Bad news was effectively translated into lower payment obligations or later due dates; risk sharing between borrowers and lenders worked. This arrangement also survived Philip II’s famous defaults. The king suspended payments no less than four times during his long reign, but defaults were excusable in Grossman and Van Huyck’s (1988) sense. Debt was de facto state contingent even under circumstances not specified in loan contracts. Defaults involved larger shocks than the contracted ones in individual contingent loans. They also entailed contingencies that were difficult to write into contracts—principally the outbreak and outcome of wars. Our data show that the pricing of loans did not change adversely after the payment stops; lenders did not update their priors about the type of borrower Philip II was. This fact along with the prevalence of contingency clauses in loan contracts offers support for interpretations of sovereign lending emphasizing implicit risk sharing in cross-border debt.
Scholars working on debt restructuring have highlighted that the process can be long and inefficient (Benjamin and Wright 2009). On the other hand, Kovrijnykh and Szentes (2007) argue that repeated cycles of borrowing and default may be an efficient outcome, with lenders having an incentive to let borrowers escape from debt overhang. Patrick Bolton and Olivier Jeanne (2009) suggest that contracts that are ex post excessively difficult to restructure can be the result of efficient bargaining ex ante. Philip II’s defaults were restructured unusually quickly and with moderate haircuts, indicating that the payment suspensions were closer to an implicit contingent contract than to full-scale breakdowns in debt markets.
CONTINGENT LENDING TO PHILIP II
In this chapter, we work with the 393 contracts for which we were able to observe every single clause and repayment scenario.7 Of these, 270 have at least one contingency clause; many contain several. There were a total of 408 different scenarios over which the king and bankers contracted. We first summarize the types of contingencies written into contracts and then analyze their economic effects.
FIGURE 23. Distribution of the major contingent scenario, by number of contracts and by principal
TYPES OF CONTINGENCIES
The loans contain a wide variety of scenarios allowing the king and bankers to deviate from the baseline payment schedule. We can distinquish five broad categories. The first two are associated with events outside the control of either king or bankers: the arrival of the fleet and the performance of specific tax streams. Two more types are actually options, given to either the king or bankers. In some cases, the king can delay payments, usually in exchange for some penalty. In others, the banker can request to be paid in juros ahead of the loan maturity date. We call these options “king’s discretion” and “banker’s discretion.” Finally, contracts involving collateral also have an “execution” clause, which specifies under what conditions the banker may seize and sell the collateral. Figure 23 plots the main type of contingency for each contract, showing both simple frequencies and the principal-weighted distribution.
FIGURE 24. Contingent and noncontingent borrowing over time
Less than a third of the contracts were issued without a contingency clause. About 31 percent of these contingent scenarios refer to insufficient revenue for the Crown—either because some taxes fall short, or because the treasure fleets do not arrive on time or in sufficient size. Slightly over 40 percent of the contingent clauses give either the king or banker discretion to change the timing or nature of the payments. The remaining ones give the banker the right to seize and sell collateral if the king fails to repay.
The use of collateral clauses was not constant over time. As figure 24 portrays, it fluctuated heavily over time. While a dominant feature during the entire period, the use of contingency clauses was curtailed between 1576 and 1585. In the run-up to both the 1575 and 1596 bankruptcies, in contrast, contingencies were quite common.
