wall street and main street: the macroeconomic consequences of new york bank suspensions,...
TRANSCRIPT
ORI GIN AL PA PER
Wall Street and Main Street: the macroeconomicconsequences of New York bank suspensions,1866–1914
John A. James • James McAndrews •
David F. Weiman
Received: 17 May 2010 / Accepted: 16 August 2012 / Published online: 31 August 2012
� Springer-Verlag 2012
Abstract Before the formation of the Federal Reserve, banking panics were
routine events in the United States. During the most severe episodes, banks in cities
across the country would often suspend or restrict the par convertibility of their
demand deposit liabilities. In diagnosing the causes of the Great Depression,
Friedman and Schwartz famously regard these local initiatives as a second best
solution, which in the absence of an effective lender of last resort would have
prevented the rash of bank failures during the early 1930s and their dire monetary
and real impacts. Recent research in macroeconomics though has raised the pos-
sibility that banks’ suspension of payments might also have negative real effects
albeit through changes in aggregate supply such as the financing of working capital.
We would expect to observe these negative shocks during the pre-Fed era, because
the decentralized, private interbank payments network was especially vulnerable to
systemic disruptions such as suspensions by New York and other money center
banks. Reports in national trade periodicals and local newspapers during suspension
periods offer many accounts of factories closing because of the inability to obtain
currency for weekly payrolls and ‘‘domestic exchange’’ to finance internal trade. We
corroborate these observations with more systematic econometric evidence at the
national and regional levels. Our results show that controlling for the overall con-
traction and bank failures, suspension periods were associated with a statistically
significant and quantitatively large decline in real activity, on the order of 10–20 %.
J. A. James (&)
Department of Economics, University of Virginia, PO Box 400182,
Charlottesville, VA 22904, USA
e-mail: [email protected]
J. McAndrews
Federal Reserve Bank of New York, New York, NY, USA
D. F. Weiman
Department of Economics, Barnard College, New York, NY, USA
123
Cliometrica (2013) 7:99–130
DOI 10.1007/s11698-012-0083-x
Keywords Financial crises � National banking system � Panics � Payments system �Suspension of payments
JEL Classification E32 � G01 � N11 � N21
1 Introduction
Under the National Banking System (1866–1914), the United States experienced
periodic financial crises, often severe enough to occasion collective action on the
part of the banking system. To relieve pressure on their reserves, in 1873, 1893, and
1907 banks in many or most cities agreed to temporarily suspend or restrict cash
payments at par. During these periods, they limited or denied altogether the par
convertibility of their deposit liabilities into cash, specie, or legal tender notes
(Calomiris and Gorton 1991, 96–100).1 Surveying the pre-1860 experience in his
masterful book on The Ways and Means of Payment, Stephen Colwell (1859, 487)
clearly viewed the ‘‘suspension of payments’’ as the preferred solution to bank
panics. The alternative, paying out specie during such a period, compelled banks
… like Samson in the temple of the Philistines, to pull down the whole fabric
of credit, public and private, about the ears of the people, to disturb and check
the progress of industry in all its departments, to make bankrupts of their
customers, and to sow pauperism broadcast in the field of labor… The triumph
of banks which pass through a season of panic and revulsion without
suspending [is] a triumph like the victory which leaves 100,000 dead bodies
on the field of battle, which makes 10,000 widows, 50,000 orphans, and
200,000 paupers!
More recently and famously, Friedman and Schwartz (1963, 163–168) contrasted
banks’ responses to panics during the national banking era and the Great Depression
under the Federal Reserve System. Like Colwell, they argued that by exercising the
suspension option banks in the earlier period could effectively stop bank runs in
their tracks. ‘‘Restriction of payments thus protected the banking system and gave
time for the immediate panic to wear off, as well as for additional currency to be
made available’’ (1963, 167). To be sure, there were some exceptions to their
generalization that in August 1893 ‘‘restriction of cash payments brought to an end
the stream of bank failures’’ (1963, 110), as some 101 banks closed their doors even
after this date (Wicker 2000, 78–80). Still, the contrast between bank failure rates
during earlier crises with restriction periods and the Great Depression before
Roosevelt’s bank holiday is stark prima facie evidence in favor of their position.
