1. Introduction
Over the last few decades, Argentina has grown slowly and unevenly. The different development strategies that followed each other have not managed to channel the country into a sustainable development path. Several authors argue that economic recessions in Argentina are directly related to the balance-of-payment (BoP) problems. From works like Diamand (Reference Diamand1983) and Azpiazu and Nochteff (Reference Azpiazu and Nochteff1995) to more recent ones like Lavarello et al. (Reference Lavarello, Montagu, Abeles, Infante and Gerstenfeld2013), Schteingart (Reference Schteingart2016), Abeles and Valdecantos (Reference Abeles and Valdecantos2016), and Rapetti (Reference Rapetti2019), the common factor in the explanation is an external constraint.
In this regard, it is worth noting that recent theoretical advances in the literature on BoP-constrained growth highlight its cyclical implications (Nishi, Reference Nishi2019; Dávila-Fernández and Sordi, Reference Dávila-Fernández and Sordi2019a, Reference Dávila-Fernández and Sordi2019b; Garcimartin et al., Reference Garcimartin, Kvedaras and Rivas2016). Moreover, Ocampo (Reference Ocampo, Damill, Rapetti and Rozenwurcel2016) stresses that, under this restriction, external shocks are the main drivers of the business cycle (BC). In turn, when this literature is combined with the structuralist approach that highlights the existence of unbalanced productive structures, vicious cyclical dynamics such as the stop-and-go and go-and-crash are expected.
The stop-and-go pattern indicates that every time the economy grows, its imports grow faster than its exports, resulting in a deficit in the Trade Balance. When the economy is not financially integrated with the rest of the world, the deficit is financed with international reserves, which, upon reaching a critical value, causes a depreciation that leads to an acceleration of inflation, a fall in real wages, and a general drop in the level of activity (stop) which, in turn, corrects the trade imbalance and restores growth (go) (Rapetti, Reference Rapetti2019).
Moreover, when the economy is financially integrated globally, there is a go-and-crash dynamic, meaning that the trade deficit can be temporarily maintained through external financing (go). However, if the trade deficit is extended and the debt burden increases, sooner or later, there will be a run on the domestic currency, leading to devaluation. After this, the sequence of events described above is the same, although ending up with larger recessionary adjustments due to the increased need for foreign exchange (crash) (Rapetti, Reference Rapetti2019; Schvarzer and Tavosnanska, Reference Schvarzer, Tavosnanska and Novick2010). Both vicious dynamics could help explain the increasing and excessive volatility documented in the literature on Argentine business cycles (Gadea and Sanz-Villaroya, Reference Gadea and Sanz-Villarroya2020; Cerro, Reference Cerro1999; Kydland and Zarazaga, Reference Kydland and Zarazaga1997), which could be damaging long-term performance.
The emergence of Argentina’s unbalanced productive structure traces back to the era of state-led industrialization (1930–1975) (Schteingart, Reference Schteingart2016). During this period, the primary sector operated at international costs, serving as a supplier of foreign exchange. In contrast, the emerging industrial sector had higher costs than the international ones and permanently demanded foreign exchange for its expansion, since many productive inputs and capital goods were not produced locally (Diamand, Reference Diamand1983). In the mid-1970s, Argentina drastically changed its development strategy, interrupting the industrialization process and radically opening up trade and finance (Rapoport Reference Rapoport2010; Bértola and Ocampo, Reference Bértola and Ocampo2013).
Therefore, by combining all these features, the objectives of this paper are two-fold. The first one is to empirically test the cyclical implications of the external constraint in Argentina from 1930 to 2018 by examining the domestic dynamics between GDP, the trade balance (TB), external debt, and the real wage. Second, considering the drastic shift in the development strategy implemented in the mid-1970s, the changes in real external vulnerability are analysed by observing the influence of the terms of trade (TOT) and the external demand on the domestic cycle. Thus, the main contribution of this paper is to the literature on Argentine BC in the long run in the understanding of their dynamic behavior and the resulting volatility. Laterally, it dialogues with works that have advanced in describing the impact of external shocks in Latin America, providing information on the particular case of Argentina.
For these purposes, autoregressive vectors (VAR) are used, as they provide a systematic way to capture rich dynamics in multiple time series. VARS are useful when there is evidence of simultaneity between a group of variables, as is the case with GDP, TB, external debt, and real wages. The introduction of real wages to the analysis implies an innovation compared to other works of this type, which usually take GDP as the only measure of economic performance.
The remainder of this paper is organized as follows: Section 2 presents a literature review, comprising theoretical and empirical background, as well as an overview of Argentina’s economic history during the period under consideration. Section 3 outlines the research methodology by explaining the data and the vector autoregression. Results are presented in Section 4, both for the whole sample and for the comparison between sub-periods. Additionally, four alternative models are presented to enhance the robustness of the results. Finally, Section 5 presents conclusions.
2. Literature review
2.1. Theoretical framework: external constraint to growth, cyclical dynamics, and volatility
The external constraint approach was first formalized by Thirlwall (Reference Thirlwall1979). The author argues that a sustainable economic growth rate in the long term depends on the ratio between the export growth rate and the income elasticity of imports (Bekerman et al., Reference Bekerman, Dulcich and Vázquez2015). In other words, the economy’s growth rate is constrained by the TB equilibrium.
