ANNUAL FISCAL ACCOUNTS
SOLVING THE PUZZLE
Before explaining how we reconstruct the fiscal position of Castile on an annual basis, it is useful to review some basic concepts of national accounting. The government’s fiscal balance is defined as the sum of all its revenues minus its expenditures. If the number is positive, the budget is in surplus; if it is negative, it is in deficit. Any deficit must be covered by the issuance of additional debt, while any surplus adds to the government’s assets and hence reduces the net stock of debt. We can therefore write
Government expenditures are divided into ordinary expenditure and debt service, while the stock of debt is composed by the sum of long- and short-term debt:
E now represents ordinary expenditures, while DS stands for debt service. The first two terms of the equation, revenue minus ordinary expenditure, represent the primary surplus, a key element in debt sustainability calculations. When the primary surplus is negative—a primary deficit—the government is borrowing just to be able to pay interest on old debts. The mechanics of compound interest mean that primary deficits can quickly cause the stock of debt to spiral out of control. Positive and large primary surpluses are needed for debt to be sustainable.
For our purposes, it is also useful to disaggregate debt service into long and short term, and ordinary expenditure into military and nonmilitary:
Here DSl and DSs denote long- and short-term debt service, ME is military expenditure, NME is nonmilitary expenditure, and PS is the primary surplus. In chapter 3, we presented series of revenue and short-term debt issues. We added series of military expenditure and short-term debt service in this chapter. We are left with three unknowns: nonmilitary expenditure, long-term debt service, and variations in the stock of long-term debt. To solve for them, we will rely on the following two assumptions:
• Assumption 1: long-term debt service grew smoothly between observations
• Assumption 2: real nonmilitary expenditure was constant throughout the period
Table 4. Juros and their service (in millions of current ducats)
Source: Debt estimates for 1560, 1565, and 1598 are from Artola 1982. The figure for 1575 is from De Carlos Morales 2008. Service estimates are from Ruiz Martín 1965; Ulloa 1977.
† Calculated using 1565 stock of juros
†† Figure from 1596
We begin with long-term debt service. For convenience, we reproduce the table from chapter 3 reporting the scant available data on juros and their service. The table reports juros service for six different years. Following our first assumption, we fill the missing years by interpolating linearly. The assumption is plausible: because of the large stock of juros, the average interest paid on them could not vary abruptly from year to year. The issuance of juros was also capped by ordinary revenue, which grew slowly and gradually. The major exception to this trend was the year 1575, when the Cortes authorized a large increase in ordinary revenue. We have an actual observation for that year, so our procedure still captures the break in the trend. While some measurement error will remain, it will be small relative to the overall size of the budget. This approach, then, gives us a usable series for DSl.
We have data for the stock of outstanding juros for 1565, 1575, and 1598. The latter two dates correspond to the third and fourth bankruptcies of Philip II. Because of the nature of the reschedulings, there was no asiento debt in these years, and so the amount of outstanding juros was equivalent to the total debt. For 1565 we have no hard data on short-term debt. Since the Fugger settlement had not yet been fully negotiated, though, and since the Genoese bankers did not enter the market in earnest until 1566, the amount of any outstanding asientos would have been small. It seems reasonable to assume that asiento debt was negligible relative to juros in 1565 and juros represented almost all the debt outstanding. Because we can guess the total debt outstanding in 1565 and 1598 with reasonable accuracy, we also know how much debt grew between those years: by 43 million ducats.
It is now useful to sum up equation (3) over time:
We just calculated the last term of this expression—the change in total debt over the entire period. Our first assumption gave us the series for long-term debt service, and we have data for revenue, military expenditure, and short-term debt service. All of these can be easily summed up as well. With this in hand, it is straightforward to solve for the sum of nonmilitary expenditure over the whole period. This turns out to be 18.7 million ducats, compared to a total of 146.2 million ducats of military expenditure—equivalent to 11.3 percent of the combined total.
Next we use our second assumption to assign yearly values to nonmilitary expenditure, spreading it out over the whole period while letting its nominal value grow at the rate of inflation. This is again plausible, as most of the expenses on the civil administration and internal law enforcement were fixed. Since nonmilitary expenditure is small compared to the size of the budget, modifying this assumption does not affect our results in any significant way.
Because we estimated military expenditure by adding up the costs of different campaigns relying on a patchwork of sources, one might worry about important outlays being overlooked. In order to allay this concern, it is worth noting that because civil expenditure is calculated as a residual, underestimating military expenditure would simply result in higher values for nonmilitary outlays. Overall expenditure would not change; “missing” military expenditures have no effect on sustainability. That said, our estimate of civil expenditure is low, and hence it is unlikely that we have underestimated military expenditure by a large margin. The government’s budget identity suggests that our series must be capturing virtually all of the outlays associated with Castile’s military campaigns.
We now have annual series for revenue, military expenditure, nonmilitary expenditure, long- and short-term debt service as well as the total stock of debt in the initial and final years. Using equation (3), it is easy to compute a yearly series of changes in the debt stock. Adding them to the initial debt gives a time series of total debt outstanding. This completes our full set of annual fiscal accounts for Castile, which we summarize in table 5.
