents – constitutions, laws, amendments, and decrees that directly refer to central
banks, and central bank charters (see online appendix). Legislation was collected mainly
from online sources, and it was coded for all countries that had available texts in English,
Spanish, French, Portuguese or Italian. This helped identifying legislation that may have
been overlooked in other datasets.6
5 I obtained primary sources for 179 countries. I did not find primary sources to
code the CWN components for three countries. However, I found reliable secondary
sources to code the existence of reforms.
6 I coded 100 documents in Spanish, 47 in French, 34 in Portuguese, and 7 in
Italian. For Suriname’s and Turkmenistan’s legislation, I used automated translations.
8
Legislation was collected and coded in four independent processes: (1) I compiled
and coded all the sample for the period 1970-2008 in 2009; (2) one research assistant
(RA) compiled and two RAs coded the period 2009-2011 during 2012, and (3) one RA
compiled and two RAs coded the period 1970-2012 in 2013-2014 as a reliability check on
the coding of overlapping years. In the first two rounds, I relied on central banks’
websites and search engines to find legislation and news about central bank reforms. The
third round of data collection included “targeted searches”: to find earlier reforms, I
searched for older laws mentioned in new laws, and used national legislatures’ search
engines and central banks’ official information services to find them by their number
and/or date. This let me find reforms previously omitted. In the third wave, 15% of the
laws coded by each of the RAs were recoded by the other RA, and we held weekly
meetings to discuss differences between the coders and to agree about criteria. This third
wave of coding produced a second score for each variable for the period that overlapped
the previous two waves (1970-2011). (4) In the final stage, I compared both scores for
each country-year. If I found a discrepancy, I went back to the laws to re-evaluate them,
and decided the appropriate coding.
The dataset relies on Cukierman (1992) and Cukierman, Webb and Neyapti’s
(1992) rules to code central bank legislation. Each piece of legislation was coded on 16
dimensions related to four components of CBI, on a country-year basis: CEO’s
characteristics (appointment, dismissal, and term of office of the chief executive officer
of the bank); policy formulation attributions (who formulates and has the final decision in
monetary policy, and the role of the central bank in the budget process); central bank’s
objectives; and central bank’s limitations on lending to the public sector. These 16
components are also combined into a single weighted index, ranging from 0 (lowest) to 1
(highest) CBI. I also computed the CWN unweighted index. See the online appendix for
coding and weighting rules.
9
If reforms were partial amendments, only the variables affected by the
amendments were recoded. If reforms did not affect CBI, they were not coded as
reforms. When legislation was not available, variables included in the CWN index were
not coded. However, if the central bank explained in its own institutional information
(or official “history”) that there was an institutional reform in a given year, and the
information clearly allowed me to determine that reform’s direction, I coded those
variables.
The dataset includes additional variables: central bank creation, a central bank
reform that affects CBI in a given year, its direction (CBI increase or decrease), and
whether the central bank is a regional entity. The dataset includes 6,764 observations for
central bank reforms, and identifies 382 reforms affecting CBI.7 Of those reforms, 276
increase CBI, 56 decrease CBI, 39 have a zero net-effect on CWN’s weighted index.8 In
11 reforms, direction was not coded. It also includes 5,866 observations with scores for
the CWN legal index of CBI.
This dataset differs from previous datasets in three aspects: First, its coverage is
significantly broader (290% larger) than the largest original publicly available dataset
(Bodea and Hicks 2015b).9 Second, I include variables that account for the existence of
central bank reforms and their direction, even when there is no information on the
specific dimensions of CBI that were reformed. Although these categorical variables do
not provide information on the magnitude of the reform, some studies can still benefit
7 Although the Panamanian National Bank’s is not strictly a central bank, other
authors consider it possible to use its legislation to code its independence.
8 See below.
9 When regional observations are excluded, the dataset is 253% larger than Bodea
and Hicks (2015b), and 270% larger than Sadeh (2011).
