Three of the MIRRs reported in table 16 are unusually high. Juan Curiel de la Torre earned over 151 percent on lending some 186,000 ducats. The Salucio and Cibo families also earned in excess of 50 percent. Curiel de la Torre achieved such a high return through a combination of factors. He had high returns on small contracts, and kept his exposure to a minimum by staggering disbursements and repayments.15 The Salucio and Cibo lent little, and hence did not obtain large absolute gains.
CORRELATES OF RETURNS
Table 16 reveals considerable heterogeneity in the rate of returns at the family level. Across individual loans, differences in outcomes are even more marked. What explains the cross-section of returns to lending? Were some bankers obtaining preferential treatment? Or does the variation mostly reflect the different characteristics of each loan?
FIGURE 17. Density of MIRRs by foreign exchange clauses
We begin by examining some of the patterns of association between contract characteristics and the agreed-on lending rate. The first factor to note is that many loans contained a foreign exchange component. These were, on average, more expensive.
Figure 17 plots the distributions of loan profitability with and without an explicit foreign currency clause. While there is a lot of variation in the interest rates charged, the difference in the means and modes is clear. Loans with a foreign exchange component had an unweighted mean (median) interest rate of 27 percent (17 percent); those without had an unweighted mean (median) interest rate of 19 percent (11 percent). At first pass, then, contracting debts that involved disbursement in foreign specie raised the cost of a loan. Depending on the method used, we find increases of up to 9 percent.
Maturity also influenced the cost of loans. Since the MIRR is not well suited to evaluate loans of long maturity, we use the PI instead.16 Loans with abovemedian maturity (more than twenty-two months) often attracted higher interest rates. Figure 18 plots the distributions. The unweighted mean (median) PI for long loans was 24 percent (13 percent), while it was 17 percent (11 percent) for short loans.
FIGURE 18. Density of PI by loan maturity
Table 17 summarizes the profitability of loans conditional on their characteristics. Those with a foreign exchange clause had 3 to 5 percent higher returns.17 Contracts with collateral did not attract a significantly higher rate of return. Transfers themselves (without accounting for the use of foreign exchange clauses) also did not attract higher returns. Large loans were somewhat cheaper than small ones, but the difference is only clear for means, not medians. Finally, lenders’ nationality affected profitability. German lenders received a lower rate of return, although their loans were frequently serviced during the defaults and hence were less risky. The Genoese earned (median) returns that were higher than the rest by about 2 percent, but their average returns were indistinguishable.18
Table 17. Conditional profitability
Loan characteristics explain few of the differences in profitability; variations between different loan types seem small compared to the standard deviation of profitability. While not satisfactory overall, these results are similar to those found in the modern finance literature.19
On average, the general settlements caused losses for the lenders. What explains the cross-section of these differences? We define the variable LOSS_MIRR as the difference between the MIRR according to the settlement and the promised MIRR, as per the original agreement. LOSS_PI is defined analogously. Table 18 regresses these measures on our standard set of explanatory variables. Loans with a foreign exchange component typically did poorly. Defaults happened in times of intense military conflict, when most loans were disbursed near foreign battlefields. Loans with a high collateral component showed higher returns; since the settlements took the presence of collateral into account when determining the haircuts, this is not surprising. Longer durations were unambiguously associated with greater losses, while larger loans did somewhat better. Nationality did not confer a clear-cut advantage when negotiating the settlements, nor were specific families favored over others. In combination, these two sets of regressions suggest that the rates of return were determined by a specific loan’s features and did not reflect the particular negotiating position of a given set of bankers. The same is true of the losses sustained during the defaults.
Table 18. Correlates of losses during the defaults
Significance levels: *10%, **5%, ***1%
PROFITABILITY OVER TIME
How did lending rates change over time? Did the defaults make lending more expensive for the king? In other words, is there evidence of mounting “debt intolerance” (Reinhart, Rogoff, and Savastano 2003)? Figure 19 plots the volume of lending and its profitability over time. The line, indexed to the left-hand axis, shows the weighted average of the ex post MIRRs of contracts according to the year in which they were signed. The bars show the volume of actual lending every year.20
FIGURE 19. Profitability and volume of lending
MIRRs fluctuated between 15 and 30 percent. The exceptions are the defaults. Only contracts signed in 1573, 1574, 1575, and 1596 failed to earn the juro rate. There is virtually no connection between MIRRs and lending volume—the correlation coefficient is −0.16. With the exception of the dips around the defaults, rates do not show major fluctuations.
