Journal of Development Economics 87 (2008) 29 – 50
www.elsevier.com/locate/econbase
The political economy of seigniorage ☆
Ari Aisen a, Francisco José Veiga b,⁎
a International Monetary Fund, 700 19th Street NW, Washington, DC 20431, USA
b Universidade do Minho and NIPE, Escola de Economia e Gestão, P-4710-057 Braga, Portugal
Received 5 July 2006; received in revised form 14 December 2007; accepted 18 December 2007
Abstract
While most economists agree that seigniorage is one way governments finance deficits, there is less agreement about the
political, institutional and economic reasons for relying on it. This paper investigates the main political and institutional
determinants of seigniorage using panel data on about 100 countries, for the period 1960–1999. Estimates show that greater
political instability leads to higher seigniorage, especially in developing, less democratic and socially-polarized countries, with high
inflation, low access to domestic and external debt financing and with higher turnover of central bank presidents. One important
policy implication of this study is the need to develop institutions conducive to greater political stability as a means to reduce the
reliance on seigniorage financing of public deficits.
© 2008 Elsevier B.V. All rights reserved.
JEL classification: E31; E63
Keywords: Seigniorage; Political instability; Institutions
1. Introduction
ences in seigniorage — real revenues a government
acquires by using newly issued money to buy goods and
The purpose of this paper is to identify the main
non-money assets.1 This is a challenge not yet
determinants of cross-country and cross-time differ-
satisfactorily confronted by the economics profession
for four reasons. First, several political and institutional
variables used as explanatory variables in earlier studies
☆ The authors acknowledge the helpful comments from Christopher
were relatively poorer measures of political instability
Bowdler, Juan Jauregui, Delfim Neto, Carlos Végh, Robert Flood,
and of the institutional environment than those available
Paolo Mauro, various staff members from the International Monetary
in new datasets such as the Cross National Time Series
Fund, two anonymous referees, and the editor, Lant Pritchett. We also
thank Reid Click for sharing his data on creditworthiness ratings. The
views expressed in this paper are those of the authors and do not
necessarily represent those of the IMF or IMF policy. Francisco Veiga
wishes to thank the Portuguese Foundation for Science and
1 Some studies, such as Buiter (2007), distinguish seigniorage
Technology (FCT) for research grant POCI/EGE/55423/2004 (par-
(change in monetary base) from central bank revenue (interest earned
tially funded by FEDER).
by investing the resources obtained through the past issuance of base
⁎ Corresponding author. Universidade do Minho, Escola de
money). This distinction is useful to study central bank operations and
Economia e Gestão, P-4710-057 Braga, Portugal. Tel.: +351
monetary policy effectiveness. For the purposes of this paper,
253604534; fax: +351 253676375.
however, it suffices to broadly define seigniorage as revenues
E-mail addresses: aaisen@imf.org (A. Aisen),
obtained by a consolidated government (treasury and central bank)
fjveiga@eeg.uminho.pt (F.J. Veiga).
from the issuance of base money.
0304-3878/$ - see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jdeveco.2007.12.006

---

30
A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
Data Archive (CNTS), Database of Political Institutions
is that it was still used to a greater extent in the 1990s
(DPI), the Polity IV Database, State Failure Task Force
than in the 1960s. Furthermore, seigniorage revenues
(SFTF) database, and the Freedom House ratings.
are on average five times higher in developing countries
Second, our analysis is based on a richer and wider
than in industrial countries for the period 1960–1999. In
dataset, covering more countries and years than those
the 1990s, average seigniorage revenues represented
used in previous studies, and includes a larger variety of
14.65% of total government revenues for developing
alternative model specifications. Third, although Aisen
countries, compared to only 1.64% for industrial coun-
and Veiga (2006) study the determinants of inflation
tries. Therefore, analyzing the determinants of seigniorage
using a similar dataset, one should not expect that
is an important endeavor, primarily for developing
variables affecting inflation should affect seigniorage in
countries.
the exact same way, since the latter might be consistent
Relying upon the theoretical literature and using a
with two different levels of the former in the presence of
dataset covering around 100 countries for the period
a well-defined Laffer curve. According to Easterly et al.
1960–1999, we estimate panel data models to investi-
(1995), studying inflation is different to studying
gate the main economic, political and institutional
seigniorage, especially for developing and high-infla-
determinants of seigniorage. After controlling for the
tion countries. Accordingly, the correlation between
countries' economic structure and for several other
inflation and seigniorage in our sample fluctuates
variables that may affect seigniorage, we find that
significantly depending on the rate of inflation (see
greater political instability leads to higher seigniorage
Table 1). While it is positive most of the time and for
levels, confirming previous results by Cukierman et al.
most of the countries, it declines with the level of
(1992) and Click (1998).
inflation and becomes negative for inflation rates above
This paper's major contribution to the literature is the
400% per year. Thus, it is misleading to assume that the
identification of the circumstances under which the
determinants of inflation are necessary the same as those
above-referred relationship is stronger. That is, we find
of seigniorage, which means that separate studies of
that political instability has stronger effects on seignio-
these variables should be made. As an example, changes
rage levels in higher inflation than in moderate and low
in inflation may result from supply-side shocks, such as
inflation countries, and also in developing than in
fluctuations in oil prices, which do not directly affect
industrial nations. In addition, this relationship is also
seigniorage. Conversely, the structure of the economy,
stronger in countries with (i) higher social polarization;
which affects the capacity to raise taxes and the reliance
(ii) a tradition of high political instability;2 (iii) higher
on seigniorage revenues, may not affect inflation in the
central bank president turnover (lower de facto central
same way. Fourth, our models are able to identify the
bank independence); (iv) lower indexes of economic
circumstances under which the relationship between
freedom; (v) more authoritarian regimes; (vi) higher
political instability and seigniorage is stronger, a central
domestic debt levels as a percentage of GDP; (vii) lower
topic of our research and virtually absent from previous
access to international financing (expressed in poorer
empirical studies on the determinants of seigniorage.
creditworthiness ratings); and, (viii) lower openness to
While seigniorage seems to be a less attractive method
international trade. It is also worth mentioning that,
of government financing for several countries, the truth
besides its effects on the relationship between political
instability and seigniorage, social polarization is by
itself a major determinant of seigniorage. Empirical
Table 1
Correlation between inflation and seigniorage
results show quite clearly that higher degrees of social
polarization (lower ethnic homogeneity) are associated
Sample
Observations
Correlation
with higher levels of seigniorage.
All
3171
.214
The paper is structured as follows. A survey of the
Inflation b 10%
1967
.102
Inflation
empirical and theoretical literature on the relationship
b100%
3083
.305
Inflation N 100%
88
.132
between seigniorage, political instability and institutions
Inflation N 200%
43
.058
is presented in Section 2. The dataset and the empirical
Inflation N 300%
34
.0001
models are described in Section 3. Section 4 presents the
Inflation N 400%
28
−.007
empirical results, and Section 5 concludes the paper.
Inflation N 500%
26
−.038
Inflation N 1000%
18
−.139
2
Notes: inflation is the annual inflation rate (IFS line 64×). Seigniorage
Expressed in a high number of state failure events in the last
is the change in reserve money (IFS, line 14a) as a percentage of
15 years, such as revolutionary wars, ethnic wars, regime crises, and
government revenues (IFS line 81).
genocides/politicides.

