• A payment of 1,950 ducats in October 1595 to cover miscellaneous transaction costs; the bankers did not have to itemize expenses
• A final payment one month after the arrival of the fleet of 1596; this payment was calculated on the basis of the outstanding 177,000 ducats, plus 1 percent monthly interest from October 1595, plus an additional 2 percent of the base amount for “other costs”
If the fleet of 1596 failed to reach Seville by December, the bankers could ask for lifetime juros of the same face value as the outstanding payment, with a maximum rate of 7.14 percent. Finally, there was a standard set of clauses allowing the bankers to export bullion (needed to disburse funds abroad).
The Maluenda contract is relatively simple. Because the deliveries were made through letters of exchange denominated in Castilian ducats and the repayments were made in Castile itself, no currency conversion was necessary. The fleet arrival was the only source of uncertainty. We assume that the bankers expected the fleets to reach Spain in September—their median arrival month. Payment therefore should have occurred in October. If the fleet arrived later, a monthly 1 percent interest charge would accrue until the payments were made or the bankers received juros. The option of receiving lifetime juros is not relevant for present value calculations under normal conditions, given that lifetime juros have a present value lower than their face value. When a default was imminent, the option could have been useful; in normal times, it would not be exercised. We therefore disregard the possibility of taking lifetime juros in lieu of payment when calculating ex ante returns.3 The cash flows implied by our method are reported in table 14.
In constructing the cash flows, we needed to adopt several conventions. The asiento described above illustrates our treatment of payments tied to the arrival of the fleets. Other assumptions relate to the valuation of juros used for repayment. As a general rule, we use the cash flows of the juros themselves and calculate their NPV.4
We use two different profitability measures to derive rates of returns: the MIRR and profit index (PI). Virtually all our findings are robust to the choice of measure.5
The MIRR is defined as the ratio between the future value of positive cash flows and present value of negative cash flows. The formula is
where n is the number of periods in the contract. If the lender receives positive cash flows before the end of the contract, the assumption is that these cash flows can be reinvested at rate rr. Negative cash flows after the start of the loan are discounted at rate rf.
Using the MIRR is attractive because of the nature of asiento contracts. The cash flow of many asientos turned from positive to negative and back several times over the lifetime of a loan. Our sample contract with the Maluenda brothers is a case in point. The obvious alternative to the MIRR is the internal rate of return (IRR), a common measure in corporate finance. The IRR is defined as the discount rate that makes the NPV of a series of cash flows equal to zero. It is unsuitable given the nature of our data. The IRR performs well only in the case of simple cash flows, with a single disbursement followed by a single repayment. Whenever there are intermediate cash flows, two problems arise. First, the IRR formula assumes that any intermediate positive cash flows can be reinvested at the same rate of return as the entire project. This is unrealistic; there was not an infinitely elastic demand for loan contracts by the Crown. The banker’s obvious alternative was to invest repayments in juros. Because juros yielded less than asientos, the IRR would overestimate the profitability of the contract. Second, intermediate negative cash flows can cause the IRR formula to yield multiple solutions or none at all. Since most asientos specified staggered disbursements and intermediate repayments, we do not use the IRR.
Table 14. Agreed-on cash flows in the Maluenda brothers’ contract
The MIRR has the advantage of yielding a unique solution. In the absence of intermediate cash flows, it is identical to the IRR. Just as the IRR, it can be interpreted as the rate of return that makes the NPV of the project equal to zero. The MIRR requires explicit assumptions about the reinvestment as well as the finance rate. For our benchmark estimates, we use the juro yield of 7.14 percent as the reinvestment rate and 5 percent as the finance rate. These are conservative choices intended to produce lower-bound estimates of profitability.6 We also conduct sensitivity analysis with alternative parameter values.
