1. INTRODUCTION

The Covid-19 pandemic caused the current economic crisis which in turn has slowed down most of the world economies’ Gross Domestic Product (GDP) growth rates. This rekindled the debate on whether strategies can positively contribute to reviving economic activity in the short and long term, as well as stimulate processes of sustained economic growth.

Currently, within the main Latin American economies, an export-led growth strategy dominates (see ^{Perrotini et al. 2008}; ^{Kristjanpoller and Olson, 2014}).¹ One option borne from the economic crisis instigated by the Covid-19 pandemic is to continue with this model; however, the first signs of recovery seem to indicate that this policy will not be enough to reach the previous levels of economic activity, let alone surpass them (see ^{Gamba, 2021}; ^{Werner et al., 2021}).²

With regard to the export growth model, the strategy itself is not considered to be the problemper se, but rather that the trade liberalization processes were not those required to stimulate the export growth rate nor its productive linkage with the rest of the sectors.³One can also argue that the region’s sluggish growth is the result of a dramatic drop in the rate of capital accumulation (cf. ^{Perrotini et al., 2008}), so maybe the time is right for a change in the growth model, for a capital accumulation-led strategy to rise as an alternative capable of better results.

On the one hand, it is important to keep in mind that the liberalization of international trade can be justified, from the orthodox approach, as a measure to better position a country’s productive resources (cf. ^{Feder, 1983}) and to solve the shortage of capital goods needed to increase aggregate productivity (cf. ^{Awokuse, 2008}). However, it can also be supported by the heterodox approach as a way to relax restrictions on effective demand by complementing the domestic market with the external market (cf. ^{Kaldor, 1970}; ^{Thirlwall, 1979}). Likewise, fostering capital accumulation can be brought about by stimulating savings (cf. ^{Solow, 1956}) or through a direct policy to drive investment, derived from post-Keynesian growth models (cf. ^{Kurz and Salvadori, 2010}). In this regard, this article’s argument is housed within the heterodox approach even though it follows a causality flow framed by an open economy which employs Harrod's postulate on the dual-role of investment (see ^{Moudud, 2000}) as a source of demand and, therefore, potentially of a demand for imports and as a mechanism to replace imports and expand the domestic market given its ability to generate productive capacity.

Thus, the objective of this work is to show that investment, not exports, is the engine of growth in Latin America, both before and after the external debt crisis of the 1980s. We also seek to show that the reduction in capital accumulation seen in the region fundamentally explains the reduction in growth. To this end, we analyze four of the largest economies in the region: Argentina, Chile and Mexico during 1961-2019, and Brazil during 1971-2019.

This article is divided into four sections, including this introduction. The second section provides a brief exposition of ^{Kaldor’s (1970}) and ^{Thirlwall's (1979}) export growth models, and a part that integrates in Thirwall's model the role capital accumulation has in determining the growth rate. The third section presents an empirical analysis showing that capital accumulation is more relevant than export growth rate in determining the study’s sample economies’ growth rate. Finally, the fourth section presents some final thoughts on the subject.

2. EXPORT GROWTH MODELS AND CAPITAL ACCUMULATION

Economic growth theories provide various explanations for the differences seen in international growth rates, both between countries and between different time periods. Among the spectrum of explanations we find at one end those based on supply, that is the accumulation of productive factors through growth models both exogenous (cf. ^{Solow, 1956}; ^{Mankiw et al., 1992}) and endogenous (cf. ^{Romer, 1986} and ¹⁹⁹⁰; ^{Lucas, 1988}; ^{Barro, 1990}; ^{Rebelo, 1991}); at the other end are those based on demand (cf. ^{Thirlwall, 1979}; ^{Clavijo and Ros, 2015}).

^{Kaldor (1957}), following a Keynesian approach at the demand end of growth theory, developed a model in which, in the short term, economic growth rates depend on capital accumulation. However, in the long run, growth depends on purely technological factors. In this case, differences seen in growth can be divided into two groups: short-term and long-term.

Thus, in the short term, different growth rates are explained by the differences in how the profits generated by income are distributed, with those economies which have the smallest share of profits growing the least. In the long run, growth rates depend on product’s elasticity in regards to capital and on the flow of new ideas and economies’ willingness to adopt them. However, regardless of its Keynesian nature, ^{Kaldor’s model (1957}) implies that in the long run growth is restricted by the accumulation of productive resources.