---

A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
31
2. The political economy of seigniorage
Cukierman et al. (1992) develop a theoretical model
whereby political instability and ideological polarization
Most economists acknowledge that differences on
determine the equilibrium efficiency of the tax system and
the way countries conduct their fiscal policies are behind
the resulting combination of tax revenues and seigniorage
the variability of the seigniorage levels they sustain.3
that governments use. Using a probit model to determine
But this explanation leads to a much deeper and
the likelihood of an incumbent government to remain in
fundamental question, which is why countries differ
power, they show evidence that higher political instability
on the way they conduct fiscal policies (see Woo, 2003,
and ideological polarization lead to higher seigniorage. In
2005). In particular, governments that are able to finance
the empirical analysis of Section 4, we employ alternative
their expenditures through taxes or debt do not need to
and more direct measures of political instability, such as
rely on seigniorage revenues. Several studies have
variables that count the exact number of cabinet changes,
explored the idea that structural features of a particular
executive changes or government crises taking place in a
economy help determine its “taxable capacity”. Chelliah
particular year. Moreover, whereas they use a dummy
et al. (1975), for example, provide evidence that
variable for democratic regimes as a proxy for ideological
countries with larger per capita non-export income,
polarization, we use the Polity Scale (ranged between −10
more open to trade and with larger mining but smaller
and + 10) to measure the degree of democracy in different
agricultural sectors have, on average, a higher “taxable
countries, and an ethnic homogeneity index as a proxy for
capacity” or ease of collection. This result leads to the
the degree of social polarization.4
conclusion that the countries' ability to tax is techno-
In line with Cukierman et al. (1992), we conjecture that
logically constrained by their stage of development and
economies with weaker institutions might be unable to
by the structure of their economies (e.g. size of the
build efficient tax systems leading them to use more
agricultural sector in GDP), and as tax collecting costs
frequently seigniorage as a source of revenue. In the next
are high and tax evasion pervasive, countries might use
sections, in addition to the effects of political instability on
seigniorage more frequently. But what if governments,
seigniorage, we also estimate the effects of institutions
independently of their countries' economic structures,
such as democracy and economic freedom. Besides
find it optimal to finance expenditures using seigniorage
structural variables accounting for the taxing capacity of
rather than levying other taxes (e.g. taxes on output)?
the economy and political and institutional variables
The Theory of Optimal Taxation (see Phelps, 1973;
affecting the use of seigniorage financing of fiscal deficits,
Végh, 1989; Aizenman, 1992) rationalizes government
we also consider, in line with Click (1998), variables that
behavior in many countries showing that it might be
measure the ability of governments to finance transitory
optimal for governments to rely on seigniorage if other
expenditures with domestic or external debt. To the extent
taxes are highly distortionary. According to this theory,
that a government is able to finance its expenditure
governments optimally equate the marginal cost of the
through debt, there is less need to rely on seigniorage.
inflation tax with that of output taxes, therefore
Our main contribution to the literature is that our
minimizing the distortions to the economy when
models not only identify the main political and economic
choosing the optimal combination of taxes to finance
determinants of seigniorage, but also reveal under which
their expenditures. Edwards and Tabellini (1991) and
circumstances the effects of political instability on
Cukierman et al. (1992) fail to find evidence that this
seigniorage are stronger. Our results indicate that the
theory applies to developing countries. Click (1998)
causal effect of political instability on seigniorage is
estimates a model using 90 countries, from 1971–90,
stronger in developing and high-inflation countries. In
and finds that only 40% of the cross-country variation in
addition, it is also stronger in socially-polarized,5 less
seigniorage can be explained with the Theory of
Optimal Taxation. The empirical failure of this theory
4 An additional shortcoming of the analysis in Cukierman et al.
to fully explain the cross-country differences in the use
(1992) is the use of a cross-sectional dataset using averages from 1971
to 1982 for only 79 countries, while we use a panel dataset covering
of seigniorage revenues motivated the use of theoretical
around 100 countries for the period 1960–99.
and empirical models focusing on the role played by
5 See Beetsma and Van Der Ploeg (1996), Bhattacharya et al. (2005)
political and institutional variables.
and Albanesi (2007) for studies presenting evidence suggesting that
inflation and income inequality are positively related. In Desai et al.
3 See Catão and Torrones (2005) for an empirical analysis on the
(2005) that relationship is conditional on the political structure. Woo
relationship between fiscal deficits and inflation and Fischer et al.
(2005) finds that social polarization is associated with fiscal instability
(2002) for a survey on modern hyper- and high inflations that includes
while generating incentives to engage in short-term policies leading to
results showing a positive relationship between fiscal deficits and
lower growth. Our findings indicate that the fiscal instability channel
seigniorage.
may also lead to higher seigniorage and inflation.

