Trade Liberalisation, the Income Elasticity of Demand for Imports and Growth in Latin America

Trade Liberalisation, the Income Elasticity of Demand for Imports
and Growth in Latin America

Penélope Pacheco-López and A.P. Thirlwall1

University of Kent



Abstract

This paper applies the balance of payments constrained growth model to seventeen
countries of Latin America over the period 1977-2002. The crucial parameter to
estimate is the income elasticity of demand for imports which is done for Latin
America as a whole, as well as for individual countries. As well as estimating over the
whole period, the technique of rolling regressions is also used to test whether a trend
increase can be discerned as a result of trade liberalisation. A trend increase is found
for Latin America as a whole and for some individual countries, and the balance of
payments equilibrium growth rate is a good predictor of growth performance in nine
of the seventeen countries. There is no evidence that the balance of payments
equilibrium growth rate has increased in Latin America as a result of trade
liberalisation.




Key words: Latin America; Trade Liberalisation; Growth; Balance of Payments.


JEL Classification: C21, C22, F13, F32, F43.



Address for Correspondence: Professor A.P. Thirlwall, Department of Economics,
Keynes College, University of Kent, CT2 7NP, Canterbury, Kent. E-
mail:a.p.thirlwall@kent.ac.uk







1 The authors are, respectively, Post Doctoral Research Associate and Professor of Applied Economics
at the University of Kent, Canterbury, UK. Professor Thirlwall is grateful to the Leverhulme Trust for
the award of an Emeritus Fellowship which provided financial support for the research project.

1

---

Trade Liberalisation, the Income Elasticity of Demand for Imports
and Growth in Latin America




1. Introduction

The countries of Latin America have undergone extensive trade liberalisation in the
last twenty years or so, but their growth record has been relatively poor, despite the
promises and expectations of the advocates of liberalisation. As Rodrik (2004)
remarks in his recent WIDER lecture: “Latin America [during the 1990s] grew more
slowly not only compared to other parts of the world… but also compared to its own
performance in the 1960s and 1970s. That is a striking empirical fact, the importance
of which it is hard to downplay. After all, Latin America of the 1960s and 1970s was
a region of import substitution, macroeconomic populism, and protectionism, while
the Latin America of the 1990s is a region of openness, privatisation and
liberalisation. The cold fact is that per capita economic growth performance has been
abysmal during the 1990s by any standards”.

One explanation is that the process of trade liberalisation in Latin America has been
too drastic and sudden without giving time, or incentives, for producers to switch to
the production of tradeable goods. The consequence has been a surge of imports,
without a corresponding increase in the growth of exports, and output has grown
below capacity which has led in many countries to a rise in unemployment and a fall
in unskilled real wages. This contrasts with the experience of South East Asia in the
1960s and 1970s where the pace of import liberalisation was much slower, and was
combined with an export-led growth strategy which kept the balance of payments in

2

---

equilibrium without the necessity for income adjustment. In their survey of balance of
payments liberalisation in a selection of Latin American and Caribbean countries,
Vos, Taylor and Paes de Barros (2002) remark: “higher import demand and typically
lagging exports meant that the trade deficit went up for a given level of output” and
“higher import propensities offset the growth impacts of export expansion that nearly
all countries witnessed. Although exports gained importance as a source of
growth…the gains do not seem to have been so strong as originally supposed by
advocates of liberalisation”.

The purpose of this paper is to examine whether trade liberalisation in seventeen Latin
American countries2 has led to an increase in the income elasticity of demand for
imports (allowing for the effect of real exchange rate changes on the growth of
imports) which would slow economic growth for any given growth of exports unless
there were compensating capital inflows. This is a test for Latin America of the
balance of payments constrained growth model, originally developed by Thirlwall
(1979).








2 The countries are: Argentina, Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic,
Ecuador, El Salvador, Guatemala, Honduras, Mexico, Nicaragua, Paraguay, Peru, Uruguay, and
Venezuela.

