We now take a closer look at the effect of contingent clauses on loan maturities. In table 24, we regress the change in the loan’s due date (in months) on its original maturity, plus a host of controls.
Longer durations are not strongly associated with maturity extensions in case a contingency clause is invoked; the coefficient is significant but small. When contingency clauses are invoked, contracts with a foreign exchange component typically offered extensions of over a month and a half. Fleet contingencies extend the due date by two months on average—roughly the time it took to tap an alternative repayment source. Tax stream shortfalls are associated with even longer extensions—four months on average. Collateral executions add only one month. King and banker discretion clauses mainly reshuffled payments before the maturity of the loan, and as such, were not significantly correlated with changes in the due date.
When was each particular type of contingency clause used? In figure 26, we examine the evolution of contingency clauses by five-year intervals. In the beginning, for the years before the 1575 bankruptcy, over half the clauses were associated with collateral executions. At the same time, almost 25 percent referred to shocks that reduce the Crown’s revenue, such as fleet arrival and revenue shortfalls. Collateral execution clauses vanished as a category after the 1575 default. There was little borrowing—and hence few contingencies—up to 1585, when the preparations for the invasion of England resulted in a new round of asientos. In the last fifteen years of the sample, banker discretion clauses became the top category, followed by revenue shortfalls.
Table 24. Loan duration modifications in case of contingency
Maturity increase
Constant
−1.97
(−2.31)**
Fleet
2.14
(2.19)**
Tax stream insufficiency
4.07
(6.19)***
King’s discretion
0.44
(0.28)
Banker’s discretion
−1.03
(−0.97)
Collateral execution
1.19
(2.02)**
Principal (real)
−1.13
(−0.96)
Duration
0.07
(1.72)*
Foreign exchange
1.58
(1.88)*
R2
0.09
N
531
Note: The standard errors are clustered at the contract level. The t-statistics are in parentheses.
Significance: *0.1, **0.05, ***0.001
The use of contingent loans was not affected by the 1575 default; their share was roughly constant before and after.18 Their composition, however, changed significantly. Before 1575, most contingent contracts had collateral clauses. During the 1575 bankruptcy, the Crown declared that asientos had been illegal and thus payments on the collateral bonds were suspended. Bankers were prevented from selling them in the secondary market. These juros were reinstated with the 1577 settlement. Nonetheless, few bankers were interested in collateral execution clauses afterward. They instead used banker’s discretion clauses, which allowed for juros to be obtained and sold off at will in advance of the contract’s maturity. Banker’s discretion clauses were also priced to offset the cost of collecting and transferring the juros. In essence, bankers learned how to better protect themselves. If before they had to wait for the king to miss a payment to switch to safe assets, now they could do it at the slightest hint of trouble.
FIGURE 26. Number of contingent scenarios by type and period
Why was it in the interest of bankers and the king to write these contracts? We emphasize three key facts. First, there was an absence of asymmetric information as a result of primitive means of communication along with the extraordinary opportunities for intertemporal barter at a time of urgent, sudden spending needs and variable revenue. Second, the pricing of loans and effect of duration on interest cost suggests that the Crown principally struggled with liquidity issues, not solvency problems. Third, the fragmented nature of fiscal authority ensured that the Crown’s borrowing could be effectively collateralized by other debts.
Both silver revenue and tax farming facilitated the writing of contingent contracts. Silver taxes were a major source of revenue for servicing asientos. Since news from the New World about the production of the mines traveled at the same speed as the galleons laden with bullion, the king had no informational advantage vis-à-vis his bankers. There was also no way to hide the arrival of a fleet from the Indies.
The key reasons that favored conditional contracts written on fleet revenues—new information about the king’s fiscal position was independently verifiable and neither party had advanced knowledge—also encouraged tax farming. Tax farmers had strong incentives to maximize revenue and were not in the king’s employ. Relevant news might have been obtained from them ahead of time, but it is highly unlikely that there was much scope for manipulation. Overall, the existence of tax farming (combined with typically low volatility in nonsilver revenue) would have made it much more likely that shortfalls reflected genuine adverse shocks to the king’s fiscal position.
