A T N A L T A of How Amsterdam Got Fiat Money Stephen Quinn and William Roberds K Working Paper 2010-17 N December 2010 A B E V R E S E R L A R E D E WORKING PAPER SERIES F FEDERAL RESERVE BANK of ATLANTA WORKING PAPER SERIES How Amsterdam Got Fiat Money Stephen Quinn and William Roberds Working Paper 2010-17 December 2010 Abstract: We investigate a fiat money system introduced by the Bank of Amsterdam in 1683. Using data from the Amsterdam Municipal Archives, we partially reconstruct changes in the bank’s balance sheet from 1666 through 1702. Our calculations show that the Bank of Amsterdam, founded in 1609, was engaged in two archetypal central bank activities—lending and open market operations—both before and after its adoption of a fiat standard. After 1683, the bank was able to conduct more regular and aggressive policy interventions, from a virtually nonexistent capital base. The bank’s successful experimentation with a fiat standard foreshadows later developments in the history of central banking. JEL classification: E42, E58, N13 Key words: Bank of Amsterdam, fiat money, commodity money, monetary policy, credit policy The authors thank John McCusker for sharing agio data, Lodwijk Petram for sharing the Deutz folios, and Albert Scheffers for help with the balance books. Also, for comments on earlier drafts the authors are grateful to Christiaan van Bochove, Pit Dehing, Marc Flandreau, Oscar Gelderblom, Joost Jonker, and Charles Sawyer as well as participants in seminars at the University of Alabama, the Bank of Canada, the Federal Reserve Banks of Chicago and New York, and Rutgers University. They are also indebted to Michelle Sloan for many hours of skilled data encoding. Generous research and travel support was provided by the Federal Reserve Bank of Atlanta and Texas Christian University. The views expressed here are the authors’ and not necessarily those of the Federal Reserve Bank of Atlanta or the Federal Reserve System. Any remaining errors are the authors’ responsibility. Please address questions regarding content to William Roberds, Research Department, Federal Reserve Bank of Atlanta, 1000 Peachtree Street, N.E., Atlanta, GA 30309-4470, 404-498-8970, [email protected], or Stephen Quinn, Department of Economics, Box 298510, Texas Christian University, Forth Worth, TX 76129, 817-257-6234, [email protected]. Federal Reserve Bank of Atlanta working papers, including revised versions, are available on the Atlanta Fed’s Web site at frbatlanta.org/pubs/WP/. Use the WebScriber Service at frbatlanta.org to receive e-mail notifications about new papers. This version: December 7, 2010 1 How Amsterdam got Fiat Money Stephen Quinn, Texas Christian University William Roberds, Federal Reserve Bank of Atlanta Abstract: We investigate a fiat money system introduced by the Bank of Amsterdam in 1683. Using data from the Amsterdam Municipal Archives, we partially reconstruct changes in the bank’s balance sheet from 1666 through 1702. Our calculations show that the Bank of Amster- dam, founded in 1609, was engaged in two archetypal central bank activities—lending and open market operations—both before and after its adoption of a fiat standard. After 1683, the bank was able to conduct more regular and aggressive policy interventions, from a virtually nonexis- tent capital base. The bank’s successful experimentation with a fiat standard foreshadows later developments in the history of central banking. 1 The authors would like to thank John McCusker for sharing agio data, Lodwijk Petram for sharing the Deutz fo- lios, and Albert Scheffers for help with the balance books. Also, for comments on earlier drafts we are grateful to Christiaan van Bochove, Pit Dehing, Marc Flandreau, Oscar Gelderblom, Joost Jonker, and Charles Sawyer, as well as participants in seminars at the University of Alabama, the Bank of Canada, the Federal Reserve Banks of Chicago and New York, and Rutgers University. We are also indebted to Michelle Sloan for many hours of skilled data en- coding. Generous research and travel support was provided by the FRB Atlanta and TCU. 1. Introduction Financial innovation consists of doing more (trading) with less (collateral). A key innova- tion, present in all modern economies, is the use of fiat money—a kind of virtual collateral whose value derives only from the force of law and custom. Conventional wisdom says that fiat money can enhance liquidity through “credit policy”—the directed relaxation of collateral constraints through a central