Lending to the Borrower from Hell: Debt, Taxes, and Default in the Age of Philip II (The Princeton Economic History of the Western World)
Page 19
To sum up, it is apparent that painful sanctions have been used to ensure the repayment of sovereign debt. This is consistent with the predictions from models in the style of Bulow and Rogoff (1989). What is less obvious is if the magnitudes and frequency of sanctions are sufficient to explain the existence of sovereign debt in the first place—and if reputation is not a better explanation for cross-border lending on average.
REPUTATION
The reputation camp argues that the threat to cut off future lending is sufficient to make sovereign borrowing feasible (Eaton and Gersovitz 1981; Eaton and Fernandez 1995). If a borrower defaults, there may be no way to punish them, but withholding future funds can be sufficient to force compliance. The need to smooth consumption is a sufficiently strong motivating force. If a borrower defaults now, their access to funds in the future will suffer. This is going to be painful because access to credit matters the most in hard times.
Jonathan Eaton and Mark Gersovitz (1981) offer one of the earliest and most influential contributions to this literature. They assume that a defaulter can never borrow again.12 The threat of exclusion from credit markets consequently can sustain lending. As the authors acknowledge, permanent exclusion is an unrealistic assumption. In the language of game theory, such a strategy is not “renegotiation proof”; both borrower and lender will probably be better off if the defaulter is permitted access to loans eventually. Defaults, in reality, have always been followed by renegotiations. While these can take a long time, there are no examples of permanent exclusion from lending markets (Benjamin and Wright 2009; Reinhart and Rogoff 2009).
With renegotiation, inefficient penalties can be avoided, but the amount of debt that can be sustained will be lower (Bulow and Rogoff 1989). The need to impose unrealistic, permanent exclusion from credit markets is also avoided if defaults are “excusable.” Herschel Grossman and John Van Huyck (1988) present a model where lenders try to determine the nature of borrowers. These borrowers come in two types: good and bad. Both can default. The former simply want to smooth consumption; if shocks are extreme, they may be better off defaulting. The latter will be inclined to default in all states of the world. Permanent exclusion makes no business sense for the “good” borrowers; they were simply unlucky. The “bad” types, on the other hand, will rationally be excluded permanently. They may have received funds in the first place, owing to uncertainty and asymmetric information, but they cannot tap the market again.
Excusable defaults are, in effect, a form of insurance against adverse shocks; lenders know in advance that there is a chance that times will turn out to be sufficiently bad for the borrower to default. In that case, they will price the loan accordingly, in the same way as an insurance company will charge a premium in exchange for having to pay out substantial amounts of money when, say, a fire devastates a building. The excusable default argument has the advantage that it can reconcile the existence of sovereign lending with the history of frequent (and repeated) defaults since 1500 (Reinhart, Rogoff, and Savastano 2003). Excusable defaults are a de facto way of making sovereign debt state dependent, thereby allowing governments to reduce the burden of debt in bad states of the world.
The existence of multiple lenders can modify incentives. A lending relationship may unravel if new lenders enter the market; without sanctions, the new lender may be able to “steal” the customer, offering mutually beneficial arrangements after a default. The old lender is not paid, and the new one splits the gains from intertemporal trade with the borrower. In a classic paper, Kenneth Kletzer and Brian Wright (2000) analyze the incentive structure in an anarchic environment when neither borrower no lender can credibly commit. There, the existence of an additional lender can be shown to produce no breakdown of the lending relationship. The reason is that there is a strong incentive for existing lenders to “cheat the cheater”—that is, offer even better terms to the borrower if a new lender enters and breaks a moratorium.
Sovereign debt theory has typically treated domestic and foreign borrowers as clearly distinct. This is not necessarily accurate; bonds are tradable, and can be passed from, for instance, a bondholder abroad to one in the same jurisdiction as the borrower. Since there are some reasons for governments to treat investors in the same country more generously, the imperfect nature of market segmentation may actually help to rationalize the existence of cross-border debt markets. If in a crisis foreigners can always sell to domestic investors, then the absence of third-party enforcement is much less important (Broner, Martin, and Ventura 2010).