---

32
A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
democratic, traditionally unstable, and highly indebted
To investigate the main political, institutional and
countries. Finally, political instability has greater effects
economic determinants of seigniorage levels across
on seigniorage in countries that have lower de facto
countries and time, we estimate panel data models,
central bank independence, lower economic freedom,
controlling for countries' fixed effects. Seigniorage
lower creditworthiness ratings and lower openness to
is defined in two alternative ways: (1) the change in
international trade. In our view, and to the best of our
reserve money (line 14a of IFS–IMF) as a percentage of
knowledge, there is no comprehensive study in the
nominal GDP (line 99b in IFS–IMF); (2) the change in
literature fully analyzing the relationship between poli-
reserve money (line 14a of IFS–IMF) as a percentage of
tical instability and seigniorage. As it will become clear in
government revenues (line 81 in IFS–IMF). Appendix
the following sections, this paper is an attempt to
A shows the number of observations, means and
contribute in this direction.
standard deviations of these seigniorage measures for
all countries for which data is available.17
3. Data and the empirical model
We hypothesize that seigniorage levels depend on the
following explanatory variables:
The dataset is composed of annual data on political,
institutional and economic variables for the years 1960 to
• A set of variables representing political instability,
1999. Although we collected data for 178 countries,
polarization and institutions:
missing values for several variables reduce the number of
o Cabinet Changes (CNTS), a proxy for political
countries in our estimations to around 100. The sources of
instability, counts the number of times in a year in
political and institutional data are: the Cross National
which a new premier is named and/or 50% of the
Time Series Data Archive (CNTS); the Polity IV dataset;6
cabinet posts are occupied by new ministers. A
Gwartney and Lawson (2002);7 the Database of Political
positive coefficient is expected, as greater instabil-
Institutions (DPI 3.0);8 the State Failure Task Force
ity should lead to greater reliance on seigniorage
dataset (SFTF);9 and the Freedom House ratings.10
revenues.
Economic data was collected from the World Bank's
Why may the number of cabinet changes be a good
World Development Indicators (WDI) and Global Devel-
indicator of political instability? First, in a country
opment Network Growth Database (GDN),11 the Inter-
characterized by frequent changes in the composi-
national Monetary Fund's International Financial
tion of government, there are also frequent changes
Statistics (IFS), the Penn World Tables (PWT 6.1),12
in macroeconomic policies, as new prime ministers
Euromoney creditworthiness ratings,13 Cukierman et al.
or ministers of finance/economics do not necessa-
(1995),14 Dollar and Kraay (2002),15 and Levy-Yeyati
rily share the views of their predecessors. Second,
and Sturzenegger (2003).16
frequent cabinet changes shorten the horizon of the
members of government, as they are not certain
that they will keep their posts during an entire term.
6 Available on the Internet (http://www.cidcm.umd.edu/inscr/polity/
The higher the probability of being replaced, the
index.htm).
7
greater will be the importance attributed to short-
Available on the Internet (http://www.freetheworld.com/release.html).
8 On this database, see Beck et al. (2001). Available on the Internet
term objectives. Then, since the costs of future
though Philip Keefer's page in the World Bank's site (http://www.
inflation are not fully internalized, it is difficult to
worldbank.org/research/bios/pkeefer.htm).
resist the temptation to finance current expendi-
9 Available on the Internet (http://www.cidcm.umd.edu/inscr/stfail/
tures with seigniorage revenues.
sfdata.htm).
10 Available on the Internet (http://www.freedomhouse.org/ratings/).
11 Available on the Internet (http://www.worldbank.org/research/
17 There is data on ΔRM/GDP for 144 countries and on ΔRM/GR for
growth/GDNdata.htm).
122 countries. These are the seigniorage measures most commonly
12 Available on the Internet (http://pwt.econ.upenn.edu/php_site/
used in the literature. We performed all estimations for both measures
pwt_index.php).
but, to make our results more easily comparable to those of
13 The data on the Euromoney creditworthiness index, raging from 0
Cukierman et al. (1992), we report in most tables those obtained
to 100, from 1982 to 1999, was kindly provided by Reid Click.
when using the change in reserve money as a percentage of
14 Underlying data available on the Internet (http://www.tau.ac.il/
government revenues. Two additional ways of measuring seigniorage,
~alexcuk/pdf/WebbPoltime2.xls).
used by Cukierman et al. (1992), are the product of reserve money by
15 Underlying data available on the Internet (http://siteresources.
the inflation rate divided by either GDP or government revenues.
worldbank.org/INTRES/Resources/469232-1107449512766/648083-
These authors have shown that these two additional alternative
1108140788422/Growth_is_good_for_the_poor_data.zip).
measures of seigniorage provide similar results for a cross-section of
16 Underlying data available on the Internet (http://www.utdt.edu/
countries. Another alternative, used by Click (1998), is the change in
~fsturzen/base_2002.xls).
the monetary base as a percentage of government spending.