Notwithstanding Dewald’s (1972) attempt to revive traditional arguments against
suspensions (see also Wicker 2000), the tide over the last half-century has run
1 Friedman and Schwartz (1963) preferred the terminology of restrictions of cash payments instead of the
more common contemporary term of suspension to distinguish it from the temporary closure of particular
beleaguered banks. But since both terms are used interchangeably in the literature, we shall do so here as
well. Only remember that by ‘‘suspension’’ we always mean a temporary restriction on par cash
payments, not a complete deferral of bank operations.
100 J. A. James et al.
123
strongly in favor of the Friedman and Schwartz view. Temporary restrictions of cash
payments are now generally regarded to have greatly reduced, if not completely
arrested, the rate of bank failures during banking panics and thereby to have
dampened the rate of monetary and real economic contraction. This argument also
has garnered theoretical support in several formal models such as Gorton (1985a)
based on imperfect information and various extensions of the Diamond and Dybvig
(1983) sequential server framework (Chari 1989; Wallace 1990; Selgin 1993).
Empirically, Dwyer and Hassan (2007) compare the effects of falling prices of
southern and border state bonds in 1861 on banks in Wisconsin and Illinois. In
Wisconsin, where there was suspension, they find failure rates (44 %) and
noteholder losses lower than in Illinois (failure rate of 87 %) where there was not.
While Friedman and Schwartz (1963, 698) considered the restriction period in
1907 as a ‘‘therapeutic measure,’’ they conceded that there were some costs—‘‘short
but brief difficulties’’—as the non-bank public adjusted to the ‘‘use of two only
partly convertible means of payment—currency and deposits.’’ Still, they (1963,
167) insisted that these seemingly dire measures ‘‘could have continued for a much
longer period, as in some earlier episodes, without producing an economic
breakdown and indeed could have continued in conjunction with economic revival.’’
We, by contrast, see a more serious downside effect of banks restrictions on
currency payments. Despite offering temporary relief to local banks, these payments
disruptions could potentially destabilize the entire interbank payments system
centered in New York and in turn temporarily derail the macroeconomy.
The payments system encompasses the complex of financial instruments and
relationships that transfer value (or good funds) between buyers and sellers to
complete their transactions. Cash is the simplest means of payment, as cash
payments directly transfer good funds from buyer to seller and so constitute final
settlement without mediation by a third party. In other words, they bundle in the
same transaction the payments order transmitting the instructions to transfer funds
from buyer to seller and the actual funds transfers themselves. The vast majority of
transactions, especially in developed economies, involve non-cash payments
instruments like checks or credit card receipts. Unlike cash, they represent payment
orders directing the transfer of good funds between intermediaries and ultimately
each party’s transactions account. Banks therefore play a vital role in mediating
payments in a world where most transactions are not based on cash on demand
(COD). Indeed, McAndrews and Roberds (1999, 32) view the linking of buyer and
seller as the ‘‘initial and key role of banks,’’ with their financial intermediation
function linking lender and borrower as a consequence of their payments system
origins. The character and reliability of the payments system therefore is clearly a
crucial support to a market economy, and even local disruptions can have
potentially severe systemic effects on economic transactions and activity.
Modern-day studies of the impacts of panics and financial crises have
concentrated on the transmission channels of changes in interest rates and/or credit
availability (e.g., Mishkin 1996). Recent research in macroeconomics though has
raised the possibility that monetary changes might affect the real economy through
changes in aggregate supply as well as changes in aggregate demand. Contemporary
studies have generally emphasized the impact on investment in working capital and
Wall Street and Main Street 101
123
consequently on the firm’s short-run ability to produce (Barth and Ramey 2002;
Christiano et al. 1997; Christiano and Eichenbaum 1992).
