34 The Fugger balance sheet for 1563 shows lending of 4.44 million florins to the king of Spain, with 1 million either doubtful or written off, for a (maximum) loss rate of 22.56 percent. We do not consider this figure comparable to our later “haircut” calculations for Philip II’s other defaults, first because it only concerns one family, and second because we do not know which share of doubtful debts was eventually recovered (Ehrenberg 1896).
35 Marie-Thérèse Boyer-Xambeu, Ghislain Deleplace, and Lucien Gillard (1994) argue that asientos were developed in response to the closing of the Laredo-Antwerp sea route in 1568–69. Our data speak against this interpretation. The Genoese system was already fully in place by 1566, and contracts signed between 1566 and 1568 already show the full range of innovations associated with Genoese finance.
36 The bankers were Constantín Gentil, Lucián Centurión, Nicolao de Grimaldo, and the Spinola family (De Carlos Morales 2008).
37 These should not be confused with the letters from various parts of Europe compiled by the Fugger.
38 Ironically, they also had to borrow from Juan Curiel de la Torre, one of the principal lenders to Philip II—at high interest rates (Ehrenberg 1896). The cost of credit for this transaction was 22 percent per year—much higher than the rates at which the Fugger had lent to various princes.
39 Chapter 5 discusses the significance of the Genoese position in the juros market for debt sustainability.
40 Among the vast literature on the Casa di San Giorgio and its central place in the history of public finance, see Greif 1994; Epstein 2001; Felloni 2006a, 2006b. For a brief overview and literature survey, see Marsilio 2013.
41 For more on the juros of the Casa de la Contratación, see Ruiz Martín 1965.
42 AGS, Contadurías Generales, Legajos 86–93.
43 We first introduced these data in Drelichman and Voth 2010, where we reported lending amounts. We added banker identities in Drelichman and Voth 2011a, full cash flows in Drelichman and Voth 2011b, and information on contingent clauses in Drelichman and Voth 2012.
44 For an example of the consolidation of asientos, see Lapeyre 1953.
45 Carande only reports the total amount disbursed by the bankers and total amount repaid. There is no information on the timing of the payments, thereby making it impossible to establish a rate of return.
46 For an account of the 1555–59 period, see Rodríguez Salgado 1988.
47 The deflator used throughout the book is the Old Castile price series from Drelichman 2005. Using alternative deflators results in immaterial changes.
48 The king did retain control over expenditure and foreign policy; while this would have important consequences for warfare and government finance, it is hard to argue that those features alone make a state absolutist.
CHAPTER 4
THE SUSTAINABLE DEBTS OF PHILIP II
For a long time, Philip II’s defaults have been blamed on a disastrous combination of the Crown’s unsustainable fiscal situation, on the one hand, and myopic lenders, on the other. Braudel famously argued that each bankruptcy ruined different bankers, who were quickly replaced by another, equally irrational wave of entrants. He concluded that “every time the state declared itself bankrupt, bringing contracts to a violent end, there were always some actors who lost, fell through a trap-door, or tiptoed away towards the wings” (Braudel 1966, 362).
Assessing lender rationality is a complex problem, which we address in the following chapters. We focus here on fiscal sustainability, the first test that any borrower must pass for a lending transaction to be sound. Did the Castilian Crown have enough resources to honor its debts in the long run? Or to put it simply: Could Philip pay?
Evaluating the health of a state’s finances requires a reliable set of long-term fiscal accounts. These include measures of revenue, expenditure, debt service, and variations in the stock of debt. Assembling them for sixteenth-century Castile is a daunting task. Like virtually every fledgling national state, the Castilian administration did not keep centralized financial records. Revenues were often spent at their collection point or transferred to other parts of the empire without passing through the central treasury. No unified tally of expenses was kept, nor were debt payments generally added up and accounted for. The treasury functioned more as a residual claimant than an efficient administrator of tax revenues. The information needed to present a comprehensive view of state finances must therefore be pieced together from a multitude of different sources, all of which vary in quality.