10
from accounting for the existence and direction of reforms. Third, I identify numerous
reforms omitted in previous datasets, including reforms restricting CBI, a possibility not
even discussed in the literature until now. This within-country variation also permits
controlling for CBI in models with fixed effects. Finally, two additional variables register
whether the central bank was created in a given year, and whether the country’s monetary
policy is in the hands of a regional monetary union (for example, the members of the
Central Bank of West African States).
The online appendix shows descriptive statistics for this dataset and other
available datasets (Bodea and Hicks 2015b, Crowe and Meade 2007, Cukierman, et al.
1992, Neyapti and Dinçer 2008, Polillo and Guillén 2005, Sadeh 2011), for comparison
purposes. Differences in the CWN indices’ sample means are misleading because the
samples vary. The correlation between this and the other five datasets ranges between .7
(Sadeh) and .92 (CWN). (See the correlation matrix in the online appendix.) This reflects
a substantial consistency in the coding criteria of overlapping observations. Differences
in coding often result from omitted reforms or coding of the year of the reforms in other
datasets.
A caveat on regional central banks: 933 observations correspond to countries that
are members of regional central banks (such as European Central Bank or the Banque
des États de l'Afrique Centrale). However, only 391 of the 2,799 country-year
observations that appear exclusively in this dataset correspond to regional central banks –
other datasets also include regional observations, but do not single them out. In order to
avoid distortions caused by the inclusion of data on regional central banks, the online
appendix reproduces all the tables and graphs presented in this article excluding regional
observations. The results are substantially similar to the results reported in the article.
11
Geographic Coverage
The broad sample is one of this dataset’s most important attributes. It codes 105
countries for the full period. In subsequent years, new countries and countries whose
legislation became available were added (see Figure 1).
Figure 1. Number of countries per year included in the different datasets
This dataset not only includes a substantial number of countries previously
omitted, but also presents a more accurate picture of regional differences. Previously
available data are not representative world or regional samples, imposing limits to the
generalizability of previous studies. Figure 2 illustrates how the new dataset more
accurately represents important groups of countries. The top panel shows that this
050 100 150 200
Number of countries
1970 1980 1990 2000 2010
Year
Garriga CWN Polillo&Guillen
Sadeh Bodea&Hicks
12
dataset greatly improves the representation of middle and lower income countries. The
bottom panel further shows countries in regions other than North America and Europe
were seriously underrepresented.
Figure 2: Countries included in different datasets, by income groups and geographic
regions
Frequent Central Bank Reforms Around the World, and in Both Directions
The identification a number of reforms previously ignored is another important
contributions. The larger number of reforms in my dataset is not due to a broader
conceptualization of “reform.” It results from coding legislation that previous
0
10
20
30
40
50
Number of countries
High income Upper middle income Lower middle income Low income
Garriga CWN Polillo&Guillen Sadeh Bodea&Hicks
0
10
20
30
40
Number of countries
E.Europe & post-sov
W.Europe & N.America
Latin Am. & Caribbean
Asia
Sub-Saharan Africa
N. Africa & Middle East
The Pacific
13
researchers apparently overlooked, and from including additional countries (see Figure 3,
top panel). 5.6% of the 6,764 observations experience reforms affecting CBI. The mean
number of reforms per country in this dataset is two in 43 years.10
Figure 3: Number of reforms affecting CBI per year. Newly coded and previously coded
countries (top panel), reforms by direction (bottom panel)
10 The online appendix shows the frequency of reforms per country and year.
0 5 10 15 20
Number of reforms
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
Previously coded countries New countries
0 5 10 15 20
Number of reforms
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
CBI_decreases CBI_increases CBI_other
14
This dataset raises questions about the conventional wisdom regarding the history
of CBI reforms. Scholars argue that most reforms occurred in the 1990s (see Fernández-
Albertos 2015), stressing “the fact that during the forty years ending in 1989 there had hardly
been reforms in [central bank] legislation” (Cukierman 2008:724, emphasis added).