Here, however, we identify two other primary channels that would have worked
through suspensions. First of all, restrictions of cash payments can cause local
‘‘currency famines,’’ as cash was the principal method of payment for wages and
most retail transactions during the pre-Fed period. Second, in non-local payments,
suspensions by banks in nodal locations can have potential network effects resulting
in coordination failures. If banks prefer to synchronize payments outflows with
payments inflows, thereby funding outlays with cash flow receipts (see McAndrews
and Rajan 2000), disruptions or delays in remittances can lead to gridlock in the
payments system.2 Whether through ‘‘currency famines’’ or the ‘‘hemorrhaging’’ of
long-distance payments flows, the resulting aggregate supply dislocations could
have had pronounced real effects.3
Shocks that disrupt payment networks moreover can affect broader measures of
financial liquidity, which today can be observed in bid-ask spreads in financial
markets for example. Empirical evidence shows that various measures of financial
market liquidity in the U.S. Treasury bond as well as in the stock market are reduced
during periods of crises. It also demonstrates that unexpected injections of money
can neutralize these shocks by enhancing liquidity in these markets (Fleming 2003;
Chordia et al. 2004).
Serious payments system disruptions are of course still possible in the modern
economy. For example, the September 11 attacks on the World Trade Center
critically impaired the Fedwire large-value funds transfer system, but gridlock was
averted only through swift and decisive intervention by the Federal Reserve
(McAndrews and Potter 2002). The analogous problem in pre-Federal Reserve
period was not so readily solved, as both local and non-local transactions were
cleared and settled through private payments networks. New York correspondent
banks were central in these networks. They constituted the ‘‘clearing house of the
country’’ in O. M. W. Sprague’s phrase (1910, 126), holding interior banks’ (excess)
clearing and secondary reserves. Consequently, payments restrictions, especially in
New York, could significantly reduce access to ‘‘good funds’’ for local and non-
local payments and thereby impede normal operations of production and exchange
(i.e., reducing the liquidity of banks’ clearing and secondary reserves).
2 This applied to intercity remittances. Within cities, interbank payments were generally accomplished
by net settlement at the end of the day at the clearinghouse. Recent theoretical work also suggests that
delayed settlement in the payment system can exacerbate the resolution of uncertainty and credit risks
which regular settlement is designed to achieve (Koeppl et al. 2005). These disruptions caused by the
banking crises we study delayed settlements for both financial and real transactions. If the counterparties
could not costlessly resort to alternative means of settlement, which we assert they could not do, then real
risks and uncertainty in the economy persisted longer than anticipated, reducing the ability of
counterparties to take on new transactions or to extend additional credit.3 Yet, another channel in modern monetary theory is based on the idea that credit instruments act as a
social memory device for economic transactions (Kocherlakota 2000). If there is a disruption in the
system by which agents in the economy make long-distance monetary transfers, as occurred during the
panics under review, they must turn to less well-suited instruments (such as cash transfers) to accomplish
the information-tracking task that the drafts and checks performed during normal times. This adjustment
constitutes a supply shock as well, because it increases producers’ information costs.
102 J. A. James et al.
123
Our argument is developed in six parts. In Sects. 2 and 3, we provide the
necessary background: a brief description of pre-Fed private payments networks and
chronology of the 1873, 1893, and 1907 panics. Section 4 details the varied impacts
of the 1893 and 1907 banking crises on the local and long-distance payments
systems, and Sects. 5 through 7 offer empirical support for this counterview,
graphically (in Sect. 5) and econometrically at the aggregate and regional levels
(Sects. 6 and 7 respectively). In the conclusion, we offer some thoughts on the
formation of the Fed and its interventions in the payments system.
2 Payments networks
After the demise of the Second Bank of the United States (in 1836), long-distance
payments posed a difficult problem in a country characterized predominately by
independent unit banks with no central monetary authority or integrated nationwide
banking system (see Knodell 1998). Buyers or their intermediaries could have
shipped specie to payees, but a system of intercity payments involving the physical
transfer of cash to settle every transaction would have been both costly and risky.
Due to the prohibition against branch banking, the clearing and collection of long-
distance financial obligations could not be internalized through the formation of
large-scale enterprises like in other key sectors of the economy (Chandler 1977).
Monetary integration, then, depended on the formation of a ‘‘national’’ banking
system to transfer deposits among independent banks without excessive cash
shipments. Instead, banks developed two types of private networks to facilitate
interbank transactions, local clearinghouses, and correspondent relationships.