Table 19 shows the average differential between ex ante and ex post MIRRs for contracts signed in the years leading up to each bankruptcy. The 1575 default was more severe than the 1596 one. Consequently, the gap between contracted and actual rates was higher in the years leading up to 1575. Many contracts of long duration were affected; maturities ranged up to seven years in 1575 versus a maximum of four in 1596. The amount defaulted on in 1575 was 14.6 million ducats and the average haircut was 38 percent; in 1596, the king stopped servicing 7 million ducats of debt and negotiated a reduction of outstanding claims by 20 percent. These numbers reflect the severity of each fiscal crisis. In 1575, two simultaneous campaigns, three unusually poor fleets, and the reluctance of the Cortes to increase taxes led to a serious cash flow shortfall. In 1596, in contrast, taxes had already increased substantially. The liquidity shortfall was caused largely by renewed military activity, while accounting errors may have contributed to the sense that the situation was critical. As is the case with any insurance contract, the losses suffered by lenders were larger when the fiscal situation was more pressing.
Table 19. Difference between average contracted and actual MIRRs
Year contract was signed
1575 default
1596 default
Year of default
30.9%
24.7%
t-1
16.8%
9.4%
t-2
11.6%
1.4%
t-3
3.1%
1.2%
t-4
1.7%
0.0%
t-5
0.7%
0.0%
t-6
0.9%
0.0%
t-7
4.3%
0.0%
NET RATES OF RETURN
In our benchmark scenario, bankers obtained an average ex post gross MIRR of 15.5 percent. To derive a net measure of profitability, we need to subtract costs. Unfortunately, we do not observe most costs at the contract level. We instead use the available information on their general range to estimate the effect on the average rate of return: the costs were not sufficient to overturn the result that lending was profitable. Nevertheless, they likely reduced some apparently high rates of return to a normal range.
The most important cost—financing—is captured by the long-term bond rate. Next, we need to consider the costs of intermediation. Lenders relied on a correspondent network. The cost of using this network is reflected in the charges for issuing letters of exchange. Between 1566 and 1575, most contracts required disbursements either in cash at the treasury or via a letter of exchange drawn on a specific fair. When the latter was requested, the king was ch
arged an additional 0.5 percent of the principal.21 In addition, most bankers did not risk their own capital but instead acted as intermediaries, splitting the contracts into smaller parts and selling them to individual investors. Their typical fee for this service was 1 percent.
Currency conversions and the need to transport bullion also added to costs. Contracts with a foreign exchange component often specified that the banker would be reimbursed for “what is customary among businessmen.”22 Sometimes the king requested an affidavit signed by three or four independent bankers attesting to the costs incurred. The king often provided free space on his ships. Other costs were either reimbursed on top of all other payments to the bankers or covered directly by the king. These costs thus do not affect the rate of return.23
Several contracts include specific allowances for other costs. Our sample contract with the Maluenda brothers is a good example. The king agreed to pay the bankers a total of 5,490 ducats to cover any costs they might incur, without demanding that they account for the expenses. This amount was 1.6 percent of total payments by the king. We don’t know whether cost allowances covered actual costs or whether they were merely used as a way to increase the rate of return. In our cash flows, we treat them in the same way as any other payment to the bankers. Hence, their effect is incorporated in the gross profitability figures. They typically amounted to 1 to 2 percent of the principal.
Finally, our rates of return are nominal. The second half of the sixteenth century witnessed the price revolution, as inflows of American silver caused a general increase in the price level. Existing price indices are not accurate enough at yearly frequencies, preventing us from calculating real rates of return for each contract. Period averages are more reliable. Between 1556 and 1600, prices rose at an annualized rate of 1.7 percent (Drelichman 2005).
We can now calculate the net rates of return. The starting point is our benchmark estimate of 15.5 percent, which already takes into account the effect of the bankruptcies. We subtract the 7.14 percent opportunity cost, up to 1.5 percent in intermediation costs, up to 2 percent in other transaction costs, and 1.7 percent for inflation. This yields a real net return of 3.16 percent in a scenario involving high costs. If intermediation or transaction costs for a specific contract were lower, the net real rates of return could have been as high as 5 percent. While stressing that these are rough estimates, we note that they remain positive, in line with our general results, and they are relatively modest values considering the risky nature of short-term lending.24
RAISING THE FUNDS
All the calculations in this chapter were presented as if Genoese bankers lent directly, with the Spinola or Maluenda effectively writing large checks to Philip II, and then waiting to be repaid. In actual fact, bankers seldom offered a loan using only their own capital. Rather, much as modern banks do, they would tap a variety of financing sources, including demand deposits and the sale of participations in the sovereign’s lending ventures. In this fashion, they could offer much larger loans than their own resources allowed, while simultaneously limiting their exposure as well as spreading the risk among their customers and equity partners. These arrangements can also have an important impact on the rate of return for the bankers.
The profitability of individual loans could differ significantly from the gains for the international bankers who were in charge of organizing large, multilayered syndicates that lent to the king of Spain. Returns for final investors could again be different: a chain of financial intermediation linked the king to investors small and large throughout Europe, and they ultimately bore the risks. The available information on this aspect of lending is much more fragmentary than the data on asientos themselves.