Recent theoretical advancements have significantly advanced the application of Thirlwall’s law to less developed countries, characterized by exchange rate fluctuations and substantial capital flows. Furthermore, these models underscore the significance of the cyclical implications of Bop-constrained growth. For instance, Dutt (Reference Dutt2002) introduces a North–South model endogenizing terms of trade within a BoP-constrained growth framework, while Garcimartin et al. (Reference Garcimartin, Kvedaras and Rivas2016) analyze the cyclical implications of this theory, revealing business cycles generated by capital flows and trade shocks. Nishi (Reference Nishi2019) also addresses off-path processes from BoP imbalances, resulting in a cyclical phenomenon. Building on Thirlwall’s law, Dávila-Fernández and Sordi (Reference Dávila-Fernández and Sordi2019a, Reference Dávila-Fernández and Sordi2019b) enhance Goodwin’s (Reference Goodwin and Feinstein1967) distributional cycle model within an open economy framework, yielding both cyclical and constrained growth.
Although all countries face external constraints by definition, this generates other types of problems when there is what Prebisch (Reference Prebisch1949) defined as structural heterogeneity. This means that in some economies, productive sectors typical of economies at different stages of development coexist, with a distinct gap in productivity between the sector with the greatest comparative advantage and that of the sector with the greatest comparative disadvantage (Schydlowsky, Reference Schydlowsky1993). This thesis is analogous to that of the Diamand’s Unbalanced Productive Structure (Reference Diamand1983), Azpiazu and Nochteff’s Heterogeneous Productivity Structure (Reference Azpiazu and Nochteff1995), or Schydlowsky’s Evolutionary Dutch Disease (Reference Schydlowsky1993).
As stressed by Spinola (Reference Spinola2021), weak economic structures, such as those found in Latin America, lead to the emergence of short cycles of significant amplitude. As Rapetti (Reference Rapetti2019) points out, within this context, the expected cyclical implications are that every time the economy grows, its imports grow faster than exports, generating a deficit in the Trade Balance and the Current Account of the BoP. When the economy is not financially integrated with the rest of the world, the external deficit is financed with international reserves, which, upon reaching a critical value, causes a depreciation that leads to an acceleration of inflation, a fall in real wages, and a general drop in the level of activity. This is the stop-and-go dynamic. The trade deficit can be temporarily maintained through external financing when the economy is financially integrated globally. However, if the trade deficit is extended and the debt burden increases, sooner or later, there will be a run on the domestic currency, leading to depreciation. This is the “go-and-crash dynamic,” in which the sequences of events that follow are the same as the ones described earlier, although ending up with larger recessionary adjustments due to the increased need for foreign exchange (Schvarzer and Tavosnanska, Reference Schvarzer, Tavosnanska and Novick2010). Testing these dynamics is the first objective of this study.
This vicious dynamic leads to extraordinary macroeconomic volatility (Rapetti, Reference Rapetti2019; Mercado et al., Reference Mercado, Kosacoff and Porta2011), contributing to the low long-term growth rate (Badinger, Reference Badinger2010; Loayza and Hnatkovska, Reference Loayza and Hnatkovska2004; Ramey and Ramey, Reference Ramey and Ramey1994). There are different channels for this to occur: in the private sector, volatility leads firms to choose short-term projects to the detriment of those that require a greater volume of investment and involve resources for innovation and the acquisition of new technologies, and in the public sector it generates a drop in resources that leads to budget cuts (Rapetti, Reference Rapetti2019).
One of the other major causes of volatility in countries facing an external constraint is external vulnerability (Manzano et al., Reference Manzano2002). External shocks do not always imply volatility; however, under the BoP constraint, external shocks, both positive and negative, are essential determinants of short-term macroeconomic dynamics (Ocampo; Reference Ocampo, Damill, Rapetti and Rozenwurcel2016). This typical behavior of developing and commodity-dependent countries prompts economic policy to behave in a pro-cyclical manner, leaving a narrow margin for buffering shock effects and addressing structural problems.
2.2. Background
2.2.1. Argentina’s historical background
Argentina seems to be a compelling case study for exploring the cyclical implications of the BoP constraint in the long term, as Thirlwall’s model approximates well its growth dynamics (Gómez et al., Reference Gómez, Álvarez-Ude and Candaudap2007; Capraro, Reference Capraro2007) and because it is possible to reconstruct long time series for the selected variables. Not only has the country failed to enter a path of sustainable growth and development, but it has moved further and further away from it. From 1930 to the present, the Argentinian economy has grown for more than five consecutive years in only four periods: 1933–1942, 1953–1958, 1964–1974, and 2003–2008, most of which coincided with trade surpluses. Figure 1 shows that from 1930 to 2018, the country experienced 19 recessive episodes, accounting for 28 years of economic contraction, or more than one recession every three years. Moreover, the intensity of these episodes increased from 1945 to 2002.
Argentina GDP growth rates and trade balance-to-GDP ratio.
However, this is not the usual performance of South American economies. Table 1 shows that Brazil, Uruguay, Chile, Colombia, Peru, and Bolivia exhibit higher average accumulated growth rates, are more stable, and have fewer recessive episodes recorded (see Figure A1 in the Appendix).
Growth and volatility; South American economies
Source: Own elaboration with Maddison Project database (2020)
During this period, Argentina underwent significant changes in its development strategy, moving from a state-led industrialization model to a more liberal one (Bértola and Ocampo, Reference Bértola and Ocampo2013). Starting in 1930, Argentina embraced an inward growth characterized by increased state intervention and incentives for manufacturing (Schteingart, Reference Schteingart2016). This shift responded to global crises and the decline of the agro-export model from the first wave of globalization, leading to a boost in the national industrial sector (Gerchunoff and Rapetti, Reference Gerchunoff and Rapetti2016; Rapoport, Reference Rapoport2010).