In the 1850s, banks in New York, Boston, Philadelphia, and Baltimore formed
clearing associations for the collection and clearing of local checks (Cannon 1910;
Gorton and Mullineaux 1987). The spatial gap in the payments system between
local clearinghouses and between city and country banks was filled by correspon-
dent relationships. Earlier on, ‘‘country’’ banks routinely maintained reserve
balances in commercial centers, notably Boston and New York, for the redemption
of their banknote issues (see Myers 1931; Weber 2003), and by mid-century, a
‘‘system’’ of bank correspondents with New York as a national center mediating
interregional payments had begun to emerge (Bodenhorn 2000, 192–198). Based on
longer-term relationships between independent banks, correspondent networks
constituted a novel organizational form, something between tighter urban clearing-
house ‘‘clubs’’ that restricted membership and actively monitored member
operations and arms-length contractual relationships in a competitive market.
As the national ‘‘jobbing’’ center in the wholesale distribution trade, New York
became the national center for clearing and settling interregional payments.
‘‘Tradesmen everywhere have dealings with New York City. There is not a store in
the country which does not receive either directly or indirectly certain of its supplies
from that city’’ (Johnson 1905, 79). New York funds were readily acceptable means
of payment everywhere because so many agents made payments there. Drafts or
other credit instruments drawn on a local bank’s correspondent account and payable
in New York City consequently became the most common medium for settling
Wall Street and Main Street 103
123
debts not just between interior cities and New York, but even between agents in
different communities. With the diffusion of deposit banking toward the end of the
nineteenth century, checks drawn on individual accounts began to displace New
York drafts as the preferred payment instrument in intercity trade. Banks still relied
on correspondents to collect checks drawn on ‘‘foreign’’ banks and to settle their
balances. These interbank settlement payments, as before, were generally accom-
plished by the issue of a draft on a financial center, usually a New York bank.
In the course of providing routine payment services to business customers, banks
would deplete and replenish their correspondent accounts. In addition to buying New
York drafts to make long-distance payments, business customers sold exchange to
their banks by depositing drafts or check payments drawn on a New York (or other
money center) bank. Banks would remit these items to their correspondent for
collection and receive payment usually in the form of ledger entries to their
correspondent account, rather than shipments of cash. In turn, at any point in time,
they could find themselves with deficient or excess funds in their correspondent
accounts. To remedy these imbalances, banks could arrange to ship cash to or from
their correspondents, but would then incur significant transactions costs and risks.
As an often cheaper alternative way to manage their portfolio of clearing reserves,
local banks developed a ‘‘wholesale’’ or interbank market where they bought and sold
surplus correspondent balances. In normal times, they would, for example, purchase a
New York deposit with vault cash and simply convert one form of excess (or clearing)
reserve into another. The price of a New York deposit or the New York exchange rate
in terms of local money was in principle fixed at one, since both locations were on the
same (gold) dollar standard and New York banks pledged to redeem their deposits for
legal tender money at par on demand. Depending on local demand–supply conditions,
the spot price of New York exchange in, for example, Chicago could deviate from
mint parity but only within bounds set by the currency points, the cost of transferring
good funds from Chicago to New York or vice versa. Analogous to the operation of
the fixed exchange rate gold standard, any larger deviation in domestic exchange rates
would elicit interregional/intercity currency flows.
Continuing our analogy to the gold standard, banks’ restriction of payments in
panic periods represented a temporary suspension of the domestic monetary ‘‘rules
of the game’’ (Bordo and Kydland 1995). The New York exchange, now the price of
a New York deposit in terms of a Chicago deposit for example, could vary more
widely, as it depended as well on the currency premiums in the two locations—the
price of legal tender money in terms of a less liquid bank deposit. Consequently,
sharp deviations in New York exchange rates outside of the normal bounds
constitute prima facie evidence of an effective restriction period.