The profit index is defined as the NPV of a contract divided by the capital at risk. Its advantage over the MIRR is that it only requires specifying one discount rate. The drawback is that the concept of capital at risk is not well defined when there are multiple staggered disbursements and repayments. Disbursements increase the capital at risk, while repayments diminish it. A long contract with a single repayment at the end exposes the lender to more risk than contracts with intermediate repayments spread across the same period. We measure the capital at risk as the total amount disbursed over the life of the contract. This overstates the true exposure, which was reduced by intermediate repayments. We also do not discount future disbursements but rather use their full value. In combination, these assumptions introduce a downward bias.
The main difference between the MIRR and PI is that the former is a gross measure, while the latter is net of opportunity cost—which we take to be the juro yield. To compare them, the juro rate must be first subtracted from the MIRR. Next, the discount rates used differ conceptually. In the MIRR, the reinvestment and finance rates refer to the yield of alternative assets. In the PI, the discount rate is a subjective measure that combines the opportunity cost of funds and risk aversion of the investor. Finally, as we discuss in the analysis, the MIRR is not well suited to evaluating long loans. When maturity is relevant, we use the PI instead.
SCENARIOS
We derive our data from the contracts as agreed between king and bankers. In many cases, the original agreement was not respected to the letter. The 1575 or 1596 bankruptcy impacted 119 contracts. Delays in both disbursements and repayments were common even in normal times. Almost 20 percent of loans contain clauses rescheduling previously unfulfilled obligations. Without observing the cash flows, we cannot derive precise measures of ex post profitability. Nonetheless, we can bound the likely returns. We do so by using our knowledge of the defaults and their settlements to approximate actual cash flows.
First, we calculate the profitability of each contract assuming that its clauses were respected to the letter. This is our upper bound. We then consider what would have happened if the king had repudiated all the outstanding debt in the 1575 and 1596 bankruptcies. This yields a (low) lower bound.7 Finally, we approximate the actual cash flows by estimating the settlement payments made by the king on each contract affected by the defaults. To illustrate the three scenarios, we return to the contract with the Maluenda brothers.
The first column in table 15 reproduces the agreed-on cash flows in the original contract. Using our benchmark reinvestment and finance rate, the expected MIRR was 12.5 percent, or a healthy 5.4 percent above the juro rate. The PI was 6.8 percent. In November 1596, however, the king issued the fourth suspension decree of his reign. The treasure that had arrived with the 1596 fleet was embargoed at the Casa de la Contratación; the final payment of the contract did not take place.8 Had the king repudiated the outstanding debt, the returns would have been strongly negative. Note that the majority of contracts would not have had such poor returns even under repudiation. Most were repaid partially or fully before the defaults took place. Bankers who had not disbursed the full loan amount could stop further payments. The Maluenda contract illustrates what could have happened in a worst-case scenario to an especially unlucky set of bankers. In actual fact, such a dire scenario did not materialize. The king agreed to repay 80 percent of the outstanding debts in October 1597. The settlement column reports our estimate of the actual cash flow. Since the language in most contracts does not distinguish between capital repayment and interest, we assume that all payments go toward capital amortization first. This produces a lower b
ound for outstanding capital at the time of the default and hence the settlement payment. By this methodology, as of October 1596, the king would have owed the Maluenda brothers 171,324 ducats from this particular contract.9 We multiply this amount by 0.8—the settlement ratio—and enter it as a positive cash flow in October 1597. This yields a MIRR of −5.3 percent (quite comparable to the PI of −14.8 percent).10
Table 15. Cash flows and profitability of the Maluenda brothers’ contract
While the 1597 settlement imposed a uniform 20 percent reduction on outstanding claims, the terms in 1575 varied according to how a contract was collateralized.11 Bankers who held standard juros as collateral recovered 70 percent of their claims; bankers holding juros guaranteed by the Casa de la Contratación received 55 percent; uncollateralized loans were granted 42 percent. For contracts affected by the 1575 default, we calculate the recovery rates for each contract based on the type of collateral used.
