Independent Currency Unions, Growth, and Inflation

Abstract

During the last few years, there has been a renewed interest in currency unions. This is the result both of the recent wave of currency crises as well as the implementation of the euro. In this paper, the authors use panel data for 1970-98 to investigate economic performance under historical independent currency unions (ICUs) along three dimensions: GDP per capital growth, growth volatility, and inflation. They use a treatment effects model that estimates jointly the probability of having a common currency and its effect on performance. The authors find that ICU countries have had a significantly lower rate of inflation, but macroeconomic volatility has been higher. Also, ICU countries have grown faster than with currency nations, but the East Caribbean Currency Area countries are found to be the driving force behind this result.

Full text

Page 1

215
Independent Currency Unions,
Growth, and Inflation
Sebastian Edwards and I. Igal Magendzo
MONETARY AND ECONOMIC STUDIES (SPECIAL EDITION)/DECEMBER 2002
During the last few years, there has been a renewed interest in
currency unions. This is the result both of the recent wave of
currency crises as well as the implementation of the euro. In this
paper, the authors use panel data for 1970–98 to investigate
economic performance under historical independent currency
unions (ICUs) along three dimensions: GDP per capital growth,
growth volatility, and inflation. They use a treatment effects model
that estimates jointly the probability of having a common currency
and its effect on performance. The authors find that ICU countries
have had a significantly lower rate of inflation, but macroeconomic
volatility has been higher. Also, ICU countries have grown faster
than with-currency nations, but the East Caribbean Currency
Area countries are found to be the driving force behind this result.
Key words: Currency unions; Dollarization; Inflation; GDP
growth; GDP volatility; Treatment effects
Sebastian Edwards: University of California, Los Angeles, and National Bureau of
Economic Research (E-mail: sebastian.edwards@anderson.ucla.edu)
I. Igal Magendzo: Central Bank of Chile (E-mail: imagendzo@bcentral.cl)
We have benefited from discussions with John Cochrane, Barry Eichengreen, Eduardo
Engel, Ed Leamer, Robert Barro, Andy Rose, Lars Svensson, and Ricardo Hausmann.
Marcela Aurelio provided wonderful assistance. We also appreciate the comments by
Robert W. Rankin and Gabriele Galati.

Page 2

I. Introduction
During the last few years, interest in currency unions has truly exploded. Consider
the following fact: according to the Social Science Citation Index, Mundell’s classical
1961 article on optimal currency areas was cited 88 times in the period 1997–2002
(up to May); in contrast, during the five years from 1982 to 1987 it was only cited
17 times. This renewed interest in currency unions has been largely the result of two
factors. First, the currency crises of the 1990s prompted a number of authors to argue
that the emerging markets should give up their domestic currencies and join currency
unions. Second, after the implementation of the euro zone in 1999, many analysts
have argued that other nations should follow suit and join a monetary union.
Others, however, have claimed that it is much too soon to evaluate the euro
experiment, and that countries in other regions should wait for hard evidence on
economic performance under the euro before joining a currency union. Still other
experts have argued that the calibration of theoretical models can shed light on
the question of whether specific countries should join a currency union (Alesina
and Barro [2001]).
Currency unions, however, have been around for quite some time, and it is
possible to use historical data to analyze economic performance in member countries.
The purpose of this paper is to use panel data for 1970–98 to investigate economic
performance under historical independent currency unions. In this analysis, we
concentrate on countries that use a currency common to the union and issued
by the union’s central bank. We refer to this type of monetary arrangement as an
“independent currency union,” or ICU. In our analysis, thus, we do not focus on
the case of “dollarized” countries, or countries that adopt an advanced nation’s
convertible currency as legal tender. There are important political and economic
differences between that type of arrangement and dollarization: under an ICU,
monetary policy is run by a common central bank, the members of the currency
union share seigniorage, and the common currency’s exchange rate may float relative
to other currencies. Under “dollarization,” on the other hand, the country in
question completely gives up monetary independence, and monetary policy is run by
the advanced nation’s central bank. Countries can “dollarize” in a unilateral fashion—
in which case they will lose the revenue from seigniorage—or they can sign a
monetary treaty with the advanced country and share seigniorage.1Also, there are
important political economy differences between dollarized and ICU countries. As
Frieden (2001) has argued, adopting an advanced country’s currency is usually
perceived as giving up sovereignty, and has serious political costs. These political
costs may be reduced, however, if the country becomes a partner in an ICU and,
consequently, has a say in the running of monetary policy. It is even possible that, by
joining an ICU, the country reaps most of the benefits of a common currency,
216 MONETARY AND ECONOMIC STUDIES (SPECIAL EDITION)/DECEMBER 2002
1. In early 2000, Florida’s senior Senator at the time, Connie Mack, introduced legislation into the U.S. Senate
aimed at sharing seigniorage with countries that decided to adopt the dollar as legal tender. The bill, however,
did not move in the legislative process.