In the sovereign debt literature, the commitment problem is typically on the borrower’s side. One alternative to servicing debt in the Bulow and Rogoff approach is to save and draw down balances in bad times. This requires access to a safe “storage technology” for savings. If bankers can also default, however, this will affect the sustainability of debt; smoothing can now only be achieved by maintaining access to credit. In this case, the reputation of the borrower matters more and may be sufficient to allow cross-border lending (Cole and Kehoe 1995). Intuitively, in a world with “Swiss bankers,” Bulow and Rogoff’s no-lending result goes through; in a world without them, only those borrowers who establish profitable relationships with their (untrustworthy) bankers can borrow.
In Kletzer and Wright’s (2000) approach, banks cannot commit to offering deposit contracts in the style of Bulow and Rogoff; hence, reputational concerns and cheat-the-cheater mechanisms can ensure that lending exists in equilibrium. Going even further, Wright (2002) shows that if banks have some degree of market power, lending can be sustained even if banks can offer any type of insurance contract. In a modification of Eaton and Gersovitz’s model, Wright demonstrates that as long as there are important benefits from syndicated lending (that is, a single bank cannot satisfy all the needs of a borrower), cooperation between banks is necessary. The more they “act as one,” the easier it is to extend credit, even if alternative ways of insuring income risks exist. Reputation, in this context, matters, yet it is the reputation of intermediaries for cooperating in punishing recalcitrant borrowers and not the reputation of borrowers that is critical.
Harold Cole and Patrick Kehoe (1998) analyze an alternative way in which reputation alone can sustain large amounts of debts. They examine a model in which reputation in one area—such as debt repayment—spills over into reputation in other interactions. The more an opportunistic default damages the ability to conduct essential business in other areas, the more lending will be feasible in equilibrium.
Both the sanctions and reputations approaches—in their simplest form—imply that defaults should never be observed in equilibrium. There are two ways to rationalize their existence. One approach uses informational frictions and contends that payment suspensions are basically “misunderstandings”—mainly because the borrower did not understand the consequences of not paying (Atkeson 1991). The second alternative is to consider defaults as a way to make lending markets “complete.” Most loan covenants do not contemplate alternative states of the world, forcing the borrower to repay regardless of their financial health. Negative shocks to a borrower’s servicing capacity can therefore make the cost of repaying debt extremely high. Under such conditions, it is often better for the borrower not to pay. Lenders realize this at the time they make the loan (Grossman and Van Huyck 1988). As long as ex ante expectations are not violated, there is no reason for the credit intermediation process to unravel. Signs of an excusable default are that they occur in verifiably bad states of the world.
So far we have only discussed those cases where borrowers unilaterally decide to default or not. Cole and Kehoe (1996) focus on a case where lenders must decide to roll over the maturing debt of a sovereign. If they decide to lend, the borrower will honor their debts, and if they do not extend fresh credit to replace the ones falling due, it can be in the sovereign’s interest to default. In particular, if the maturity of debt is low, countries will find it optimal to stop payments if there is a “run” on their de
bt. This gives rise to multiple equilibriums. If all lenders are happy to extend credit, a country will not default; if some are worried that they will not be repaid and refuse to roll over debt, a generalized run can break the incentive-compatibility of the initial contract.
IMPLICATIONS FOR EMPIRICAL RESEARCH
Both the reputation and sanctions views can marshal a measure of empirical support, but many doubts remain. If the search for a single theoretical model of sovereign debt and default that fits all facts in all time periods is in vain, we should study what made sovereign lending feasible during key episodes. Confronting the predictions of theoretical models of sovereign debt with data is, by its very nature, difficult; many of the predictions concern incentives for players, not their observable actions.