OVERALL PROFITABILITY OF LENDING
Our first question is whether bankers on the whole made money by lending to the king. To this end, we aggregate all lenders into a fictitious single financial entity for the years 1566–1600. Contractually agreed-on (ex ante) rates of return are only a first step in assessing the profitability of bank lending, as the case of the asiento of the Maluenda brothers demonstrates. We can learn about actual cash flows from three types of evidence. First, we have detailed information on the settlements after the defaults: we know who the king defaulted on and how the impasse was resolved. Second, the contracts themselves are meticulous in recording the king’s payment behavior on earlier contracts. When an old loan was not paid in accordance with the letter of the original contract, the next one would often provide compensation. Third, when the same bankers offered loan after loan, it is unlikely that they received returns far below their opportunity cost of capital.
The king could deviate from loan agreements in two different ways. For one, he might fall behind on payments on a particular loan. The payment in this case of the arrears would be rescheduled in a new contract with the banker. Although the return might not be as high as originally agreed, bankers seldom lost part of the principal and frequently received some compensatory interest. Second, the king could declare a bankruptcy and suspend payments on all outstanding loans at the same time. Philip did so four times during his reign, and our data cover the last two. Defaults like these would be renegotiated with all bankers in a general settlement, which specified principal and interest write-offs. The total ex post returns can therefore be written as
R = Re – prLr – pdLd,
where R is the total ex post return, Re is the contracted rate, pr is the proportion of debt rescheduled in individual contracts, Lr is the loss rate for rescheduled debt, pd is the proportion of debt defaulted on in general bankruptcies, and Ld is the loss rate in the defaults.
Based on the expected returns, Re is 20.3 percent. Obligations from earlier loan contracts were rescheduled in ninety-six cases. The king typically acknowledged the earlier debt and then offered various sweeteners in the new loan contract. This procedure affected 10 percent of the total amount lent—hence pr is 0.1. Rescheduling earlier obligations typically increased returns for the new contract by 2 to 3 percent.12
How high was the recovery rate on rescheduled loans? The most optimistic interpretation implies that the additional returns to subsequent lending fully compensated lenders for what they had lost. A more cautious approach would assume that lenders received no interest on their earlier loans. This would reduce average profitability linearly, in line with the proportion of loans that were rescheduled. Lr would be 0.203, the same as the average return on loans. Hence, returns would have been 0.1*0.203 = 0.0203 lower than the ex ante contracted rate because of the subsequent recontracting.
Next, we need to derive values for the proportion of loans defaulted on and the recovery rates. Philip’s four defaults were not of equal magnitude. The two earlier ones, in 1557 and 1560, mainly involved German bankers. They largely concerned debts contracted by Philip’s father, Charles V, and were settled by transferring revenue-yielding assets. The famous quicksilver mines of Almadén, for instance, were given to the Fugger in exchange for debt cancellation. Since the original loans were not part of our data set, we were not able to examine the revenue impact of these two payment stops.
In 1575, the king suspended payment on 14.6 million ducats of outstanding loans. The majority of bankers negotiated a comprehensive settlement with the Crown. It resulted in write-offs of 30 to 58 percent. On average, the king agreed to honor 62 percent of the outstanding principal of short-dated loans and the associated interest payments. Long-term bonds escaped unscathed. In 1596, the king defaulted on 7 million ducats of debt, and the haircut imposed was 20 percent. We know that total asiento lending was 99.7 million nominal ducats over the period of these last two defaults, and that no more than 21.6 million worth of loans were affected by them—just over 21 percent of all contracts. The weighted recovery rate for the third and fourth defaults is 68 percent. The cost of the defaults to lenders is thus pdLd = 0.21*0.32=0.067.
The average write-offs from the defaults on loans amounted to less than 7 percent of lending over the period. Defaults hence reduced profitability twice as strongly as our pessimistic calculations for ordinary reschedulings suggest.