---

A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
33
o Ethnic Homogeneity Index (SFTF): ranges from 0 to
• Variables accounting for fixed effects of countries
1, with higher values indicating ethnic homogeneity,
and time:
and equals the sum of the squared population
o Country dummy variables;
fractions of the seven largest ethnic groups in a
o Dummy variables for each decade: 1960s, 1970s,
country. For each year, it takes the value of the index
1980s and 1990s.
in the beginning of the respective decade. According
to Woo (2003, 2005) higher social polarization,
Appendix B presents the descriptive statistics for the
which can be proxied by ethnic heterogeneity, leads
above-described dependent and independent variables
to higher polarization of preferences for different
and for additional/alternative explanatory variables that
types of government spending and to public deficits.
appear in the tables shown in the paper.
Thus, a negative coefficient is expected;
The empirical model for seigniorage levels can be
o Polity Scale (Polity IV): from strongly autocratic
summarized as follows:
(−10) to strongly democratic (10). Although the
economic theory is not conclusive, we anticipate
Sit ¼ aPIi;tÀ1 þ bSPit þ dPSit þ EcoitVj þ EcPitVg ð1Þ
that democracy is associated with lower reliance
þm
on seigniorage (negative coefficient);18
i þ eit ;
i ¼ 1; N ; N
t ¼ 1; N ; Ti
• A set of economic structural variables that reflect
where S is seigniorage, PI is a proxy for political
characteristics of the countries that may affect their
instability, SP is a proxy for social polarization, PS is the
capacity to control inflation:
Polity Scale, Eco is a vector of economic structural
o Agriculture (%GDP): share of the value added of
variables, EcP is a vector of variables accounting for
agriculture in GDP (WDI, WB). According to
economic performance and external shocks, ν
Chelliah et al. (1975), a positive coefficient is
i is the
fixed effect of country i, and ε
expected. An alternative proxy for the structure of
it is the error term.
It is worth noting that seigniorage is not persistent (its
the economy is Urban Population (% of total), the
first lag is never statistically significant when included
urbanization ratio (WDI, WB), which according to
as an explanatory variable) and that the error term of
Edwards and Tabellini (1991) should have a
Eq. (1), ε
negative sign;
it, is not serially correlated. Fisher type unit
root tests for panel data reject the null hypothesis that
o Trade (%GDP): openness to trade (WDI, WB).
seigniorage is non-stationary in all countries.20 Dickey–
Since it is associated with larger revenues of
Fuller and Augmented Dickey–Fuller tests performed
import duties, we expect that countries more open
on each individual country reject unit root behavior of
to trade rely less on seigniorage revenues (a
seigniorage for all countries that have at least ten
negative coefficient is expected);19
observations (in 15 countries, a drift term has to be
o Real GDP per capita (PWT 6.1). Richer countries
included). These results, which are available upon
have more efficient tax systems and, thus, have a
request, are consistent with those of Click (2000), who
lesser need for seigniorage (negative coefficient
rejected a unit root behavior of seigniorage in the four
expected);
•
countries considered in his study (USA, UK, Brazil, and
Variables accounting for economic performance and
Argentina).
external shocks:
The proxy for political instability (PI
o % Change in Terms of Trade (WDI, WB). Favorable
i,t − 1) is lagged
one period for two reasons. First, political instability
evolution of terms of trade provides greater tax
may translate into higher seigniorage only after some
revenues (negative coefficient expected);
time. Furthermore, if a cabinet change occurs in the end
o Growth of real GDP (WDI, WB). Higher growth rates
of one year, it is very likely to lead to higher seigniorage
are associated with increasing tax revenues, reducing
the need for seigniorage (negative coefficient);
20 The results of three of those tests are presented below:
18 Although ethnic homogeneity and the polity scale may also be
related with political instability, we see them more as institutional
Fisher Test (0 lags, no drift). Ho: unit root
variables than as indicators of political instability.
19
chi2(244) = 1964.3487
Prob N chi2 = 0.0000
The outcome on seigniorage may be similar, even if more open
countries are imposing lower tariffs. These countries may rely less on
Fisher Test (1 lag, no drift). Ho: unit root
seigniorage in order to avoid the real appreciation of the home
chi2(240) = 1360.5939
Prob N chi2 = 0.0000
currency associated with higher inflation. We owe this rationale to an
Fisher Test (0 lags, with a drift term). Ho: unit root
anonymous referee.
chi2(240) = 2095.2873
Prob N chi2 = 0.0000.