In chapter 3, we introduced our new series of short-term debt as well as some estimates of long-term debt for benchmark years. We also pieced together a new series of revenues, constructed on the basis of individual income streams. Here we add calculations of military expenditure and short-term debt service. In combination with carefully chosen assumptions, we can then estimate the missing data and correct for the effect of the bankruptcies. The result is a full set of annual fiscal accounts for Castile between 1566 and 1596, the earliest such reconstruction for any sovereign state. While the exercise is conjectural rather than precise, it nonetheless will allow us to explore the financial health of the Crown in more depth than has previously been possible.
Sustainability means that a country is able to service its debts indefinitely, given current fiscal parameters. Solvency is a less stringent criterion; it simply implies that there are plausible future paths of revenue and expenditure that will lead to sustainable outcomes. To assess if debts are sustainable or a country is solvent thus requires a number of assumptions about the evolution of fiscal variables. These in turn depend on economic growth, policy measures, and unforeseen events. Not surprisingly, there is little consensus about how to exactly determine whether a sovereign is likely to honor their debts. We adopt the most common approach—that used by the IMF. At every step, we use the most conservative assumptions possible, and further complement the analysis with a number of robustness checks and counterfactual thought experiments.
MORE DATA
MILITARY EXPENDITURE
Civil expenditures in early modern states were small. The payroll of the central administration was tiny. The administration of justice was largely funded at the local level, frequently through user fees.1 Poor relief was the domain of religious and charitable organizations. To be sure, the pomp and circumstance of monarchy—the palaces and lavish feasts, royal visits, and upkeep of the royal household in general—cost money. Yet they absorbed only a small share of government revenues. Rising national states spent the overwhelming majority of their resources on a single endeavor: war.
Philip II was at war in every single year of his reign, as described in chapter 2. Because of their geopolitical significance, each of Castile’s campaigns has been thoroughly studied by military historians.2 We use their work to reconstruct the impact of warfare on the finances of Castile.
Military operations were funded at the local level as far as possible. Flemish and Italian revenue covered part of the large military costs in those territories. Shortfalls were met with transfers from Castile, usually through a financial intermediary. These transfers could grow quickly whenever things went awry, as they did repeatedly during the Dutch Revolt. The total cost of war is equivalent to the amounts spent from both sources. Since Castilian debt was paid only from Castilian revenues, however, only the transfers from Castile matter when evaluating fiscal sustainability. Given this, our first step is to classify military expenditures according to the source of the funds. This is fortunate, as the Castilian accounts are remarkably complete, while data on costs defrayed from local sources are considerably spottier.
To construct an annual series of military expenditures paid by the Castilian treasury, only one strong assumption is needed. The contributions to the Army of Flanders between 1580 and 1596, reported by Parker (1998), are only available as quinquennial totals. In this case, though, we have good yearly data for the contributions paid by the Flemish treasury. To apportion the quinquennial contributions
from the Castilian treasury to individual years, we assume that they followed the same annual pattern as Flemish contributions; the latter are therefore expected to capture short-term fluctuations in total expenditure on the Dutch War. To check for robustness, we dropped the Flemish expenditures as a source of variation and used the alternative assumption that the yearly contributions of the Castilian treasury were one-fifth of the quinquennial totals. This did not alter the sustainability results in any significant way.
Figure 7 presents our estimates of military expenditure between 1565 and 1596. Where the cost of an individual campaign differs across sources, we chose the estimate supported by better documentation. Of the 146 million ducats spent on war during Philip’s reign, fully 77 million (some 53 percent of the total) were spent on the Army of Flanders.3 Because the Dutch Revolt was such a large item in Castile’s military budget, we differentiate it from the other campaigns.
In the early 1570s, the War of the Holy League and growing intensity of the Dutch Revolt led to a spike in military outlays, which peaked in 1574. The 1575 bankruptcy followed immediately after. The following decade saw relatively limited military expenditure. This changed with the resumption of hostilities in the Netherlands in 1583. Expenditure continued to rise in the run-up to and aftermath of the Armada. Outfitting it cost approximately 10 million ducats, roughly two years of total revenue. Following the disaster, a similar sum was spent on rebuilding the fleet to defend Spain against French and British attack. The 1596 peak—the last year for which our sources allow a comprehensive assessment—reflects the continuing expenses in Flanders and the response to the English naval threat.