Table 5. Fiscal accounts, 1566–96 (period averages)
BASIC TRENDS
Before assessing whether Castile’s debt was sustainable, it is useful to discuss the basic trends. Large issuance of short-term debt did not always coincide with major increases in the total debt stock. Nominal debt increased by 40.9 million ducats between 1565 and 1596. Over the same period, the Crown entered into asiento loans for 92.1 million ducats. Thus, on average, around half of asiento borrowing was either rolled over into new short-term loans or consolidated into long-term debt; the rest was used to repay lenders. A large portion of short-term borrowing covered transitory fluctuations in income and expenditure rather than adding to overall indebtedness. As a robustness check, our total debt series closely matches the intermediate estimates for individual years in table 4.
The next key result is that revenues throughout Philip’s reign were markedly higher than military and nonmilitary expenditure combined—that is, the Crown on average ran a primary surplus. Spending excluding debt-servicing costs amounted to 76 percent of revenue in the 1560s and early 1570s, fell to 46 percent in the late 1570s and early 1590s, and then increased to 79 percent. Once we take debt-servicing costs into account, the budget was on average in deficit during Philip’s second and fourth decade on the throne, and in surplus during the third one.
In panel B of table 5, we report period average fiscal accounts in real terms. Castilian economic performance was strong in the third quarter of the sixteenth century. This allowed the Crown to push through a remarkable fiscal expansion. Revenues grew by 41 percent in real terms between 1566–74 and 1575–84. Even more surprisingly, revenues increased by an additional 8 percent during the last decade of Philip’s reign, despite the marked deceleration in Castilian economic performance. Over the entire period, Crown income grew by 53 percent, while nondebt
expenditure increased by 62 percent.
In 1575–84, real military spending had fallen 14 percent relative to 1566–74. Philip earned a “peace dividend” after the successful Battle of Lepanto and the lull in the Dutch Revolt. Castile’s budget swung into surplus as a result, having been in deficit during 1566–74. This surplus gave way to annual deficits of more than 2 million ducats (in 1565 prices) in the period 1585–96. Military spending then more than doubled, driven by the Armada and renewed fighting in the Low Countries. In real terms, Philip’s overall debts rose by 47 percent between the second and fourth decade of his reign—less than the increase in revenues.
Panel C of table 5 shows our fiscal accounts as a percentage of annual revenue. Our choice of scaling variable requires some explanation. Arguably, the right way to measure the burden of military commitments and debt is to scale them by the economy’s total output. Yet estimating sixteenth-century GDP is difficult; the latest published estimates differ by more than 200 percent between their upper and lower bounds (Alvarez Nogal and Prados de la Escosura 2007). In addition to the substantial uncertainty surrounding Castilian GDP, there are reasons to use revenue as a scaling variable. While modern states exert control over large portions of GDP, early modern ones did not. To assess sustainability is to examine potentially available resources for servicing debts; therefore, actual fiscal revenues are a better indicator in the early modern period than GDP. Creditors, for good reason, would have cared more about the Crown’s revenue than about the economy’s total output when assessing a country’s creditworthiness.
Military spending fluctuated strongly from year to year, but its long-term share was broadly stable. The debt burden displayed an overall negative trend. The total debt-servicing cost amounted to 60 percent of revenue in the first decade. This fell to 44 percent in the second one, and rose slightly to 49 percent in the last one—still finishing with a lower value than it started. For the period as a whole, Philip II ran average fiscal deficits of approximately 20 percent of revenue. While the average deficit in the first period amounted to 37 percent, the second one witnessed surpluses of 10 percent of revenue. The decade of the Armada saw a return to deficits of, on average, 28 percent.
Figure 8 shows the primary surplus and fiscal balance side by side. The run-up to the bankruptcy in 1575 and the Armada are associated with primary deficits. After the rescheduling in 1577 and the big tax hike agreed to by the Cortes, surpluses became substantial, varying between 50 and 70 percent of revenue. Lower military expenditure helped with the return to large (primary and overall) surpluses. Similarly, the introduction of the millones in the 1590s improved Castile’s fiscal position. During Philip’s reign as a whole, Castile ran primary surpluses equivalent to 32 percent of revenues. Despite almost continuous warfare, Philip II almost never borrowed to pay interest. A substantial proportion of his revenue was instead available for servicing his debts, year after year. The only exceptions to this were periods of exceptional military effort—the great Dutch offensive of the early 1570s and the Armada.8
War did not only dominate overall spending; it also cast a long shadow over Castile’s fiscal balance. Revenues could fluctuate from year to year, and did so largely as a result of silver windfalls or shortfalls. Debt-servicing costs also fluctuated, depending on the mix of short-and long-term debt along with the financing conditions in each market. Yet the prime determinant of the Crown’s fiscal position was the scale of its military effort.9 Our international comparisons, which we present in chapter 8, further suggest that long-term sustainability and growth were tied more to success on the battlefield than to “responsible” fiscal behavior.