---

34
A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
only in the following year. Second, since from Aisen
are associated with lower use of seigniorage, which is
and Veiga (2006) higher seigniorage leads to higher
consistent with the findings of Cukierman et al. (1992) 23
inflation, which may affect political instability, using the
and Woo (2003), and with the theoretical model of Woo
contemporaneous value of political instability could
(2005). Democracy does not seem to affect seigniorage, as
create simultaneity/endogeneity problems. Taking the
the Polity Scale is not statistically significant.24 Regarding
first lag avoids these problems, as current seigniorage
the economic variables, only Agriculture (%GDP), Real
does not affect past political instability. Since current
GDP per capita, and Growth of Real GDP(−1) are
seigniorage can affect current economic growth, Growth
statistically significant, with the expected signs. Finally,
of GDP is also lagged one period.21
the coefficients on the decade dummy variables are all
positive and statistically significant.
4. Empirical results
Since Trade (%GDP) and %Change in Terms of
Trade are not statistically significant in the first column,
The first objective of our empirical analysis is to
they are excluded from the model of column 2.25 Results
identify the main political, institutional and economic
remain practically the same. Then, in column 3, Agri-
determinants of seigniorage levels across countries and
culture (%GDP) was replaced by an alternative proxy
time. Then, after finding strong support for our hypothesis
for the structure of the economy, Urban Population
that greater political instability leads to higher seignio-
(% of total), for which there is a higher number of
rage, we try to determine under which circumstances or
observations. The negative coefficient conforms to the
country characteristics this relationship is stronger.
idea that greater urbanization ratios are associated with
Finally, we perform a sensitivity analysis that checks
greater ease to collect taxes and, thus, with lower
whether or not the main results hold for alternative proxies
seigniorage (see Edwards and Tabellini, 1991). The only
of political instability, for an alternative definition of
changes in results are that the Ethnic Homogeneity Index
seigniorage, for a sample that only includes developing
becomes highly statistically significant, and the coeffi-
countries, when our main proxy for political instability
cients of the decade dummies indicate that seigniorage
(Cabinet Changes) is defined in a different way, for a
increased until the 1980s and slightly decreased in the
cross-section and for samples of 5-year and 10-year
1990s. Since this specification of column 3 increases the
periods, when outliers are controlled for, and when
number of observations by 324 (or 16.3%) and the
instrumental variables are used to account for the pos-
number of countries by 7 (or 7%) relative to that of
sibility that some explanatory variables are endogenous.
column 2, it will be used as our reference model.
Results regarding political instability26 conform to
4.1. Main determinants of seigniorage levels
our expectations and are consistent with those found by
Aisen and Veiga (2006) for inflation levels, and with
The estimation results of the model described in the
those of Cukierman et al. (1992) using cross-sectional
previous section, using a fixed effects specification,22 are
data. Those concerning economic variables are consis-
shown in Table 2. The dependent variable is the change in
tent with the findings of previous studies, such as
reserve money as a percentage of government revenues,
Chelliah et al. (1975), Edwards and Tabellini (1991),
and all explanatory variables described in the previous
and Click (1998), indicating that larger agricultural
section were included in the estimation reported in column
1. Results confirm the hypothesis that greater political
23 Although Cukierman et al. (1992) refer to ideological polarization,
instability leads to higher seigniorage levels, and show that
the crucial factor in their model is the polarization of preferences for
the effects are sizeable: an additional cabinet change
different types of government spending, which can also result from
increases seigniorage as a percentage of government
social polarization. Furthermore, higher social polarization is
revenues by 4.15 percentage points. Higher values of the
generally associated with higher ideological polarization.
24
Ethnic Homogeneity Index (lower social polarization)
This is not surprising, as Aisen and Veiga (2006) found that
democracy marginally affects inflation and the effect is very small.
25 They are never statistically significant when included in the
models of the following columns of Table 2 or in those of the
21 The contemporaneous values are used for the remaining
following tables. Wald tests allow for the exclusion of these variables
explanatory variables, since they are taken as exogenous.
from the model.
22 Hausmann tests indicate that the fixed effects specification is
26 The results obtained when using three alternative proxies of
preferable to a random effects model, and the joint statistical
political instability also available in the Cross National Time Series
significance of the country dummies implies that a fixed effects
Data Archive – Government Crises, Executive Changes, and the
model is preferable to a simple pooled OLS model. These results are
Weighted Conflict Index – are very similar. These results are not
available from the authors upon request.
shown here, but are available from the authors upon request.

