2.1 The first wave

In the 1820s, the new independent governments of Latin America approached the burgeoning international capital markets of London and Amsterdam, as shown in Table 1. Funding was sought to establish security and infrastructure, and on a smaller scale the private sector went in search of development finance. British investment dominated the first wave.

TABLE 1 DEFAULT HISTORY OF LATIN AMERICAN GOVERNMENT BONDS ISSUED IN THE 1820S

Source: Rippy (Reference Rippy1959, pp. 26-28).

In 1822, government bond issues were floated by Colombia, Chile, and Peru with a face value of £3.65 million; in 1824, there were new issues by Colombia and Peru, plus Buenos Aires, Brazil and Mexico to the tune of £10.4 million; and in 1825, Peru (yet again) plus Brazil, Mexico and Central America issued bonds for a further £7.1 million. Sold at an average discount of almost 25 per cent, these £21 million in government bonds realised on net only £16 million for the borrowers. As investors soon discovered, these issues were at best risky, at worst (in the case of the fictional Poyais) a fraud. When fiscal burdens escalated with the wars of independence and subsequent civil wars, a wave of defaults ensued, with all bond issues in default by 1827 (Rippy Reference Rippy1959, pp. 17–24; Stone Reference Stone1977, p. 692; Marichal Reference Marichal1989, pp. 13–14).

New loans were not extended to the region until the defaults were resolved and political and economic stability seemed more assured, a process that took years and, in some cases, decades, as noted in Table 1. Of the various 1820s sovereign issues that quickly failed, only Brazil, after a short moratorium in 1826-1829 could avoid default, but most remained in default for decades, with restructuring attempts frequently subject to failure as well (de Paiva Abreu Reference de Paiva Abreu2006). Here was a seemingly clear case where reputation mattered: the bad debtors paid for their defaults by being excluded for a long period from the financial marketsFootnote ².

2.2 The second wave

Starting in the 1850s, there was a marked renewal of interest in Latin America in the London capital markets. By 1880, these new investments had accumulated into a sizeable stock that dwarfed the previous boom in the 1820s, and by then a total of £179 million was outstanding to Britain, £123 million in government bonds (69 per cent) and £56 million in private enterprise debts (Table 2). The new surge was driven in part by a global trade boom from the 1850s until the onset of the Great Depression of the 1870s. More exports and imports meant more revenues (principally from customs duties) that governments could use to amortise loans. These new debts constituted a major increase in public borrowing and a test of the governments’ creditworthiness after decades of «financial hibernation». A total of fifty major foreign loans were negotiated from 1850 to 1873, most in London, and a few in Paris and other European markets (Marichal Reference Marichal1989).

TABLE 2 BRITISH INVESTMENTS IN LATIN AMERICA AT THE END OF 1880

Source: Rippy (Reference Rippy1959, pp. 25, 32).

But the extension of credit to sovereigns was more selective in the second wave as compared with the first — investors avoided the riskier locations and started to follow the signals given by the few countries that had shown some dedication to debt service. With respect to sovereign loans, Brazil had worked harder than other countries to honour debts and was duly rewarded with the largest share of the new flows. Other countries took longer to re-establish their creditworthiness. Argentina did not fully resolve internal disputes and old debts until 1857, and only then did new loans appear. Paraguay borrowed in London in 1871. Uruguay and Bolivia could do likewise in 1872 (the first Bolivian issue in 1864 had failed). Chile floated issues in 1858, 1865, 1866, 1867, 1870 and 1873 totalling £8.5 million. Costa Rica, Guatemala and Honduras all issued non-refinancing debt (new net inflows) in the peak of the investment boom from 1867 to 1872 (Rippy Reference Rippy1959; Marichal Reference Marichal1989).

As might be expected, risk premiums paid by countries varied over a wide range. Good risks like Brazil or Chile could float loans with 5 per cent coupons at a price of 80 or 90, for a yield under 6 per cent. The exceptionalism of the Peruvian case also deserves mention. As stated by Vizcarra (Reference Vizcarra2009), Peru's experience with sovereign debt was an unusual case where, in spite of ongoing political instability and a poor capital market reputation, due to the enormous guano boom, whose large export proceeds served as a quasi-backstop to national solvency, the country was accorded a low credit risk until the 1870s (Vizcarra Reference Vizcarra2009). Argentine coupons ran to 6 or 7 per cent, and the issues sold at around 90. Costa Rica floated 6s and 7s and sold them for about 70. But war-torn Paraguay offered 8s, and Honduras 10s, and these bonds still could not be sold for more than 80 (Marichal Reference Marichal1989).

But a global macroeconomic and financial crisis was stirring yet again, and a second wave of defaults soon spread over the region in the 1870s. By the end of 1880 of the £123 million of British capital invested in Latin American government bonds, more than £71 million (58 per cent) were in default (see Table 2). Some of these loans had been a bad idea in the first place, and some were again tainted by fraud. But even the genuine loans in the larger republics ran into servicing problems as the global depression spread. Credit conditions suffered. A much wider global debt crisis was under way of which Latin America was only a small part: by 1876 fifteen non-European nations had defaulted to the tune of £300 million. Global capital flows again ground to a halt and irate bondholders chased down the insolvent republics long into the 1880s. Settlements were again drawn out and defaulting governments were shut out of new borrowing during negotiations and often for many years beyond.

2.3 Third wave, crash and fourth wave

An even bigger borrowing boom began in the 1880s as global economic activity, and especially trade, recovered. Defaulting governments gradually straightened out their fiscal problems and sought access to credit again. The overall flows were massive, and by the end of 1890 total British investments in the region were £426 million, more than double the 1880 total. Of this, £194 million sat in government bonds, now for the first time surpassed by a slightly higher amount, £231 million, in securities issued by private enterprises (Rippy Reference Rippy1959).