Following World War II, despite the global reintegration of trade and capital flows, Argentina (and most of Latin America) continued to experience inward growth, state dirigisme, and industrialization (Edwards, Reference Edwards2009; Clemens and Williamson, Reference Clemens and Williamson2011). However, Azpiazu and Notcheff (Reference Azpiazu and Nochteff1995) highlight that the industrial sector leaned toward an import-oriented trajectory, surpassing the growth rate of the primary sector. This shift accelerated demand for foreign exchange, contributing to a stop-and-go dynamic. Schteingart (Reference Schteingart2016) points out that Argentina developed an unbalanced productive structure during the state-led industrialization phase (1930–1975) because of the limited depth of the import-substitution process and the country’s technologically adaptive behavior.
In the mid-1970s, Argentina underwent a significant shift in its development strategy, driven by a combination of external and internal factors. Regarding the former, the oil shock, the exit from the Bretton Woods agreements, and the expansion of the global financial sector contributed to a general change in conditions. On the domestic front, the internal political–social conflict and recurrent BoP crises led to the military coup d’état of 1976, which was the longest in duration and the one with the deepest reforms, aimed at deregulation and drastic opening of the economy (Bértola and Ocampo, Reference Bértola and Ocampo2013).
This period witnessed the abandonment of capital controls, protective policies, and export promotion, along with the crawling peg exchange rate system (Ocampo, Reference Ocampo, Damill, Rapetti and Rozenwurcel2016; Frenkel and Rapetti, Reference Frenkel and Rapetti2016). Between 1976 and 1979, the military government drastically reduced tariffs from nearly 100% to around 30% (Bambrilla et al., Reference Brambilla, Galiani and Porto2018). The resulting surge in trade flows is depicted in Figure 2, while financial openness crystallized in the Financial Reform of 1977.Footnote 1 The exchange rate emerged as the primary tool to absorb external shocks, increasingly adopting flexibility to accommodate fluctuations in the capital account. However, at times, it was also used to stabilize price levels in anti-inflationary programs, precluding its active management of balance of payments shocks.
Trade openness (%), 1930–2018.
The subsequent military governments continued to neglect productive policies, leading to a resurgence of a primary sector-oriented economy, while their exposure to external shocks increased (Rapoport, Reference Rapoport2010). Although some of these measures were reversed with the return to democracy in 1983, the trade and financial openness model was deepened in the 1990s during the Menem administration. During this period, the capital controls that the government had reestablished upon returning to democracy were removed, and the average tariff declined from 30% to 18% (Arza and Brau, Reference Arza and Brau2021; Bambrilla et al., Reference Brambilla, Galiani and Porto2018).
Given the transformative nature of this period, the second objective is to compare the cyclical dynamics and external vulnerability before and after 1975. An analysis of structural breaks exceeds the study’s scope, emphasizing the unique and pivotal nature of this period in Argentine economic history (Sanchez, Reference Sánchez, Cortés Conde and Della Paolera2018).Footnote 2 Additionally, the annual frequency of the data hinders direct comparisons between shorter periods due to the limited number of observations.
2.2.2. Empirical background
This study empirically contributes to understanding long-term cyclical dynamics in Argentina, anchored in the theoretical framework of growth constrained by the Balance of Payments and structural heterogeneity. While primarily engaging with existing literature on Argentina’s business cycles using long and detrended series, this work extends the dialogue to studies of Latin America and references works employing similar analytical techniques, thereby enhancing the overall comprehension of the subject.
Several studies explore long-term cyclical dynamics in Argentina. Sturzenegger and Moya (Reference Sturzenegger and Moya2003) examine the basis for a general theory of Argentine cycles by analyzing series from 1884 to 1990, identifying patterns across Keynesian, Monetarist, and Real Cycle theories. Gadea and Sanz-Villaroya (Reference Gadea and Sanz-Villarroya2020) contribute by comparing the cycles of Argentina, Australia, and Canada from 1870 to 2015, arguing that the short-term has conditioned the relative and long-term evolution of the Argentine economy. Cerro (Reference Cerro1999) describes BC over 178 years, linking the slowdown in the current account balance to cycle peaks, especially during periods of an open economy. For a shorter span (1970–1995), Kydland and Zarazaga (Reference Kydland and Zarazaga1997) stress similarities in Argentine BC with the United States and OECD countries, noting high GDP volatility relative to the United States. In fact, this excess volatility of the Argentine economic growth is also documented by Gadea and Sanz-Villarroya (Reference Gadea and Sanz-Villarroya2020) and by Cerro (Reference Cerro1999), who also make international comparisons.
However, these studies take a static approach and do not advance in providing a model of the joint dynamics of multiple variables over time, nor do they analyze the response of fundamentals to external shocks. Partially filling this gap, Lanteri (Reference Lanteri2009) uses a structural VAR to investigate the Harberger–Laursen–Metzler (HLM) effect, i.e. the positive relationship between the terms of trade and the trade balance. Nevertheless, the study covers a short period (1980–2007) and neglects other trade shocks.
Studies by Ahmed (Reference Ahmed2003), Bertholet (Reference Bertholet2021), and Campos (Reference Campos2023) use VARS (recursive and Bayesian) to analyse the effects of devaluations. Despite their valuable insights into short-term contractionary devaluations, these studies do not address the broader cyclical dynamics associated with the BoP constraint nor the impact of external shocks. Studies by Izquierdo et al. (Reference Izquierdo, Romero-Aguilar and Talvi2008) and Osterholm and Zettelmeyer (Reference Zettelmeyer and Österholm2007) also employ techniques similar to those used in this study to examine the impact of external shocks on Latin America. Izquierdo et al employ a restricted Vector Error Correction Model to analyze the effects of terms of trade movements and reversals in external financial conditions on a Latin American output index from 1991 to 2006. Osterholm and Zettelmeyer (Reference Zettelmeyer and Österholm2007) employ a Bayesian VAR model to examine the sensitivity of Latin American GDP to external shocks, including financing shocks, external growth shocks, and commodity price shocks, during the period 1994–2006. Additionally, Spinola (Reference Spinola2023) and Abeles and Valdecantos (Reference Abeles and Valdecantos2016) focus on Latin American volatility within the theoretical framework of BoP-constrained growth, associating it with commodity cycles and identifying terms of trade movements and variations in the growth of main trading partners as the channels of real vulnerability.