---

A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
35
Table 2
Results for seigniorage
Seigniorage
1
2
3
4
Cabinet Changes (−1)
4.149
3.688
4.282
4.309
(2.52)⁎⁎
(2.45)⁎⁎⁎
(3.01)⁎⁎⁎
(2.99)⁎⁎⁎
Ethnic Homogeneity Index
−22.776
−22.419
−24.054
−24.747
(−1.78)⁎
(−1.86)⁎
(−2.65)⁎⁎⁎
(−2.78)⁎⁎⁎
Polity Scale
.380
.379
.300
.306
(1.44)
(1.55)
(1.45)
(1.50)
Agriculture (%GDP)
1.748
1.594
(3.62)⁎⁎⁎
(3.57)⁎⁎⁎
Urban population (% of total)
−.486
−.565
(−2.39)⁎⁎
(−2.58)⁎⁎⁎
Trade (%GDP)
.013
(.20)
Real GDP per capita
−.001
−.001
−.002
−.002
(−3.77)⁎⁎⁎
(−4.23)⁎⁎⁎
(−5.32)⁎⁎⁎
(−5.11)⁎⁎⁎
% Change in terms of trade
.89e−07
(1.32)
Growth of real GDP (−1)
−.467
−.432
−.664
−.655
(−2.97)⁎⁎⁎
(−3.05)⁎⁎⁎
(−3.85)⁎⁎⁎
(−3.87)⁎⁎⁎
Dummy1970s
10.247
8.779
7.088
(3.88)⁎⁎⁎
(4.09)⁎⁎⁎
(3.83)⁎⁎⁎
Dummy1980s
18.575
16.998
13.448
(3.97)⁎⁎⁎
(4.17)⁎⁎⁎
(3.85)⁎⁎⁎
Dummy1990s
19.476
17.651
12.367
(3.34)⁎⁎⁎
(3.56)⁎⁎⁎
(2.80)⁎⁎⁎
Trend
1.622
(4.67)⁎⁎⁎
Trend2
−.026
(−4.06)⁎⁎⁎
# Observations
1836
1982
2306
2306
# Countries
97
101
108
108
Adjusted R2
.25
.25
.22
.22
Adjusted R2 (without fixed effects)
.07
.07
Notes: Panel regressions with fixed effects of countries and a constant. T-statistics based on heteroskedastic consistent standard errors are in parenthesis;
Significance level at which the null hypothesis is rejected: ⁎⁎⁎, 1%; ⁎⁎, 5%, and ⁎, 10%; Seigniorage, the dependent variable, was defined as the change in
reserve money (IFS, line 14a) as a percentage of government revenues (IFS line 81).
sectors, lower urbanization ratios, lower GDP per capita
start reducing seigniorage sooner, several developing
levels, and slower economic growth are associated with
countries still had high inflation (or even hyperinflation)
greater reliance on seigniorage revenues.27
and seigniorage in the late 1980s and in the beginning of
The time-dimension of seigniorage is captured by the
the 1990s.28 It is also interesting to note that most
decade dummies (column 3) and by a quadratic trend
explanatory variables, with the exception of Cabinet
(column 4). These indicate that seigniorage increased
Changes, exhibit relatively low time-series variation
until the 1980s, and declined during the nineties. In fact,
within each country. In fact, while Cabinet Changes has
the estimated coefficients of Trend and Trend2 indicate
an average coefficient of variation within countries of
that seigniorage hit its peak in 1990, and declined
1.48, those of the other explanatory variables are all
afterwards. Although one would expect the increased
below .25 (the lowest is .065 for the Ethnic Homo-
independence of central banks in industrial countries to
geneity Index, which varies very little over time).
As mentioned above, the country dummy variables
are always jointly statistically significant. They account
27 The first three variables were not statistically significant in Aisen
for a considerable part of the adjusted R2 of .22 reported
and Veiga (2006). That is, those structural variables help explain
seigniorage but not inflation, supporting our assertion in the
introduction that their determinants are not the same and that separate
28 For example, Argentina had hyperinflation in 1989, Brazil in 1990
studies for inflation and seigniorage should be implemented.
and 1994, Peru in 1990, etc.