The regional distribution of the new wave of investment favoured countries that prospered the most in the new trade boom. In the 1880s, capital flows were concentrated in just five countries: 37 per cent in Argentina, 17 per cent in Mexico, 14 per cent in Brazil, 7 per cent in Chile and 5 per cent in Uruguay. Government loans were more skewed, with 60 per cent of new loans going to Argentina and Uruguay. Economic divergence was starting to be seen: Foreign capital — which sought out the most profitable investment, the most dynamic economies, and the most creditworthy countries — played a part in furthering the economic divergence in the region (Marichal Reference Marichal1989).

As financial development and monetisation in Latin American economies grew in the late 19^th century, the consequences of government-induced macroeconomic crises became deeper and more far-reaching. With any increase in the probability of default, sovereign spreads widened and the capital market tightened. Domestic banks found themselves in distress, and a credit crunch followed that squeezed local borrowers. Whereas government defaults in the 1820s and 1870s could bypass pre-modern economic modes of production that relied more on retained profits and less on financial intermediation, by the 1890s the region's more modern economies risked more resounding economic crises after a default. The major crises in the 1890s for two large capital recipients, Argentina and Brazil, illustrated these new financial risks.

The first crisis was in Argentina — arguably the world's first example of a modern «emerging market» crisis, combining debt crisis, bank collapses, maturity and currency mismatches, and contagion. Argentina's bold development strategy of the 1880s rested on a highly leveraged parastatal banking sector, which borrowed in gold, and lent in pesos. When the economy faltered and the fiscal gap widened, it was covered by money printing, which broke the exchange rate peg and unleashed inflation. A generalised financial and banking crisis ensued. Stabilisation and debt restructuring took the better part of a decade. Foreign capital flows dried up as a consequence of the regional impact of the Baring crash, and a global recession contributed to a delayed recovery (della Paolera and Taylor Reference della Paolera and Taylor2001; Mitchener and Weidenmier Reference Mitchener and Weidenmier2008).

The other major crisis then hit, in Brazil. Political and economic instability was high in the 1890s following the proclamation of the Republic: the country was adjusting to the abolition of slavery, the gold standard had been abandoned, and inconsistent monetary and fiscal policies had the printing presses running at full speed. The currency steadily devalued by a factor of 3.5 from 1890 to 1898, adding to the domestic costs of debt service. Default was put off for a time, but was unavoidable in 1898-1900, and again in 1902-1909. By then, the real economy was by in deep recession, having never really recovered from the financial instability of the early 1890s (Cardoso and Dornbusch Reference Cardoso and Dornbusch1989; Fishlow Reference Fishlow1989; Triner Reference Triner2000; Abreu Reference de Paiva Abreu2006).

Global capital markets quickly recovered from the crisis of the 1890s, although countries badly affected, notably Argentina, took longer to recover. However, compared with the 1870s boom and bust, this one was not associated with widespread default in the region, but rather a more general and global increase in country risk that slowed foreign capital flows for the better part of a decade. Inflows to Argentina and Uruguay were sluggish in the 1890s, but in other countries in the region the tap was still somewhat open as shown in Table 3. As Table 3 also shows, capital flows resumed in full force from 1900 to 1914, when, as is well known, capital flows surged in the great fourth wave which represented the culminating years of the first age of globalisation, an era only brought to an end by a political shock, the outbreak of the Great War.

TABLE 3 CUMULATIVE GROSS CAPITAL FLOWS FROM BRITAIN TO LATIN AMERICA, 1880-1913

Note and Source: Millions of pounds, from Stone (1999).

***

Latin America's experience was characterised by successive booms and busts, as four great waves of foreign capital flowed into the region; in the booms both private investment and sovereign lending tended to surge, but these episodes were separated by periods of near autarky in sovereign debt markets and a retrenchment in private flows. Each wave generally ended with a crisis in one or more countries. Very often it was a twin or triple crisis that damaged the value of the currency, the solvency of public debt, and (especially in later times) the health of the financial sector. The inflows also became rather more geographically concentrated over time. Markets tended to pick out «winners» among these emerging markets, and fund them well, even after crises; other countries gained access once or twice, but spent most of their time excluded from large-scale external capital flows.

Still, the reach of the global capital market should not be overstated — the country risk perceived by foreign investors remained high and only a handful of countries had reputation enough regularly to issue external debt. Even as late as 1913, the point of deepest maturation of the global capital market before the 1980s and 1990s, the five countries of Argentina, Brazil, Chile, Mexico and Uruguay accounted for 90 per cent of outstanding Latin American issues in London. Countries like Peru and Venezuela made occasional issues, often merely rollovers of earlier defaults. Others were out of the game entirely. In the first era of globalisation, just as today, capital flows were highly volatile. To better show their stop-and-go nature Figure 1 presents annual data on capital flows to the region. If we compare trends in capital flows with the country risk data in Figure 2, for then a more complete picture of the global crises emerges. Booms were typically associated with a convergence in bond spreads; defaults were associated with a sudden stop of capital flows and dramatically increased country risk.

FIGURE 1 BOOM AND BUST CYCLES, 1850-1914

Source: Taylor (Reference Taylor2005).

FIGURE 2 COUNTRY RISK, 1870-1914

Source: Taylor (Reference Taylor2005).

3.1 Variations in investment needs

Overall investment financed from abroad is not necessarily a good predictor of overall government spending financed from abroad. That is, the economy-wide stock or flow of foreign investment need not correlate with the subsets of flows and stocks involving the government sector. However, in the 19^th century there were powerful reasons to expect a correlation between the two and this logic underpins a discussion of the links between the two. The basic reason for the high correlation of the two measures was the strong complementarity between private and public sector investments in this period.

The case of transport and infrastructure is the obvious example. As we have noted, foreign finance of railways was a dominant category of foreign capital flows in this period (Twomey Reference Twomey2000). When the railroads were publicly operated, the lending was directed via government borrowing. But even when the railroads were privately owned, the construction of the railroads was often accompanied by significant public expenditure: related infrastructure, guarantees and subsidies, and so on. The same was true of ports, canals, and other large projects. At a more general level, a fast growing economy generally needed broad investments in a whole range of public goods as part of a «balanced growth» strategy, so that utilities, education, and so on, would all be higher in a high investment path.