We now illustrate the working of contingency clauses using specific examples from the primary sources. In 1566, the king entered into a contract with Lucián Centurión and Agustín Spinola. They agreed to disburse 38,000 and 57,000 ducats in Flanders in, respectively, May and September of that year, and were meant to receive one repayment in August from the first silver fleet.8 If the fleet did not arrive by the end of July, the king promised to pay a penalty rate of 1 percent per month until the full repayment was made. The bankers also received a juro that covered the full value of the contract, which they were allowed to sell in case the king failed to meet his obligations. The original contract without contingency clauses produced an annualized MIRR of 24.1 percent. If the contingency clauses were invoked, this fell to 15.6 percent. The king thus could insure part of the income risk that came from the highly volatile silver revenue stream. At the same time, the bankers’ financial position was largely safeguarded against the risk that the king could or would not pay through the use of collateral juros.9
Another asiento demonstrates how variable tax revenue could also trigger contingency clauses. In October 1581, Juan Ortega de la Torre lent 60,000 ducats to the king. He was to be repaid from the second payment of the excusado.10 De la Torre was not first in line; the contract specified that Baltasar Cattaneo should collect his money first. Importantly, the banker would have to collect payments himself, to which end the king provided him with the necessary documentation. Should the revenue from the excusado be insufficient, the banker had the right to be repaid from the perlados y cabildos (a minor revenue stream levied on ecclesiastical rents). Other contracts in this category specified that if the tax revenue in one year was insufficient, the king would pay a penalty interest rate until he could repay with the following year’s taxes.11
We now turn to a more complex asiento. We already met the lender—Tomás Fiesco, the Genoese banker waiting for the silver fleet to arrive at the beginning of this chapter. His contract with the Crown showcases how multiple contingent clauses could be used to provide insurance for both king and bankers under a variety of scenarios. In 1591, Fiesco had agreed to provide 300,000 Flemish ecus.12 Of these, 200,000 ecus were paid out in Flanders, while the rest were delivered at the payment fairs of Besançon. The king advanced 75,000 ducats at the contract’s signing, which meant that the actual loan was for 218,000 ducats (the rest was a mere transfer of funds). The first disbursement by the banker, also in April, was for 61,500 ecus. It was followed by nine equal monthly payments of 26,500 ecus each.
The king promised to repay the loan from a variety of sources. Several of these payments were not contingent: 84,700 ducats from the new millones excises in November 1591 and May 1592; another 60,000 from the cruzada ecclesiastical tax in October and November 1592; 12,000 ducats from the sale of vacant lands; and 30,000 ducats from the extraordinary service. The single-largest payment, for 90,100 ducats, was to come from the proceeds of the silver fleet of 1591, which was expected in late summer or early fall. This was followed by a fleet contingency clause: if the silver did not arrive by October, a penalty of 1 percent per month would apply until the banker was repaid from alternative tax streams—specifically the subsidio, excusado, and ordinary and extraordinary services. The treasury disbursed payments from these sources every four months, in March, July, and November.
Even if the fleet arrived on time, the king could unilaterally choose to delay repayment until the maturity of the loan, twelve months later. This is what we have labeled king’s discretion. It came at a steep cost: if the contingency was invoked, the banker had the right to stop the remaining disbursements (for a total of 53,000 ecus) while still being entitled to collect all the promised repayments on earlier disbursements. Finally, from January 1592, the banker had the right to request repayment of up to 100,000 ducats of principal and interest in perpetual juros—a banker’s discretion clause. This contingency allowed the banker to receive safe bonds instead of promised cash payments.
Table 22 shows the cash flows for the Fiesco contract under the baseline scenario, and under each of the fleet’s and king’s discretion contingencies. Becau
se the banker’s discretion contingency only affects the payment instruments but not the actual timing or values, we do not report it in a separate column.
Table 22. Net cash flows from the Fiesco contract under three repayment scenarios
Note: Figures are in ducats; 1 ducat = 1.023 ecus.
The baseline scenario gave bankers a 23.2 percent return. While higher than average, this was not an unusual cost for a contract that included transfers to multiple locations, deliveries in several currencies, and repayments sourced from many different tax streams. Under the fleet contingency scenario, the king misses the largest part of the November payment (some 90,100 ducats) in 1591, and makes up the shortfall in March, July, and November 1592 with 1 percent monthly interest. In this case, the rate of return on the contract increases slightly, to 24 percent.
The king’s discretion scenario is markedly different. The king additionally misses the payment of 90,100 ducats in November 1591, causing the banker to cancel the December and January disbursements. The king is still obliged to make all promised repayments, including the 90,100 ducats, which are now due in November 1592. The banker gets paid much later than promised, but since he skips two disbursements totaling over 50,000 ducats, his rate of return increases to 39.8 percent.