---

36
A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
Table 3
Additional determinants of seigniorage
Seigniorage
1
2
3
4
5
6
7
Cabinet Changes (−1)
4.638
4.372
4.299
5.686
5.965
3.150
1.253
(2.62)⁎⁎⁎
(3.07)⁎⁎⁎
(3.03)⁎⁎⁎
(2.59)⁎⁎⁎
(2.78)⁎⁎⁎
(2.76)⁎⁎⁎
(1.51)
Ethnic Homogeneity Index
−56.688
−23.074
−23.869
−86.308
−74.736
−22.404
−6.727
(−3.22)⁎⁎⁎
(−2.62)⁎⁎⁎
(−2.68)⁎⁎⁎
(−1.89)⁎
(−3.29)⁎⁎⁎
(−2.15)⁎⁎
(−.63)
Polity Scale
.529
.266
.313
.550
.148
.121
.178
(1.74)⁎
(1.35)
(1.49)
(1.32)
(.31)
(.49)
(1.18)
Urban population (% of total)
−.573
−.430
−.548
−1.144
−.654
−.502
−.033
(−2.19)⁎⁎
(−2.29)⁎⁎
(−2.46)⁎⁎
(−2.52)⁎⁎
(−1.64)
(−2.21)⁎⁎
(−.22)
Real GDP per capita
−.002
−.001
−.001
.001
−.001
−.001
(−5.15)⁎⁎⁎
(−5.62)⁎⁎⁎
(−1.13)
(1.41)
(−1.94)⁎
(−4.11)⁎⁎⁎
Growth of real GDP (−1)
−.568
−.617
−.616
−.701
−.624
−.510
−.380
(−2.89)⁎⁎⁎
(−3.88)⁎⁎⁎
(−3.85)⁎⁎⁎
(−3.03)⁎⁎⁎
(−2.97)⁎⁎⁎
(−3.43)⁎⁎⁎
(−2.72)⁎⁎⁎
Index of economic freedom
−9.381
(−5.27)⁎⁎⁎
Revolutionary war
12.561
(1.86)⁎
Civil/ethnic conflicts in border states
5.530
(1.99)⁎⁎
Exchange rate regime
−2.416
(−2.91)⁎⁎⁎
Creditworthiness
−.309
(−2.40)⁎⁎
Deposit money bank assets/central bank assets
−32.155
(−1.95)⁎
Liquid liabilities (%GDP)
−3.325
(−.41)
# Observations
1758
2295
2293
1433
1168
2182
1688
# Countries
93
108
108
101
106
107
94
Adjusted R2
.24
.22
.22
.20
.34
.25
.25
Notes: Panel regressions with country fixed effects. T-statistics based on heteroskedastic consistent standard errors are in parenthesis. Significance
level at which the null hypothesis is rejected: ⁎⁎⁎, 1%; ⁎⁎, 5%, and ⁎, 10%; Seigniorage, the dependent variable, was defined as the change in reserve
money (IFS, line 14a) as a percentage of government revenues (IFS line 81); Models estimated with a constant and 3 decade dummies (1970s, 1980s,
and 1990s). Their estimated coefficients are not shown in order to economize space.
in columns 3 and 4. Since a pooled OLS, without fixed
to exchange with foreigners, and more flexible regula-
effects, would only have an adjusted R2 of .07, roughly
tions of credit, labor, and business. Since these are
.15 of the variation in seigniorage is not explained by
characteristics of more advanced economies with lesser
independent variables listed. This also means that more
need of seigniorage financing, the negative coefficient
work needs to be done in this topic in order to improve
found conforms to our expectations. Revolutionary wars
the explanatory power of our models.
in the country and civil/ethnic conflicts in Border States
The results of robustness tests based on the model
(columns 2 and 3, respectively) lead to higher reliance on
of column 3 are shown in Table 3. Those reported in
seigniorage. This result is intuitive, since these occur-
column 1 indicate that higher economic freedom is
rences are associated with larger military spending,
associated with lower reliance on seigniorage. A higher
which may be at least partially seigniorage-financed. The
Index of Economic Freedom29 is associated with smaller
model of column 4 indicates that fixed exchange rates30
governments, stronger legal structure and security of
property rights, access to sound money, greater freedom
30 The result reported in column 7 is for the 5-way classification system
29 Gwartney and Lawson's (2002) data on the Index of Economic
of de facto exchange rate regimes of Levy-Yeyati and Sturzenegger
Freedom starts in 1970 and has a 5-year frequency. In order to avoid
(2003). Results are the same when their 3-way classification system is
missing values, straight line interpolation was used to generate annual
used instead. Since their data starts only in 1974, the inclusion of this
data. Since Access to Sound Money is affected by seigniorage, we
variable originates a large number of missing values. That is why it was
avoided eventual endogeneity problems by using a transformed index
not included in the models of the previous columns. When included, it is
that excludes that area (Area III).
always statistically significant, with a negative sign.

---

A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
37
lead to lower seigniorage levels. A possible explanation
include in our estimations.32 Nevertheless, considering
is that fixed exchange rates constrain monetary policy to
the persistence of our main results across a vast array of
the defense of the fixed parity and, thus, make the
alternative specifications, it might be safe to argue that
collection of seigniorage revenues harder. The results of
they are robust.
column 5 confirm Click's (1998) result that seigniorage
will be higher when the international creditworthiness of
4.2. Circumstances under which the effects of political
the country is lower. That is, when external borrowing is
instability on seigniorage are stronger
less available (or costlier), the government has to rely
more heavily on seigniorage revenues. Finally, the last
Although our results regarding the relationship
two columns test the effects of financial depth, which
between political instability and seigniorage are clear,
Woo (2003) found to be positively related with fiscal
it is possible that they are stronger in some circum-
deficits. Two proxies taken from the database of financial
stances or in countries with specific characteristics.
development and structure of Beck et al. (2000) are used:
Aisen and Veiga (2006) found that political instability
the ratio of deposit money bank assets to central bank
affect inflation levels especially in high-inflation and
assets, and liquid liabilities as a percentage of GDP.
developing countries, whereas that relationship was
Although both have the expected negative sign, indicat-
practically non-existent in low inflation and industria-
ing that countries with more developed financial markets
lized countries. In order to check if the same happens
are more capable of financing higher deficits without
with seigniorage, we performed estimations based in the
resorting to seigniorage, only the first of these variables
model of column 3 of Table 2 in which Cabinet
is statistically significant.31
Changes was interacted with dummy variables account-
Despite all the tests implemented, which involved
ing for annual inflation rates above and below 50% and
regressing seigniorage on a vast array of potential deter-
for developing and industrial countries. Results, illu-
minants, robustness may still be a concern. As the
strated in Fig. 1,33 are consistent with those of Aisen and
empirical economic growth literature has shown (see
Veiga (2006). That is, greater political instability,
Durlauf et al., 1995, and Sala-i-Martin et al., 2004) the
expressed in a higher number of cabinet changes,
parameter estimates obtained in growth regressions
leads to higher seigniorage levels only in high-inflation
are often sensitive to the inclusion of other conditional
and developing countries.
variables. Unfortunately, to our knowledge, there are no
According to Woo (2003, 2005), social polarization,
studies of the robust determinants of seigniorage that
which can be proxied by income inequality and ethnic or
can be used to guide the decision of which variables to
religious heterogeneity/fractionalization, and the quality
of institutions are important determinants of budget
deficits. In highly polarized societies, the high hetero-
geneity of preferences may translate to political parties
31 A series of additional robustness tests, whose results are not
and interest groups lobbying for different types and
shown here, were also performed. First, the Freedom House ratings of
amounts of government spending. Then, high polariza-
Political Rights and Civil Liberties were used instead of the Polity
tion of interests may lead to higher seigniorage, in the
Scale. None was statistically significant. The same result was obtained
when using indicators of Executive Constraints (CNTS) and of
presence of political instability.34 The quality of
Checks and Balances (DPI). Second, indicators of Ideological
institutions is also very important because more
Polarization (DPI), Ideological Orientation (DPI) and Religious
stringent and transparent budgetary procedures, inde-
Homogeneity (SFTF) were added to the reference model, but were not
pendence of the central bank, and greater parliamentary
statistically significant. Third, we also found that trading partners
influence in the budgetary process can reduce the
GDP growth (GDN), external debt (WDI), domestic debt (IFS), de
jure central bank independence (CW), U.S. Treasury Bill rates (IFS),
government's ability to increase budget deficits and
real effective exchange rates (WDI), current account balance (IFS),
extract seigniorage revenues.
government revenues as a percentage of GDP (IFS), and dollarization
ratios (share of dollar deposits) do not affect seigniorage in a
statistically significant way. All results not shown in the paper
32 Implementing an analysis such as that of Sala-i-Martin et al.
are available from the authors upon request. Although the indicator of
(2004) to determine the robust determinants of seigniorage is beyond
Ideological Polarization taken from the DPI was not statistically
the scope of this paper.
significant, we should not interpret this result as a rejection of the model
33 The coefficient obtained for Cabinet Changes (Pol.Instability) in
of Cukierman et al. (1992) in which greater ideological polarization leads
column 3 of Table 2 is shown in the first bar of Fig. 1. The estimation
to higher seigniorage. Since this indicator only takes the values of 0, 1 or
results for the interactions of Cabinet Changes considered in Fig. 1
2, it does not satisfactorily represent the wide differences in ideological
are reported in Appendix C.
polarization among countries. These may be better proxied by the
34 In the model of Cukierman et al. (1992), this high polarization of
indicators of social polarization used in this paper.
interests results in higher seigniorage.