Can simple one-sector macroeconomic models help us here? These models do not make a distinction at a fine level between types of capital, but we need to keep such complementarities in mind when understanding the motives of governments to borrow. We do know that the countries in the region had very different investment needs in the 19^th century and this certainly affected their overall need to draw on foreign capital inflows, and hence their need to engage in infrastructure-led public borrowing.

In the simplest macroeconomic sense, we could speak of some countries being much more capital scarce than others, and therefore having a much higher marginal product of capital (MPK). In a standard neoclassical model, capital would tend to flow to the countries with the highest MPK. However, in what is known as the so-called «Lucas paradox», the highest MPK need not be found in the poorest country (Lucas Reference Lucas1990).

Consider a standard production function where output per worker y depends on productivity A and capital per worker k. If all countries have the same productivity level A, then log MPK is perfectly inversely correlated with log k, and classic diminishing returns means that capital will flow to the poorest country. In general, this is not true: countries do not have the same A and this offers the simplest resolution of the Lucas paradox. Much of the great divergence in incomes between rich and poor countries today is explained by A and not k (Hall and Jones Reference Hall and Jones1999; Easterly and Levine Reference Easterly and Levine2001). Explaining A (its level and/or its growth rate) has led to debate concerning the importance of deeper determinants such as colonial history, human capital, legal origins, institutions, policies or geography (Sachs and Warner Reference Sachs and Warner1995; Sachs et al. Reference Sachs, Gallup and Mellinger1998; Acemoglu et al. Reference Acemoglu, Johnson and Robinson2001; Glaeser et al. Reference Glaeser, La Porta, Lopez-de-Silanes and Shleifer2004).

How well do the insight from this contemporary debate translate to the 19^th century? And how well do any lessons apply to Latin America and the heterogeneous experiences within the region?

At the most general level, recent historical research suggests that the Lucas paradox was less apparent in the first era of globalisation than it is today. Figure 3 shows Obstfeld and Taylor's (Reference Obstfeld and Taylor2004) simple tabulation of the distribution of foreign investments by country in 1913 and in 1997. In this «then» vs. «now» comparison it is evident that the global capital markets of a century ago were much more successful at directing financial flows to very poor countries as compared with the capital markets of today. Moreover, these flows included significant investments in Latin America, for example by 1913 cumulative gross capital flows from Britain to Latin America accounted for a lower bound of 23 per cent of total British overseas investments. This begs the question as to what fundamental forces led to such an outcome. To do this one may analyse the flows themselves, the quantities or the prices at which the flows moved, which in the case of government debt means looking at the risk premia charged to sovereign borrowers, a challenge we take up below.

FIGURE 3 FOREIGN CAPITAL IN RICH AND POOR COUNTRIES: THEN VS. NOW

Source: Obstfeld and Taylor (Reference Obstfeld and Taylor2004).

3.2 Variations in saving scarcity

Since capital inflows are, by definition, the difference between domestic investment and savings, it does not suffice to look only at investment-based (demand side) determinants of capital flows, but also at the saving-based (supply side) determinants. This applies as much to aggregate flows of saving, investment and foreign capital as it does to public (government) debt. Two important aspects of the Latin American economies on the supply side should be noted, factors that have persisted to some extent up to today: financial underdevelopment and demography.

According to financial historians like Davis (Reference Davis1965), the function of capital markets in the aggregate can be broken down into two functions: the mobilisation of savings supply resources and the allocation of these resources to competing investment demands. If the system is efficient, then the maximal resources are mobilised, and they are allocated efficiently across sectors and firmsFootnote ⁴. This ideal is never reached, and reality falls particularly short in developing countries.

In a closed-economy view of the world, these failures of financial development could place a serious brake on economic growth. (Schumpeter Reference Schumpeter1911; Gurley and Shaw Reference Gurley and Shaw1955; King and Levine Reference King and Levine1993). Some evidence has been adduced to this effect, although controversy still swirls around the finance-development nexus given the possible reverse causality running from growth to finance.

In an open economy world, however, the inefficiency of financial markets may place a tax on long-run national wealth, but it need not slow economic growth. In a hypothetical frictionless world, small open economies lacking the means to mobilise or allocate their own savings could still draw on the pool of global savings available at the real world interest rate. Foreign banks and financial firms might then enter the domestic market to allocate and monitor the capital so raised. In line with this logic, in the 19^th century, differences in domestic financial development did not necessarily hinder countries’ ability to accumulate capital, even if they did affect subsequent development paths once the global capital market collapsesFootnote ⁵.

For government finance, the effect was to supply at low transaction costs a large group of counterparties in London, and elsewhere, who were willing to subscribe to government debt issues, for whom the investments in any one periphery country were a small part of their portfolio. In contrast, working through the domestic financial market meant high transaction costs using inefficient banks to reach a small pool of domestic investors for whom a debt issue would be a large share of their portfolio. It is easy to understand why the former route would be preferable. Indeed, the latter route was often ignored, unless there were telling reasons to use it — for example, a loss of access to external credit after a default, or a desire to engage in financial repression or «forced savings» and thus tax the domestic bond holders (whether institutions or individuals) either through below market nominal interest rates or via the inflation tax on paper debt, tactics that could not be exploited too often without totally destroying what little financial development there was.

In essence then, open capital markets created the possibility of «trade diversion» in asset trade. We can recast this in the language of gravity models, which have been successfully applied to financial flows also. In a country with poorly developed capital markets the intra-national transaction costs were high. If the country were closed, however, «multilateral resistance» against all foreign countries would be high, sustaining the domestic market. But if inter-national transaction costs suddenly fell — globalisation — the country would be expected to divert financial flows from domestic to foreign markets.