3

---

2. The Balance of Payments Constrained Growth Model

The central proposition of the balance of payments constrained growth model is that
no country can grow faster than that rate consistent with balance of payments
equilibrium on current account unless it can finance ever-growing deficits, which, in
general, it cannot. There is a limit to the current account deficit, or external debt, of a
country relative to its gross domestic product (GDP). Thirlwall’s original model
assumed strict current account balance over the long term, but the model can be
extended to include permanent, sustainable capital inflows (Thirlwall and Hussain,
1982) or a fixed current account deficit, or debt, to GDP ratio (McCombie and
Thirlwall, 1997; Barbosa-Filho, 2001; Moreno-Brid, 1998-99). It turns out that for
most countries these extensions make very little difference empirically to the long run
predictions of the model which is dominated by the rate of growth of export volume
(x) of a country (determined by world income growth and the income elasticity of
demand for exports) relative to the income elasticity of demand for imports (π), i.e.:
yb = x/ π








(1)
where yb is the growth of income consistent with balance of payments equilibrium on
current account.3 This result can be shown (Thirlwall, 1982) to be the dynamic
equivalent of the static Harrod foreign trade multiplier (Harrod, 1933).

To date, there has been a paucity of studies applying the balance of payments
constrained growth model to Latin American countries. There are some individual

3 With a fixed debt to GDP ratio (McCombie and Thirlwall, 1997) or a fixed export to import ratio
(Moreno-Brid, 1998), the growth rate can be shown to be: y *
b = θx / [π – (1-θ)]. θ = X/(X+F) =
(X/Y)/[X/Y + F/Y], where X/Y is the ratio of exports to GDP and F/Y is the current account deficit to
GDP ratio financed by capital inflows (F). This makes very little difference to the basic result for
reasonable values of the ratios. Suppose, for example, that the export ratio is 30 per cent, the current
account deficit is 4 per cent of GDP, and π = 1.5, then y *
b = 0.64 x, compared with 0.67 x, assuming
balance of payments equilibrium.

4

---

case studies, which are supportive, such as for Mexico (Moreno-Brid, 1998, 1999;
López and Cruz, 2000; Pacheco-López, 2005); for Brazil (Ferreira and Canuto, 2003);
for five Central American countries (Moreno-Brid and Pérez, 1999); and a wider
study of 34 developing countries, including some in Latin America (Perraton, 2003),
but there is no study for Latin America as a whole either taking all the major countries
individually, or pooling the data. We have assembled a consistent data set of GDP
growth; export growth; import growth; real exchange rate changes, and ‘world’
income growth (proxied by the growth of the US economy) for seventeen countries of
Latin America over the period 1977 to 2002, which we use to do a number of
estimations.4

Firstly, we pool the data; estimate an income elasticity of demand for imports for the
whole region, and fit the basic model of yb = x/ π (equation 1 above). We then
compare yb and actual GDP growth (y) over the whole period. Secondly, we do the
same exercise as above for individual countries using appropriate time series
econometric techniques. The McCombie test (McCombie, 1989) is then used to see
whether the estimated income elasticity of demand for imports (πˆ) is significantly
different from the income elasticity (π*) that would make the actual GDP growth (y)
equal to the balance of payments equilibrium growth rate (yb). If it is not significantly
different, then yb will be a good predictor of y.

Thirdly, we run rolling regressions (first used in this field by Atesoglu, 1993, 1994)
using pooled data taking overlapping periods. Thirteen overlapping periods are taken
starting with 1977-1990 and finishing with 1989-2002. Trade liberalisation may be

4 The data sources are: the World Development Indicators CD-Rom 2004, World Bank; Insituto de
Pesquisa Econômica Aplicada (www.ipea.gov.br); Banco Central de Nicaragua (www.bcn.gob.ni).

5

---

expected to gradually raise the income elasticity of demand for imports through time,
and to constrain growth unless export growth improves. The basic model is then
applied to each overlapping period and the predicted growth rate compared with the
actual growth rate, also using the McCombie test.

Fourthly, we do the same rolling regression exercise as above for each individual
country in the sample.

We conclude the paper with a brief commentary on the process of trade liberalisation
in Latin America, and the lessons to be learned.