FIGURE 7. Military expenditure paid from Castilian sources
SHORT-TERM DEBT SERVICE
Another important component of national accounts is the cost of servicing debts. We presented what little information is available on juros in chapter 3; we will derive estimates of the yearly interest paid on them in the next section. Our data on asientos, on the other hand, is detailed enough to allow us to calculate servicing costs with reasonable precision.
Asientos were convenient as a short-term borrowing device; they allowed the Crown to obtain money quickly and transfer it to virtually any point in its European dominions. They were also expensive. Their median gross rate of return was 13.8 percent, and many contracts cost more than 20 percent. This included compensation for currency conversions, overseas deliveries, transportation costs, and the risk of late payment and subsequent renegotiation. Many asientos used convoluted contractual forms. A good part of the return resulted from exchange transactions at favorable rates, advance payments by the Crown without interest, and swaps of financial instruments. Further complicating matters, scheduled payments seldom specified whether they were for interest or repayment of capital. Debt service, then, is not observable directly. We use an indirect estimation methodology instead.
First, we transcribe every clause in each asiento contracted between 1566 and 1596, and use them to reconstruct the agreed-on monthly cash flows. From the overall set of cash flows, it is possible to calculate the rate of return of each contract (chapter 6 discusses the methodology in detail). We then estimate the total interest for each asiento by multiplying its loan component by the rate of return. Because we do not know which payments are for interest and which ones are for capital repayment, we assume that each payment from the king to the bankers is composed of interest and capital amortization in constant proportions throughout the life of the loan. This is in line with what the few asientos separating interest and principal repayment specify.4 Annual short-term debt service of the Crown is then the sum of the interest payments for all asiento contracts in force in any single year. The default of September 1575 stopped payments on all asientos. The settlement was only reached in late 1577, and there were no new loan repayments until 1578. Therefore, we assign a value of zero to 1576 and 1577 asiento service. The default of 1596, in contrast, lasted less than a year, and so there are positive amounts of debt service in both 1596 and 1597.
THE BANKRUPTCIES
Each of Philip II’s defaults resulted in a negotiated reduction of outstanding debt. As such, they affected the fiscal position directly, and thus need to be taken into account when calculating the balance of revenues and expenditures. We have little data on the 1557 and 1560 episodes—dates for which we also do not know revenues or expenditures. Yet we can assess the scale of the 1575 and 1596 suspensions as well as the conditions of their settlements in detail.
A bankruptcy always started with a decree of suspension, the decreto. After some legal skirmishes and the gathering of relevant information, both the Crown and bankers created negotiating groups. In the case of the bankers, the goal was to have the delegates speak for as large a share of outstanding debt as possible, and hence confront the king with substantial market power. When an agreement was reached, a settlement—called medio general—was subscribed. This document specified the amounts to be repaid, payment instruments to be used, timing of repayments, and conditions of any fresh loans supplied by the bankers.
The first default covered by our data started with a decree of suspension issued on September 1, 1575. A settlement was reached with the medio general in late 1577.5 The king recognized outstanding obligations for 15.1 million ducats, divided into 14.6 million ducats of outstanding principal as of September 1, 1575, and 584,000 ducats in interest accrued between September 1 and December 1, 1575. It is not clear why this interest was added; in any event, the first provision of the settlement was to write it off. We work from the outstanding capital at the time of the suspension: 14.6 million ducats.