3 The panics of 1873, 1893, and 1907
We offer here the tersest chronologies of the panics and suspensions of 1873, 1893,
and 1907. More detailed accounts may be found in Sprague (1910), Kemmerer
(1911), and Wicker (2000). In 1873, the period of financial turbulence began with
the failure of the New York Warehouse and Security Company on September 8,
104 J. A. James et al.
123
which was ‘‘not regarded at the time as having any general significance’’
(Commercial and Financial Chronicle, September 20, 1873, 382). Conditions
thereafter deteriorated, leading to the hallmark collapse of the (until then) esteemed
private banking house of Jay Cooke & Co. on September 18. Plummeting stock
prices led to the closure of the New York Stock Exchange for 10 days. In response,
the New York Clearing House Association on September 20 authorized the issue of
clearinghouse loan certificates, which was accomplished on Monday, September 22.
Devised in the Panic of 1857, these loan certificates were issued in every
significant financial crisis thereafter (Dunbar 1906, 78–94; Cannon 1910, 75–136;
Myers 1931, 97–102; Timberlake 1984; Gorton 1985b; Roberds 1995).4 Application
was made to the Loan Committee with collateral offered in 1857 of New York state
bank notes, later of specified bank portfolio assets. If the security was deemed
acceptable, loan certificates were issued up to a percentage of the market value of
the securities. They typically carried a 6 % interest charge and could be used to
settle internal clearing balances within the clearinghouse, preventing currency
drains to other local clearinghouse banks. Banks could thereby be freed to pay out
currency to their depositors.5
The panic nevertheless spread quickly down the east coast and into the interior. New
York banks suspended cash payments several days later on Wednesday, September 24,
and they were soon followed by banks in cities across the country (Sprague 1910,
65–66). Typically, suspension was a last resort after loan certificates had already been
issued—only in 1907 did the issue of loan certificates and suspension coincide in New
York. The suspension period lasted just over a month. Particulars are displayed in
Table 1 (Sprague 1910, 35–71; Roberds 1995, 21–23; Wicker 2000, 16–26).
In 1893, the New York stock market collapsed in early May, with the failure of the
National Cordage Trust. Most of early financial disruptions, however, occurred in
interior cities rather than in New York, with panics and bank runs in several interior
cities. These continued through July in such cities as Portland, Kansas City, Denver,
and Louisville (Wicker 2000, 65–77). In view of the resultant cash drains to the interior,
the New York Clearing House on June 15 authorized the issuance of clearinghouse loan
certificates as a precautionary measure. But with continued external drains on reserves,
New York banks restricted cash payments on August 3, strongly limiting but not
completely prohibiting cash payments to liability holders. This decision was followed
immediately by banks throughout the country. The restriction in New York was not,
however, complete, and banks continued to ship cash to some degree to interior banks
4 Loan certificates were issued in 1860 through 1864 in New York; in 1873 in New York, Boston,
Philadelphia, Baltimore, Cincinnati, St. Louis, and New Orleans; in 1879 in New Orleans; in 1884 in New
York; in 1890 in New York, Boston, and Philadelphia; and in 1893 in New York, Boston, Philadelphia,
New Orleans, Buffalo, Pittsburgh, Detroit, Atlanta, and Birmingham (Cannon 1910, 75–116). In 1907,
according to Andrew (1908b), clearinghouses in 41 of the 145 cities surveyed issued these large-
denomination clearinghouse loan certificates.5 The decision to issue loan certificates involved significant conflicts of interest within the clearinghouse,
among other things between considerations of avoiding the effects of panics and resultant decreases in
profits on the one hand and giving aid to commercial rivals on the other (Goodhart 1988, 38–39; Roberds
1995, 19; James and Weiman 2005). In July 1893, for example, 78 % of interbank settlements in the New
York Clearing House were accomplished with loan certificates; in August, it was 95 % (Sprague 1910,
182).
Wall Street and Main Street 105
123
drawing down their bankers’ balances (Sprague 1910, 177–178, 182). The period of
restriction for New York banks lasted around 1 month with resumption there beginning
on September 2. Again, Table 1 summarizes the details.
The 1907 panic, in contrast to 1893, was more similar to those in 1873 and 1884
(though the latter did not result in general suspension and so is not discussed here). The
initial crisis occurred among New York banks and then radiated out to the interior.6 On
October 16 after the collapse of a copper corner by Augustus Heinze and of two of the
brokerage houses involved, runs developed on three banks associated with Heinze.