bank’s lending operations, and through “monetary policy”—the beneficial ma- nipulation of economic aggregates through variation of the money stock.2 Fiat money, and its implications for policy, are usually seen as the twentieth-century devel- opments. This paper analyzes an earlier and less well known experiment with fiat money, under- taken by the Bank of Amsterdam (Amsterdamsche Wisselbank, henceforth AWB or simply “bank”). The Amsterdam experience with fiat money is noteworthy for its originality, its promi- nence in European financial history, and its compatibility with price stability over a long period (roughly a century: 1680 through 1780). The AWB opened in 1609 as a municipal exchange bank, an institution for facilitating settlement that was common in Early Modern Europe. Our focus is on the period around 1683 when the bank limited its depositors’ ability to withdraw coin, and so effectively became a fiat money provider. The fiat money regime remained in place until the bank’s collapse in 1795.3 The AWB’s transition from exchange bank to fiat bank has been described by economic historians (e.g., Mees 1838, van Dillen 1934, Neal 2000, Gillard 2004, van Nieuwkerk 2009), but these contributions do not fully explain the motivation for the transition. If fiat money did indeed lower and smooth the costs of collateral in Amsterdam markets, how were these changes mani- 2 In its pure form credit policy does not change the stock of money; see e.g., King and Goodfriend (1988). 3 The bank was not fully dissolved until 1819. 2 fested and who benefited? To lapse into modern terminology, how did an early central bank alter its monetary and credit policies after limiting the right of withdrawal? To shed some light, we examine historical data on the AWB. Using ledgers available from the Amsterdam Municipal Archives, we have compiled partial balance sheets, at a daily fre- quency, for the AWB from 1666 through 1702, a period centered on the fiat money transition. When combined with information from other sources, these data present a revealing picture of the bank’s activities. First, the data clearly show that the fiat money regime facilitated the AWB’s lending to a preferred customer, the Dutch East India Company (Vereenigde Ostindische Compagnie or VOC, a government-sponsored enterprise employing approximately 50,000 people during our period of interest). The bank lent to the Company both before and after 1683; but afterward this lending becomes more seasonal and regular in nature. Seasonality means that this lending often does not show up in the annual AWB balance sheets assembled by van Dillen (1925) nor in the annual balance sheets of the VOC assembled by de Korte (1984). Lending was cheaper and less risky for the AWB after 1683 because liquid claims on the bank were limited and chances of a run were ameliorated. Lending activities were extensive but, over the period considered, never exposed the bank to substantial credit risk. We find that the 1683 changes also freed the City of Amsterdam to frequently take the bank’s retained earnings from this profitable activity. Secondly, our analysis indicates that both before and after 1683, the AWB regularly en- gaged in open market operations. Again, however, the character of this intervention evolves un- der the fiat regime, as the bank more often chose to “drain funds” by selling off its metal stock. Indirect evidence suggests that an objective of these operations was to smooth short-term fluc- tuations in the stock of base money. 3 To summarize, the data we analyze show that by the time of 1683 transition, the AWB managers had ample experience with both lending and open market operations. The move to fiat money simply allowed for more vigorous pursuit of these same activities. The markets seem to have applauded the change: following the 1683 reorganization, there was widespread agreement that trading had been enhanced by this new, if puzzling, kind of money. Writing in 1767, James Denham-Steuart offered the following explanation: The bank of Amsterdam pays none in either gold or silver coin, or bullion; conse- quently it cannot be said, that the florin banco [bank money] is attached to the met- als. What is it then which determines its value? I answer, That which it can bring; and what it can bring when turned into gold or silver, shows the proportion of the metals to every other commodity whatsoever at that time: such and such only is the nature of an invariable scale.4 The rest of the paper is organized as follows. Section 2 sets the historical stage for the 1683 policy change. Section 3 describes and presents the data. Section 4 offers some interpretations of the data. Section 5 discusses related literature, and Section 6 concludes. 