T H IR L W A L L 'S L A W W IT H A N E M P H A S IS

O N T H E R A T IO O F E X P O R T / IM P O R T IN C O M E
E L A S T IC IT IE S IN L A T IN A M E R IC A N E C O N O M IE S
D U R IN G T H E T W E N T IE T H C E N T U R Y ¤

C a r lo s G u e r r e r o d e L iz a r d i

In stitu to T ecn o l¶o gico y d e E stu d io s S u perio res d e M o n terrey

R esu m en : C o n b a se en u n a esp eci¯ ca ci¶o n esto c¶a stica q u e en fa tiza el p a p el d e la

rela ci¶o n d e la s ela sticid a d es in g reso d e la s ex p o rta cio n es e im p o rta -
cio n es, a p lica m o s el m o d elo d e crecim ien to restrin g id o p o r la b a la n za
d e p a g o s a 1 9 p a¶³ses d e A m ¶e rica L a tin a en tre 1 9 0 0 y 2 0 0 0 . P o sterio r a
la p resen ta ci¶o n d e la \ L ey d e T h irlw a ll" v eri¯ ca m o s la ex isten cia d e u n a
rela ci¶o n d e la rg o p la zo en tre ca d a u n a d e la s eco n o m ¶³a s seleccio n a d a s
y la d e E sta d o s U n id o s, p a ra a n a liza r el co rto p la zo estim a m o s u n
m o d elo d e p a r¶a m etro s ca m b ia n tes m ed ia n te el F iltro d e K a lm a n . L o s
resu lta d o s su g ieren u n a red u cci¶o n d el ta m a n~ o d el co e¯ cien te relev a n te
a lo la rg o d el sig lo X X , lo q u e rep resen ta u n a ca ra cter¶³stica in esp era d a
y n eg a tiv a d el m o d elo d e d esa rro llo im p la n ta d o en la reg i¶o n .

A bstra ct: U sin g sto ch a stic sp eci¯ ca tio n s th a t em p h a size th e ro le o f th e ra tio o f

ex p o rt/ im p o rt in co m e ela sticities, th is p a p er a p p lies th e b a la n ce-o f-
p a y m en ts co n stra in t m o d el to n in eteen L a tin A m erica n co u n tries fro m
1 9 0 0 to 2 0 0 0 . T h e p a p er b eg in s w ith a b rief p resen ta tio n o f T h irlw a ll's
w ell-k n o w n m o d el. Im m ed ia tely fo llo w in g th is, w e v erify th e ex isten ce
o f a lo n g ru n rela tio n sh ip b etw een d ev elo p in g eco n o m ies o n o n e h a n d ,
a n d th e U S eco n o m y o n th e o th er. T o ex p lo re th e sh o rt term ev o lu tio n
o f th e q u a n tita tiv e lin k b etw een eco n o m ies, a tim e v a ry in g m o d el is
estim a ted b y m ea n s o f a n a lg o rith m k n o w n a s a K a lm a n ¯ lter. M a in ly,
th e resu lts sh o w a d im in ish in g th e ra tio o f ex p o rt/ im p o rt in co m e ela s-
ticities o v er th e y ea rs, w h ich rep resen ts a n u n ex p ected a n d serio u s
fea tu re o f th e n ew eco n o m ic stra teg y th a t h a s a lrea d y b een im p la n ted
in th e reg io n .

C la si¯ ca ci¶o n J E L : C 2 2 , E 1 2 , F 0 0

P a la bra s cla ve: ley d e T h irlw a ll, L a tin oa m ¶e rica , K a lm a n , T h irlw a ll's la w , L a tin
A m erica , K a lm a n .

F ech a d e recepci¶o n : 6 X 2 0 0 4 F ech a d e a cep ta ci¶o n : 2 6 V II 2 0 0 5

¤
ca rlo s.g u errero .d e.liza rd i@ itesm .m x

23
24 E S T U D IO S E C O N O¶ M IC O S

1 . In tr o d u c tio n

The aim of this study is to test the balance of payments constrained

growth model using stochastic speci¯cations that emphasize the role
of the ratio of export/import income elasticities (Davidson, 1990-91)
for nineteen Latin American economies from 1900 to 2000. The pa-
per begins with a brief presentation of Thirlwall's well-known model.
Immediately following this, using the Johansen procedure we identify
the long-run value of the ratio of export/import income elasticities
and, recognizing that we are in the presence of a time varying pa-
rameter, we draw on a state-space representation to determine its
evolution over time. Concluding remarks are in the ¯nal section.