3. Data and research methodology
3.1. Data description
Table 2 presents the labels and definitions of the variables used in the VAR model, along with the sources from which they were obtained (see Table A1 and Figure A3 in the Appendix for descriptive statistics and individual graphs of the variables). The data are annual and cover the period from 1930 to 2018. All variables are expressed in logs, except for the Trade Partner’s growth and Trade Balance, which are GDP ratios.
Variables
Since there are no official sources that have the complete series used here, the “backwards splicing” methodology has been used to obtain a homogeneous series of the variables. The procedure involves “stretching” the most recent series based on the variation rate of the previous series (Graña and Kennedy, Reference Graña and Kennedy2008). Although not ideal for preserving data quality, it is a standard practice in those studies that seek a long-term approach for Argentina, as can be seen in the works of Campos (Reference Campos2023), Gadea and Sanz-Villaroya (Reference Gadea and Sanz-Villarroya2020), Sturzenegger and Moya (Reference Sturzenegger and Moya2003), Cerro (Reference Cerro1999), and Kydland and Zarazaga (Reference Kydland and Zarazaga1997).
Considering the influence of the BoP on the short-term macroeconomic dynamics of developing countries (Ocampo, Reference Ocampo, Damill, Rapetti and Rozenwurcel2016) the focus is on the interrelation between the variable’s cycles. To obtain them, the Hodrick–Prescott filter (Reference Hodrick and Prescott1997) is applied to the series. It consists of a linear filter that breaks down the time series into two components: the long-term trend and a stationary cycle (the fluctuations around the long-term trend).Footnote 4
This filter is selected to ensure comparability, as previous studies have consistently examined Argentinean BC (Sturzenegger and Moya, Reference Sturzenegger and Moya2003; Kydland and Zarazaga, Reference Kydland and Zarazaga1997; Cerro, Reference Cerro1999). However, considering that it may exaggerate the volatility and autocorrelation of the series because it passes much of the undesirable lower-frequency content in the series outside the BC range (Basu & Taylor, Reference Basu and Taylor1999), robustness tests are conducted. These tests involve detrending the series through both the first differences and the application of the Christiano-Fitzgerald asymmetric bandpass filter. The outcomes of these tests are shown in Section 4.3.
To consider real external vulnerability, we follow Abeles and Valdecantos (Reference Abeles and Valdecantos2016) and Osterholm and Zettelmeyer (Reference Osterholm and Zettelmeyer2008) who, in general terms, argue that the channels that shape real external vulnerability in Latin America are two-fold: the movements in terms of trade—understood as the relative price of exports in terms of imports—and the variation of external demand. To acknowledge the latest, the growth rate of the main trading partners, weighted by exported value each year, is considered. Two criteria were followed to build the variable: represent at least 50% of exports in each year—the average is 78.9% for the entire period—and include at least the first 14 export destinations of the corresponding year (see Figure A2 in the Appendix). Information on the export percentage to each destination is obtained from two sources: up to 1996, Ferreres’ (Reference Ferreres2005) data are used, while as of 1997—the first year for which they are available—data from the Foreign Trade Office of the National Institute of Statistics of Argentina are aggregated.
3.2. Recursive VAR analysis
To describe the impact of external shocks and certain endogenous dynamics with which they are related, a recursive VAR is performed:
\begin{equation*}{\textit{Y}_{\textit{t}}} = \mathop \sum \limits_{\textit{j} = 1}^3 {\textit{A}_\textit{j}}{\text{ }}{\textit{Y}_{\textit{t - j}}} + {\textit{u}_\textit{t}}\end{equation*} where 
${\textit{Y}_\textit{t}}$ is a column vector of the variables listed in Table 2 and 
${\textit{u}_\textit{t}}$ is a column vector of innovations (Novales, Reference Novales2011). Based on the Akaike information criterion, a three-lag VAR is performed for the baseline model. The model does not present autocorrelation in the residuals, they follow a normal distribution and present homoscedasticity. For details of the diagnostic tests of this and the remaining models, see Table A3 in the Appendix.
The model is identified by imposing contemporaneous restrictions using the Cholesky method. This means that the first variable cannot respond to contemporaneous shocks (within the year) of any other variables. In contrast, the second variable can respond to contemporaneous shocks in the first variable but not in the subsequent variables, and so on. This strategy is preferred over a structural VAR, which would require theoretical assumptions. Moreover, as demonstrated by Sturzenegger and Moya (Reference Sturzenegger and Moya2003), no singular theory comprehensively explains Argentina’s cyclical regularities. Cholesky identification proves especially advantageous when a clear theoretical structure is lacking to justify a structural model.
The variables are ordered as follows. The main trading partners’ growth rates and the terms of trade are ordered in the first place, respectively. Therefore, they cannot be contemporaneously affected by the subsequent variables, which makes sense since Argentina is a price-taking country for the products it sells to the rest of the world and does not represent more than 6% of the export basket of any of the countries. These two variables are followed by GDP, TB, External debt, and Real wages. Considering that the trade balance is a component of GDP, it appears immediately after it in the ordering. The external debt is expected to depend on the country’s economic performance and its trade surplus or deficit. Real wages are placed at the end, as they are one of the variables that adjust most quickly, so they can respond contemporaneously to any variable. In any case, it is corroborated that none of the main results discussed below vary significantly from changes in the order of the variables (see Table A2 in the Appendix).