2. Historical prologue For Amsterdam, the original purpose of its exchange bank was to protect commercial credi- tors from the unreliable commodity money in general circulation. Modest debasement and resul- tant inflation was ubiquitous in the Early Modern Netherlands, so the AWB was to be an island of debt settlement backed by high-quality coins (Quinn and Roberds 2009b). To support settle- ment, the bank needed to attract metal deposits, get debtors to internally transfer payments to creditors, and deliver out metal of an assured quality. The Dutch chose to follow the model of Venice’s Banco di Rialto and make the AWB an exchange bank that provided only payment and 4 (Steuart 1805, 75-76). For another favorable review of the Dutch monetary system see Adam Smith, Wealth of Nations, Book IV, Chapter 3. 4 settlement services (Dehing and ‘t Hart 1997, 45-6).5 With no lending, the bank was to cover operating expenses with fees. Asymmetric rules promoted metal inflows and debt settlement but discouraged metal out- flows. On the accommodating side, the AWB had no fees on deposits or internal transfers.6 Also, one could present the AWB with precious metal in any form. If a coin had a price assigned by statute, then the bank honored that price. Metal in other forms was valued by precious metal con- tent. And once created, a balance could settle a debt through transfer to the creditor’s account. Creditors gained finality and a trusted general collateral claim. Similar to modern large-value payment systems (e.g., Fedwire), the AWB created finality through gross settlement, meaning that the bank payments could credibly be viewed as final because the bank avoided extending credit and never (explicitly) adopted netting of payments.7 Withdrawals, in contrast, were costly. The bank was obliged to supply high-quality Dutch coins at official prices, but the bank was allowed to charge a withdrawal fee of up to 2 percent for silver coins and 2.5 percent for gold coins, though under normal conditions, fees averaged 1.5 percent or less (Van Dillen 1964a, 348; see also Table 2 below). The fees compensated the bank for minting costs and helped cover operating expenses. Most important to our story, however, is that the fees discouraged withdrawals. Some uncertainty also existed, for the bank had discretion regarding which of those Dutch coins it offered at withdrawal. If a customer desired a different 5 Unlike later central banks, the AWB did not issue circulating banknotes. 6 The bank was permitted to charge transfer fees but chose not to until 1683 (van Dillen 1934, 85). 7 Some qualifications are necessary. The bank cleared payments once every day (Mees 1838, 124-5) so there was in principle scope for multilateral netting at a daily frequency, i.e., the practical seventeenth-century definition of “real- time” gross settlement was probably once per day. Also, an examination of AWB account positions every half year indicates that despite rules to the contrary, some accounts were in an overdraft position during the summer months of peak market activity, particularly before the 1683 transition (Willemsen 2009). 5 coin, then the bank could charge an additional premium based on its role as a moneychanger. Moneychanger fees of some level were necessary to prevent coin-to-coin arbitrage.8 This paper focuses on the consequences of withdrawal structure, yet we stress that the ef- fects of the early AWB’s high withdrawal fees varied by customer. Unlike a modern central bank, anyone could open an account, so customers ranged from foreign merchants to financial intermediaries. Among merchants who routinely operated within the bank’s internal payment system, fees were a negligible concern, for they did not expect to withdraw balances. Of far greater moment to them was that the city of Amsterdam required all large bills of exchange to be settled at the AWB. The requirement created demand for deposits, for bills of exchange were the primary means of commercial credit. The bank’s total balances reached 925,562 guilders after one year (van Dillen 1934, 117), and grew to 8.3 million guilders by 1683, approximately 5 per- cent of the coin stock of the Dutch Republic (De Vries and van der Woude 1997, 90).9 In contrast, customers who did expect