3. Fitting the Balance of Payments Constrained Growth Model Using
Pooled Data

The derivation of the balance of payments equilibrium growth rate (or the dynamic
Harrod trade multiplier result) of yb = x/ π is well known, and need not be repeated
here (see the collection of essays in McCombie and Thirlwall, 2004, for the original
derivation of the model and its application). The crucial parameter to be estimated for
testing the model is the income elasticity of demand for imports (π). To do this, we
specify a conventional multiplicative import demand function of the form:
M
A RER ψ Y π
=







(2)
t
t (
t )
t
where M is the volume of imports; RER is the real exchange rate;5 ψ is the price
elasticity of demand for imports (<0); Y is domestic income (as a proxy for

5 Measured as the domestic price of foreign currency (the nominal exchange rate), multiplied by the
ratio of foreign to domestic prices. A rise in RER represents a depreciation of the currency.

6

---

expenditure), π is the income elasticity of demand for imports (>0), and t is a time
subscript. Taking logs, differentiating with respect to time, and adding a constant term
gives an estimating equation of the form:
mt = a + ψ (rer) + π (y) + et





(3)
where lower case letters represent the rate of growth of the variables and et is an error
term with the usual properties. Since the number of countries (17) is less than the
number of time series observations (26), we use a pooled time-series/cross section
estimate to determine the value of π, allowing for groupwise heteroscedastic and
correlated regressions with group specific autocorrelation.6 The fitted result with
t-values in brackets is:

mt = – 0.32 – 0.069 rer + 2.29 y




(4)
t-statistic (-0.92) (-5.74 ) (32.79)

Diagnostic statistics
Likelihood Ratio Statistic: 199.2
Number of Observations: 425

Both variables have the expected sign and the income elasticity of demand is well
determined with an estimate of 2.29. The average growth of exports (x) over all
countries and all years is 5.49 per cent per annum. Fitting the basic model of yb = x / π
gives a predicted growth rate for the Latin American region as a whole of
5.49/2.29 = 2.40 per cent. This compares with the actual growth rate of the region of
2.67 per cent. The actual growth rate and balance of payments constrained growth rate
are clearly very close. Using the McCombie test, the income elasticity of demand for
imports that would make y = yb is 2.05. This is not significantly different from the
calculated π of 2.29.


6 Using the econometric software package LIMDEP.

7

---

4. Fitting the Balance of Payments Constrained Growth Model to
Individual Countries

In this section we estimate π for each of the 17 countries using equation (3). First we
test the order of integration of the variables using the Augmented Dickey-Fuller test.
For ten countries all the variables are I(0): Argentina, Costa Rica, Dominican
Republic, Ecuador, El Salvador, Honduras, Mexico, Nicaragua, Paraguay, Peru, and
Venezuela. For the other seven countries, some of the variables are I(0) and others are
I(1). Before proceeding to estimation for these countries we therefore use the Pesaran
et. al. (2001) test to assess whether there is a long run relationship between the
variables regardless of their order of integration. Six of the seven countries (Bolivia,
Brazil, Chile, El Salvador, Guatemala, and Uruguay) pass the test either at the 95 or
90 per cent confidence level.7 The exception is Colombia, and is excluded from the
sample.

The results of fitting equation (3) to the data for individual countries are shown in
Table 1. The real exchange rate parameter (ψ) is significantly negative in five
countries −Argentina, Chile, Costa Rica, Mexico and Peru− but the magnitude of the
coefficient is very small. In the remaining countries, however, the coefficient is
positive or insignificant suggesting that the exchange rate is not an efficient balance
of payments adjustment weapon at least for curbing imports. By contrast, the income
elasticity of demand for imports (π) is a well-determined parameter in fourteen out of
the sixteen countries, which enables the balance of payments constrained growth
model to be fitted to these countries. The two countries where the income elasticity of