Standard practice in VAR analysis is to report the results of Granger-causality tests, impulse responses, and variance decomposition. From the reduced form VAR, Granger causality contrast examines whether past values of a given variable help predict the behavior of another variable. The accumulated impulse response functions (AIRF) and variance decomposition are obtained from the recursive VAR. AIRF measures the sum of each variable’s reaction to innovation in one variable across time. They are represented in several graphs, each of which includes the accumulated responses over 4 years of a given variable to an impulse in each of the innovations.Footnote 5 In turn, the decomposition of the variance allows us to divide the variance of the prediction error of each variable into the components that are attributable to the different shocks that the system may experience (Novales, Reference Novales2011).
4. Results
4.1. Full sample: 1930–2018
Table 3 presents the results of the Granger-Causality test. A subset of key impulse responses is reported in the text, and the complete set of AIRF is reported in Figure A4 in the Appendix. The shock of each variable is set as one standard deviation of that variable, and the accumulated responses are traced through four periods. The red dotted lines represent confidence bands.
Granger-causality tests*
* In bold are indicated P-values that allow rejecting the null hypothesis of the regressor not causing the dependent variable in Granger’s sense. Source: Author’s estimation.
Figure 3 shows that when Argentina begins to grow, it automatically activates the mechanisms that block its future growth possibilities by increasing imports faster than exports. Moreover, GDP causes the Trade Balance in the Granger sense, and 18% of its variance is explained by GDP at the 4-year time horizon, demonstrating the importance of this mechanism (Table 5).
Trade balance response to a shock in GDP.
This reaction is also consistent with Cerro’s (Reference Cerro1999) and Rapetti’s (Reference Rapetti2019) results. However, unlike the latter, the correction of the external imbalance does not occur in the fourth year; instead, a persistent trade deficit is observed. Moreover, this could be related to Chena (Reference Chena2008), who stresses the consequences of external constraints in the case of countries that export foodstuffs. He argues that when GDP increases in countries with high poverty levels that export food to the rest of the world, part of the supply is consumed internally, which causes exports to fall or grow below their potential.
Also, in Table 4, it can be seen that GDP reacts directly to TB shocks, increasing by 2.4% 4 years later. These results confirm the stop-and-go dynamic: when the economy grows, the shape of its productive structure leads to a bottleneck in the trade balance, resulting in a drop in activity (stop). In turn, this fall produces a contraction in imports, which adjusts the TB and restores the activity level (go) (Schvarzer and Tavosnanska, Reference Schvarzer, Tavosnanska and Novick2010). Notably, the real wage response to TB shocks is negative. This could indicate that the productive sectors that drive exports are not labor-intensive, which would generate downward pressure on real wages as the economy shifts to these sectors. However, given that the upper confidence band is above 0 for this response, it can be considered a negligible effect (Figure A4 in the Appendix).
Accumulated impulse responses after 4 years
Source: Author’s estimation.
Moreover, the expected inverse reaction of the external debt to the increase in the TB is corroborated, along with a significant persistence effect (Table 4). In addition, it is worth noting that almost 30% of the variation in external debt is explained by movements in the trade balance, and that trade causes external debt in the Granger sense, which highlights the significance of debt as a mechanism to meet trade deficits (Table 5).
Variance decomposition from the recursive VAR after 4 years
Source: Author’s estimation.
More striking is that increases in debt lead to a decline in output and real wages. Figures 4 and 5 indicate that, instead of contributing to growth or improving living standards, external debt has a persistent negative effect in the short term, particularly on real wages. This is consistent with Canitrot’s (Reference Canitrot1983) statement: the real wage compatible with the payment of external debt commitments is lower than the equilibrium real wage freed from such commitments. In the case of GDP, although the inverse relationship is inconclusive, it is worth noting that GDP has no positive response to external debt, at least for 2 years after the shock.
The response of GDP to a shock in the external public debt.
The response of real wage to a shock in the external public debt.
This is consistent with the go-and-crash dynamic previously mentioned, in which the borrowing of external debt serves mainly to correct trade imbalances whose origin is not addressed. As Schvarzer and Tavosnanska (Reference Schvarzer, Tavosnanska and Novick2010) and Rapetti (Reference Rapetti2019) explain, external debt allows economic expansion to continue despite the external imbalance but, when a crisis occurs, the adjustment must be such that it not only adjusts the trade balance but must also lead to a trade surplus of sufficient magnitude to cover the interest payments. In other words, the larger the current account deficit and the more indebted the economy is, the greater the recessionary adjustment imposed by the external constraint, which in turn could explain the greater depth and frequency of the crises observed since the stage of financial openness in the mid-1970s.
Regarding the cyclical implications of external vulnerability in the context of the aforementioned constraint, Figures 6 and 7 present the output responses to shocks in trade partners’ growth and the terms of trade, respectively. As expected, both responses are positive, although GDP appears slightly more sensitive to trade partners ‘growth (first two columns of Table 4). Moreover, 4 years after the shock, 19% of GDP variance is explained by the trade partners ‘growth, while 9% is explained by the TOT shock (Table 5). Therefore, it can be stated that the real external vulnerability explains more than 28% of its variability. These results are highly in agreement with Osterholm and Zettelmeyer (Reference Zettelmeyer and Österholm2007), who find that, for Latin America, the forecast error variance of GDP is explained by 18% of world growth shocks and 6% of commodity prices 5 years post-shock.
GDP response to shock in trading partner’s growth.
GDP response to shock in terms of trade.