These forces were something of a double-edged sword in countries that plugged into the global capital market before 1914. On the plus side they made the supply of capital much more elastic, and this could help explain the very positive association between foreign investment and growth, notwithstanding any claim about causality (Schularick and Steger Reference Schularick and Steger2006). On the minus side, these pressures possibly deterred domestic financial development. The latter proved to be a major weakness once global capital supply dried up after 1914. Countries had to turn to the domestic supply of capital, and domestic intermediaries. As argued, the intermediaries were often few and inefficientFootnote ⁶. In addition, supply of capital was low too. Why?

The dominant explanation argues not only that savers were cutoff from investors by high transaction costs, but that savers were themselves rather scarce for demographic reasons. Taylor (Reference Taylor1992) made the argument for Argentina, and it would apply to some other countries too. In many developing countries then, as now, fertility and population growth rates were very high. The standard life cycle argument would predict that such countries would tend to save less, as compared with countries with a more mature population with greater numbers in high-saving mid-life cohorts. Taylor and Williamson (Reference Taylor and Williamson1994) show how these effects could explain a fair portion of the capital flows from Britain to the settler economies before 1914.

This argument offers another explanation for the need for governments to borrow from abroad — at home, there were few savers in the population. This may well have been efficient for other reasons. Much domestic saving in Latin America was within the firm or the household, and this was probably important for the financing of many small firms and businesses. We know, for example, that domestic stock markets were often quite thin and banks were often financing only large firms with some type of «connection» (Haber Reference Haber1991; Triner Reference Triner2000; Regalsky Reference Regalsky2002; Hanley Reference Hanley2005). Thus, small enterprises, if they relied on domestic saving were liable to feel crowding out if government borrowing went directly to bond issues to the domestic market, simply because such small firms were likely to have severe asymmetric information problems that would preclude them issuing debt or equity abroad. On the other hand, the government was rather better placed to issue debt overseas as compared with such small firms, so it did them a favour by doing so, leaving most of the market and non-market domestic capital flows untouched.

***

So far we have examined government (and aggregate) borrowing as driven by the forces of demand and supply. However, a third and final set of factors must be considered — frictions or wedges, in the form of transaction costs. These costs raise the expected return of the marginal dollar invested above the expected returns of the marginal dollar saved, and thus serve to inhibit transactions between borrowers and lenders.

For 19^th century Latin American governments borrowing in the world capital market, these frictions took two principal forms: exclusion from the market, leading to quantity rationing; or market access subject to a risk premium, a sovereign spread over the market's benchmark bond yield (in those days the British consol). Market exclusion, as we have seen, was typically a result of unresolved past defaults. What drove risk premia? A considerable body of research in recent years has explored this topic, and several pertinent findings can be summarisedFootnote ⁷.

3.3 Gold standard

In a seminal paper, Bordo and Rockoff (Reference Bordo and Rockoff1996) offered evidence that sovereign borrowers received a lower risk premium when they adhered to the gold standard — the authors referred to this as a «seal of approval» whereby the sound policies necessary to achieve successful (i.e., credible) commitment to the gold standard also operated to lower the macroeconomic risks that were of concern to holders of sovereign debt, namely the risks of default because of economic underperformance or fiscal crisis. This hypothesis was lent further support by Obstfeld and Taylor (Reference Obstfeld and Taylor2003) who studied much larger sample of countries in the 1870-1913 period. The risk premium appeared to be 40 or more basis points lower for a country after it went on gold, and this effect remained after many other controls were added.

Of course, being on the gold standard was an endogenous outcome and was also correlated with many other macroeconomic variables, so this effect needs to be carefully interpreted. More generally, attempts to extend the Bordo-Rockoff analysis by adding right-hand side variables to produce a «kitchen sink» regression have to be interpreted with care, and if the right-hand side variables are collinear with the gold standard then the results may be meaningless.

To illustrate the correlation problem, we can examine results in Flandreau and Zumer (Reference Flandreau and Zumer2004) where exchange rate volatility is added to most regressions and the authors find that the inclusion of this variable reduces or negates the impact of gold standard adherence. This is not surprising since going on gold was the way in which countries lowered their exchange rate volatility in this period: hence these results, which ignore the fact that exchange rate volatility depends on the exchange rate regime, offer no evidence for or against the «seal of approval» hypothesis. A more robust attack on the gold standard effect was offered by Ferguson and Schularick (Reference Ferguson and Schularick2006, Reference Ferguson and Schularick2012): they have a larger sample that includes many African and Asian poor countries. In poorer countries (where the cutoff was in the US$1,000-2,000 per capita range in 1900, roughly 20-40 per cent of the British level), the impact of the gold standard on spreads appears weak: perhaps their politico-economic institutions were too weak to make gold standard commitments credible. However, in samples that focus more on the rich Atlantic economies, and include the main Latin American borrowers the gold standard effect remains. We might infer that, as a region of middling incomes and middling institutional quality, Latin America was poised on the periphery of an Atlantic club where gold standard adherence could, if credible, offer some benefits.

Even if that is accepted, however, the endogeneity of the gold standard is a more difficult nut to crack. Econometrically, Obstfeld and Taylor (Reference Obstfeld and Taylor2003) used GMM techniques to address this issue, but other approaches still need to be explored, for example the use of a first-stage model of gold standard adoption along the lines proposed in an innovative paper by Meissner (Reference Meissner2005). These results matter for the Latin American countries in the 19^th century because they were generally among the weakest of countries at maintaining gold standard adherence. However, returning to the endogeneity issue, this really begs the question: why? What was it about the region's economies that made it so difficult for them to stick to a hard monetary regime?