The idea that the long-term growth rate can be determined by demand and not by the accumulation of productive factors can be traced back to ^{Kaldor's (1970}) export growth model,⁴ which was formalized by ^{Thirlwall (2003}). This model is based on the “Verdoorn Law” and on the belief that exports are the only truly autonomous component of demand (cf. Thirlwall, 2003), thereby producing a virtuous cycle of more exports, more growth, more productivity, and more exports. The process described can be expressed in the following manner, assuming that the growth rate of product (g) depends on the growth rate of exports (x):

where γ is output elasticity in regards to exports; while assuming thatxdepends on the percentage variation rate of the real exchange rate (e+ π*- π) and the foreign product growth rate (z):

whereeis the nominal depreciation/appreciation rate, π* and π are the external and domestic inflation rates respectively, ε _x is the price elasticity of export demand and Ψ* is the income elasticity of the country’s export demand with regards to the world or relevant trading partners. In turn, π depends on the change percentage in nominal wages (w), labor productivity (l) and profit margin (τ):

The importance of the Verdoorn coefficient lies in the idea thatlis endogenous tog:

Wherel ₀ is the autonomous growth rate of labor productivity. The equilibrium growth rate is obtained by substituting equation (4) in (3), the result in equation (2) and the new result in equation (1), from which the long-term growth rate (ց _LP ) is obtained:

Equation (5) shows that when the Verdoorn coefficient is equal to zero, ց _LP is less than when it is between zero and one; therefore, equation (5) also shows a positive relationship between domestic growth rate and foreign growth rate, or in other words, between domestic growth rate and export growth rate.

According to ^{Thirlwall (1979}), ^{Kaldor's (1970}) export growth model suffers from omitting imports and their effect on the trade balance in the economic growth process. In this regard, Thirlwall (1979) argues that while long-term growth depends onx, the trade balance’s dynamic equilibrium is the main constraint to economies’ growth, which he demonstrates based on ^{Harrod's (1933}) theory of an external multiplier. Thirlwall’s model (1979) can be formalized on the basis that economies must maintain trade balance’s dynamic equilibrium in the long term; this condition, in terms of domestic goods, can be expressed as follows:

whereθis the variation percentage in the real exchange rate (e+ π*- π) andmis the import growth rate. We likewise assume that the behavior ofxcan be expressed by equation (2) and that, symmetrically,mdepends onθand on ց:

where ε _m is the price elasticity of import demand and Ψ is the income elasticity of the domestic country's import demand. Substituting equations (2) and (7) in (6) and solving forgyields the long-term growth rate consistent with the trade balance’s dynamic equilibrium (ց _tb ):

In equation (8) we see that ց _tb has a positive relationship withθandz. Nevertheless, ^{Thirlwall (1979}) emphasizes the fact that even if the Marshall-Lerner condition is met (ε _x + ε _m - 1 > 0), it is impossible to use the real exchange rate as a permanent method for increasing ց _tb To this we add the fact that it is normal for price elasticities to be very low and/or for relative prices to be more or less constant,⁵with which Thirlwall (1979) discards the effect of θ on ց _tb . On the other hand, given that Ψ *z represents x, equation (8) can be restated as what is called the weak version of Thirlwall's Law:

so, according to equation (9), givenΨ,long-term growth depends onx. Now, according to ^{Vázquez-Muñoz (2018}), the long-term predictions provided by the weak version of Thirlwall's Law are not plausible as they imply that whenΨis greater than one, the economy tends to only produce for the external market and, when it is less than one, tends to become a closed economy, which is not feasible in the real world.

Likewise, according to ^{Vázquez-Muñoz (2018}) and ^{Ibarra (2015}), the traditional specifications of the equations ofxandmyield biased estimates ofΨandΨ*, given that they do not take into consideration the effects of capital accumulation, such as whether an expansion into the foreign market is accompanied by domestic capital accumulation or not. In the former, the increase in export demand will be accompanied by a greater supply of domestic goods, while in the latter it will occur without a response from domestic production. This means that in the first case the estimate ofΨ* is greater than in the second case. Likewise, if domestic expansion occurs without capital accumulation there will be an increase inmwhile, if it happens in the presence of capital accumulation, the response of imports will be low, meaning that in the first case the estimate ofΨwill be higher than that in the second case.

Thus, ^{Vázquez-Muñoz (2018}) expands ^{Thirlwall’s model (1979}) in order to include the role of capital accumulation in import demand. According to Vázquez-Muñoz (2018), capital accumulation has two effects onm, one direct and positive stemming from the need to import capital goods, and another negative and indirect resulting from the possibility of replacing imported goods with domestically produced goods. The function of the import growth rate is then respecified as:

whereΨ _I is import demand’s gross capital stock elasticity,Iis gross investment andceis the growth rate of economic capacity, understood as the desired level of production, given capital stock (K) (cf. ^{Shaikh, 2016}). That is,ceis a variable which depends on I/K:

whereais the growth rate of capital productivity. We likewise omit in equation (10) the effect of the real exchange rate on import demand. Therefore, assuming thatθis equal to one and remains constant, and considering thatΨ*zis equal tox, substituting equations (10) and (11) in (6), one obtains the long-term growth rate consistent with trade balance’s dynamic equilibrium which considers the role of capital accumulation (ց _tbI ):

according to equation (12), the long-term growth rate in a model constrained by the trade balance’s dynamic equilibrium not only depends onx, but also onI/K.