Based on the figures just derived, we calculate
R = Re – prLr – pdLd = 0.203 – 0.1*0.203 –
– 0.21*0.32 = 0.203 – 0.0203 – 0.067 = 0.116
How profitable was lending? The fiscal turmoil that characterized Philip II’s reign cost lenders less than half their potential profits, according to our calculations. Their average rate of return was 4.43 percent above the juro rate, indicating that they earned profits over and above their opportunity costs.
This result is derived from a calculation with many unknowns. We have tried to err on the side of caution, using estimates that are, if anything, too pessimistic about bankers’ profits. How robust is our finding? Since the amount of rescheduled debt is relatively well established, we examine what happens when we vary the write-off rates. To reduce average profitability to zero, given the losses on ordinary reschedulings, the write-off during the defaults would have had to be 87 percent instead of the 32 percent actually suffered. Alternatively, write-offs on the reschedulings would have to be greater than 100 percent (135 percent) instead of the 20.3 percent we calculated (taking the estimated losses during defaults as given). Only extremely large deviations from the estimated loss rates and rescheduled amounts could reduce ex post rates of return to zero.
PROFITABILITY BY FAMILY
While the results presented so far show that lending to Philip II was profitable on average (even under unfavorable assumptions), this could mask considerable variation across lenders. We now examine rates of return by family. Between 1566 and 1596, 145 different bankers belonging to 78 families engaged in business with Philip II. Still, only 127 bankers, belonging to 60 families, ever risked capital. The rest provided intermediation services without putting their own resources on the line. We focus on the 60 families that extended credit.
Table 16 reports the MIRR by family for the period 1566–1600. The families are ranked by the total amount lent over the period as a whole. The provision of credit was heavily concentrated. The Spinola family, which counted 12 active members, lent over 20 percent of all funds. The top 10 families provided just short of 70 percent of all loans, and 19 families lent over 1 million ducats each.
The rates of return varied considerably. No family agreed to compensation below the 7.14 percent juro rate.13 In the event of a complete repudiation, 18 families would have lost money. The remaining 42 families, however, would have realized positive rates of return; 37 of them would have earned more than the juro rate.14
According to our best estimate of actual profitability, reported in the settlements column, only 9 families failed to earn their opportunity cost; fully 51 earned more than the long-term
bond yield. Of the 5 families that actually lost money, 3 invested little—2,080, 6,110, and 28,601 ducats respectively. All 5 entered into one or two contracts with the king, closely before the defaults. The Galletto and Salinas families sustained losses on somewhat-larger contracts, but their rates of return, −11.3 and −10.5 percent, respectively, are hardly catastrophic. In fact, the Galletto family signed its only contract just four days before the 1596 bankruptcy. In all likelihood the disbursement was never made and the family did not suffer any losses. We nonetheless assume the contract was fulfilled in order to bias our results against finding profitability. The repudiation scenario therefore shows a profitability of −100 percent; the family would have lost the entire amount disbursed. The absolute losses of these 5 families amounted to just over 75,000 ducats. This is less than 0.1 percent of the total short-term lending to Philip II.
Table 16. MIRR by family (1566–1600, annualized rates)
Note: The reinvestment rate is assumed to be 7.14 percent, and the finance rate 5 percent. The amounts disbursed are expressed in ducats. We use the Spanish spellings of the family names, as they appear in the archival documents.
According to the settlement scenario, 4 families did not lose money in absolute value, but failed to earn the juro rate. One of these was the Fucar (Fugger) banking dynasty. As a matter of fact, the Fugger were the only family exempted from the provisions of the 1575 bankruptcy. Their actual rate of return was the originally contracted 11.4 percent. The other 3 families were the Lercaro (3.1 percent), Serra (2.9 percent), and Sauri (5.8 percent). The last 2 families had only a sporadic relationship with the king and happened to lend just prior to the defaults. The Lercaro lent somewhat-larger amounts throughout the entire period—just over 550,000 ducats. These loans were provided in the run-up to the bankruptcies, and the reduction in payment obligations caused the Lercaro to earn less than they would have by investing in juros.
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 23