2 . T h e o r e tic a l F r a m e w o r k

Thirlwall's model can be represented by three equations:

x = ´ (p d ¡ p f ) + ¼ w (1)

m = ' (p f ¡ p d ) + » y (2)

(p d + x ) = (p f + m ) (3)
with ´ ;' < 0 and ¼ ;» > 0. Here, x ;m ;w , and y are the growth
rates of real exports, imports, rest of the world income, and domes-
tic income respectively, and (p d ¡ p f ) is the growth rate of relative
prices measured in a common currency. Equations (1) and (2) are
standard export and import demand functions. Price elasticities of
exports and imports are ´ and ' respectively. Income elasticities of
exports and imports are ¼ and » respectively. It is worthwhile to
highlight that both income elasticities re°ect the non-price aspects
of competition (McCombie and Thirlwall, 1994, p. 265; Bairam and
Dempster, 1991, p. 1720) . Equation (3) assumes that the current
account is continuously balanced.
Substituting equations (1) and (2) into (3) gives the balance of
payments constrained growth model rate of real domestic income,
designated by y b :

¼ w + (´ + ' + 1) (p d ¡ p f )
yb = (4)
»
 T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 25

As Thirlwall (1979, p. 49) points out, if the assumption can

be made that the Marshall-Lerner condition is exactly satis¯ed or if
relative prices measured in a common currency do not change over
the long run, equation 4 can be reduced to:
¼w
yb = (5)
»
In reference to equation (5) , Davidson clari¯es:

\ ...th e ra te o f g ro w th a n a tio n ca n m a in ta in w ith o u t ru n n in g in to a ceteris

pa ribu s b a la n ce o f p a y m en ts p ro b lem d ep en d s o n th e rest o f th e w o rld 's rea l
eco n o m ic g ro w th a n d th e relev a n t in co m e ela sticities fo r im p o rts a n d ex p o rts..."
(1 9 9 0 -1 9 9 1 , p . 3 0 0 ).

The policy implications of equation (5) are relevant in the sense

that in an open economy, pertinent economic management is the one
that manipulates the income elasticities of exports and imports.

\ ...A su ccessfu l eco n o m ic p o licy th a t in crea ses th e v a lu e o f ¼ a n d / o r re-

d u ces th e v a lu e o f » , rela x es th e b a la n ce o f p a y m en ts co n stra in ts a n d , ev en tu a lly,
a ccelera tes eco n o m ic g ro w th ..." (B a ira m a n d D em p ster, 1 9 9 1 , p . 1 7 2 0 ).

Allowing a stochastic residual term (u t ) , equation (5) can be

written as:

y b;t = ® w t + u t (6)
Here ® = ¼» is the coe±cient. It is expected to be positive, and
the larger it is, the better. Its vector autoregressive ( V A R ) form is the
following:

y j;t = ® 0 + ® i y U S ;t¡ i + ® i y j;t¡ i + u j;t

(7)
yU S ;t = ¯ 0 + ¯ iy U S ;t¡ i + ¯ i y j;t¡ i + u U S ;t

Where j represents each of the 19 Latin American countries se-

lected, i indicates the number of lags required, y is the natural log-
arithm of the real gross domestic product, and u t are iid N (0;−)
processes. The following section presents results of econometric esti-
mations of equation (6) under the form of vector autoregressive ( V A R )
models with cointegration, and of state-space representations, for the
selected developing economies.
26 E S T U D IO S E C O N O¶ M IC O S