3.3 War

Obstfeld and Taylor (Reference Obstfeld and Taylor2003) also included controls for the occurrence of military conflict involving the sovereign borrower, both civil war and inter-state war. Perhaps surprisingly, both of these variables turned out to be statistically insignificant as determinants of country risk. Ferguson and Schularick (Reference Ferguson and Schularick2006) found mixed results. However, the arguments of Bordo and Kydland (Reference Bordo and Kydland1995) that the gold standard was a «contingent rule» caution against a simple interpretation of these weak findings. Wars tended to mean going off gold, so indirectly, via the gold standard effect, we may still be capturing an effect of conflict on sovereign risk. Returning to the last point, we can see here one endogenous source of weak gold standard adherence. In this sense, we do find support for the notion that the prevalence of conflict in the region may have had something to do with Latin America's high sovereign risk. Again, this cautions against using the kitchen sink regression as a means to refute the gold standard as a seal of approval.

3.4 Institutions and empire

Looking at institutional determinants, it is not clear that most parts of Latin America could have been expected to attract large-scale capital flows. Stressing extractive institutions arising from colonial origins, Acemoglu et al. (Reference Acemoglu, Johnson and Robinson2001) reach a conclusion that builds on the insights of North and Thomas (Reference North and Thomas1973) concerning the legacy of Iberian rule in Latin America. Spain and Portugal did not establish colonies that were characterised by good political and economic institutions. Power was concentrated in privileged elites, democracy never flourished, property rights and rule of law were weak (except where needed to protect the elite). In the AJR account these flaws persisted after independence, and slowed economic growth, keeping the region poor. Here, institutions are the fundamental driver of A, which in a Solovian model determines a country's level of productivity and hence income per capita in the long run as k converges to a steady state level that depends itself on A.

The Acemoglu et al. view is often — wrongly — seen as purely deterministic, but of course there is still an error term in their model, and residual variation is significant. Between the colonial past and present outcomes, history reveals even bigger residual variance. In a challenge to the persistence story Prados de la Escosura (Reference Prados de la Escosura2005) notes that the «great divergence» between Latin America and the core economies was even more a 20^th century than a 19^th century phenomenon. If Prados de la Escosura is correct, then, perhaps despite political and institutional weaknesses, the region did manage to sustain growth in the 19^th century, and hence became attractive to foreign capital, except where the worst political and institutional failures could not be contained.

This may call for future research to assess the more detailed mechanisms needed for an «institutional explanation» — that is, why some institutions matter more than others, and at some times more than others. For example, Vizcarra (Reference Vizcarra2009) explains clearly that the vast amount of exportable resources (the guano) and a well-structured debt service scheme could overcome the complete lack of Peruvian stable political institutions. Some countries obviously surmounted their colonial legacies, if only for a time, and were able to get onto a reasonably fast growth track. The most dramatic example would be Argentina, which by 1900-1913 was one of the five richest countries in the world and a prime destination for foreign capital. Argentina's subsequent growth failures put the country back in the ranks of the less developed a century later, so perhaps its colonial legacy finally caught up with it. Brazil, and several other countries have also witnessed ups and down in growth performance, yet without sustained convergence. These perturbations along the development path deserve greater scrutiny, as we need to better understand the extent to which countries are «prisoners of history» and how many degrees of freedom they have that allow them to escape from their past.

In this respect Latin America again stands as a crucial test case. In other areas of the periphery still under colonial rule, it would be simple to attribute the large flows of capital in the 1870-1913 period to institutions, such as rule of law and property rights that were enforced by the colonial powers themselves. As Lucas himself noted:

The same idea supports the theory of «anglobalization» that underpins Ferguson's (Reference Ferguson2003) qualified rehabilitation of the benefits of the British Empire. Of course, empires have since collapsed and it can be argued that this crucial difference explains the absence of a Lucas paradox then, but its presence today (Schularick and Steger Reference Schularick and Steger2006).

Whether one views the financial and institutional exports of the European imperial epoch as a plus or a minus, it is fair to say that Latin America's post-independence experience remains relatively neglected in this explanatory framework, and fits none to well either with the pro- or anti-colonial view. The region was politically independent then, as now; and some countries enjoyed respectable economic growth and capital market access; we still have to answer how they pulled it off. On the one hand, defaults were undoubtedly higher «on average» than in the Empire group. On the other hand, the region still managed to attract significant capital flows despite those risks. The benefits (returns) must have outweighed the costs (risks), which would be consistent with the Prados de la Escosura (Reference Prados de la Escosura2005) story that, despite its flaws, some parts of the region had good growth potential. Colonial origins did not doom the region to failure, at least up to 1914.

The empire thesis was examined by Obstfeld and Taylor (Reference Obstfeld and Taylor2003) and Ferguson and Schularick (Reference Ferguson and Schularick2006, Reference Ferguson and Schularick2012). Both found little evidence for an «average» discount based on British empire membership using the industry-standard (Bordo-Rockoff Reference Bordo and Rockoff1996) CAPM framework, but the latter found stronger support for a lower «beta» among Empire borrowers when an Empire-beta interaction was added. Ferguson and Schularick (Reference Ferguson and Schularick2006, Reference Ferguson and Schularick2012) have the advantage of an even bigger sample than Obstfeld and Taylor (Reference Obstfeld and Taylor2003), including many more British colonies in Asia and Africa, although such a sample obviously weights Latin American observations a lot less. Both also found evidence of noticeable convergence of risk premia over the 1870-1914 period. In the Ferguson-Schularick framework, this trend is mopped up either by time dummies or by the CAPM Empire-beta interaction, and in both cases it negates the gold standard effect: in the former case «something else» explains convergence and in the latter case it is the gradual diminishing amplitude of the global weighted risk premium. Of course, either or both of these effects are correlated with the global spread of the gold standard, so inference is still unclear. We need to remember that the gold standard was not viewed as a fundamental by Bordo et al.: merely as a «seal of approval» which, if credible, was indicative of other fundamentals. Such fundamentals could, of course, include the Empire — where the gold standard was virtually a sine qua non of macroeconomic regime design.