Sanctions models at least offer easily testable predictions once defaults occur. Reputation models, in contrast, are even harder to examine in light of empirical evidence. Recent models of contracting under anarchy, where enforcement and punishment reflect the incentive structure among lenders, are equally difficult to test. The case of Habsburg Spain is especially important because the quality of information allows us to examine the implications of reputation models in detail. Uniquely, we are able to observe incentives directly for a borrower whose case has historical significance.
Imperial Spain has already served as a testing ground for different sovereign debt models. The school of thought that sees banker folly behind defaults claims Philip II’s case as key evidence. Much work on Philip II’s borrowing has emphasized the Crown’s hopeless financial position.13 Braudel famously argued that the king’s payment stops resulted in major losses for his lenders. To this day, journalists use Castile’s bankruptcies to illustrate banker irrationality (The Economist, September 23–29, 2006). Also, Reinhart and Rogoff (2009) argue that lending to serial defaulters may not be fully rational, and cite the repeated defaults of Spain as a case in point.14
In contrast, Conklin (1998) concluded that effective sanctions sustained lending to Philip. In his view, Imperial Spain’s debts in the second half of the sixteenth century provide important support for the sanctions view of international borrowing. When Philip II stopped payments in 1575, his bankers halted all transfers. This stopped wage payments to the Army of Flanders, which mutinied. A sharp setback for Spain ensued and forced the king to settle.
Table 8. Key predictions of debt models
The purpose of this chapter is to confront theoretical predictions with the Spanish evidence. There are a number of clear implications about what we should expect to see; table 8 gives an overview of the approaches guiding our analysis.
In our view, the data do not support earlier interpretations of Philip’s bankruptcies. Sanctions were not effective, and bankers were not irrational in lending to the king. Before we explain why these approaches fall short, we review the specific aspects of the data that we use in building our argument.
DATA
To better understand why Philip II received and retained access to credit, we turn to the new series of asientos we introduced in chapter 3. The series starts in 1566, which means that we have no data on the first two defaults (1557 and 1560).15 We use the complete set of existing contracts until 1600, two years after Philip’s death. While earlier authors used information on lending volume (summarized on the first page of each document), the actual loans contain a wealth of additional information that has never been exploited: the identity of lenders, services performed, and other contractual arrangements.16 We utilize this detailed microdata later in the chapter.
Financial transactions between the bankers and king involved transfers, loans, or exchange operations—usually in combination. Transfers allowed funds to be disbursed in distant locations. Exchange operations normally specified the currencies involved, the exchange rate to be used, and permits for exporting specie. Other details include the place of delivery and repayment, the tax stream for repayment, and transfer and exchange fees. Occasionally, the king posted collateral in exchange for a loan. Other benefits could include lifetime pensions or noble titles. The time of repayment frequently depended on the king’s fiscal position (for example, as a function of the silver fleet’s annual arrival).17
Regular borrowing in Philip’s reign started after the resolution of his second bankruptcy. After 1566, the king concluded an average of 12.5 asientos per year—sometimes none, and in other years as many as 38. Their duration varied between a few months and several years (with a maximum of 134 months). The largest contract was for 2.1 million ducats (equivalent to 30 percent of the annual fiscal revenue at the time).18 The smallest contract was for a mere 1,663 ducats. Table 9 summarizes the key features of our data.
Foreign exchange transactions appear in 42 percent of all contracts. The nominal interest rate stated in the loan documents averages 9.9 percent. It could be as low as 0 percent (in special cases where the loan funded the construction of ecclesiastical buildings) or as high as 16 percent. The actual yield of contracts could sometimes be much higher, depending on the exchange rates used, the valuation of payment instruments, and ad hoc clauses. In one-third of all cases, the king posted collateral (typically juros).
Philip II borrowed from several banking families throughout his reign. No fewer than nine members of the Lomelín family lent to him. The Spinola family contributed twelve lenders, the Gentil ten, the Centurión six, and the Fugger five.19 Often, several members of the same banking family lent in a single contract. On March 13, 1572, for instance, we find Gerónimo and Esteban Grillo providing a loan of 100,000 ecus.20 The brothers Augustín, Tadeo, and Pablo Gentil join forces in several contracts between 1567 and 1569.21 Lending in small syndicates composed of members of different families was also common. Of 438 total transactions, 141 had multiple lenders. They account for 30 percent of all money lent.