Assistance from the New York Clearing House preserved the Heinze banks, but the
financial disturbance began in earnest less than a week later with runs on New York
trust companies. While a money pool organized by J. P. Morgan fended off disaster for
the trust companies in the near term, interior bank withdrawals from national banks led
the New York Clearing House to issue clearinghouse loan certificates and suspend cash
payments on October 26. Virtually a nationwide restriction of payments followed the
New York banks’ decision (see Chicago Tribune, October 27, 1907, 1). They did not
resume cash payments until after the first of the year, a period of suspension more than
twice as long as in 1893. Yet again, the chronology appears in Table 1.
4 Payments disruptions
Decisions regarding when to restrict and resume cash payments were generally made
by local clearinghouses.7 In the great majority of cases (of which we know),
Table 1 Panics of 1873, 1893, and 1907
1873 1893 1907
Panic onset Sept 18 May 1 Oct 21
Issue of clearinghouse loan certificates in NYC Sept 22 June 21 Oct 26
Aggregate loan certificate issue, NYCH ($ million) 26.6 41.5 101.0
Maximum amount outstanding ($ million) 22.4 15.2 38.3
Bank reserves of NYCH members ($ million)a 57.1 95.6 121.0
Ratio of maximum certificates issued to reserves (%) 39.2 15.9 31.7
Loan certificate issue nationwide ($ million) – 69.1 238.1
Restriction of cash payments in NYC Sept 24 Aug 3 Oct 26
Resumption of cash payments in NYC Nov 1 Sept 2 Jan 1
Sources: Andrew (1908b, p. 507), Sprague (1910, pp. 34, 145, 163, 261–262, 432–433), Roberds (1995),
Wicker (2000, pp. 9, 121)a On May 6, 1893; Oct 19, 1907
6 Hanes and Rhode (2011) argue, however, that the fundamental roots of financial crises in the United
States under the gold standard lay in fluctuations in the cotton harvest.7 There are some exceptions, of course. In 1907, for example, the governors of Oklahoma, Nevada,
Washington, Oregon, and California declared legal holidays, allowing individual banks to choose whether
to restrict payments. In several mid-western states, bank commissioners and superintendents made it
known that they would not look askance if banks decided to restrict payments; some even suggested it as
an option (Andrew 1908b, 498–500).
106 J. A. James et al.
123
however, they followed the timing of the New York Clearing House. To be sure,
restriction in New York did not instantly result in uniform restriction of payments
across the country. Even in 1907, when restrictions were said to have been the most
widespread, there were pockets of cash payments. Among banks in financial centers,
only those in Washington, DC, did not restrict payments. And according to Andrew’s
survey of 145 independent cities with a population (in 1900) over 25,000, banks in
only 53 (or just over one-third) did not restrict cash payments (1908b, 438–446).8
Just as there was variation in the geographical extent and timing of cash
restrictions, so also did the restrictions practices differ across cities. In New York in
1907, as in earlier episodes, restrictions on cash withdrawals were ‘‘discretionary.’’9
This seems to have been the standard policy in major financial centers including
Boston, Chicago, Philadelphia, St. Louis, and San Francisco. In other cities, explicit
limits were often placed on the magnitude and frequency of cash withdrawals such
as $50 per day or $100 per week in Atlanta, $50 per day in New Orleans, and $200
per customer in Peoria (Andrew 1908b, 440–443).
Such constraints on deposit customers might have had relatively little effect on
local transactions, if agents were content simply to deposit check payments into
their account. During suspensions, banks remained open and performed regular
functions (except paying cash for bank liabilities at par on demand). Checks were
routinely stamped ‘‘payable through the clearing house’’ and credited to individual
accounts. But to the extent that checks were not readily acceptable everywhere and
that recipients were not content simply to deposit them into their accounts, such
limitations most probably had serious and immediate effects.10 Bank deposit
liabilities then did not fully fulfill their medium of exchange role. Rockoff (1993)
argued that this fact alone represented a decrease in the quality-adjusted money
supply and could have disrupted planned spending.