Could such an alternative explanation exist for interest rate convergence after 1870? Perhaps «institutional convergence» was afoot in the global economy in this period? Schularick and Steger (Reference Schularick and Steger2006) argue that smaller property rights differentials could explain the absence of a Lucas paradox before 1914. However, with no data on property rights before 1914, this hypothesis awaits further testing, for example, in the manner of Alfaro et al. (Reference Alfaro, Kalemli-Ozcan and Volosovych2005). But why would such a convergence have occurred? And when? One might appeal to the notion of «informal empire» or «soft power» to argue that, in effect, British (or at the least, European) capitalistic norms were being spread around the globe under the influence of the colonial police powers — that «anglobalization» had spillovers beyond the formal empire itself by setting global «rules of the game» in the shape of norms for good behaviour, that is, global public goods (Ferguson and Schularick Reference Ferguson and Schularick2006, Reference Ferguson and Schularick2012). Such an argument would say that, for all its notional independence, Latin America could not have chosen to abrogate property rights before 1914 without facing deep shame and (more importantly) stiff penalties. There is some evidence that the flexing of U.S. power in the Latin American region may have been instrumental in this regard (Mitchener and Weidenmier Reference Mitchener and Weidenmier2005). However, with no way to measure informal empire or its influence, these conjectures remain largely untestable. Indeed, if the entire globe is under the sway of an informal empire, the hypothesised effect cannot be identified at all.

3.5 Default history

Another place to look for an explanation of high sovereign risk is in the default record. Obstfeld and Taylor (Reference Obstfeld and Taylor2003) found a high penalty for contemporaneous default: maybe 100 basis points for full default and 50 basis points for partial default. But what about the longer term record? Ferguson and Schularick (Reference Ferguson and Schularick2006, Reference Ferguson and Schularick2012) also find a strong impact of past default on current spreads.

The role of past default in determining current spreads is central to the «reputation» model of the sovereign debt market. The 1820s and 1870s crises started to cement in investors’ minds the untrustworthiness of Latin American sovereign borrowers, a reputation that was to expand in the years ahead and which persists even to this day. According to Tomz (Reference Tomz2001), of the seventy-seven government defaults from 1820 to 1914, fifty-eight involved Latin American countries (75 per cent). Compared with other periphery countries, the economic potential and sovereign independence of the region obviously encouraged this outcome: the potential for high returns encouraged more borrowing ex ante and the independence from Empire gave more freedom to default ex post.

Clearly the borrowers in the region could not manage their fiscal affairs with anything approaching the prudence of most borrowers in the core countries. Figure 4 shows the incidence of sovereign default in the region from 1820 to 1940 and the fraction of years that debtors spent in default status is impressive, 38 per cent on average. The better-behaved borrowers like Uruguay (12 per cent) or Brazil (17 per cent) managed to maintain a pretty clean sheet, but the odds of getting repayment from others like Honduras (79 per cent) or Mexico (57 per cent) were no more favourable than a coin toss.

FIGURE 4 DEFAULT, 1825-1940

Note: Fraction of years in default shown in parentheses. Source: Taylor (Reference Taylor2005). Default data from Tomz (Reference Tomz2001), issue dates from Marichal (Reference Marichal1989).

We can assess the importance of reputation for capital market access in the first era of globalisation by following Reinhart et al. (Reference Reinhart, Rogoff and Savastano2003, «RRS») and Reinhart and Rogoff (Reference Reinhart and Rogoff2004, Reference Reinhart and Rogoff2009 «RR»). We look at the relationship between default history and country risk. The country's default history is captured by the number of defaults or the fraction of years spent in default over a specified prior period, in this case 1801-1900. Country risk is measured by the spread of the country's long-term gold bond yield over the benchmark yield of the British consol, using data collected by Obstfeld and Taylor (Reference Obstfeld and Taylor2003), in this case for 1900-1913.

We might expect that a country with a record of «serial default» ought to be penalised with higher country risk, and «RRS» show that this is the case in the present era. «RR» show that this behaviour matters for capital market access today. Was it true in the past also? Figure 5 shows that more defaults and longer defaults in the 19^th century were associated with higher spreads in 1900-1913. In other words, the problem of serial default is an old one, and there appears to be evidence that even in the 19^th century, global capital markets were quick to identify and punish a serial defaulter.

FIGURE 5 SERIAL DEFAULT AND COUNTRY RISK

Source: Obstfeld and Taylor (Reference Obstfeld and Taylor2004), Reinhart and Rogoff (Reference Reinhart and Rogoff2004).

3.6 Fiscal burdens: levels and volatility

Obstfeld and Taylor (Reference Obstfeld and Taylor2004) found that debt/GDP ratios had very little impact on sovereign risk in the 1870-1914 period, suggesting that so long as the debt path was consistent with gold standard adherence, creditors were willing to take the latter signal as credible and allow quite wide movements in public debt to GDP ratios, a pattern that was to change dramatically in the interwar period. Nonetheless, alternative models of risk premia in the 1870-1914 period have given different results: Flandreau and Zumer (Reference Flandreau and Zumer2004) and Ferguson and Schularick (Reference Ferguson and Schularick2006, Reference Ferguson and Schularick2012) found evidence that debt/revenue ratios did matter for sovereign risk, and that in some specifications these debt burden measures could sometimes lessen the impact of gold standard adherence. Empirically, the question is likely to remain unresolved until we can better model the endogeneity of debt levels, which were surely correlated with the ability to make a credible gold standard commitment. But in theory the result is clear: unsustainable debt paths will usually break any commitment to gold, since they are almost surely indicative of fiscal weaknesses that will, at least in some states of nature, cause explosive budget dynamics. It is quite possible that the gold standard seal would not be apparent once such a control was added.