7 The F statistics for each country are available on request.

8

---

Table 1
Import Demand Equation, 1977-2002
Country Constant
π
ψ
Argentina
4.66 (1.68)
3.66 (7.58)
-0.13 (-1.91)
Bolivia
-0.16 (-0.06)
1.82 (2.25)
0.00 (0.25)
Brazil
-0.26 (-0.08)
1.59 (2.25)
-0.31 (-1.82)
Chile
0.04 (0.01)
2.03 (6.17)
-0.54 (-3.91)
Costa Rica
-1.55 (-0.80)
2.27 (6.34)
-0.20 (-3.25)
Dom. Rep.
3.16 (0.48)
0.92 (0.76)
0.04 (0.26)
Ecuador
-1.17 (-0.42)
1.83 (2.62)
0.00 (0.09)
El Salvador
1.65 (0.86)
2.47 (6.68)
0.09 (0.68)
Guatemala
-7.34 (-1.81)
3.78 (3.60)
0.18 (0.77)
Honduras
-1.69 (-0.83)
1.41 (3.22)
-0.01 (-0.12)
Mexico
0.06 (0.01)
3.17 (5.40)
-0.18 (-2.71)
Nicaragua
5.08 (1.06)
0.97 (1.59)
-0.00 (-0.28)
Paraguay
0.29 (0.04)
2.48 (2.53)
-0.24 (-0.99)
Peru
0.45 (0.22)
1.56 (4.85)
-0.37 (-2.98)
Uruguay
1.42 (1.09)
2.13 (6.27)
-0.07 (-0.79)
Venezuela
-0.39 (-0.10)
3.76 (4.59)
0.15 (0.74)
Note: t-values are in parentheses.


demand for imports is not significant are the Dominican Republic and Nicaragua. The
magnitudes of the estimated elasticities are generally higher than previous studies
have found for earlier time periods. Senhadji’s, (1998) study of import demand
functions for 66 countries contains twelve of the countries taken here and covers the
period 1960 to 1993. Perraton’s (2003) study of 51 developing countries found 34
countries with significant income elasticities of demand for imports over the period
1973 to 1995 and contains nine of our sample of countries. Their results and ours are
compared in Table 2. Our generally higher estimates for most countries no doubt
partly, if not mainly, reflect the process of trade liberalisation which gathered
momentum in many countries in the 1990s.




9

---

Table 2
A Comparison of the Estimates of the Income Elasticities
of Demand for Imports
Country
Our Estimates
Senhadji Estimates
Perraton Estimates

1977-2002
1960-19931
1973-19952
Argentina 3.66
1.21 3.01
Bolivia 1.82 n.a. 0.50
Brazil 1.59 1.24
1.77
Chile 2.03 1.87
n.a.
Costa Rica
2.27
1.21
1.76
Dom. Rep.
0.92
0.86
1.15
Ecuador 1.83 n.a. 0.24
El Salvador
2.47
1.47
n.a.
Guatemala 3.78
n.a.
n.a.
Honduras 1.41
0.74 0.56
Mexico 3.17 1.31 n.a.
Nicaragua 0.97
0.57 n.a.
Paraguay 2.48
1.58 n.a.
Peru 1.56
0.50
0.94
Uruguay 2.13 5.48 n.a.
Venezuela 3.76
n.a. 2.78

Notes: 1 Long-run estimate using the Phillips-Hansen Fully Modified Estimator.
2 Estimated using an Autoregressive Distributed Lag Model.

The results of fitting the balance of payments growth model for each of the 16
countries are shown in Table 3. The first column gives the growth of exports (x); the
second column gives the estimated income elasticity of demand for imports (π); the
third column gives the estimated balance of payments equilibrium growth rate (yb);
the fourth column gives the actual growth rate (y) for comparison; the fifth column
gives the deviation between the actual and balance of payments equilibrium growth
rate; the sixth column gives the calculated income elasticity of demand for imports
(π*) that makes y = yb; and the last column gives the McCombie test of whether the
estimated π is equal to π*. It can be seen that the McCombie test is passed for nine
countries, and fails for seven. For the countries that pass the test the actual growth rate
is very close to the balance of payments equilibrium growth rate. For those that fail
the test, some have growth rates below yb, suggesting the accumulation of balance of

10

---