3 . E c o n o m e tr ic R e su lts a n d D isc u ssio n

As a preliminary step for evaluating equation (6) , table 1 provides

background information about selected countries. Certainly, the an-
alyzed time period for each economy depends on the availability of
information. According to the World Bank, two countries listed be-
long to the category of \low income" (less than 735 U S dollars) , nine
to the \lower middle income" (between 736 and 2 935 U S dollars) ,
and eight to the \upper middle income" (between 2 936 and 9 075 U S
dollars) .
During the 20th century we found two speci¯c developing strate-
gies implemented in Latin America. The ¯rst one corresponds to a
model based on protecting national markets and on state interven-
tion. Under the second one, the tendency has been for the market
to replace regulation, private ownership to replace public ownership,
and competition, including that from foreign goods and investors, to
replace protection. For each economy, the precise cut-o® year was
determined based on a trade index {see Lora (2001) for details{ and
considering the trough of its business cycle.
The question about the order of integration of the variables is
relevant in order to apply the Johansen procedure into a balanced
V A R model. The state of the art econometrics does not accept the
use of the Augmented Dickey-Fuller test and the Phillips-Perron tests,
because of their size distortion and their low power (Maddala and
Kim, 2002, chapter 4) . Thus, we applied the D F -G L S test (Elliott,
Rothenberg and Stock, 1996) and the Perron-Ng test (Perron and
Ng, 1996) . As expected, the results show that the Latin American
and U S economies are integrated of order one.
Once it is clear that the analyzed variables are non stationary, we
can proceed with the estimation of V A R models for real gross domestic
product of the selected countries and of U S . As was established in
the previous section, the relevant parameters represent elasticities.
Therefore, a log-log functional form was chosen. In order to check
their statistical adequacy, misspeci¯cation tests were applied, and
their stability was veri¯ed by means of the long run matrix. 1
1
\ C o n sid er th e im p a ct o f a sh o ck to in n o v a tio n in o n e o f th e eq u a tio n s o f a
V A R ; d o es th e resp o n se to th is sh o ck (ev en tu a lly ) d ie o u t a s w e g et fu rth er a w a y
in tim e fro m th e d a te o f th e sh o ck ? If th e a n sw er is y es th e m o d el is sta b le; if n o t
th e m o d el is u n sta b le...A sta b le m o d el is th u s w ell b eh a v ed in th e sen se th a t th e
im p a ct o f a sh o ck is ca lcu la b le a n d ¯ n ite. T h is is p a rticu la rly im p o rta n t in th e
co n tex t o f w h a t is k n o w n a s m u ltip lier a n a ly sis" (P a tterso n , 2 0 0 0 , p p . 6 0 1 -6 0 2 ).
A cco rd in g to L uÄ tk ep o h l (1 9 9 1 ), th e estim a ted V A R is sta b le if a ll ro o ts o f th e
 T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 27

ch a ra cteristic A R p o ly n o m ia l h a v e a m o d u lu s o f less th a n o n e a n d lie in sid e th e