CONTINGENT DEBT AND EXCUSABLE DEFAULTS
Having analyzed the conditional nature of many loans to Philip II, we now shed light on sovereign lending and the nature of defaults. King and bankers contracted over a large number of different states of the world; they found an effective way to share risks. Nevertheless, some eventualities could not be written into loan covenants.19 Default under these conditions was a form of risk sharing, but it did not differ fundamentally from the contingencies foreseen in loan documents; it simply extended the sharing arrangements already in place to a different situation. We argue that the defaults of Philip II were excusable (in the sense of Grossman and Van Huyck 1988).20 While violations of the letter of lending contracts occurred, these were anticipated by both sides and did not violate lenders’ original expectations. To support this argument, we need to demonstrate that defaults took place in times of exogenous, independently verifiable adverse shocks and that there were no significant negative changes in loan conditions after an actual default.21
Modern-day sovereign bonds issued in New York or London are said to be in default when the borrower misses a single agreed-on payment. No such definition existed in sixteenth-century debt markets. At the time, neither banker nor Crown could firmly commit to servicing debts or taking deposits. Actual outcomes could fall somewhere on a spectrum between full compliance and default. It helps to think in terms of five possibilities: full compliance with the baseline scenario, as detailed in the original contract; use of one or more of the contingency clauses; violation of one or more of the clauses, followed by a rescheduling; full suspension of payments to all creditors, followed by a general settlement; and outright repudiation.
We argue that because contracts already considered a wide range of states of the world, both parties were aware that repayment according to the original agreement was not always possible. At the same time, the fact that contingencies were written into loan documents suggests that the Crown attached considerable importance to not violating explicit loan conditions. If defaults are excusable, financial outcomes for the lenders should reflect the borrower’s fiscal position. Crucially, differences in outcomes should be driven by exogenous shocks—that is, events that are beyond the borrower’s control. The king should live up to the letter of his obligations during normal times. When some minor shocks occur, he will invoke some of the emergency clauses in the contracts, which we document extensively. Larger shocks will see him violate some of these clauses, only to compensate lenders later. Full-blown moratoriums reflect even larger negative shocks and in this sense are driven by events that cannot be contracted over ex ante. Finally, for defaults to be excusable, outright repudiation should never be observed, unless it was preceded by a negative shock so large that no further payments can happen. This is the easiest
part of the argument; there is no single case of outright debt repudiation in our database.
We first demonstrate that payment stops occurred in poor states of the world for Philip II, which were caused by exogenous shocks. Excusable defaults also imply that lenders’ expectations are not disappointed. These are not directly observable; instead, we analyze how the pricing of loans changed after the 1575 default.
FIGURE 27. Military expenditure, actual and HP filtered, 1566–96
DEFAULT IN VERIFIABLY BAD STATES OF THE WORLD
For the Grossman and Van Huyck interpretation to be correct, defaults have to take place in verifiably bad states of the world. This was true in the case of Philip II. Twin shocks hit the Spanish monarch’s finances in both 1575 and 1596—military expenditures surged and revenue from the New World was below trend.
In figure 27, we show military expenditures during the period 1566–96. From year to year, expenditures on armies and fleets varied considerably. Three spikes are clearly visible: 1572–75, 1587–88, and 1596. These reflect the escalation of the Dutch Revolt, the Armada, and the outbreak of war with Britain. In two of these cases, the king defaulted. The exogeneity of these shocks varied. The Dutch Revolt had been simmering for a few years, and responding to it was a deliberate policy choice. Its escalation into a full-blown—and costly—war of independence, however, was beyond the imagination of Spain’s ruler (and arguably among European observers). Retreat was not a realistic option, as control over important revenue-generating territories was at stake.22
The Armada was a deliberate choice, planned and budgeted for over a period of years. Its failure was always a possibility, and in line with the predictions of the excusable defaults literature, it did not lead to a payment stop. The 1596 escalation of the Anglo-Spanish War was initiated by Britain; it required a major military effort on the part of Spain. The expenditure shocks were also large. In 1574, military spending accounted for 93 percent of all expenditure (without debt-servicing costs); it exceeded Crown revenue by 25 percent. In 1588, it also exceeded revenue by 16 percent (while staying below total revenue in 1596).23 Figure 28 shows revenues compared to an HP-filtered trend during the final thirty years of Philip II’s reign.
Revenue did not fluctuate as much as military expenditure. The king only defaulted in those years when revenue was markedly below trend and expenditures were simultaneously above trend. This confluence of expenditure and revenue shocks occurred in the mid-1570s, for several years in a row. As figure 28 shows, there were also many years when revenue was significantly below trend and the king did not default. This does not contradict our hypothesis that the king’s defaults were excusable because they occurred in bad states of the world. For it to be correct, the king does not have to default in all bad states; it is enough that he never defaults in good times. The observation is also easy to rationalize: silver revenues contributed substantially to volatility in the 1580s. Years of low revenue typically alternated with years of high revenue. Extra asiento borrowing smoothed over normal fluctuations. Combined with risk-sharing elements in the loan contracts (such as the one with Fiesco), the Crown coped with most fluctuations. In years of extraordinary pressure, a payment stop was declared and a general renegotiation became necessary.