Table 9. Descriptive statistics
Note: Principal is given in constant 1565 ducats; FX is a dummy variable for the presence of a foreign exchange transaction; duration is given in months, r is the nominal rate stated in the contract; and collateral is a dummy variable for the presence of collateral.
a The minimum value for principal is calculated excluding nine contracts that merely restructured old loans; because they did not result in fresh cash for the king, they are deemed to have a principal of zero.
b The maximum loan corresponds to a portion of the general settlement of 1577, which was apportioned between four banking syndicates. The largest contract excluding the settlement was for 2.08 million ducats.
Lending was heavily concentrated. While 130 individuals from 63 families lent to Philip II, a few account for the bulk of the funds. The top 10 banking families were responsible for more than 70 percent of all money lent. The Spinola, Grimaldo, and Fugger families extended 40 percent of all loans. In contrast, the bottom 48 lenders combined provided less credit than the Spinola family alone. Figure 10 plots the cumulative percentage of the total amount lent to the Crown against the rank of the banking family. A Gini coefficient of 0.73 indicates a highly unequal distribution.
FIGURE 10. Cumulative lending to Philip II by rank of lending family, 1566–1600
Lending relationships proved to be enduring. The Fugger started lending to Charles V early in the century and continued to do so until the death of Philip II in 1596. Jakob Fugger lent in 1519. His nephew, Anton Fugger, did the same in the 1550s, and in 1590 we find Jakob’s great-grandson, Marcos Fugger, also supplying credit to the king of Castile.22 The Grimaldo family lent 27 times between 1566 and 1589. The record holders in terms of frequency were the Spinola, whose members participated in a total of 98 loan contracts over the period 1566–99.
Table 10 summarizes the place of delivery of funds by bankers before and after the 1575 default. Fully 62 percent of the amount borrowed was delivered outside Castile. Flanders was the most important foreign destination for funds. Italy was a distant second, as the Mediterranean fleets were partly funded by local revenue (Parker 1998, 135). Repaym
ent typically took place in Castile. The Spanish Empire, for all its size and might, was mainly financed by the Castilian economy—the strongest in Europe at the time (Alvarez Nogal and Prados de la Escosura 2007).
Table 10. Place of delivery of asientos
WRONG EXPLANATIONS
Before we put forward our interpretation, we will discuss alternative solutions to the problem of “Why lend to a monarch?” Two are particularly prominent in the literature: banker irrationality and the idea that penalties were crucial for sustaining lending.
IRRATIONALITY AND BANKER TURNOVER
Braudel (1966) argued that Philip II managed to borrow massively, stop payments often, and pay back little because he succeeded in fooling one group of bankers after another.23 Sequential default and financial ruin of this kind require that each group of financiers thought they would be treated better than the last ones—a form of banker irrationality. Journalistic references to Philip’s defaults frequently make this point, arguing that “Genoese lenders’ indulgence of Philip II of Spain’s expensive taste for warfare caused not only the first sovereign bankruptcy in 1557, but the second, third and fourth as well” (The Economist, September 23–29, 2006).
Did successive waves of lemming-like lenders—first from Germany, then from Italy, and finally from Portugal and Spain—enter the borrowing game? We examine the nationality of bankers and turnover ratios in our database, taking the default of 1575 as a potential breakpoint. The 438 loan transactions demonstrate that the composition of financiers remained stable throughout. After 1575, the share of Spaniards declined from 28.8 to 25.6 percent. The German bankers, who were allegedly burned by the first bankruptcy, acted as a continuous source of funding. Their share more than doubled—from 4.3 to 10.9 percent—after the third bankruptcy. The Genoese provided 67 percent of the loans before the 1575 bankruptcy and 64 percent after it. Thus there is little to suggest that the king’s access to credit depended on the repeated fooling of bankers from different countries.