u n it circle.
28 E S T U D IO S E C O N O¶ M IC O S

.
 T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 29

.
30 E S T U D IO S E C O N O¶ M IC O S

.
 T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 31

Speci¯cally, the following tests were applied: multivariate resid-

ual autocorrelation tests based on Box-Pierce/Lj ung-Box Q-statistics
(LÄu tkepohl, 1991) and on L M statistics (Johansen, 1995) , the mul-
tivariate extension of the Jarque-Bera residual normality test using
{as factorization matrix{ the inverse square root of residual covari-
ance matrix (Urz¶u a, 1997) , and extensions of White's (1980) test for
systems of equations as discussed by Kelej ian (1982) . 2
To determine the lag length of the V A R models, we weighed
the information criterions {Akaike, Schwarz, and Hannan-Quinn{,
and positive/negative results of diagnostic statistical tests. Speci¯-
cally, the lag length for Ecuador was one, for Argentina, Chile, El
Salvador, and Mexico it was two, for Colombia, Costa Rica, Haiti,
Panama, Peru and Uruguay it was three, for Brazil, Dominican Re-
public, Guatemala, Honduras, Nicaragua and Venezuela it was four,
and for Paraguay it was ¯ve. Table 2 shows relevant outcomes.
The estimated ratios of export/import income elasticities are
consistent with ¯gures obtained by other authors. Bairam (1993)
reports 1.20 for Colombia. Senhadj i (1997) , and Senhadj i and Mon-
tenegro (1998) reveal 1.008, 1.275 and 0.566 for Argentina, Colombia
and Haiti respectively. Lopez and Cruz (2000) , and Guerrero (2003)
state 1.692 and 1.640 for Mexico, respectively. Lastly, Moreno-Brid
and Perez (2003) report 1.94 and 1.50 for Costa Rica and El Salvador,
respectively.
The content of table 2 represents positive evidence for Thirlwall's
Law under the form of expression (6) . The long run co-movement
between domestic and U S economies reinforces our belief that the
balance of payment constraint represents a mayor force driving the
economic growth process. In other words, results positively support
{let me propose{ \the rest of the world lead growth hypothesis..." . As
a matter of fact, statistical tests (Johansen, 1995, chapter 8) indicate
{as expected{ that the U S economy is weakly exogenous with respect
to Latin American countries. 3
Noticeably, the estimated ratios of export/import income elas-
ticities were useful to simulate the observed economic growth rates.
Certainly, the di®erence between observed and balance of payments
2
E co n o m etric tests w ere im p lem en ted u sin g a u to m a tic ro u tin es o f E -view s
a n d P cF IM L . In d eed , th e co m p lete set o f sta tistica l resu lts - w h ich in clu d es u n it
ro o t tests, sy stem tests, a n d in fo rm a tio n tests fo r 1 9 V A R m o d els - is h u g e. It is
a v a ila b le fro m th e a u th o r.
3
T h e a ccep ted restrictio n s w ere im p o sed o n th e ® m a trix { w h ich in clu d es
th e a d ju stm en t co e± cien ts. T h u s, if w e so w ish ed , it w o u ld b e v a lid to estim a te
sin g le-eq u a tio n erro r co rrectio n m o d els.
32 E S T U D IO S E C O N O¶ M IC O S

constrained economic growth rates is a consequence of the in°uence of

terms of trade and capital in/out°ows (Thirlwall and Hussain, 1982) .
To examine the annual behavior of the export/import income
elasticity ratios, we split equation (6) under the form of state-space
representation (Hamilton, 1994, chapter 13) . In the observation equa-
tion (8) , the export/import income elasticity ratios are speci¯ed as
time-varying coe±cients, and in the state equation (9) as ¯rst order
autoregressive processes.

T a b le 2
N o rm a lized co in tegra tin g coe± cien ts, U S eco n o m y a vera ge
gro w th ra te, a n d o bserved a n d sim u la ted a vera ge
G D P gro w th ra tes fo r selected co u n tries
 T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 33

y j;t = » t y U S ;t + u j;t (8)

» t+ 1 = ¸ 0 + ¸ 1 » t + u t+ 1 (9)
The system of equations (8) and (9) was estimated using a fore-
cast recursive algorithm known as a Kalman ¯lter. Figure 1 shows
results for the nineteen economies.
It is plausible to group selected economies into four categories.
One distinctive country is Chile. It seems that there is a link be-
tween the early implementation of economic liberalization policies
and the increasing ratio of export/import income elasticities. Dur-
ing the eighties and nineties, which include periods of armed con°ict,
sensibility to U S economy growth rate increased in Central American
countries and the Dominican Republic. For the ¯rst four economies
of the region, Brazil, Mexico, Argentina and Colombia, the external
performance was negative. Unfortunately, Bolivia, Ecuador, Haiti,
Paraguay, Peru, Uruguay and Venezuela, all exhibited the same re-
sult.
F ig u r e 1
S h o rt-ru n beh a vio r o f th e expo rt/ im po rt
in co m e ela sticities ra tio s
34 E S T U D IO S E C O N O¶ M IC O S

F ig u r e 1
(co n tin u ed )
T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 35

F ig u r e 1
(co n tin u ed )
36 E S T U D IO S E C O N O¶ M IC O S

F ig u r e 1
(co n tin u ed )
T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 37

F ig u r e 1
(co n tin u ed )
38 E S T U D IO S E C O N O¶ M IC O S

F ig u r e 1
(co n tin u ed )
T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 39

F ig u r e 1
(co n tin u ed )
40 E S T U D IO S E C O N O¶ M IC O S

F ig u r e 1
(co n tin u ed )
T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 41