The events that caused fiscal difficulties were easy enough to confirm and identify. Only one or two silver fleets reached Spain every year. The cargo of the arriving ships was a key determinant of Crown revenue. Once the ships had arrived, it proved impossible to suppress information on the size and value of the fleet (Morineau 1985). Commercial gazettes all over Europe carried details on the value of treasure brought from the Indies; a major determinant of the king’s fiscal position became public knowledge almost instantly. Military events such as the escalation of fighting in the Netherlands after 1568 were also simple to verify. While not all years of high military expenditure or revenue shortfalls led to payment stops, every default during Philip II’s reign happened when the king’s fiscal position was poor. Strained finances always reflected exogenous events and not poor fiscal policy. They were caused by the Dutch rebellion flaring up as well as Caribbean storms.
FIGURE 28. Crown revenue, actual and HP filtered, 1566–96
NO ADVERSE CHANGE IN LOAN CONDITIONS
We have established that defaults occurred in bad times and that bankers to Philip II shouldered some fiscal risks that the monarch was exposed to, thereby effectively providing insurance. In interpreting the defaults, the key question is the extent to which these were de facto anticipated. If so, they were simply another instance of claims falling due on an insurance policy; with the Crown’s finances stretched due to a lack of liquidity, contracts could not be honored to the letter. We now demonstrate that this is what happened.
FIGURE 29. Rate of return on short-term loans, before and after 1575
If lenders did not understand that they were de facto holding contingent debt, and if the defaults were not excusable, then loan conditions after 1575 should have changed for the worse.24 This is the null hypothesis that we examine. Figure 29 presents the distribution of interest rate (MIRR) on asientos before and after the 1575 default. As is readily apparent, there was no systematic shift in the cost of loans. The range, means, and medians of both distributions are virtually identical. A t-test finds no evidence that the king’s access to credit became more expensive.25
Alternatively, the quantity lent could have gone down even while the cost stayed the same. This is not what we find. The principal of loans made did not decline. In figure 30, we plot the log size of loans before and after 1575. The average dropped from a log value of 11.86 to 11.83 (p-value 0.78). While the distributions are not identical, there is no evidence that loans are systematically smaller.26
FIGURE 30. Size of loans, before and after 1575
Next we examine the cost of borrowing more formally. In the first column of table 25, we regress rates of return on the principal lent, foreign exchange clauses, and duration as well as a dummy for lending after 1575.27 We find that longer-duration lending, on average, was less expensive—a result that is consistent with the fixed cost of underwriting asientos and relatively cheaper alternatives available to the king for long-term borrowing. The size of a loan was not a significant determinant of its cost. Foreign exchange transactions raised the cost of borrowing by over 6 percent on average—a result of both the cost of operating overseas and use of foreign exchange transactions as a way to circumvent usury laws. Lending became 5 percent cheaper on average after the 1575 default. If we estimate with banker fixed effects (column 2), we find that the dummy variable for post-1575 loses its significance. This implies that the same bankers lent at the same rates as before, but that the number of bankers willing to only provide “dear” credit now played a smaller role. The post-1575 dummy becomes significant if we use robust regression estimation instead, and the size and significance of the coefficient is similar to simple OLS (column 3). The results so far do not suggest that the default caused bankers to suddenly update their beliefs about the riskiness of lending to Philip II. They did not begin charging more to compensate for higher perceived risk. If anything, the cost of borrowing dropped substantially after the resolution of the bankruptcy.
Table 25. Correlates of borrowing costs (dependent variable: MIRR)
Notes: The standard errors are clustered at the year level in columns (6) and (7). The t-statistics are in parentheses.
Significance: *0.1, **0.05, ***0.001
In specifications (4) and (5), we estimate the basic regression for the period before and after 1575. In both periods, lending for a longer period was associated with a significantly lower cost of financing. Before 1575, extending a contract’s duration by one year (roughly one standard deviation) was associated with a 2.4 percent fall in the cost of borrowing; thereafter, the predicted decline was by 4 percent. These two coefficients are statistically different from each other (beta coefficient −0.015; t-statistic −3.05). Column 5 also shows a significant c
oefficient for principal. This is in contrast to the result prior to 1575. At the same time, if we interact the post-1575 dummy with the size of principal offered, we do not find a statistically significant shift in terms of loan pricing in response to lending volume. Overall, the only major changes in loan pricing after 1575 favored the Crown.
In columns (6) and (7), we add two fiscal indicators as explanatory variables: the debt-to-revenue ratio and fiscal balance.28 In column (6), the debt-to-revenue ratio is significant at the 5 percent confidence level. The coefficient suggests that a unit rise of the ratio resulted in an additional 1 percent in borrowing costs for the king. Once banker fixed effects are added, however, the coefficient is no longer significant. This implies that as debts mounted, only a subset of “premium” bankers continued to operate in the short-term debt market. The fiscal balance, in contrast, is never significant; bankers did not take year-to-year fluctuations in the budget into account when pricing loans.29
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 28