Examining a more representative global sample casts doubt upon this assertion: this
dataset identifies 113 reforms between 1970 and 1989 (75 of them increase CBI), usually
ignored in the literature.11 On average, in the 1970s 5.3% of the sample experienced
reforms affecting CBI. The percentage of observations coded as reforms is 3.2% for the
1980s, 7.9% for the 1990s, 5.5% for the 2000s, and 5.9% for the first three years of the
2010s.12 This contrasts with other data, as shown in Figure 4.13 This picture is similar in
the light of the number of countries included in different samples (Figure 6.1 in the
online appendix plots the proportion of observations included coded as experiencing CBI
reforms by year in different datasets).
11 In the same two decades, Bodea and Hicks identify 17 reforms (nine of them
increasing CBI). 84 of the 113 reforms I identify are in newly coded countries, and 29 in
countries that were coded by Bodea and Hicks.
12 The magnitude of the reforms also varies through the sample. In absolute
terms, the average reform changes the index by .112 before 1989, and by .206 after that
year. In relative terms, the average percentage change in CBI before 1989 is 40%
(excluding Iran, a significant outlier), and 63%, between 1990 and 2012.
13 Daunfeldt, et al. (2013) coded central bank reforms in a sample of 132
countries between 1980-2005. Their data is not public, so these data comes from their
figure 1 (Daunfeldt, et al. 2013:431).
15
Figure 4: Number of reforms affecting CBI per year. Different datasets
This dataset thoroughly identifies not only numerous reforms, but also their
direction (see Figure 3, bottom panel). In particular, it identifies 56 reforms restricting
CBI, a movement not discussed by the literature possibly because it was considered an
exceptional event. Table 1 compares the number and direction of reforms identified by
three datasets. I also code 39 instances in which reforms to different aspects of CBI do
not affect the scores based on the CWN index, or the direction of different amendments
offset each other (as in Slovenia 2007). In other eleven cases, missing data on the
regulation before the reform does not allow me to code with certitude the reform’s
direction; therefore, direction is missing.
0 5 10 15 20
Number of reforms
1970 1980 1990 2000 2010
Year
Garriga Polillo&Guillen
0 5 10 15 20
Number of reforms
1970 1980 1990 2000 2010
Year
Garriga Bodea&Hicks
0 5 10 15 20
Number of reforms
1970 1980 1990 2000 2010
Year
Garriga Daunfeldt&al
16
Table 1: CBI reforms, by direction
Garriga
Polillo&Guillen
Bodea&Hicks
Number of countries
182
91
81
Period
1970-2012
1989-2000
1972-2008
Observations
6,764
1,004
2,314
Number of reforms
Total (%)
Reforms increasing CBI (%)
Reforms decreasing CBI (%)
Zero effect or no direction coded (%)
382 (100%)
276 (72.2%)
56 (14.7%)
50 (13.1%)
58 (100%)
57 (98%)
1 (2%)
113 (100%)
95 (84%)
18 (16%)
CBI in the World: A Different Picture?
This dataset shows that the global dynamics towards CBI may have been
overstated as an artifact of sample selection. This section suggests that this picture is
mainly a product of the overrepresentation of higher-income and post-communist
countries in the samples.
The high correlation of my coding with other datasets suggests that coding
criteria were consistent. However, sample selection has significant effects on our
understanding of CBI. The first difference refers to the worldwide levels of CBI (see
Figure 5, top panel). It is a common practice to compare the CWN data world average
for 1989 and Crowe and Meade’s from 2003 (Crowe and Meade 2007, Fernández-
Albertos 2015). This suggests an 80% increase in the global level of CBI between 1989
and 2003. However, the global effect of central bank reforms seems less dramatic on
samples including more countries, especially because the previously excluded countries
show less variance in CBI (see Figure 5, middle panel). Bodea and Hicks register a 67%
increase, but this article’s dataset shows a more modest 40% increase in the world average
CBI between 1989 and 2003.