FIGURE 8. Fiscal balance and debt relative to revenue
SUSTAINABILITY
For public debt to be sustainable, revenues and expenditures have to allow all debts to be serviced in perpetuity. Similarly, the ratio of debt to income should not rise above sensible levels, defined by the development of the tax system and public debt administration. Beyond this broad definition, however, there is little theoretical consensus on how to calculate the maximum sustainable debt level or how to establish whether a country is solvent (that is, if its debts, while unsustainable under current parameters, could be paid with appropriate adjustments in the fiscal accounts). The assessment of fiscal sustainability remains largely the domain of practitioners, with significant variation across the major commonly accepted techniques. To evaluate whether Philip II’s debts could be serviced in the long run we first employ a mainstream approach, and then complement it with a number of robustness checks and a counterfactual analysis.
Before we start, it is worth emphasizing one important fact. Between 1566 and 1596, Philip II’s debts did not increase relative to revenue. Taking period averages, they fell from 5.9 times annual revenue in 1566–74 to 4.8 times in 1575–84, before rising to 5.7 times in the final decade. There is therefore no prima facie evidence of a growing fiscal crisis; revenues rose faster than debt. While this observation does not prove that debts were sustainable, it is a first test that the Castilian fiscal data pass with ease.
THE IMF APPROACH
A more systematic approach, favored by the IMF (2003), calculates the primary surpluses necessary to stabilize the debt-to-GDP ratio. The basic idea is that debts will continue to be serviceable as long as the growth rate of debt does not exceed the growth rate of output. This requires a primary surplus—keeping expenditure (net of the cost of debt service) below revenue. Thus, revenue growth combined with cheap borrowing can lead to a favorable outcome even if debts continue to increase.
We explore this approach more rigorously using the debt accumulation equation from Joshua Aizenman and Brian Pinto (2005),
where Δd is the change in the debt as a percentage of GDP, r is the (nominal) rate of interest, g is the growth rate of GDP, and pd is the primary deficit as a percentage of GDP. Equation (5) says that the increase in the debt-to-income ratio equals the current period primary deficit plus the interest on previous period debt, adjusted by the growth rate of the economy. Because sustain-ability requires that debt does not increase as a percentage of GDP, we set Δd to zero and obtain
where ps* is the primary surplus that will hold the debt-to-GDP ratio constant and hence meet the sustainability requirement. Lagging the equation one period, setting dt−1 equal to dt, and assuming that r and g do not change over time yields
where d* is the sustainable debt level relative to GDP. The right-hand side is simply the discounted value of future primary surpluses, where the discount rate is calculated as the difference between the interest and growth rates of the economy. The higher the primary surplus, and the higher the growth rate of income, the larger the debt that can reliably be serviced.
This approach to sustainability assumes that the state can lay claim to a constant share of GDP in the form of taxation. This was definitely not the case in sixteenth-century Castile. As the Crown consolidated its power in the early sixteenth century, its tax revenues grew faster than the economy. Negotiations between the king and Cortes of 1575 and 1591 led to large tax hikes. One may wonder how sustainable a fiscal position can be that depends on government revenues growing faster than the economy at large. Taxing income above subsistence income is markedly easier than taxing incomes below a minimum consumption threshold. As total incomes increased, surplus incomes grew much faster—merely because average income at the beginning of the period was close to subsistence. Fast revenue growth after 1500 simply meant that early modern states laid claim to a high share of rapidly growing “surplus” income (Voigtländer and Voth 2013).
In the case of Castile, revenues from the Indies were also rising rapidly. Overall, in the thirty-one years covered by our data, revenues increased by 53 percent—a growth rate that output cannot possibly have matched. This reinforces our choice of revenue as a scaling variable instead of GDP; lenders surely cared more about the actual income of the Crown rather than a notional upper limit of national production. Given this, we perform
all our sustainability calculations scaling our variables by revenue. We will nonetheless repeat the analysis using various estimates of GDP when discussing robustness.
Table 6 shows our baseline sustainability results, comparing required and actual primary surpluses as well as possible and actual debt levels. The analysis is performed by decade and for the entire period covered by our data. Primary surpluses for the period as a whole were sufficient to keep upward pressure on the debt-to-revenue ratio in check. The primary surplus ps* required to stabilize the debt-to-revenue ratio was 35 percent of revenue, which is only slightly higher than the 31.5 percent actually attained. At the time of Philip’s death, the Crown’s debt in relation to revenue (d) stood where it had been thirty-three years earlier—at a multiple of less than 6. The average sustainable debt (d*) was 5.2 times revenue, and actual levels stood at 5.5 times, which is a minor difference.
Table 6. Sustainability calculations: Baseline results
Note: G is the growth rate of revenue, r is the average interest rate on government debt, ps is the actual primary surplus relative to revenue, ps* is the surplus required for stabilizing the debt-to-revenue ratio, d is actual debt to revenue, and d* is the debt-to-revenue ratio that can be sustained given actual primary surpluses. Growth rates are calculated as annualized compounded rates of growth between benchmark dates. Hence, the overall rate is not equivalent to the weighted average of the growth rates in subperiods.
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 16