to withdraw specie learned to skip the primary ac- count-to-coin process offered by the bank. One could avoid bank fees by paying for coins outside the bank with free transfer inside the bank. Fee avoidance also meant that potential deposit cus- tomers did not bring metal to the bank. By 1650, the outside market in bank balances had deep- ened as private bankers, called cashiers, emerged as dealers who specialized in holding AWB balances and various coins (Van Dillen 1964a, 366-7). The secondary market lived on margins within the bid-ask spread of the AWB’s primary coin-account facility, and the expected costs of the primary market were particularly high for short-term deposits. For example, someone who deposited metal and withdrew it one month later at a 1 percent fee had, in effect, borrowed funds at a simple annualized rate of 12 percent. The 8 Arbitrage is discussed in more detail in Section 4. 9 The guilder, also known as the florin, was the unit of account in the Dutch Republic. At the time of the AWB’s founding, the guilder did not correspond to an actual coin in circulation. 6 AWB was thus an expensive place to “park” specie. Relative costs fell with time, and long-term participants in the Amsterdam payment system, like cashier-bankers, could recoup these “bor- rowing” costs through their secondary market operations. As a result, the short-term metal mar- ket stayed outside the bank, and little metal routinely flowed in or out of the bank. Instead, de- posits waited for periods of cheap metal and withdrawals for expensive metal. 2.1 Lending Lending was the first major deviation from the bank’s original plan. The bank soon began lending to the city, the province of Holland, the Republic, government sponsored entities like the VOC, and select individuals such as mint masters and officers of the Admiralty (Van Dillen 1934, 94-100).10 After a turbulent half century, however, the bank limited new lending to Am- sterdam and the VOC. Table 1 gives the bank’s balance sheet at the end of January 1669. The bank’s metal-to-deposit ratio is 74 percent. While not a reckless position, the bank needed to be mindful of the threat of a run. Table 1. Bank of Amsterdam Balance Sheet as of January 31, 1669 (Millions of Bank Guilders) Assets Liabilities 4.5 Metal 6.1 Deposits 2.1 Loan to Amsterdam 0.2 Loan to Holland 1.1 Loan to VOC 1.8 Capital 7.9 Total 7.9 Total Source: Amsterdam Municipal Archives, 5077-1314. 10 The bank’s lending activities were widely rumored, but the bank did not publicly acknowledge these until much later. See, e.g., Steuart (1805, 403). 7 Indeed, the French invasion of the Dutch Republic triggered a run in June 1672, during which (our calculations find) the bank lost 34 percent of its balances in two weeks.11 Both the Province of Holland and the VOC suspended debt payments, 12 but the bank successfully passed this test, partly because withdrawal fees had kept the large yet volatile short-term specie flows out of the bank. The absence of “hot money” directly reduced the scale of the run and spared the bank the adverse signals produced by the sudden flight of short-term capital. Evidence also suggests that the bank adjusted fees to affect withdrawal rates, for the bank raised fees in 1672 and kept them high for years afterward. Average fees can be estimated from the ratio of the bank’s non-interest revenues as a percentage of total withdrawals from 1666 to 1681; these ratios are reported in Table 2. The calculation is possible because the bank reported its revenue for these years.13 From total revenue, we subtract interest from loans to get a numera- tor that is an imperfect proxy for fee revenue because we do not know the extent of non- withdrawal revenue from sources like overdraft fees, bullion trading, etc., so we cannot explain what loss adjustment created an outlier like the 1676 observation. The denominator we have con- structed from the AWB’s ledgers, and we are missing complete withdrawal information for three of the years. Peering through noise and missing years, fees rose in 1671 as war fears and with- drawals mounted, fees jumped in 1672 with the panic, and fees remained high until at least 1675. 11 On June 14, 1672, the AWB’s total balances were 7.6 million guilders. Balances had fallen to 5.0 million by June 30 with a metal stock at an estimated 4.5 million. 12 For sovereign debt, see Gelderblom and Jonker (2010). For the VOC, see de Korte (1984, 66). 13 After 1683, the AWB reported only profit: revenue less expenses. 8