17
Figure 5. CBI world average. Different datasets (top panel), different subsamples
(middle panel), and by income groups (bottom panel)
.3 .4 .5 .6 .7
CBI World Average
1970 1980 1990 2000 2010
Yea r
Garriga CWN Polillo&Guillen Sadeh Bodea&Hicks Crowe&Meade
.3 .4 .5 .6 .7
CBI World Average
1970 1980 1990 2000 2010
Yea r
Garriga (full dataset) Garriga (new data) Garriga (recoded data) Bodea&Hicks
.3 .4 .5 .6 .7
CBI Group Average
1970 1980 1990 2000 2010
Yea r
High income Upper middle income Lower middle income Low income
18
The bottom panel in Figure 5 shows that this misleading picture was mainly
driven by the central bank reforms in higher-income countries, which are
overrepresented in other samples. However, lower-middle income and low-income
countries had more stable levels of CBI through the period. This also contrasts with the
assertion that “central banks in emerging market and developing economies have seen an
even more impressive shift towards independence over the past two decades than their
advanced-economy counterparts” (Crowe and Meade 2007:73, emphasis added).
Although that is certainly the case for Eastern European and post-Soviet countries, it is
not an accurate description of CBI in most developing countries (see Figure 6).
Figure 6. CBI regional averages
.2 .4 .6 .8
CBI group Average
1970 1980 1990 2000 2010
Year
E. Europe and post Soviet
W. Europe and N. America
.2 .4 .6 .8
1970 1980 1990 2000 2010
Year
East Asia S&SE. Asia
Latin Am. & Caribbean
.2 .4 .6 .8
CBI group Average
1970 1980 1990 2000 2010
Year
N. Africa & Middle East The Pacific
Sub-Saharan Africa
19
This dataset unveils important regional dynamics. During the period covered by
this dataset the most dramatic increase in CBI occurred among post-communist countries
in the 1990s until the mid-2000s, followed by Western European countries. These
countries rarely had CBI reversals. Latin American and Asian countries also increased
their CBI, but gradually, through a much longer period. Also, Latin American countries
restricted CBI seven times during the 2000s. Finally, the dynamics were very different in
Africa and the Pacific: these countries’ CBI average was similar to Western Europe’s in
the 1970s, but did not change substantially throughout four decades. This new
information casts doubt upon statements like “most central banks in today’s world enjoy
substantially higher levels of […] legal […] independence that twenty years ago or earlier”
(Cukierman 2008:723, emphasis added).
The components of CBI
The CWN criteria permit the analysis of different components or dimensions of
CBI, which show distinctive patterns. Figure 7 plots world averages of the CWN
composite indices, and of their four components (see online appendix for variables
included in each component). This figure shows a general tendency to convergence
among the four components of the CWN index. Also, the data suggests that the
weighting rules to combine the components do not alter significantly the index. Figure 8
plots the yearly average of the components, by income group.
20
Figure 7. CBI indices and their components. World averages
Figure 8. CBI components, income group averages
.3 .4 .5 .6
World Average
1970 1980 1990 2000 2010
Year
Weighted index Unweighted index Personnel indep.
Policy indep. CB objectives Financial indep.
.2 .3 .4 .5 .6 .7
Group ave.
1970 1980 1990 2000 2010
Year
Financial indep.
.2 .3 .4 .5 .6 .7
1970 1980 1990 2000 2010
Year
Policy indep.
.2 .3 .4 .5 .6 .7
1970 1980 1990 2000 2010
Year
Low income Lower middle income
Upper middle income High income
Personnel indep.
.2 .3 .4 .5 .6 .7
1970 1980 1990 2000 2010
Year
CB objectives
21
Regarding the main components of the CWN measure, personnel independence
(CEO variables) has been the most stable throughout the sample and subsamples –
consistent with Crowe and Meade’s (2008) account. Financial independence (the ability
of the government to use central bank credit to finance itself) exhibits the most dramatic
changes through time. However, this dynamic particularly characterizes reforms in
higher-income countries. Central banks in lower-income countries gained relatively more
independence in policy matters and from the redefinition of the bank’s objectives.