Such external shocks have a positive but negligible impact on real wages, as seen in Figures 8 and 9. It is worth noting the real wages sensibility to TOT shocks: the latest cause real wages in Granger’s sense (Table 3) and, as can be seen in Table 5, at the 4-year time horizon the Terms of trade explain 10% of the Real Wage variance, while only 4.8% is explained by the Trade Partners growth.
Real wage response to shock in trading partner’s growth.
Real wage response to shock in terms of trade.
Remarkably, positive TOT shocks generate a decrease in the Trade Balance. This could be because the increase in the terms of trade causes the exchange rate to appreciate, making exports less competitive internationally, or because higher export income generates an increase in the quantities imported. In any case, this constitutes evidence against the HLM effect in the short run, contrary to what Lanteri (Reference Lanteri2009) finds for the long term (Table 4).
So far, evidence of the Trade balance bottleneck has been found between 1930 and 2018, which imposes structural constraints on growth. The cyclical implications of this limitation are corroborated: assuming symmetry in the effects, it can be stated that a trade deficit ultimately results in a GDP decline, and when the deficit is financed through external debt, the final outcome remains the same. This means external public debt does not constitute a way out of the fall in output and living standards. Regarding real vulnerability, the impact of the growth of trading partners and movements of TOT on GDP and real wages was characterized. Notably, trade shocks account for 28% of GDP variability and nearly 15% of real wages. Additionally, evidence is found against the HLM effect for the short term.
4.2. Period’s comparison: 1930-1975 and 1976-2018
From the aforementioned change in the development strategy in Argentina in the mid-1970s, the question arises as to whether this affected the country’s cyclical dynamic and external vulnerability. To compare the short-run impact of external shocks on the Argentinian economy in the periods 1930–1975 and 1976–2018, a VAR is made for each of them. The results indicate substantial changes in the impact of shocks, which increase the country’s external vulnerability.
A recursive VAR (2) is configured for the same six variables as in the previous model. The order of the variables is the same, following the criteria for the whole sample. The systems satisfy the stability condition; no autocorrelation exists in the residuals.Footnote 6
At first glance, Figures 10 and 11 indicate that the trade balance bottleneck is not only observed in both periods but is also exacerbated in the most recent one, with a stronger initial reaction and greater persistence. Moreover, the TB variance explained by GDP increases significantly between periods, from 11.1% to 36.6% (Table 8). This is consistent with the worsening of the unbalanced productive structure resulting from the abandonment of productive policy and trade liberalization, which aggravates the external constraint and narrows the economy’s margin of action to buffer external shocks. As Schvarzer and Tavosnanska (Reference Schvarzer, Tavosnanska and Novick2010) point out, the augury that the stop-and-go cycles would be overcome through openness and capital inflows did not come true, which the authors associate with a new dynamic (the go-and-crash) that managed to extend the upward phase of the cycles at the cost of leading to crises of greater magnitude.
1930–1975 trade balance response to shock in GDP.
1976–2018 trade balance response to shock in GDP.
At the same time, trade balance shocks increase their incidence on GDP between sub-periods, which completes the picture of even worse stop-and-go dynamics from the mid-1970s onwards (Table 7). Furthermore, the trade balance becomes more relevant in explaining real wage movements (Table 8), indicating that problems on the external front affect the living conditions of the population. In turn, the AIRF shows that until 1976, the reaction of the real wage to positive trade balance shocks was positive, while from that year onwards, the response was negative during the three post-shock years (Figures A5 and A6). This could indicate that the above-mentioned mechanism of export growth supported by non-labor-intensive sectors emerges in the second period.
Moreover, although both GDP and real wages respond negatively to increases in external debt in both sub-periods, debt shocks do not seem to exacerbate their effects on either variable (Figures 12, 13, 14, and 15).
The 1930–1975 GDP response to shock in external debt.
The 1976–2018 GDP response to shock in external debt.
The 1930–1975 real wage response to shock in external debt.
The 1976–2018 real wage response to shock in external debt.
The results of the real external vulnerability are described in the following paragraphs. In Figures 16 and 17, it can be seen that the GDP response to shocks in external demand becomes slightly stronger in the second period, with a faster initial reaction, especially in the first two years following the shock.
The 1930–1975 GDP response to shock in main trading partner’s growth.
The 1976–2018 GDP response to shock in main trading partner’s growth.
What is striking is the increased sensitivity of GDP to movements in TOT, as shown in Figures 18 and 19. Not only is the response in the short term larger, but also the positive reaction in subsequent periods and its persistence over 4 years. Moreover, in the first period, the TOT shocks explained 8.8% of the prediction error in GDP, while in the second, they explained more than 30% (Table 8).
The 1930–1975 GDP response to shock in TOT.
The 1976–2018 GDP response to shock in TOT.
Thus, we observe that since the change in the development strategy in the mid-1970s, the cumulative impact of external trade shocks on the Argentine GDP’s error variance increased significantly, rising from 29.8% to 43%. These results indicate the end of the state-led industrialization stage and the beginning of an era of greater trade openness, with the corresponding increase in external vulnerability that this implies. As Ocampo (Reference Ocampo, Damill, Rapetti and Rozenwurcel2016) explains, until the mid-1970s, the primary macroeconomic policy instruments focused on managing external shocks, particularly those arising from the current account. During the trade and financial liberalization stage, many instruments were abandoned, except for the exchange rate, which became increasingly flexible to accommodate external shocks coming through the capital account. The exchange rate will be discussed again in the following section.
Regarding the reaction of real wages to shocks, there are some differences concerning GDP. While real wages increase when the external demand grows, they do so less strongly in the 1976–2018 period, which can be seen in (Figures 20, 21, 23). However, in the case of the real wage response to movements in the TOT, despite the AIRF confidence bands do not allow for conclusive results about the response (Figures 22, 23), the evidence is indicative of greater external vulnerability: TOTs appear to cause real wages in the Granger sense from one period to the next, and the portion of the real wage prediction error explained by TOTs grows between the two sub-periods (Tables 6 and 8).