However, debt levels are not the only matter of concern for creditors. The volatility of the fiscal situation is also important. The unusual volatility of public debt in the region is shown in Table 4. More than the core or the periphery as a whole, the Latin American economies seem to have been more susceptible than any other group of countries to extreme fluctuations in public debt to GDP ratios. The region's governments experienced big run ups in debt levels during periods of easy credit, followed by big crashes during tighter times or after a default/repudiation episode. This led to higher means and standard deviations of public debt to GDP ratios as compared with other countries, especially core countries and even other periphery countries. In an institutionally sound and financially mature environment, such data might reflect the decisions of a strong optimising government taking maximal advantage of the flexibility provided by a debt-financed buffer. But in a financially and fiscally backward periphery environment these data tell a different story — a tale of fiscal «snakes and ladders» where governments were poorly equipped to cope with tax and spending shocks, forced to use debt (when not using the printing press), and eventually crashing into debt ceilings, followed by default, and austerity.

TABLE 4 VOLATILITY OF PUBLIC DEBT TO GDP RATIO, 1870-1913

Source: Obstfeld and Taylor (Reference Obstfeld and Taylor2004).

One might view these higher moments of the fiscal deficit as exogenous «fundamentals» linked to political and institutional weakness. In this view, Latin American countries were burdened with fiscal volatility either because their tax revenues were volatile (e.g., due to trade volatility and terms of trade shocks affecting customs revenue) or because spending was volatile (e.g., due to wars and military spending caused by internal/external political instability). Alternatively, shocks may have been no different but the governments’ propensity to use external borrowing may have simply been higher. The latter might reflect institutional weakness of a different sort, namely the sub-optimally short time horizons of sovereigns. Whatever the origin, it is clear that Latin America governments lived in a more fiscally volatile world and witnessed more dramatic fluctuations in their debt positions than other countries in the core or periphery.

3.7 Access to capital markets: the microeconomic explanation

A recent, novel, and compelling explanation addresses the issue of how sovereign debt emerges, subsides and then again re-emerges, as advanced by Flandreau and Flores (Reference Flandreau and Flores2009). The central problem they analyse is a scenario where there are severe information asymmetries, where countries renege on their contractual arrangements, and where financial intermediaries have incentives either to make debt sustainable against all macroeoconomic odds or to cheat investors. In their explanation it is the reputation or «brand» of the financial intermediary or underwriter which is the key mechanism that helps overcome the enormous information asymmetries and market imperfections. This approach addresses two related stylised facts to solve the sovereign debt puzzle: (1) the existence in general of overshooting in lending and (2) the heterogeneous performance of countries seen during each of the capital flow waves.

Their study differentiates between wildcat underwriters and top-rank well-reputed ones; evidence suggests that more reputable banks with a longer time horizon supported borrowers with the intention to help maintain their access to capital markets. In contrast, wildcat underwriters would engage in selling the bonds with a short horizon. They observe that many sovereign defaults followed and yet did not precede intermediaries’ failures. According to this market structure analysis the underwriting business was a hierarchy — not a competitive perfect market. And the champions of all were the Rothschilds who always signalled the best of the best investment-grade securities. A contrast is the Barings house that was involved with issuers of much more risky bonds, as became clear. When the Rothschilds acted as underwriters of Brazil, for example, they put at stake their own reputation but also this acted as a transfer of information about the fundamentals of the country via the underwriter. In this view, even if investors could not learn directly very much about borrowers and their fundamentals, they could learn more easily about underwriters, the more reputable of which could then give a bond issue a «seal of approval». The hierarchy of underwriters was thus a proxy for the hierarchy of users. On the other hand, the fact that prominent houses did act as agents, but not as underwriters, could have signaled «risky bets» and hence might have fed sizeable periods of overborrowing.

4.1 Vignettes: sovereign borrowing episodes in Latin American economic history

4.1.1 The characteristics of sovereign bonds during the 1820-1930 period

In the 19^th century, sovereign bonds typically had a very long maturity. They averaged more than 20 years while in the current globalisation of the 1990s and 2000s the issue of Eurobonds by emerging market sovereigns was at maximum maturities of 7-10 years. Some of the contemporary bonds may have had clauses of semi-automatic renewal, but we have still not observed anything like the 25-year bonds that were typically seen in the 1820s-1930s epoch. Also, in particular, the 1870-1913 period witnessed a situation in which early redemption clauses were the norm in the structuring of the public debt issues (the so-called «lottery clause», allowing partial repayment and conversion). This tells us that the international capital markets of the 19^th century (notably, the London Market) took a «friendly» approach towards debtor countries, allowing them to refinance and swap long-term debt instruments for comparable instruments at lower interest or coupon rates to exploit favourable liquidity conditions (Mauro et al. Reference Mauro, Sussman and Yafeh2006).

As Bordo and Meissner (Reference Bordo and Meissner2005) have shown, most of the sovereign bonds floated by Latin American countries in the period were denominated in foreign currency or in terms of gold (or else had «gold clauses»). The limitation for an emerging country to issue a bond only in a hard or key currency was referred by Eichengreen and Hausmann (Reference Eichengreen and Hausmann1999) as «original sin» and they viewed this as a symptom of extreme fragility that might explain the recurrence of interruptions in the original contractual debt obligations of the emerging market countries. Here, the linkages between monetary regimes, public finances, the level of economic activity, and the sovereign debt issued in foreign currency provides important insights that help explain recurrent problems in the management of the sovereign debt in Latin America countriesFootnote ⁹.

Owing to the acute credibility problems posed by the conduct of volatile monetary and fiscal policies, the Latin America countries also issued domestic debt with gold clauses. This is shown in Figure 6 taken from Bordo and Meissner (Reference Bordo and Meissner2005), we can see that two very big Latin American countries, Brazil and Argentina, consistently issued public debt by resorting to gold clauses or hard currency denominations.

FIGURE 6 ORIGINAL SIN

Source: Bordo and Meissner (Reference Bordo and Meissner2005).

Finally a notable difference with today's international markets was that in many debt issues export revenues and tax revenues were earmarked as collateral to guarantee the servicing of the debt. In other words, it seems that some public bonds had explicit «seniority» over the same type of bonds issued by the same national political entity. Clauses linking different bonds — to prevent selective default — were not a central feature as they are in the present day international bond market.