Finally, the variables reflecting central banks’ policy independence show the largest
variance depending on the income-groups (see upper-right panel in figure 11).
Simple Tests: CBI, Inflation, Unemployment, and Growth
Table 2 shows the results of regressing inflation, unemployment and GDP
growth on their lagged values and on CBI (with fixed effects). These models do not
intend to test whether CBI has a causal effect on those variables, but to show the
potentially important effects of sample selection (and in some cases, of measurement) on
the association between CBI and variables of interest.
When inflation is regressed on CBI in the full sample, the coefficient is negative
and highly statistically significant. I obtain similar results if the sample is divided between
high-income and middle- and lower income countries.14 Although these are very simple
models, the fact that in a larger sample I find a negative relationship between CBI and
inflation for both developed and developing countries is noteworthy – and contrasts with
previous findings (see Arnone et al. 2007, Bodea and Hicks 2015:40). The relationship
between CBI and inflation is very sensitive to the sample. When the same model is run
14 The substantive magnitude of the coefficient is larger in the middle- and lower-
income countries.
22
on the CWN sample, there is no statistically significant relation between inflation and
CBI – either with CWN’s measure or my measure of CBI. I run the same model on two
additional samples: Using Polillo and Guillen’s data, the coefficient does not achieve
statistical significance. If I replace their data with mine, on the same sample, the
coefficient becomes statistically significant. This suggests that differences in coding also
play a role. Finally, I replicate the exercise using Bodea and Hicks’ data. In their sample,
both their variable and mine are negative and statistically significant.
Table 2. Association between CBI, inflation, unemployment and growth. Different
datasets and samples
DV: Inflation
DV: Unemployment
DV: GDP growth
Model
CBI measure
(sample)
Coefficient
N
Coefficient
N
Coefficient
N
1
Garriga
(full)
-174.62
(-4.01)***
5672
-.328
(-1.66)*
3443
.587
(1.02)
5474
2
Garriga
(high income)
-26.50
(-3.72)***
1897
-.670
(-2.41)**
1083
-1.310
(-1.73)*
1819
3
Garriga (middle &
lower income)
-329.87
(-4.20)***
3775
.183
(0.68)
2360
3.267
(3.86)***
3655
4
CWN
-292.01
(-1.03)
1470
-1.688
(-1.30)
176
3.696
(1.10)
1364
5
Garriga
(model 4)
-173.85
(-0.70)
1470
-1.154
(-0.99)
176
5.307
(2.01)**
1364
6
Polillo & Guillen
-293.19
(-1.29)
985
-.684
(-1.47)
902
1.993
(1.60)
965
7
Garriga
(model 6)
-427.53
(-2.10)**
985
-.068
(-0.16)
902
2.698
(2.46)**
965
8
Bodea & Hicks
-241.45
(-3.80)***
2305
-.307
(-1.06)
1455
1.864
(3.22)***
2273
9
Garriga
(model 8)
-273.81
(-4.09)***
2305
-.226
(-0.77)
1455
1.669
(2.77)***
2273
Notes: DV: dependent variable. N: sample size. Coefficients after panel regression with
fixed effects. Constant and lagged dependent variable omitted, t-values between
parentheses.
23
This exercise is more interesting for the other two dependent variables. The
(marginally significant) negative association between CBI and unemployment in the full
sample disappears when the same model is run using any of the other datasets’
subsamples (t-values<1). Separating this article’s sample between developed and
developing countries show a significant negative relationship for the first group of
countries, and positive but insignificant coefficient for developing countries. The same
analysis run on the Polillo and Guillen and the Bodea and Hicks samples also shows
opposite directions for both groups of countries, but these coefficients are highly
significant (not shown in table).