The 1930–1975 real wage response to shock in trading partners’ growth.
The 1976–2018 real wage response to shock in trading partners’ growth.
The 1930–1975 real wage response to shock in TOT.
The 1976–2018 real wage response to shock in TOT.
Granger causality test; comparison between periods
Source: Author’s estimation.
Besides, the division between periods confirms the absence of the HLM effect in both samples, and even in the second period, the negative impact of the TOT shock on net exports is stronger (Table 7). The idea that higher export revenues resulting from the increase in their relative price entail a proportionally higher increase in imports is consistent with a re-primarized production structure and greater heterogeneity across sectors after 1976, as evidenced by Schteingart (Reference Schteingart2016).
Accumulated impulse responses after 4 years: 1930–1975 and 1976–2018
Source: Author’s estimation.
Variance decomposition after 4 years. 1930–1975 and 1976–2018
Source: Author’s estimation.
In summary, the worsening of the trade balance bottleneck stands out, indicating that the external constraint to growth worsened between 1976 and 2018. The cyclical implications of this constraint are corroborated for both sub-periods, highlighting the higher sensitivity of GDP and real wages to movements in the trade balance. Contrary to expectations, a lower response to movements in external debt is found. However, it is corroborated that in both periods, it does not contribute to short-term growth or improve the population’s purchasing power. As for real external vulnerability, the greater sensitivity of GDP to shocks, especially to the terms of trade, stands out: trade shocks go from explaining 29.8% to 43% of GDP variability. In the case of wages, they explained 26.3% of their variability, while after 1976 they explained 14,8%. Both GDP and real wages show a greater vulnerability to terms of trade shocks, especially in output.
4.3 Robustness checks
In this section, four alternative models are run to check the robustness of the results, in addition to considering different variable orderings as previously mentioned. The first pair of models incorporates the Real Exchange Rate and the International Reserves variables, which were excluded from the baseline model to maintain an adequate number of degrees of freedom. Consequently, it is impossible to directly compare results across different periods due to variations in the number of variables and the frequency of the data. The second pair of models is practically the same as the baseline model but differs in the techniques used to detrend the series. All the models pass the same tests as the initial one.Footnote 7
The complete AIRFs of the models that include the RER and International Reserves are shown in Figures A7 and A8 in the Appendix, together with the Granger causality test and the variance decomposition (Tables A4 and A5 in the Appendix). It is noted that none of the results listed for the baseline model change significantly. Regarding the RER model, in line with Ahmed (Reference Ahmed2003), we find that adverse external shocks generate a real exchange rate depreciation (i.e., an increase in the RER). Nevertheless, in contrast to Campos (Reference Campos2023) and Ahmed (Reference Ahmed2003), these depreciations do not have conclusive effects on GDP, the trade balance, or the real wage, at least in the short run. Only a positive, significant, and persistent impact of real depreciation on external debt is observed, consistent with the fact that the RER depreciation makes debt more costly to repay in local currency terms. If we consider the feedback of these effects, we observe that adverse shocks are followed by a depreciation, which increases external debt and which, in turn, undermines GDP and erodes the purchasing power of the wage (the inverse response of these two variables to the increase in external debt is more robust in this model).
The model that includes International Reserves (IIRR) corroborates the results of the baseline model and those expected by including the variable. For instance, a direct response of IIRR to movements in TB is observed, while the TB explains a significant portion of the variability of IIRR 4 years post-shock (14%). Moreover, the inclusion of IIRR in the model improves the significance of the negative response of GDP and wages to increases in external debt. The cyclical implications of the BoP constraint are observed in that increases in GDP led to a decrease in Reserves. This is relevant because, in turn, it is observed that accumulating reserves has a positive effect on economic growth (Granger causality is observed), while increasing reserves has a negative impact on the trade balance (this relationship also shows causality). In other words, accumulating reserves stimulates growth, but, in turn, growth causes them to fall because of the trade bottleneck. Moreover, it is noteworthy that movements in the trade balance do not cause IIRR in the Granger sense. This could be related to the structuralist thesis regarding the weakness of the productive structure to be a genuine generator of foreign exchange.
The models with the variables detrended with the first differences and the Christiano–Fitzgerald filter also show that the results are robust to the change in the method for detrending the series.Footnote 8 The detailed results are presented in Figure A9 and A10 and Tables A4 and A5 in the Appendix. Some differences between the baseline model and these models are detailed below. First, the negative response of the trade balance to GDP shocks loses statistical significance after the third post-shock period (the upper confidence band is above zero). Second, differences are observed between models in the response of the debt to shocks in the TB: while in the first-differences model a negative response is observed as in the baseline model that only acquires statistical significance in the fourth post-shock period, in the model with the Christiano-Fitzgerald (CF) filter the response is positive and only becomes negative in the fourth post-shock period. Third, it is noted that the negative response of wages to increases in external debt becomes stronger with the model using the asymmetric bandpass filter. Fourth, the direct response of GDP to TOT shocks loses significance in these alternative models.
However, when comparing periods, it can be observed that the greater sensitivity of GDP to variations in terms of trade after 1976 is even stronger than in the baseline model. As for the response of real wages to external shocks, it is noteworthy that their response to external demand becomes smaller (although it is never statistically significant), while the response to changes in the terms of trade becomes sharper. It should be noted that the comparison between periods can be performed using the first-differenced variables, but not with the asymmetric bandpass filter, as the models for the sub-periods with the CF-filtered variables fail to pass the autocorrelation test.
