The Argentina Paradox: microexplanations and macropuzzles

There is an old saying among economists, possibly apocryphal, and of unclear attribution: “throughout history there have been only four kinds of economies in the world: advanced, developing, Japan, and Argentina”.

This idea can be more concretely grasped by looking at evidence on the long run levels of income per capita in a broad range of countries over the last two centuries in Fig. 1. Material living standards have advanced across the entire world, but the well-known Great Divergence is quite apparent. A few rich countries have become much richer; a larger group of poorer countries have grown more slowly average. Within each group are notable exceptions, with some very poor countries making little progress at all. However, most striking is those countries witnessing a reversal of fortune, moving from one group to the other.

The set of once-poor countries that are now rich include Japan, where the transition began more than 100 years ago, and some other East Asian countries such as Korea and Taiwan, whose transition started only 50 years ago and is now almost complete. But going the other way there is only one notable case of a country that started life relatively rich and ended up comparatively poor: this is the great puzzle or paradox of Argentina. In the late nineteenth century, it was among the top five countries in income per capita, richer than all European countries except Britain and on a par with other rich settler societies such as the United States, Canada, and Australia. Its status among the richest economies reached a peak around 1913 when income convergence stopped. But even until 1929 it still clung on in this club, only to face a slow and widening relative decline thereafter. It is now close to the average country in the world given its level of income per capita, and its citizens enjoy only 40% of the average income per capita of the 12 core countries of Western Europe.

For most of the twentieth century, Argentina’s trade volumes (as a fraction of GDP) have been very low, whether relative to their initial levels in the 1900–14 period, or relative to the trade levels one might predict in a similar economy of comparable size and geographic remoteness.

Figure 2 traces the evolution of the trade share over time in Argentina, measured by exports plus imports divided by GDP. From a high of 80% or more on the eve of World War One, this ratio fell to levels below 20% in the 1920s and 1930s and has remained there ever since (Berlinski 2003). Even with the push towards liberalization in the 1990s this ratio barely ticked up during that decade. This pattern constitutes the main empirical fact about trade in Argentina in the twentieth century.

One question is how much of this trend is explicable in terms of (potentially changeable) trade policy frictions? And importantly, in addition, how much of those frictions reflected policies in Argentina as compared to the Rest of the World? From the 1930s to the 1960s, trade barriers were high almost everywhere in the global economy. As Díaz Alejandro and others have noted, Argentina’s policy decisions were, therefore, not that unusual, and, as in many other countries, the shift towards import substitution industrialization (ISI) was a reflection of path dependence and political economy dynamics: when the world’s rich, core economies chose autarky in the 1920s and especially 1930s—forces then amplified by wartime frictions—the countries of the periphery had to seek another economic strategy in the short term, and having been exposed to the downside risks of openness and specialization in the primary-product “lottery,” they were understandably wary of again placing large bets predicated on the assumption of a stable, globalized, liberal world order with open markets for trade. Argentina fits this pattern.

However, while Argentina’s stance was not that peculiar by the standard of developing countries, where inward-looking development strategies were the norm, it was unusual by the standard of the rich countries, the club to which ostensibly Argentina wished to belong, or rather remain. After 1945, the gradual progress of GATT (and in Europe of the EU project after 1957) carried trade integration rapidly forward. But until the Uruguay Round (circa 1990) of negotiations, Argentina, like most other developing countries, stood apart from this process, and policies remained strongly protectionist.

Data on the distortionary impact of quotas are scant, but these barriers were often very significant in the Argentine context, whether imposed directly or by the quota rationing of foreign exchange (as in the 1940s and early 1950s). Trade taxes are easier to document, and Fig. 3 shows what we know about average import and export taxes in the long run (Berlinski 2003). Import taxes were not trivial prior to World War One as they were a key revenue source, but export taxes were zero. Subsequently, in the 1920s and 1930s, average trade taxes began to climb. They abated during World War 2 and the early postwar exchange control epoch. Then import and export taxes climbed rapidly after 1960, to about a 15% level for each, or a 30% distortion total. Judging from the timing of two asymmetrical spikes in the 1980s, export taxes tended to evaporate in hyperinflation episodes, while import taxes tended to rise in an offsetting fashion, but these figures may also reflect accounting problems. In the liberalization period of the 1990s, export taxes were lifted, but import taxes remained high, although trade policy becomes somewhat more liberal on other dimensions (e.g., quota removals for GATT/WTO compliance and an attempt to start a regional trade area, MERCOSUR).

What would be the likely impact of these trade barriers on income levels? We cannot hazard a precise answer but we can use some simple impacts based on either calibrated models or econometric estimates. In this setting, I will neglect the standard dead-weight loss considerations since utility losses arising from static consumption and production distortions are typically an order of magnitude too small to be useful in discussions of the Great Divergence (usually 1–2% at most). I narrow the focus further by examining the impact of trade frictions on two of the most widely discussed channels through which protectionist policies might lower incomes.