It is useful to contrast the two contingencies. The cash flows are identical up to and including November 1591. The missed payment is exactly the same. In one case, however, the reason is an exogenous, verifiable event: the fleet has not arrived. The banker continues to make the disbursements as scheduled, while the repayments are delayed. The cost of the contract rises marginally. In the second case, though, the fleet has arrived. If the king fails to pay in this particular instance, he has much less of an excuse. The cost of the contract in this scenario goes up substantially.
The fleet contingency insures the king against factors outside his control, such as adverse Caribbean weather and disruptions to silver production. Because these factors are self-equilibrating in the medium term, the bankers do not charge high insurance premiums. The king’s discretion contingency is different. It gives the king the option to extend the maturity of his debts without having to borrow fresh funds, even if the fleet arrived in time. The king is now protected against an unexpected need for liquidity or the prospect of a rollover crisis. Because these situations would signal mounting pressure on the king’s finances (or a lower willingness to use available funds for repayments), the banker demanded a hefty premium in exchange for providing that insurance. Finally, the banker’s discretion contingency insures the lender against a downturn in the king’s ability to pay. After the first eight months of the contract, the banker can swap almost all his remaining claims for relatively safe long-term bonds of the same present value.
THE ECONOMIC IMPACT OF CONTINGENCY CLAUSES
What is the economic purpose (and impact) of the different contingencies? In this section, we examine the effect of contingent clauses on cash flows. We also analyze cost and maturity modifications as a function of the Crown’s and bankers’ interests as well as in response to the arrival of new information. Contingent clauses provided ample, bidirectional risk sharing between the king and his bankers. Their use is interesting: they reveal a preference to deal with eventualities ex ante, before they materialize, instead of having to renegotiate ex post. This implies that frequent recontracting (in the spirit of Bulow and Rogoff 1989) was not costless in the eyes of Crown and financiers.
How exactly did contingencies influence cash flows? For each contingent scenario, we compare the rate of return to that of the noncontingent cash flow scenario. Loans that feature contingent scenarios have a baseline return of 20.5 percent. In aggregate, contingencies do not affect the returns substantially. The median change in the cost of a contract is close to zero under the average contingent scenario.13 Panel A in figure 25 plots the distribution of cost changes. While some contracts saw their cost rise or fall by 20 percent or more annually, most changes were much smaller; the bulk of observations involve changes of around 5 percent per year.
Contingent scenarios also affected the maturity of the loans (panel B). On average, the maturity of loans changed little: an increase of two months with a standard deviation of nine. One hundred and twenty one scenarios allowed for a longer maturity, giving the king an average of 9.7 additional months to repay. In 18 cases, there was an early termination date, either because the king could exercise an early repayment clause or because a missed payment allowed bankers to cancel future disbursements. In these situations, the average termination date preceded the original one by 17.6 months. The remaining 269 scenarios do not affect the maturity date either because they shuffle intermediate repayments or because they specify a swap of payment instruments.
Table 23 summarizes the contingent changes in the MIRR and the maturity of loans. The modifications reflect the changes in the king’s fiscal position associated with each contingency type. In case of a fleet-related event, the maturity of the loan was extended for an average of 2.6 months. The cost increased only by a small amount.14 On average the bankers received some compensation, reflecting only a minor increase in risk. Fleets would eventually arrive, and delays did not convey new information about the king’s solvency. At the same time, bankers on average required some additional compensation entering into contracts that had a fleet contingency written into them; on average, the baseline cost was 4.1 percent higher (even if variability was high and the difference is not statistically significant).