In our opening chapter, we recounted the story of the Di Negro–Pichenotti partnership—the small merchants from Genoa who purchased shares in two asientos extended by Agustín Spinola and Nicolás De Negro in 1596. Here we take a closer look at their transactions and show how the risks of lending to the king were spread through several layers of financial intermediation. In good times, a large number of investors, big and small, benefited from the asientos. When payments were suspended, many parties shouldered the losses with diversified portfolios, ensuring that most would weather the storm unscathed.
In February and July 1596, Agustín and Nicolás jointly underwrote the two asientos we analyze here.25 As discussed above, the Spinola were the largest lenders to Philip II. They alone accounted for over 20 percent of the total short-term borrowing over the period. While the De Negro lent money on a more modest scale, amounting to some 770,000 ducats in total, they were also among the leading business families in Genoa.26 Agustín and Nicolás lived permanently in Madrid, and were in charge of managing the financial operations that their families entered into with the king. This management included negotiating new loans, arranging the disbursements promised in Madrid, and issuing the necessary letters of exchange to authorize disbursements abroad. They were also responsible for collecting the repayments, which required skill at navigating the royal bureaucracy and trustworthy agents in many places where treasurers in charge of different royal revenue streams were stationed. Finally, the bankers had to obtain the necessary permits to remit the proceeds back to their families in Italy or wherever else they were needed, and had to ensure that the bullion was delivered to a port of exit and shipped safely.
The first asiento was concluded on February 24, 1596. Spinola and De Negro initially agreed to deliver 90,000 ecus in Milan. Half the amount was due immediately and payable on presentation of the letters of exchange by the royal officials. The other half would be disbursed in three equal payments over the months of April, May, and June. In addition, the bankers promised to deliver 112,500 ducats in Madrid in six equal payments. The first two payments had already been made on January 1 and February 1, 1596; the remaining four installments were to be paid once a month.27 The contract valued the Italian ecus at 404 maravedíes each, which represented a 1 percent premium over their gold content of just under 400 maravedíes (1.067 ducats). The combined principal of the contract therefore amounted to 209,460 ducats.
The king promised to repay the capital using the proceeds of the three graces from the years 1597 and 1598 as well as those from the ordinary and extraordinary servicios.28 The contract stipulated that the proceeds of these taxes would be disbursed to bankers in six installments, starting in July 1598, and every four months thereafter. The interest rate was 1 percent per month, not compounding; each capital repayment would also be accompanied by the accrued interest on that part of the capital only. The first installment would include an extra two months of interest as well. As additional compensation, the bankers were allowed to swap juros worth up to 485 ducats for other bonds of their choice. They could therefore purchase nonperforming bonds at bargain prices, exchange them at the treasury for choice securities, and net a substantial profit.29
The contract included a number of additional provisions. First, the bankers were allowed to export bullion equivalent to the value of the principal. Although 112,500 ducats were to be delivered in Castile, the bankers would be raising the necessary funds outside the kingdom and hence would need to export the repayments to satisfy their own liabilities. The bankers were also given permission to export another 60,000 ducats to Portugal. These export licenses were valuable, as they allowed their holders to arbitrage between different currency markets. Bankers could sell them to other businessmen. If a license went unused, the treasury would on occasion buy it back.
Spinola and De Negro were given the option of collecting their repayments from alternative income streams, too. In particular, they were allowed to choose to be repaid from the fleets of 1596 and 1597. In this fashion, they could have begun to collect funds a few months earlier, at the cost of forfeiting the extra 2 percent on the first installment. Alternatively, the bankers could request that repayment be made in the form of lifetime juros. This would have allowed them to receive payment almost immediately, but at a higher cost.
30 The bankers, according to the contract, also could opt for perpetual juros, but they would have to wait until the originally promised repayment dates to receive them. This last option would only be valuable if, for some reason, the original income streams failed. Finally, the contract allowed the bankers the use of one or two royal galleys to convey bullion to Italy.
Table 20 shows the agreed-on cash flows from the asiento of February 24. All the disbursements took place in the first six months of the contract, and with the exception of the small profit from the juro operation, no repayments were promised until July 1598, a full thirty months after the beginning of the contract. In laying out the cash flows, we abstract from the several options that the bankers could exercise, such as choosing different repayment streams or converting part of their credits into juros. Most of these would have resulted in some small modification of the contract’s profitability. The actual sign and magnitude of the change depended on unobservable conditions. In order to produce a conservative estimate of the rate of return, we also omit the value of the license to export bullion.31
Had the contract been honored as originally signed, the bankers would have realized a yearly rate of return of 10.4 percent.32 If they chose to exercise some of the built-in options—for example, requesting payment from the fleets while forfeiting the extra months of interest—the returns could have climbed to 11.7 percent per year. The bankruptcy decree of November 1596 came once the bankers had disbursed the entire principal, but had not yet received a single repayment. In terms of timing, this is the worst scenario that bankers could find themselves in. The settlement of 1597 gave the bankers juros worth 80 percent of the outstanding debt. The promised profits evaporated. Evaluated at its terminal date of March 1600, considering the reduction in principal and yield of the juros, the operation resulted in a loss of 1.08 percent annually.33
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 24