Furthermore, we must add that according to ^{Perrotiniet al.(2019)}, given that the adjustment variable in the model isg, it is important to differentiate between its components, that is between the growth of domestic demand for domestic goods (id) andx, as the adjustment will be verified in the former. This distinction is important because, contrary to the traditional postulate thatximproves trade balance, it can happen thatxworsens it due to the multiplier effect on income, givenΨ,and thatxtherefore negatively affectsid.

Next, we present empirical evidence on the importance ofxandI/Kin determining growth in economies constrained by the trade balance.

3. CAPITAL ACCUMULATION AND EXTERNAL CONSTRAINT, THE CASES OF ARGENTINA, BRAZIL, CHILE AND MEXICO

In this section we analyze the role of capital accumulation in determining GDP growth rate and in variations in the trade balance, as a proportion of GDP, in the cases of Argentina, Brazil, Chile and Mexico. The data used for the analysis was obtained from databases from CEPALSTAT,⁶ World Penn Table version 10.0 and the World Bank’s World Development Indicators. The period analyzed was that of 1961-2019 for the cases of Argentina, Chile and Mexico, and 1971-2019 for Brazil; all series used were measured in domestic currency at constant prices.

It is important to begin the analysis by highlighting that while ^{Thirlwall (1979}) argues that the main constraint to economic growth is the trade balance’s dynamic equilibrium, ^{Moreno-Brid (1998-1999)} argues that the real constraint lies in maintaining the trade balance constant in proportion to GDP. In this regard, and taking into account the external debt crisis which affected the Latin American countries in the early 1980s, Figure 1 shows the relationship between the annual variation in the trade balance as a percentage of the GDP (xmy) and the GDP growth rate (g) for the four economies in the sample before and after their external debt crisis.

Figure 1 shows, on the one hand, that the four economies exhibit a negative relationship betweenxmyandg, and, on the other hand, that after the external debt crisis, every case exhibited a decrease in the growth rate that keeps the trade balance constant as a proportion of the GDP.⁷ In the case of Argentina it went from 3.3 to 2.1%, in Brazil it went from 8.6 to 2.3%, Chile from 4.2 to 4.1%, and Mexico from 6.8 to 2.4%. This intrinsically means a tightening of the external constraint on economic growth⁸ and, according to what we have shown in the previous section, this phenomenon can be attributed to a drop in the export growth rate or a slowdown in capital accumulation.

Note: the line indicates g’s estimated values with regards toxmy using a linear relationship acquired using the Ordinary Least Squares method. Source: created by the authors using the CEPALSTAT’s and the World Bank’s World Development Indicators databases.

Figure 1 Annual variation in trade balance as a percentage of GDP (horizontal axis:xmy) and GDP growth rate (vertical axis:g) 

In order to analyze the effect capital accumulation has on the study’s sample economies’ growth rate, we first calculate the gross capital stock using the perpetual inventory method, for which in accordance with ^{Berlemann and Wesselhöft (2014}) we estimate the trend growth rate of the net capital stock as corresponding to that of gross investment (I) using the following equation:

wherelndenotes the natural logarithmic operator,tis a trend variable,uIare the estimation errors,Θ _i are the parameters to be estimated and, in particular,Θ ₁ is the estimated value of the trend growth rate ofI.⁹

The gross investment trend growth rates are assumed to be equal to the trend growth rates of their net capital stocks. Once these were obtained for each country in the study’s sample, the initial net capital stock for each country was calculated as follows:

whereI ₁ andδ ₁ are the initial values of gross investment and the capital stock’s depreciation rate respectively. Then we obtained the series of net capital stock¹⁰ for each of the countries in the sample using the perpetual inventory method:

Using the net capital stock series, Table 1 shows a closer relationship between GDP growth rates and gross capital accumulation rates than between GDP growth rates and export growth rates, both before and after the external debt crisis. In fact, according to what was described regarding the relationship betweenxmyandg, with the exception of Chile, all economies exhibited a reduction in their gross capital accumulation rate. Furthermore, it is worth pointing out that export growth rates fluctuated to a smaller degree than capital accumulation rates.