F ig u r e 1
(co n tin u ed )
42 E S T U D IO S E C O N O¶ M IC O S

F ig u r e 1
(co n tin u ed )
 T H IR L W A L L 'S L A W IN L A T IN A M E R IC A N E C O N O M IE S 43

4 . C o n c lu d in g R e m a r k s

The ¯ndings suggest the existence of a link among Latin American

and U S economies during the twentieth century. This represents a
piece of empirical evidence in favor of \the rest of the world lead
growth hypothesis" .
During the analyzed period, the annual evolution of the ex-
port/import income elasticities ratio is somewhat di®erent for the
nineteen selected economies. Nevertheless, for the four major coun-
tries of the region -and others-, this ratio declines noticeably over the
last two decades. On the other hand, Chile {the ¯fth major economy{
was the ¯rst country in Latin American to implement economic lib-
eralization measures and, fortunately, has relaxed its balance of pay-
ments constraint. In this sense, and looking at what is happening
right now around the world, the region needs to make a second e®ort
in order to improve its e±ciency.
Finally, from a broader perspective equation (6) emphasizes the
dependency of Latin American countries on U S . Indeed, it is not de-
sirable for the region to discontinue or even to reverse the economic
reform process, but recent disillusionment with the results of this pro-
cess has been growing in our countries, see Lora and Panizza (2002) ,
among others.

R e fe r e n c e s

B a ira m , E . I. (1 9 9 3 ). In co m e E la sticities o f E x p o rts a n d Im p o rts: A R e-ex a m i-

n a tio n o f th e E m p irica l E v id en ce, A p p lied E co n o m ics, 2 5 , 7 1 -7 4 .
| | a n d G . J . D em p ster (1 9 9 1 ). T h e H a rro d F o reig n T ra d e M u ltip lier a n d
E co n o m ic G ro w th in A sia n C o u n tries, A p p lied E co n o m ics, 2 3 , 1 7 1 9 -1 7 2 4 .
D a v id so n , P . (1 9 9 0 -1 9 9 1 ). A P o st K ey n esia n P o sitiv e C o n trib u tio n to `T h eo ry ',
J o u rn a l o f P o st K ey n esia n E co n o m ics, 2 , 2 9 8 -3 0 3 .
D o o rn ik , J . A . a n d D . F . H en d ry (1 9 9 5 ). P cF IM L 8 .0 , C h a p m a n & H a ll.
E llio tt, G ., T . J . R o th en b erg , a n d J . H . S to ck (1 9 9 6 ). E ± cien t T ests fo r a n
A u to reg ressiv e U n it R o o t, E co n o m etrica , 6 4 , 8 1 3 -8 3 6 .
G u errero , C . (2 0 0 3 ). M o d elo d e crecim ien to eco n ¶o m ico restrin g id o p o r la b a la n -
za d e p a g o s: ev id en cia p a ra M ¶e x ico , 1 9 4 0 -2 0 0 0 , E l T rim estre E co n ¶o m ico ,
L X X (2 ), n o . 2 7 8 , 2 5 3 -2 7 3 .
H a m ilto n , J . D . (1 9 9 4 ). T im e S eries A n a ly sis, P rin ceto n U n iv ersity P ress.
J o h a n sen , S . (1 9 9 5 ). L ikelih ood -ba sed In feren ce in C o in tegra ted V ecto r A u to re-
gressive M od els, O x fo rd U n iv ersity P ress.
44 E S T U D IO S E C O N O¶ M IC O S