4.1.2 The cost of capital for a small independent nation: moral hazard and collusion among lenders and borrowers or the price to pay for the reputation of a newcomer?

It is well known — notwithstanding the frictions and productivity differences underlying the Lucas paradox — that when the yield of available investments in the core countries drops, the international capital market has a strong tendency to funnel liquidity to more risky prospects in the periphery. This is evident today, but was just as true in the 1820s, 1870s and 1880s. Hence, the possibility of «overborrowing» arises when extremely high liquidity in international capital markets becomes an equilibrium possibility. An instance of «irrational exuberance» in the case of some of the riskier speculations in the wave of the 1820s cannot be totally ruled out. It is implicit in the arguments of Flandreau and Flores (Reference Flandreau and Flores2009), and the following example illustrates.

A basic guiding principle can be set forth: potential lemons behave always like «seasoned» public bonds in good times — they just become junk bonds in bad times. The example of the 1824 £1 million so-called «Baring loan» to the Province of Buenos Aires is a very interesting case that illuminates the behaviour of the various actors involved in a sovereign debt placement. As Peters (Reference Peters1934) explains, the loan was placed with Baring Brothers at 70 per cent. Barings then deducted 13 per cent of the par amount (£130,000) so the loan yielded the Province of Buenos Aires a mere £570,000. Then the bonds were placed in the London market at 85 per cent of par, yielding to Baring Brothers a spread of 15 per cent, and a further £150,000 profit. With total profits of £280,000 on a loan of £570,000, the merchant house achieved a return of 49 per cent! This is one of the many instances noted by Fodor (Reference Fodor2002, p. 35): he argues that data usually used for the Latin American loans cannot be accepted at face value and should be treated with many caveats. As far as the loans are concerned, perhaps there was no boom at all. While it is true that as market makers and subscribers, the merchant house had their reputation at stake, the initial phases of outburst of financial liquidity in the London market were exploited to the full by the same houses as well as by the policymakers in the recipient countriesFootnote ¹⁰.

While the net proceeds to the recipient were very low in this first experiment of the Province of Buenos Aires, and indeed very profitable for the merchant houses, the gross yield to maturity of the bond approached 7 per cent, equivalent to a 4.5 per cent spread over the British consol. The «risk premium» or spread is traditionally viewed as the «price of access» to capital markets for emerging market borrowers and is in accord with Tomz (Reference Tomz2007) who argues that it is not the particularities of the international capital market that characterise the interactive behaviour of investors and borrowers but rather the formation of reputationFootnote ¹¹. And reputations form in Bayesian way: creditors update the debtor's creditworthiness based on performance. Tomz considers two periods: the first wave (1820s) and the second wave (1870s). Then he identifies the differential premium between seasoned borrowers and new entries; in the second wave he also distinguishes between the good, the bad and the worse: that is, the payers, the resettlers and the defaulters.

As we can see from Figure 7, in the first wave, as in the example of Argentina given above, the Latin American economies were new borrowers par excellence and spreads were around 350 basis points. In the second wave of the 1870s, the market attached «reasonable» prices to the seasoned players, the new borrowers and the proven lemons or junk bonds (the latter having an average yield of 2,700 basis points).

FIGURE 7 REPUTATION: FIRST AND SECOND WAVE SPREADS

Source: Tomz (Reference Tomz2001).

The fall in yields from the 1820s to the 1870s marked an important development in the region's access to global markets. The high cost of capital in the first wave might have been associated with the building-up of reputation for the early borrowers, but, in addition, genuine asymmetric information problems were surely quite acute during the 1820-1870 period which would also drive up the required yield on a sovereign bond.

4.1.3 Availability of information and the importance of news

Paucity of information was a major issue, especially until the second wave in the 1870s. In the 1820s there were in London several very important newspapers that compiled quite sophisticated data: on bond pricing and volumes traded, and also quotes on the political economy events of different countries. The Colonist, Common Sense, The Times and Course of Exchange followed closely the Latin American debt during the first phase on a daily basis until the generalised defaults of 1826-1827. From the information we could gather from the six years 1822-1828 were data on a good portion of the sovereign bonds outstanding, allowing us to construct a Latin American bond composite index that is quite comparable with the current EMBI index, as seen in Figure 8Footnote ¹².

FIGURE 8 LONDON LATIN AMERICAN BOND INDEX FOR THE FIRST WAVE

Source: See text.

During the second wave, however, the availability of news was much more fluid. Information on macro variables such as outstanding debt per nation, trade flows, fiscal positions, population, railway construction as a proxy for investment, prices and quotations of sovereign bonds were readily available from additional sources such as Investor's Monthly Manual, The Economist, the Palmer's Index and from the Annual Reports of the Corporation of Foreign Bondholders that was created in the mid 1860s as an association of British investors holding bonds issued by the emerging economies.

Whether the availability of this information, plus the Consular Reports being drafted by the different Creditor Nations, meant a better understanding of the political economy realities of the Latin American countries is, we think, debatable. In the case of Argentina, Peters (Reference Peters1934) noted that:

A federal union was again formed in 1861, without definitely solving many of the controversial issues between Buenos Aires and the other provinces….the lack of knowledge of and confidence in the Argentine Republic, as distinct from Buenos Aires, led to the failure of the attempt [to float a national bond in 1866]…The Buenos Aires bonds remained about ten points above a national issue, indicating that the investors feared a possible dissolution of the union, and the possible repudiation of the debt by the province. (Peters Reference Peters1934, p. 24)

He then shows clearly that the messy arrangements of fiscal federalism gave rise to asymmetric information for foreign investors, coordination problems at the national level, moral hazard and free riding. These problems have always been present and have made it very difficult to achieve a full understanding of the presence of different political subunits as major actors in the sphere of fiscal policy and debt management. These problems played a huge role in the Baring crash of 1891 and, again, in the Argentine crash of 2001 (della Paolera and Taylor Reference della Paolera and Taylor2003).