5. Conclusions
This paper tests the cyclical implications of the BoP constraint in Argentina from 1930 to 2018 and explores the effects of external trade shocks on domestic variables. Additionally, it scrutinizes changes in real external vulnerability post the mid-1970s shift in the development strategy towards trade and financial openness, deregulation, and the abandonment of productive policies. The main contribution is to the literature on Argentine business cycles in the long run, from the approach of growth restricted by BoP and structuralism. The main results are presented below.
First, a stop-and-go dynamic is identified as a key feature, illustrating that economic growth leads to faster-growing imports than exports, resulting in a trade bottleneck and subsequent activity drop. This decline effectively restores the trade balance, fostering renewed growth.
Second, another crucial finding supports the go-and-crash dynamic: increases in external debt led to declines in both output and real wages. This dynamic refers to the extension of the expansionary phase of the cycle through external debt without addressing the original cause of the structural foreign exchange shortage. The go-and-crash approach potentially contributes to the observed increase in the frequency and depth of post-1976 crises, as further adjustments following BoP mismatches must be such that they not only correct the trade balance but also lead to a trade surplus of sufficient magnitude to cover interest payments.
Third, it is observed that with the change in development strategy in the mid-1970s, external vulnerability increased. The stop-and-go dynamic intensified, and the economy became more sensitive to external shocks, particularly to fluctuations in trade. The cumulative impact of external trade shocks on Argentina’s GDP error variance increased significantly after 1976, accounting for approximately half of output variability. This aligns with Ocampo’s (Reference Ocampo, Damill, Rapetti and Rozenwurcel2016) idea that economies constrained by the BoP have their cycles mainly determined by external shocks.
In this way, it is demonstarted that the BoP-constrained growth approach and the structuralist view provide key elements to understand the Argentine cyclical dynamics in the long run, contributing valuable insights to address the country’s excessive volatility. In line with Gadea and Sanz Villaroya’s (Reference Gadea and Sanz-Villarroya2020) interest in the short- and long-term relationship, a potential mediating mechanism arises. The existence of vicious cyclical dynamics, such as the stop-and-go and go-and-crash patterns, can erode sustained growth through increased volatility. Such dynamics require a comprehensive strategy to synchronize the path of productive policy, leading to structural transformation in the face of the opportunities and restraints presented by external shocks in macroeconomic dynamics.
Furthermore, considering the effect that foreign debt has on growth and real wages and the nine sovereign debt crises that the country has had, it is appropriate to refer to the recent debate in the Argentine Congress on the need to reach a social and political consensus on indebtedness.Footnote 9 It is necessary not only for its quantitative consideration but also for qualitative aspects, which concern the destination of these foreign currencies: for the indebtedness to be efficient, the destination of the foreign currencies must necessarily be oriented to the productive transformation necessary to break the trade bottleneck.
Finally, during the course of the research, some questions arose that are beyond the scope of the study and are pending for future work. The first is to explore the asymmetric effects of external shocks, i.e., whether the magnitude of the response to shocks is the same when they change sign. Secondly, the financial side of external vulnerability remains to be assessed. This means analysing the impact of, for example, movements in the international interest rate or country spreads. Third, it would be opportune to extend the analysis to other South American countries to compare the cyclical dynamics associated with external constraints in similar countries. Fourth, while incorporating real wages is innovative, the variable has inherent limitations in capturing real fluctuations in the purchasing power of the population because of the expansion of the informal sector in the labour market. For this reason, obtaining informal wage data would provide a more accurate understanding of the phenomenon studied, which is still pending.
Acknowledgements
The author gratefully acknowledges helpful comments from Esteban Nicolini, Francisco Barberis Bosch, Sebastián Valdecantos, Santiago Graña, and from participants in the Congreso Nacional de Estudiantes de Posgrado en Economía (Argentina), the Uppsala Seminar in Economic History, and the Virtual Economic History Seminar of Warwick University. The author also thanks the three anonymous referees for their insightful suggestions. Support from the Asociación Universitaria Iberoamericana de Posgrado (AUIP), which enabled a research stay that contributed to improving this article, is also gratefully acknowledged.
Appendix
GDP growth rates—selected South American economies (1930–2018).
Evolution of export destinations from 1930 to 2018: percentage distribution of the export basket.
Variables included in VAR.
Descriptive statistics
Source: Own elaboration.
VARs with different ordering of variables and fulfillment of hypotheses
Source: own elaboration
Accumulated responses: whole sample 1930–2018.
Baseline model- accumulated responses—first sub-sample (1930–1975).
Baseline model-accumulated responses—second sub-sample (1976–2018).
VAR tests (P-values)
Notes: AIC: Akaike information criterion
VAR Residual Serial Correlation LM Tests. Null hypothesis: No serial correlation at lag h
Jarque-Bera Residual Normality Test. Joint probability. Null hypothesis: Residuals are multivariate normal
VAR Residual Heteroskedasticity Tests. Joint test. Null hypothesis: There is homoskedasticity
a The non-normality of the residuals, while not desirable, does not represent problems for the consistency of the estimators and allows for inference in an asymptotic sense. For the models with no-normally distributed residuals confidence bands are obtained with Montecarlo simulations.
Model including RER. Accumulated responses: whole sample 1930–2018.
Model including international reserves. Accumulated responses: whole sample 1930–2018.
Granger-causality tests. Alternative models
Source: Author’ estimation
Variance Decomposition from the Recursive VAR after 4 years. Alternative models
Source: Author’s estimation.
VAR model with first-differenced variables, 1930–2018.
VAR model with variables filtered with christiano-fitzgerald, 1930–2018.
Open access