K elejia n , H . H . (1 9 8 2 ). A n E x ten sio n o f a S ta n d a rd T est fo r H etero sk ed a sticity

to a S y stem s F ra m ew o rk , J o u rn a l o f E co n o m etrics, 2 0 , 3 2 5 -3 3 3 .
L o p ez, J . a n d A . C ru z (2 0 0 0 ). T h irlw a ll's L a w a n d B ey o n d : T h e L a tin A m erica n
E x p erien ce, J o u rn a l o f P o st K ey n esia n E co n o m ics, 3 , 4 7 7 -4 9 5 .
L o ra , D . (2 0 0 1 ). S tru ctu ra l R efo rm s in L a tin A m erica : W h a t H a s B een R e-
fo rm ed a n d H o w to M ea su re It, In ter-A m erica n D ev elo p m en t B a n k , W P n o .
466.
L o ra , D . a n d U . P a n izza (2 0 0 2 ). S tru ctu ra l R efo rm s in L a tin A m erica U n -
d er S cru tin y , In ter-A m erica n D ev elo p m en t B a n k a n d In ter-A m erica n In -
v estm en t C o rp o ra tio n , p rep a red fo r th e sem in a r: R efo rm in g R efo rm s.
L uÄ tk ep o h l, H . (1 9 9 1 ). In trod u ctio n to M u ltip le T im e S eries A n a ly sis, S p rin g er-
V erla g .
M a d d a la , G . S . a n d I. M . K im (2 0 0 2 ). U n it R oo ts, C o in tegra tio n , a n d S tru ctu ra l
C h a n ge, C a m b rid g e U n iv ersity P ress.
M cC o m b ie, J . S . L . a n d A . P . T h irlw a ll (1 9 9 4 ). E co n o m ic G ro w th a n d th e
B a la n ce-o f-P a y m en ts C o n stra in t, S t. M a rtin 's P ress.
M o ren o -B rid J . C . a n d E . P ¶e rez (2 0 0 3 ). L ib era liza ci¶o n co m ercia l y crecim ien to
eco n ¶o m ico en C en tro a m ¶e rica , R evista d e la C E P A L , 8 1 , 1 5 7 -1 7 4 .
P a tterso n , K . (2 0 0 0 ). A n In trod u ctio n to A p p lied E co n o m etrics: a T im e S eries
A p p roa ch , M a cm illa n P ress.
P erro n , P . a n d S . N g (1 9 9 6 ). U sefu l M o d i¯ ca tio n s to S o m e U n it R o o t T ests
w ith D ep en d en t E rro rs a n d T h eir L o ca l A sy m p to tic P ro p erties, R eview o f
E co n o m ic S tu d ies, 6 3 , 4 3 5 -4 6 5 .
S en h a d ji, A . (1 9 9 7 ). T im e-series E stim a tio n o f S tru ctu ra l Im po rt D em a n d E qu a -
tio n s: a C ro ss-co u n try A n a ly sis, IM F , W P / 9 7 / 1 3 2 .
| | a n d C . M o n ten eg ro (1 9 9 8 ). T im e-S eries E stim a tio n o f E xpo rt D em a n d
E qu a tio n s: a C ro ss-C o u n try A n a ly sis, IM F , W P / 9 8 / 1 4 9 .
T h irlw a ll, A . P . (1 9 7 9 ). T h e B a la n ce o f P a y m en ts C o n stra in t a s a n E x p la n a -
tio n o f In tern a tio n a l G ro w th R a te D i® eren ces, B a n ca N a zio n a le d el L a vo ro
Q u a rterly R eview , 1 2 8 , 4 5 -5 3 .
| | a n d M .N . H u ssa in (1 9 8 2 ). T h e B a la n ce o f P a y m en ts C o n stra in t, C a p ita l
F lo w a n d G ro w th R a te D i® eren ces B etw een C o u n tries, O xfo rd E co n o m ic
P a pers, 3 , 4 9 8 -5 1 0 .
U rz u¶ a , C . M . (1 9 9 7 ). O m n ib u s T ests fo r M u ltiv a ria te N o rm a lity B a sed o n a
C la ss o f M a x im u m E n tro p y D istrib u tio n s, in T h o m a s B . F o m b y, a n d R .
C a rter H ill (E d s.), A p p ly in g M a xim u m E n tro p y to E co n o m etric P ro blem s,
A d v a n ces in E co n o m etrics, v o l. 1 2 , 3 4 1 -3 5 8 .
W h ite, H . (1 9 8 0 ). A H etero sk ed a sticity -C o n sisten t C o v a ria n ce M a trix a n d a
D irect T est fo r H etero sk ed a sticity, E co n o m etrica , 4 8 , 8 1 7 -8 3 8 .