Based on the descriptive analysis of the data, we apply a more rigorous analysis of the relationship between GDP growth rate and export growth rate and the capital accumulation rate. To this end, and based on the theory for determining growth in an economy constrained by trade balance developed in the previous section, we calculate the following equation, based on equation (12):

whereΘ _i are the parameters to be estimated andu _g is an error term. The estimation of equation (16) is based on the panel data constructed with the available data for the sample countries. In this regard, this data allowed the construction of two panels, one balanced for the period of 1971-2019 and an unbalanced one covering the period of 1961-2019.

Table 2 presents the unit root tests for the series used. As one can see, all are stationary, allowing us to avoid the problem of spurious regression.

Table 3 shows the results of the estimation of equation (16) for the methods of Fixed Effects, Random Effects, Feasible Generalized Least Squares and Panel Corrected Standard Errors. The use of four estimation models is due to two important reasons. First, we seek to obtain a robust estimate of equation (16) which guarantees the consistency of the estimated parameters in their statistical significance, whether the effect of dependent and independent variables is positive or negative in value and the magnitude of the estimated coefficient. In other words, we seek to demonstrate that even when changing the estimation technique, the estimations of the effect had by exports and capital accumulation are consistent. Second, estimation using Feasible Generalized Least Squares and Panel Corrected Standard Errors models allows one to control the estimation considering the panel type data models’ problems of autocorrelation, contemporaneous correlation and heteroscedasticity (^{Beck and Katz, 1995}).

According to the results acquired, there is a positive and significant relationship between the GDP growth rate and the capital accumulation rate; in contrast, the export growth rate does not seem to affect it, or to do so positively by a fairly small magnitude, as it was only statistically significant in the Pane Corrected Standard Errors and Feasible Generalized Least Squares models in the balanced panel, which implies exports have a reduced but positive effect on product growth.

Now, in order to present even more refined evidence of the comparative importance of capital accumulation versus export growth, we proceeded to substitute product growth rates as a dependent variable in the estimates with the growth of domestic demand for domestic goods.¹¹ Therefore, the next equation for estimation is specified below:

whereΘ _i are the parameters to be estimated andu _id is an error term. In Table 2 we find the unit root tests for theidseries which is shown to be stationary. Meanwhile, Table 4 shows the results of equation (17)’s estimation; as one can see,I/Kconsistently exhibits a positive and significant relationship withid; the estimated coefficient is highly significant in all models and its size barely changes. Likewise, x is not found to have a consistent relationship withidas it is only significant for the unbalanced panel’s estimation with Panel Corrected Standard Corrected Errors, the estimated coefficient is very small however.

Finally, in relation to the above, considering the fact that the trade balance as a proportion of GDP is the main constraint to economies’ growth, we pose the next equation for estimation in order to analyze the effect ofxandI/Konxmy:

whereΘ _i are the parameters to be estimated andu _xmy is an error term. Thus, Table 5 presents the unit root tests of the series used in equation (18)’s estimation, by which we conclude that the series are stationary.

Table 6 shows the results of the estimate and one can see thatxpositively affectsxmy; however, the estimated values of the effect ofxonxmytend toward zero, while on the other hand, we obtained negative coefficients corresponding toid. This implies thatxdoes not significantly improvexmy, while the effect ofI/Konxmy, obtained by adding the estimated coefficients ofid-I/Kandx-I/K¹³ is positive and greater in absolute value than the effect ofx.

4. FINAL THOUGHTS

According to the empirical evidence presented, in the context of the need to reconfigure growth strategies when faced by both old and new needs which have arisen in economies due to the current pandemic, economic policy needs to promote an accelerated rate of capital accumulation.

This may seem paradoxical given that, as has been mentioned, it is assumed that the study’s sample economies are constrained by the trade balance’s dynamic equilibrium, or rather, by the need to keep the trade balance constant as a proportion of GDP. This would make it seem that the natural candidate for leading growth is the export growth rate as it positively affects both growth as well as the trade balance. However, we must emphasize that imports depend on income which in turn depends on exports and if, as is typical, economies exhibit an income elasticity of import demand greater than one,¹⁴ the response of imports to exports may be more than proportional or quasi-proportional. In addition to that demonstrated in the previous section, this means that exports would not be as relevant in the behavior ofxmy, or at least they would not have the importance which they are normally attributed.

In contrast, capital accumulation has two effects on import demand. One positive effect is related to the import of capital goods and another negative, related to the production of goods and services which can lead to substituting imports, thereby improving the trade balance. In other words, capital accumulation increases income and relaxes external constraints on growth, allowing for greater long-term growth as seen in the empirical analysis for the panel composed with data from the four largest Latin American economies.