Of the total outstanding asientos, 5.6 million ducats were collateralized by perpetual juros with a yield of 7.14 percent, guaranteed by ordinary revenues. The holders of these juros were allowed to keep them, but their annuity rate was reduced to 5 percent. Compared to the 7.14 percent that had been contracted, the reduction in the annuity rate amounts to a write-off of 1.6 million ducats. A further 4.4 million ducats worth of asientos were collateralized by perpetual juros with a yield of 5 percent guaranteed by the revenues of the Casa de la Contratación. Because too many juros had been issued against these revenues, they were often not serviced; in the secondary market they traded at around 50 percent of their face value.6 The Crown recognized 55 percent of juros de contratación at face value by converting them to 5 percent perpetuities. The remaining 45 percent, 1.96 million ducats’ worth, were treated as uncollateralized debt.
Uncollateralized debt, which amounted to 6.6 million ducats including the juros de contratación, suffered the harshest treatment. Two-thirds of it was converted into perpetuities of the same face value with a yield of 3.3 percent. The remaining third was converted into tax farms on small towns (vasallos) with a nominal yield of 2.3 percent. The write-off on this portion of the debt relative to a 7.14 percent interest rate amounts to 3.8 million ducats. In total, the 1575 medio general rescheduled a total of 14.6 million ducats of short-term debt. It imposed a 5.5-million-ducat haircut in present value terms, or 37.7 percent of the loans in default.
The 1596 bankruptcy, which we described in chapter 1 following Ulloa (1977, 823), Enrica Neri (1989, 109), and Sanz Ayán (2004), was mild in comparison. The 1597 settlement rescheduled a total 7.05 million ducats. Two-thirds, or 4.7 million ducats’ worth, were converted into 5 percent perpetual juros. Using the same interest rate assumption as for the 1575 settlement, this would imply a haircut of 1.41 million ducats. The remaining third was guaranteed by 12.5 percent lifetime juros in possession of the bankers; these lifetime bonds had been issued in 1580, and hence were halfway through their accounting life expectancy of thirty-three years. The settlement stipulated that they were to be swapped for 7.14 percent perpetual juros; the bankers would be given enough perpetual juros so as not to alter the present value of the principal. In short, this portion of the outstanding debt suffered no write off; the king lengthened the repayment schedule in exchange for a higher face value of t
he bonds. The total write-off of the 1597 settlement amounted to 1.41 million ducats, or exactly 20 percent of the amount defaulted on.
Christophe Chamley and Carlos Alvarez Nogal (2012) offer a different interpretation of Philip II’s bankruptcies. They point out that juros were redeemable at the request of the king. So if prevailing interest rates fell, it was in the king’s interest to buy back all juros and issue new ones at lower rates. In their view, the defaults were just a way to accomplish this; the 1575 and 1596 settlements accomplished nothing more than to swap asientos collateralized by high-interest juros for new, lower-interest perpetuities. There was no actual haircut, as the king was merely exercising an option to which he had a right all along. The historical record strongly contradicts this explanation; the bankruptcies were not consensual events and involved acrimonious negotiations. Equally important is the fact that most redeemable juros—those that were not attached to asientos—were not converted to lower-interest ones.7 We nonetheless note that should Chamley and Alvarez Nogal’s interpretation be correct, this would actually reduce Castile’s debt burden, and the results we detail below would be strengthened.
DOWNWARD BIAS
Before we proceed, it is worth noting that all the assumptions we made in constructing the fiscal database are as cautious as possible. Whenever it was necessary to make a choice regarding the collection methodology or fill in missing data, we chose the path that resulted in lower revenues or higher costs. This introduces a bias against our argument, reducing the chance that we will find sustainability. In particular, we worked with confirmed revenues (as opposed to agreed-on ones). When facing a missing observation, we imputed the lowest of the closest values available. Conversely, we use the agreed-on asiento servicing costs, even though we know that many contracts were subject to individual renegotiations that lowered their yield. In the next section we will use a similar approach to estimate the debt service on juros, applying the average yield to every bond even though some of them may not have been paid in full because they were attached to an underperforming revenue stream. We cannot be sure that our military expenditure series captures every last ducat spent on military operations, but the logic of fiscal accounting implies that whatever we missed in this category will be added to civil expenditure. Overall, our data reflect a conservative take on Castile’s financial position. The calculations will therefore yield a lower bound on the sustainability of sovereign debt.
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 15