First, higher trade costs raise the costs of imported capital goods and intermediate inputs. These costs are nontrivial, and they matter all the more in countries that are both very open and have comparative disadvantage in these products. Argentina is a classic example of such a country, and the understanding of this type of drag on economic performance dates back to the classic analysis of postwar underperformance by Díaz-Alejandro (1970), who noted that many key imported capital inputs (e.g., farm or industrial machinery) bore domestic prices that were very high indeed compared to world prices.

Let us now feed some numbers into a model, backed by econometric support, that can capture this effect. Suppose, as noted above, trade costs increase by 50% due to trade barriers (the rough magnitude of the trade tax burden since 1950). Also suppose that imports in the initially open economy would be 40% of GDP (the figure last seen in Argentina circa 1910, the last date when both it and the rest of the world were close to fully open). Let us assume that intermediate inputs account for 50% of imports, and capital goods account for 25% the roughly stable figures seen in decades of historical data in Argentina (Berlinski 2003).

Using the Estevadeordal and Taylor (2013) standard open-economy neoclassical model, we could conclude that the trade taxes would lower GDP by roughly 0.200 log points or 20% in the long run steady state, given the import shares as above. Two-thirds of this would arise due to higher costs of intermediates (an effect analogous to a negative productivity shock) and one-third would arise from the higher cost of capital goods (an effect analogous to a negative savings rate shock). These are quantitatively large effects when the full gap to be explained is 1.000 log points, since they explain one-fifth of Argentina’s decline.

A further place to look for an impact of trade frictions on output is in the process of technology transfer. Here there are plenty of candidate theoretical models, but no consensus on the structure and calibration that best fits the data, nor is there solid statistical evidence for this channel. Accordingly let us rely on recent empirical estimates and, since the effects will turn out to be small anyway, allow ourselves to compute an upper bound for this effect. In recent work, Acharya and Keller (2008) examine the impact of expanded imports from the “technology leader” country on the TFP levels in follower countries, controlling for import levels and R&D intensity in the leader, and interactions between the two. For their analysis, based on mostly developed countries, and the US is the leader.

Here, we consider how the same analysis might apply to Argentina as a follower, where the OECD might serve as the R&D source. One of the upper bound results in Acharya and Keller (2008, Table 8) suggests that a “high” estimate (the 95th percentile) for the elasticity of local TFP with respect to import volume is about 0.06 for the case of R&D intensive sectors. (For many sectors the effect is small or negative, reflecting the possibility of countervailing forces where, say, import competition is destructive of an industry that cannot catch up.)

Applying these estimates, if we apply the 0.06 elasticity to the post-1914 halving of Argentina’s trade volumes, then this implies a reduction of TFP due to weaker technology transfer of about only 3%. In steady state, given endogenous capital accumulation, the impact on income would be somewhat larger and might account for an overall income effect of 5% or just 0.050 log points. So technology transfer via imports would appear to be a very small part of the overall story: the statistical evidence for the channel is quite weak in aggregate, even if we make several calibration assumptions designed to make its impact as large as we dare. To sum up, in contrast to income losses due to inhibited technology transfer (about − 5% or − 0.05 log points of income) the bulk of the income losses due to trade policy frictions (about − 20% or − 0.200 log points of income) would seem to derive from the direct effect of increased input costs for capital and intermediates.

A second area we might examine as an explanation for Argentina’s low income is capital scarcity. By this we mean, in a standard neoclassical growth model, a suboptimal capital/labor ratio, denoted k = K/L. In the simplest model, output per worker y = Y/L is expressed as y = A k ^a , where A is productivity (total factor productivity or TFP) and a capital share of a = 1/3 is the typical exponent used in modern empirical work (Gollin 2002).

As regards the Great Divergence in incomes between rich and poor countries, the consensus since Hall and Jones, has been that the A/A ₀ term above explains much more of the divergence than the MPK/MPK₀ term (e.g., see Easterly and Levine 2001; Gourinchas and Jeanne 2006; Caselli and Feyrer 2007, inter alia). Indeed, for Argentina, Hall and Jones used 1988 data to compute that the MPK term above explained about 5% of the income difference between Argentina the United States. Does this mean that the MPK explanation is dead?

Not quite. Ideas from recent empirical research can provide us with an improved understanding of the evolution of the Argentine capital stock. Properly computed, MPK distortions make a significant contribution to the income gap. For example, Fig. 4 plots the implied MPK for the United States and Argentina using the Hall–Jones method based on installed capital derived from a perpetual inventory method (PIM). Their estimates stopped in 1988 since that was the last year of PWT 5.6, their data source. But we now have PWT 6.2, with coverage until 2004, and we can see that 1988 was quite an unusual year.