Finally, the opposite happens with GDP growth: CBI is far from achieving
statistical significance in the full sample. However, CBI becomes significant at
conventional levels in all the other datasets’ samples. If the analysis is run dividing each
of the samples between developed and developing countries, it is evident that the lack of
significance in my full sample is a consequence of divergent relationships between CBI
and growth in these two groups of countries (in the full sample, these opposite effects
cancel each other). If I split the other datasets’ samples, developing countries also show a
positive relationship between CBI and growth. However, I find a non-statistically
significant positive relationship for developed countries, suggesting that the results in the
aggregate are driven by developing countries (and differences in coding or sample
selection for developed countries).
These simple regressions show that differences in coding and sample selection
may have important effects on relationships of interest for the study of CBI. Differences
in coding are a main source of variance with the decade-invariant CWN data, and with
Polillo and Guillen’s data. Sample selection is especially problematic for smaller samples
analyzed here. Differences with Bodea and Hicks’ data are smaller because of their larger
24
sample and similar coding, but they appear when the analysis is broken down in sub-
samples of countries.
Final Remarks
This article introduces an original dataset coding central bank reforms and CBI in
182 countries between 1970 and 2012. The correlations with previous data suggest
consistent criteria when analyzing the legislation of interest. However, the importance of
this dataset derives from innovations over previously available data. First, this dataset has
a substantially broader coverage that will allow scholars to examine important research
questions in larger and more representative samples. Descriptive data presented here
shows that different samples offer different pictures of the worldwide dynamics of CBI
and central bank reform. Furthermore, non-representative samples may have affected
previous results, suggesting that there might be limits to the generalizability of some
empirical results in the literature.
The second feature of this dataset is a finer-grained analysis of the legislation
affecting CBI. A meticulous search of documents, together with the coding of sources in
multiple languages, made it possible to identify numerous central bank reforms previously
overlooked. Additionally, the fact the dataset’s coded reforms include both increases and
decreases in CBI opens new avenues for researching the determinants and consequences
of monetary institutions. For example, they suggest the possibility of developing a theory
to explain CBI restrictions or, more generally, liberalizing reforms reversals.
Furthermore, the careful identification of reforms also results in data with within-country
variance that can be exploited to answer different research questions, using CBI or
central bank reforms as explanatory variables.
25
Indices of legal CBI have been criticized because they may not accurately reflect
actual independence from the government. Furthermore, other aspects regarding the
design and actual operation of central banks, such as their transparency or accountability,
can be equally or even more important than CBI for certain research questions.
Nonetheless, de jure measures are suitable to explore the determinants of monetary
institutions. Of course, other factors such as regime type or rule of law need to be taken
into account to fully understand the effects or even the meaning of CBI in different
countries. CBI is seldom a consequence of merely monetary logics, and it may proxy
other domestic dynamics of interest for political scientists, such as executive powers,
institutional hurdles for reform, difficulties for reform implementation, or diffusion of
particular policies. The new dataset described here will permit researchers to address
these important questions in different fields with more certitude than was possible
before, harnessing in-depth data from a globally representative sample.
Acknowledgement
This dataset was presented at the workshop “Monetary Policy and Central Banking:
Historical Analysis and Contemporary Approaches,” Princeton University, February 6-7,
2015, at the REPAL annual conference, Montevideo, Uruguay, July 2-3, 2015, and at the
57th ISA Annual Convention, Atlanta, March 16-19, 2016. I thank the participants at
these events, David Bearce, Julia Gray, Brian Phillips, and the reviewers for their helpful
comments. Marc Grau, Santiago Minor, María Fernanda Nieto, Mauricio Ochoa, José
Manuel Toral, and Ludwig van Bedolla helped collecting legislation and coding different
waves of data. María Fernanda Porras provided excellent research assistance.
26
References
Acemoglu, Daron, Simon Johnson, Pablo Querubin, and James A. Robinson. 2008.
"When Does Policy Reform Work? The Case of Central Bank Independence." In
NBER Working Paper no. 14033. Cambridge, MA: National Bureau of Economic
Research.
Adam, Antonis, Manthos D. Delis, and Pantelis Kammas. (2011) Are Democratic
Governments More Efficient? European Journal of Political Economy 27:75-86.