Page 3

without incurring the costs associated with this measure.2Although as mentioned,
our analysis deals mostly with ICUs, we do discuss briefly, in Section IV, some results
pertaining to strictly dollarized countries.
To be more specific, in this paper we ask the following important question: how
have the ICU countries performed relative to countries that have their own currency?
That is, we are interested in evaluating economic performance along three dimen-
sions: inflation, GDP per capita growth, and growth volatility. Performing this type
of international comparison, however, is not easy. The problem is how to define
an appropriate “control” group with which to compare the ICU nations. Since
membership in an ICU is not a “natural experiment,” using a broad control group of
all countries with a domestic currency is likely to result in biased estimates. In this
paper, we tackle this issue by using a treatment effects model that estimates jointly the
probability of having a common currency and its effect on performance (Maddala
[1983], Heckman et al. [1997], and Green [2000]).
Before proceeding, it is useful to point out the ways in which our analysis differs
from other related work in this general area. First, we have sought to include data
on as many ICU countries as possible. We were able to obtain data on GDP per
capita growth and inflation for 34 ICU countries. Second, we focus directly on the
most important macroeconomic variables—real GDP per capita growth, inflation,
and growth volatility. Other studies, in contrast, have analyzed performance in an
indirect fashion, and have focused on ancillary variables such as the level of inter-
national trade and/or interest rates. For instance, Frankel and Rose (2002) have
analyzed the way in which currency unions affect bilateral trade and, through this
channel, economic growth.3Edwards (1998), and Powell and Sturzenegger (2000)
have investigated the way in which the exchange rate/monetary regime affects interest
rate behavior, and the cost of capital. Third, we are particularly interested in estimat-
ing as precisely as possible the actual magnitude of the “ICU effect.” That is, we want
to know, in the most accurate possible way, by how many percentage points countries
under a certain regime have outperformed (or underperformed) countries with an
alternative regime. Obtaining precise estimates of the “ICU effect” is important
for any cost-benefit analysis of a common currency regime. And fourth, we use a
“treatment effects model” to estimate the way in which dollarization affects the
macroeconomic variables of interest.
The rest of the paper is organized as follows. In Section II, we provide a prelimi-
nary analysis of historical experience with ICUs. In Section III, we use treatment
regressions to analyze the effects of “common currencies” on a group of macro-
economic variables. In Section IV, we undertake a robustness analysis and, finally, in
Section V we provide some concluding remarks.
217
Independent Currency Unions, Growth, and Inflation
2. In our analysis, we consider countries that adopt a nonconvertible currency as their own as an ICU. If these few
countries are excluded from the analysis, the results reported in this paper do not change, however.
3. See Klein (2002) for a discussion on dollarization and trade, including a comprehensive bibliography on the
subject.