Tax revenue shortfalls were a different matter. Most taxes were collected directly by cities or tax farmers, which had an incentive to maximize revenue. The performance of revenue streams was independently verifiable by the lenders. The incentives of bankers and those of tax collectors were compatible, and there was no possibility for the king to manipulate the total yields.15 Tax stream insufficiencies were bad news on the fiscal front, but they did not convey information about what type of borrower the king was. The associated contingent scenarios gave the king an extra 4.6 months to repay, while reducing the rates of return by 1.7 percentage points. Consistent with a risk-sharing arrangement, a negative shock in terms of fiscal revenue resulted in a reduction of borrowing costs. The baseline cost of loan agreements with a tax shortfall clause was 1.6 percentage points below the average; bankers were willing to offer this type of “insurance” without a premium.
FIGURE 25. Changes in the cost of contracts and in maturity after contingency clauses are invoked
Table 23. Baseline MIRR, differential MIRR, and maturity (in months) by type of contingency
Note: P-values in parentheses.
*The coefficient is from a regression of MIRR on contingency type dummy, use of foreign exchange clause, duration, and loan size. The standard errors are clustered at the contract level.
King’s discretion scenarios involve nonpayment without an externally verifiable trigger. This was followed by a rearrangement of cash flows. The arbitrary postponement of payments by the king was undoubtedly bad news, either because of new, urgent spending needs or because other loans were receiving priority for repayment. Unlike the case of tax stream insufficiencies, the cause of the need for extra liquidity was uncertain, and moral hazard could not be ruled out. The risk to the bankers was increasing compared to the original contract. Risk sharing implies that this additional risk should be associated with a transfer to the bankers. The large increase in returns—4.1 percent on average—is consistent with this interpretation. Contracts with a postponement option for the king were also more expensive in the baseline scenario, by an average of 4.3 percent.
The effect of banker’s discretion is more difficult to evaluate, as there are liquidity considerations to take into account. These clauses typically allowed bankers to collect part or all repayments in juros instead of cash. There was often no reduction in the amount payable, and bankers were allowed to collect the entire current-term interest of the juros.16 This accounts for the increase of 1.5 percent in the rate of retu
rn. In practice this accounting profit probably did not translate into an actual cash flow advantage. Bankers would have had to sell juros on the secondary market—a costly operation that could nullify the 1.5 percent gain. If they chose to keep the bonds, they would have had to wait for and oversee the collection of the coupon payments. A reasonable guess is that banker’s discretion clauses allowed lenders to switch their repayments to safer assets without a substantial impact on their cash flow. Bankers typically asked for these clauses when they had reason to be concerned about the future yield of extraordinary revenues.17 Whenever asientos contained banker discretion clauses, they were on average 1.6 percent more expensive initially for the king.
The final contingency type was collateral executions, which were triggered by the king missing the final loan payment. Because the event was predefined in the contractual clauses, it was not considered a default. When it was exercised, the cost of the contract fell. Sometimes the contracts specified that the bankers had to wait before being able to sell collateral juros—hence the two-month average maturity increase, which reduced profitability. Bankers also lost because the collateral was not always sufficient to cover the last repayment. When bankers and the king’s representatives entered into contracts with a collateral execution clause, the cost was on average 2.1 percentage points lower. This reflects the additional security of holding a collateralized loan.
Our findings indicate that the king was mainly concerned with the risk of a liquidity shortfall; the cost of borrowing mattered relatively less. The majority of short-term loan contracts envisaged the option to postpone the payment or swap payment instruments. Return differences by the type of contingency strongly suggest that these options allowed for effective risk-sharing arrangements. Instead of having to find fresh funds to redeem maturing debt, the king had the right to extend the maturity of his borrowing either by delaying payments or swapping short-term debt for perpetual bonds. At the same time, the bankers reduced the risks from the king changing his spending priorities. The most costly eventualities—combining higher baseline cost and the expense of the contingency cost—were those triggered at the king’s discretion. Here, bankers received an extra 8.4 percent in interest. In other words, when the Crown postponed payments without “just cause” in the form of late fleet arrivals or tax insufficiencies, the increase in borrowing costs was greater than average. Bankers realized that writing an option on such eventualities did not imply good news about the king’s finances and demanded to be compensated accordingly.
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 27