Alesina, Alberto, James Mirrlees, and Manfred J. M. Neumann. (1989) Politics and
Business Cycles in Industrial Democracies. Economic Policy 4:55-98.
Arnone, Marco, Bernard J. Laurens, Jean-Francois Segalotto, and Martin Sommer. 2007.
"Central Bank Autonomy: Lessons from Global Trends." In IMF Working Paper,
WP/07/88. Washington, DC: International Monetary Fund.
Banaian, King, Richard C. K. Burdekin, and Thomas D. Willett. (1998) Reconsidering
the Principal Components of Central Bank Independence: The More the Merrier?
Public Choice 97:1-12.
Barro, Robert J., and David Gordon. (1983) Rules, Discretion and Reputation in a
Model of Monetary Policy. Journal of Political Economy 91:101-21.
Bendor, Jonathan, A. Glazer, and Thomas H. Hammond. (2001) Theories of
Delegation. Annual Review of Political Science 4:235-69.
Berggren, Niclas, Sven-Olov Daunfeldt, and Jörgen Hellström. (2014) Social Trust and
Central-Bank Independence. European Journal of Political Economy 34:425-39.
Bernhard, William. (2002) Banking on Reform. Political Parties and Central Bank Independence
in the Industrial Democracies. Ann Arbor, MI: University of Michigan Press.
Bernhard, William, J. Lawrence Broz, and William Roberts Clark. (2002) The Political
Economy of Monetary Institutions. International Organization 56:693-723.
27
Bernhard, William, and David Leblang. (2002) Political Parties and Monetary
Commitments. International Organization 56:803-30.
Blinder, Alan S. (2000) Central-Bank Credibility: Why Do We Care? How Do We Build
It? American Economic Review 90:1421-31.
Bodea, Cristina, and Raymond Hicks. (2015a) International Finance and Central Bank
Independence: Institutional Diffusion and the Flow and Cost of Capital. Journal
of Politics 77:268-84.
———. (2015b) Price Stability and Central Bank Independence: Discipline, Credibility,
and Democratic Institutions. International Organization 69:35-61.
Broz, J. Lawrence. (2002) Political System Transparency and Monetary Commitment
Regimes. International Organization 56:861-87.
Clark, William Roberts. (2002) Partisan and Electoral Motivations and the Choice of
Monetary Institutions under Fully Mobile Capital. International Organization
56:725-49.
Clark, William Roberts, Sona N. Golder, and Paul Poast. (2013) Monetary Institutions
and the Political Survival of Democratic Leaders. International Studies Quarterly
57:556-67.
Crowe, Christopher, and Ellen E. Meade. (2007) The Evolution of Central Bank
Governance around the World. Journal of Economic Perspectives 21:69-90.
———. (2008) Central Bank Independence and Transparency: Evolution and
Effectiveness. European Journal of Political Economy 24:763-77.
Cukierman, Alex. (1992) Central Bank Strategy, Credibility and Independence: Theory and
Evidence. Cambridge, MA: The MIT Press.
———. (2008) Central Bank Independence and Monetary Policymaking Institutions -
Past, Present and Future. European Journal of Political Economy 24:722-36.
28
Cukierman, Alex, Geoffrey P. Miller, and Bilin Neyapti. (2002) Central Bank Reform,
Liberalization, and Inflation in Transition Economies. An International
Perspective. Journal of Monetary Economics 49:237–64.
Cukierman, Alex, and Michael C. Webb. (1995) Political Influence on the Central Bank:
International Evidence. The World Bank Economic Review 9:397-423.
Cukierman, Alex, Steven B. Webb, and Bilin Neyapti. (1992) Measuring the
Independence of Central Banks and Its Effect on Policy Outcome. The World
Bank Economic Review 6:353-98.