---

38
A. Aisen, F.J. Veiga / Journal of Development Economics 87 (2008) 29–50
Fig. 1. Interactions of political instability. Notes: the grey bars show estimated coefficients of panel regressions: see column 3 of Table 2 for the
coefficient of “Pol.Instability” (Cabinet changes), and in Appendix C, for the remaining coefficients (each pair, separated by vertical lines,
corresponds to a separate estimation). 2-standard error bands are shown on top of the bars. In the horizontal axis, “H.” stands for High, and “L.” stands
for Low. Seigniorage, the dependent variable, was defined as the change in reserve money (IFS, line 14a) as a percentage of government revenues
(IFS line 81). The proxy used for political instability was Cabinet Changes (CNTS).
The hypothesis that the relationship between
politicides. Although both dummies turned out as
seigniorage and political instability is affected by
statistically significant, the magnitude of the coeffi-
social polarization is tested interacting Cabinet
cients implies that the number of cabinet changes in
Changes with dummy variables for average Gini
the previous year (our proxy for political instability)
coefficients above and below 40,35 for high and low
has greater impact on seigniorage in traditionally
ethnic homogeneity,36 and for high and low religious
unstable countries.38
homogeneity. Results clearly support the hypothesis
The hypothesis that institutions affect that relation-
that political instability has stronger effects on
ship was tested interacting Cabinet Changes with
seigniorage in countries with large social polarization
dummy variables for high and low turnover rates of
(high income inequality and low ethnic or religious
central bank presidents,39 for high and low economic
homogeneity). Finally, we test the hypothesis that
freedom,40 and for Polity Scale below and above
political instability will have greater effects on
zero. The results, illustrated in the second and third
seigniorage in countries that have traditionally been
bars of Fig. 2,41 imply that greater political instability
more unstable. Two dummy variables were created
using the variable Upheaval from the SFTF,37 which
indicates the sum of the maximum magnitude of
38 When Cabinet Changes is interacted with regional dummy
events in the prior 15 years, including revolutionary
variables, the positive effect of political instability on seigniorage is
wars, ethnic wars, regime crises, and genocides/
statistically significant only for Western Hemisphere (Latin American)
and African countries. These results are not shown here, but are
available upon request.
39 Cukierman et al. (1995) use this turnover rate as an indicator of de
35 The dummy Gini N40 takes the value of one for countries whose
facto central bank independence. The dummy High Turnover takes
average Gini coefficient is above 40, and equals zero for the
the value of one when the turnover rate is above the sample median of
remaining countries (Gini ≤ 40) = 1 − (Gini N 40).
0.20, and is zero otherwise (Low turnover = 1 − High Turnover).
36 The dummy Low Ethnic Homogeneity takes the value of one for
40 The dummy variable High Economic Freedom takes the value of
countries whose respective index is equal to or lower than the 25th
one when the Index of Economic Freedom is greater than 5, and
percentile, and equals zero for the remaining countries (High Ethnic
equals zero otherwise (Low Economic Freedom = 1 − High Economic
Homogeneity = 1 − Low Ethnic Homogeneity). The same procedure
Freedom). Again, we used a transformed index that excludes Area III
was adopted for the religious homogeneity dummies.
(Access to Sound Money).
37 High Upheaval equals one when the value of Upheaval is above
41 The estimation results for the interactions of Cabinet Changes
3, and equals zero otherwise (Low Upheaval = 1 − High Upheaval).
considered in Fig. 2 are reported in Appendix D.

---

Document Outline

• • 
  • ﾿
  • 
    • ﾿
    • 
  • ﾿
  • 

---