Page 4

II. Independent Currency Unions during 1970–98:
A Preliminary Analysis
In Table 1, we present a list of 33 ICU countries with available data for the period
1970–98.4In addition, we present a list of 21 strictly dollarized countries, or
countries that have used an advanced country’s currency as legal tender. Two ICUs
dominate the list in Table 1: the Communauté Financière Africaine (CFA) franc
zone, and the East Caribbean Currency Area (ECCA), with 15 and seven members,
respectively. Both of these ICUs have a central bank of their own, and in an effort to
boost credibility, both of these areas have pegged their exchange rates to an advanced
nation. The CFA franc zone is pegged to the French franc, and has an agreement
with France to finance balance of payments disequilibria. In 1994, and after years
of overvaluation and external imbalances, the CFA was devalued and repegged to
the French franc. Until 1975, the ECCA’s East Caribbean dollar was pegged to the
British pound; since that year, it has been pegged to the U.S. dollar.
218 MONETARY AND ECONOMIC STUDIES (SPECIAL EDITION)/DECEMBER 2002
Table 1 Monetary Unions with Available Data
CFA Franc Zone
Benin
Burkina Faso
Cameroon
Central African Republic
Chad
Comoros
Congo
Côte d’Ivoire
Equatorial Guinea
Gabon
Guinea-Bissau
Mali
Niger
Senegal
Togo
United States
Liberia (D)
Marshall Islands (D)
Micronesia, Fed. States of (D)
Palau (D)
Panama (D)
Puerto Rico (D)
France
Andorra (also Spanish peseta) (D)
French Guiana (D)
French Polynesia
Guadeloupe (D)
Martinique (D)
Monaco (D)
New Caledonia
Reunion (D)
ECCA
Antigua and Barbuda
Dominica
Grenada
Montserrat
St. Kitts and Nevis
St. Lucia
St. Vincent and the Grenadines
South Africa
Lesotho
Namibia
Swaziland
New Zealand
Cook Islands (D)
Italy
San Marino (D)
Australia
Kiribati (D)
Nauru (D)
Tonga (D)
Tuvalu (D)
West Africa
Kenya
Tanzania
Uganda
India
Bhutan
Singapore
Brunei
Denmark
Greenland (D)
Switzerland
Liechtenstein (D)
Belgium
Luxembourg (D)
Note: (D) corresponds to a dollarized country.
4. These countries have data for a long enough period for at least one of two variables: GDP per capita or inflation.
In the rest of the paper, we will use the term “countries” to refer both to independent countries and to territories.

Page 5

In Table 2, we present comparative data on inflation, per capita GDP growth, and
the standard deviation of growth for our ICU countries. To put things in perspective,
we also present data on these three variables for an “unadjusted” control group that
includes all countries with a currency of their own. In this table, we include data
on the mean and median for the three macroeconomic variables. In column (C), we
present data on mean and median differences between the common currency
countries and the “with currency” control group. The numbers in parentheses are
t-statistics for the significance of these differences. The test for the mean differences
is a standard t-statistic, while the median differences test is a t-test obtained using a
bootstrapping procedure. In making the computations for inflation differentials, we
have followed Engel and Rose (2002) and have excluded countries with hyperinfla-
tion.5However, excluding these observations only affects the calculation of the mean
differences; it has no discernible effect on the computation of median differences.
The results reported in this table indicate that the difference in inflation means is
quite sizable and statistically significant; on average, inflation in ICU countries as a
group (Panel A) has been 7.7 percentage points lower than in countries with their
own currency.6The difference in inflation medians is still negative, much smaller
(–2 percentage points), and still statistically significant. In terms of real per capita
GDP growth, the results in Table 2 show that there are no significant differences in
219
Independent Currency Unions, Growth, and Inflation
5. More specifically, we excluded from the control group those observations with a rate of inflation in excess of
200 percent per year. This resulted in 80 observations being dropped from the control group of countries with a
currency of their own. See Section IV for results under alternative definitions of “very rapid inflation.”
6. When hyperinflation countries are not excluded, the means difference in inflation is a staggering 62 percent.
Table 2 Inflation, Growth, and Volatility
ICUs versus control group
Independent
currency unions1
(A)
Other countries2
(B)
Difference3
(C) = (A) – (B)
A. Inflation
Mean8.9016.59–7.69
(–7.07)
–2.24
(–4.86)
Median7.229.46
B. Per capita GDP growth
Mean1.361.160.20
(0.76)
–0.58
(–3.88)
Median1.301.88
C. Volatility of growth
Mean6.235.310.92
(7.80)
0.73
(7.20)
Median5.314.58
Notes: 1. Number of observations with data for inflation is 533. There are 804 observations with data
for per capita growth.
2. Number of observations with data on inflation is 2,831. There are 3,933 observations with
data for per capita GDP growth.
3. Numbers in parentheses are t-statistics.