Daunfeldt, Sven-Olov, Jörgen Hellström, and Mats Landström. (2013) Why Do
Politicians Implement Central Bank Independence Reforms? Atlantic Economic
Journal 41:427-38.
de Haan, Jakob, and Clemens L.J. Siermann. (1996) Central Bank Independence,
Inflation, and Political Instability in Developing Countries. Journal of Policy Reform
1:135-47.
Debelle, Guy, and Stanley Fischer. 1995. "How Independent Should a Central Bank
Be?" In Goals, Guidelines and Constraints Facing Monetary Policymakers, ed. Jeffrey C.
Fuhrer, Conference Series No. 38. Boston, MA: Federal Reserve Bank of
Boston. 195-221.
Dreher, Axel, Jan-Egbert Sturm, and Jakob de Haan. (2008) Does High Inflation Cause
Central Bankers to Lose Their Job? Evidence Based on A New Data Set.
European Journal of Political Economy 24:778-87.
Eijffinger, Sylvester C. W., and Jakob de Haan. 1996. "The Political Economy of Central
Bank Independence." Paper presented at the Princeton Special Papers in
International Economics, No. 19, Princeton, NJ.
Fernández-Albertos, José. (2015) The Politics of Central Bank Independence. Annual
Review of Political Science 18: 217-37.
29
Fry, Maxwell J., Charles A. E. Goodhart, and Alvaro Almeida. (1996) Central Banking in
Developing Countries. Objectives, Activities and Independence. London and New York:
Routledge.
Grilli, Vittorio, Donato Masciandaro, and Guido Tabellini. (1991) Political and
Monetary Institutions and Public Financial Policies in the Industrial Countries.
Economic Policy:341-92.
Hallerberg, Mark. (2002) Veto Players and the Choice of Monetary Institutions.
International Organization 56:775-802.
International Monetary Fund. 1999. "Code of Good Practices on Transparency in
Monetary and Financial Policies: Declaration of Principles." Washington, DC:
International Monetary Fund.
Kydland, Finn E., and Edward C. Prescott. (1977) Rules Rather Than Discretion: The
Inconsistency for Optimal Plans. Journal of Political Economy 85:473-91.
Lohmann, Susanne. (1998) Federalism and Central Bank Independence: The Politics of
German Monetary Policy. World Politics 50:401-46.
Neyapti, Bilin, and Nergiz Dinçer. (2008) What Determines the Legal Quality of Bank
Regulation and Supervision? Contemporary Economic Policy 26:607-22.
Persson, Torsten, and Guido Tabellini. (1990) Macroeconomic Policy, Credibility and Politics.
London: Harwood Publishers.
Pistoresi, B., F. Salsano, and D. Ferrari. (2011) Political Institutions and Central Bank
Independence Revisited. Applied Economics Letters 18:679-82.
Polillo, Simone, and Mauro F. Guillén. (2005) Globalization Pressures and the State:
The Global Spread of Central Bank Independence. American Journal of Sociology
110:1764-802.
Reenock, Christopher, Jeffrey K. Staton, and Marius Radean. (2013) Legal Institutions
and Democratic Survival. The Journal of Politics 75:491-505.
30
Rogoff, Kenneth S. (1985) The Optimal Degree of Commitment to an Intermediate
Monetary Target. Quarterly Journal of Economics 100:1169-89.
Rosas, Guillermo. (2006) Bagehot or Bailout? An Analysis of Government Responses to
Banking Crises. American Journal of Political Science 50:175-91.
Sadeh, Tal. (2011) Hard Currencies for Hard Times. Terror Attacks and the Choice of
Monetary Anchors. Defence and Peace Economics 22:367-92.
Siklos, Pierre L. (2008) No Single Definition of Central Bank Independence Is Right for
All Countries. European Journal of Political Economy 24:802-16.
Wessels, Buks. (2006) Are African Central Banks Sufficiently Independent for Monetary
Convergence? South African Journal of Economics 74:230-47.
World Bank. 1992. "How the Independence of Central Banks Affects Policy
Outcomes." In World Bank Policy Research Bulletin. Vol. 3. Washington, DC:
World Bank.