Argentina had over-borrowed and over-invested prior to the debt crisis, and then in 1988–90 output was depressed as the economy slumped into recession and hyperinflation, with installed capital suffering heavy underutilization. If one wanted to pick a moment to make Argentina’s APK, and hence MPK look as low as possible, that would have been the year to choose, suggesting a small MPK distortion in total, and none at all after a price adjustment, and hence minor capital scarcity problems. Mismeasurement is, therefore, a potentially serious issue for this calculation.

And as we can see, for most of the last three decades the story has been very different. Using data back to 1960 and the Hall–Jones PIM standardized depreciation rate of 6%, the Argentine MPK level appears to be on average 50% higher than the US level, a considerable wedge. I would argue that the deviations from this pattern in the 1980s and in 2000–03 are easily understood and should be discounted: these were periods of severe economic downturn when measured installed capital is not the same as capital in use. Were it possible to further refine Argentina’s measured capital input time series every year for capacity utilization levels—something no statistician has yet done—we would probably discover similar gaps even in the recession periods.

Are these wedges entirely due to a factor we have already considered, the relative price of capital? If so, we must not double count, which necessitates evaluating MPK at local rather than world prices. The chart shows that this does make a small difference. Evaluated at local prices the gap is clearly not so large, but it is still significant, and it matches up with other recent capital stock estimates using different methodologies. For example, Coremberg (2003) pegs the Argentine and US capital–output ratios in 1998 at 2.85 versus 1.95, respectively, translating into APK levels of 0.351 versus 0.513, and in turn (assuming a = 1/3) MPK levels of 0.117 and 0.171.

These independent country-specific estimates very closely match the rough estimates in Fig. 1 after applying the domestic price correction (where the 1998 MPK levels are 0.129 and 0.180). These gaps have factored in the trade distortions considered above: these are, in other words, evidence of additional capital wedges, beyond barriers to trade in capital goods. And they still show MPK 50% higher in Argentina than in the United States.

These data push back a little against the “it’s A not k” line of argument commonly applied to developing country underperformance. Even researchers working in traditions traditionally sympathetic to TFP-based explanations have had to concede that the large MPK gaps in the 1990s are clear evidence of “capital shallowing” in Argentina. That is, even in the most dramatic period of economic success in recent years, there was a pronounced failure of capital accumulation to keep pace with the path one might expect during a productivity boom (Kydland and Zarazaga 2002). These findings suggest that Argentina does have some difficulty in mobilizing adequate capital accumulation, even when profitable conditions appear. Perhaps from the 1960s to the 1980s slow investment was the counterpart of decelerating productivity, and Argentina could coast along with a depreciating capital base and modest net capital stock additions, but then in the 1990s, the scope for TFP led growth appeared and capital was apparently not adequately mobilized.

The income implications of these gaps are nontrivial. Suppose MPK in Argentina is, on average, 50% or 0.500 log points above the US level as is suggested in the above estimates from the 1990s, from either the PWT or Coremberg. Then in the above expression for income differences, applying the exponent of ½, this capital accumulation friction or wedge explains 25% or 0.250 log points of the overall income difference between the two countries, and we have explained another one-quarter of the Argentine puzzle.

If capital is low, and MPK is high, compared to the neoclassical benchmark, this begs the question: why has Argentina under-invested to such an extent that the marginal product of capital has found itself, so often, stuck far above reference levels? What is the nature of the investment wedge? What underlying factors cause this distortion? I cannot quantify every possible channel, but I propose several candidate explanations which center on factors that either raise the cost of capital or the risk of investment, and all may warrant further scrutiny.

First, there is the problem of risk due to macroeconomic rare events. As is well known, returns to risky investments often appear excessive given what seem like plausible models of risk aversion (Mehra and Prescott 1985). However, the possibility of rare “crash” states or valuation jumps, which wipe out significant wealth through large capital losses, may well be sufficient to resolve this puzzle (Rietz 1988; Barro 2005). And undoubtedly, Argentine history is filled with many examples of crashes that severely damaged many kinds of investment returns. High or hyperinflation events eroded nominal debts on several occasions. These and other major economic crises have often left the banking sector in ruins, causing credit crunches and broader losses on a wide range of financial instruments. If, as a result, investment returns are more crash prone in Argentina then investors may demand a higher return as compensation for volatility and/or skew, implying a higher equilibrium MPK in aggregate. These risks may also be manifest in a repressed financial system with lower money multipliers and leverage, further tightening credit.

Second, there is the problem of default risk and property rights. In addition to rare events driven by market fluctuations, possibly in response to macroeconomic policies, we also have to recognize that explicit confiscation or redistributions of wealth, or other failures of property rights, have often figured in Argentina’s history. Beyond the serial pattern of default (Reinhart and Rogoff 2004), we would include bank suspensions, forced debt conversions, specifications, and other expropriations. Although on occasion, ex post, these events were discriminatory as to locals versus foreigners, on most occasions, and in general ex ante, such differential treatment may not have been expected.

If capital price distortions (e.g., trade policy) explain 0.250 log points of income difference, and capital accumulation frictions (high MPK) explain another 0.250 log points, we have explained one half of the 1.000 log point income difference. This is not trivial. A 50% increase in income per person would lift Argentina from the $8000 level to the $13,000 level (roughly on a par with Greece, Portugal, and approaching South Korea). And even in 1913, at its relative peak, Argentina’s income was at most 70–80% of US or UK levels, so were even half of today’s gap closed like this we would probably not speak so much of an Argentine puzzle.

Still, can we explain any more of the OECD-Argentina gap? There is a reason to think that we can, for various reasons, given several empirically important factors we have not yet accounted for. Three such factors could be very important.

First, there is the issue of investment quality. All calculations of MPK rely on calculations of capital stocks based on PIM or HV methods and many standardized assumptions. But capital “quality” may be generally lower in poorer countries. Public investments are often more dilapidated in poor countries with low quality of governance, and where large fractions of public investment spending are lost to bribery and corruption (Tanzi and Davoodi 1997). Firm data from some countries suggest that the same may be true of private sector investments (Bu 2006), perhaps due to private sector corruption; or due to high costs or barriers to technical maintenance; or due to capital complementarities with adversely maintained public capital, leading to premature discard or under maintenance. Capital is thus less productive and of lower capacity than its vintage alone would suggest and some correction for higher rates of depreciation is warranted. For example, the Hall–Jones method assumes a 6% depreciation rate on all capital. But these rates may be far too low for uniform application to rich and poor countries. Bu (2006) estimates “low” median firm depreciation rates for all fixed assets as 16% (Philippines) or 12% (South Korea) in the 1990s. In Indonesia and some African countries, the reported median depreciation rates are higher still, between 25 and 60%. This poses a profound problem for capital stock and MPK estimates because the results are highly sensitive to the depreciation parameter: increase this parameter by 1% and the implied PIM measure of the capital stock falls by 1%, and implied MPK rises by 0.67% (if a = 1/3). If capital quality is lower and depreciation higher than typically assumed, Argentina could be even more capital scarce than has been commonly thought.

Second, there is the problem of investment misallocation. The MPK calculations also rely on the assumption that capital is efficiently allocated within the economy, or that MPK is equalized across sectors. But a contrary view with a log tradition maintains that this is unlikely to be the case in developing countries. Instead, investment may be misallocated for a variety of reasons—such as corruption and inefficiency in the private financial sector or the role of the state in allocating finance. Work by Hsieh and Klenow (2009) on China and India suggests that, compared to the US, an efficient re-allocation of capital could be the equivalent of a 50% or larger increase in TFP. It is quite plausible that similar misallocation problems, although perhaps not as grave, could affect Argentina and would go along way to explaining any remaining income gaps, over and above what we have measured so far. This is likely to be a productive area for future research, using industrial census data and other measurements.

Third, and finally, there is the role of investment variety. Input price distortions were probably the main trade-related drag on Argentina’s growth in the twentieth century. After 1914, and particularly from the 1930s to the 1950s, this scenario could be ascribed in some large part to highly unfavorable global conditions for open trade; but once global trade started to boom thereafter, self-inflicted trade policy distortions would remain as the principal cause of the problem. The estimate presented above (0.200 log income points) may also be an understatement since it focuses only on the so-called “intensive margin”—the quantity of a given set of goods imported. But recent empirical research in the trade literature suggests that comparable economic costs may be inflicted by input tariffs on the “extensive margin”—by limiting the variety of inputs that are imported. If these results carry over to intermediate and capital inputs, as they well might, then we would have identified yet another trade-related barrier to investment. Quantifying that impact for a broad range of countries, as well as for Argentina itself, remains an important goal of future research.

These three additional factors—investment quality, allocation, and variety—represent additional barriers to efficient investment which have also probably acted as a drag on Argentine economic performance, even if the magnitudes in question remain open frontiers for research.

The discussion so far of likely “microexplanations” suggests that we know, within some approximate bounds, how various economic policies and institutional deficiencies in the Argentine economic environment might have contributed to economic underperformance. And indeed these contributions appear to be empirically large, sufficient to explain much of the divergence witnessed. But this only pushes the question deeper: why have such choices been made and what can account for them?

For all countries, not just Argentina, economists and historians have grappled with this question in a bid to explain the deep and exogenous origins of the Great Divergence. The problem, as I argue in this section of the paper, is that in the particular case of Argentina the explanations that have been proposed—and which may seem to work quite convincingly in many other countries—do not appear half as persuasive when applied to Argentine economic history.