The impact of restrictions in paying out cash would of course have fallen most
heavily on deposit-holders who needed currency to meet obligations. Wages were
typically paid in cash. Indeed, most households at this time did not maintain
checking accounts. Business firms therefore, unable to procure cash to meet
payrolls, were forced to lay off workers and shut down plants. In July 1893,
newspapers published many accounts of factories closing due to failures, inability to
make collections or to obtain credits from banks, but by August, after restrictions
were imposed, the most frequently cited cause had become the inability to procure
8 In Andrew’s ‘‘roll of honor,’’ those with populations greater than 100,000 included Rochester (NY),
Toledo (OH), Worcester (MA), Patterson (NJ), and Scranton (PA). He found only six states (Maine,
Vermont, South Dakota, Montana, Idaho, and Wyoming) in which there was no record of restrictions of
payments, but in several of those, there was no city of greater than 25,000 population and hence no
information on doings there (p. 446).9 Refusals by New York banks to pay out cash for their interior correspondents in 1907 are described in
Senate Document No. 435 (U.S. Senate 1908). As for 1893, ‘‘the majority of New York institutions
continued to pay cash on demand to all depositors, and those which did refuse cash payments not only
offered to such depositors checks on other banks, but cashed small checks without inquiry,’’ but ‘‘the
banks which did shut down on cash payments to depositors included several of the soundest institutions in
the city’’ (Noyes 1894, 26–27).10 Kroszner (2000, 162), au contraire, argued ‘‘the temporary suspension of cash payments in late 1907,
while causing some inconveniences, allowed the banks to continue to provide payments functions….’’
Wall Street and Main Street 107
123
cash to make payrolls (Sprague 1910, 202–203; see also Bradstreet’s, August 12,
1893, 511; Stevens 1894, 137; and Noyes 1894, 28). Speaking of August 1893, the
Commercial and Financial Chronicle (September 16, 1893, 446) observed, ‘‘Never
before has there been such a sudden striking cessation of industrial activity. Nor was
any section of the country exempt from the paralysis: mills, factories, furnaces,
mines nearly everywhere shut down in large numbers, and commerce and enterprise
were arrested in an extraordinary and unprecedented degree.’’ Toward the end of the
month, however, some factories began to reopen and cash payments were restored
in September.
Difficulties in meeting cash payrolls, however, appeared to have been less
pronounced in 1907 than in 1893 (Sprague 1910, 290), even though Andrew judged
the 1907 panic as the more disruptive to local payments.11 ‘‘Probably the most
extensive and prolonged breakdown of the country’s credit mechanism which has
occurred since the establishment of the national banking system… Even during the
critical periods of 1873 and 1893 it is unlikely that as many banks limited the
payment of their obligations in cash’’ (Andrews 1908b, 497). The 1907 panic was
characterized by the extensive issue of local emergency currency or currency
substitutes.12 Banks in many localities issued small-denomination clearinghouse
certificates, obligations of the clearinghouse which could circulate from hand to hand
(as opposed to the traditional large-denomination ones used in interbank settlement);
clearinghouse checks, again typically in small denominations and payable through
the clearinghouse (i.e., not convertible into cash) but drawn on particular banks;
cashier’s checks in convenient denominations which were ‘‘practically circulating
notes’’; New York drafts in denominations of $1 up (in Birmingham); negotiable
certificates of deposit to be used in local payments; and finally, pay checks drawn by
bank customers upon their banks in small denominations and used for payments of
wages (widely used in Pittsburgh) (1908b, 506–512). These currency substitutes,
although of dubious legal status, were not just tolerated but widely embraced.
Andrew (1908b, 516) concluded that they ‘‘worked effectively and doubtless
prevented multitudes of bankruptcies which otherwise would have occurred.’’
Such issues, designed to be held by the public, appeared first in 1893 and
flowered in 1907. They supplemented in local payments the large-value clearing-
house loan certificates described in the previous section, which provided liquidity
for settling internal interbank transactions. Andrew estimated the total volume of all
cash substitutes outstanding during the 1907 restrictions at over $500 million, as
11 Although the Chicago Tribune (November 23, 1907) noted that ‘‘some plants are idle because of the
difficulty experienced in obtaining cash with which to pay employees…’’ Sprague (1910, 71) also
suggested that payroll difficulties in 1873 were less pronounced than in later suspensions because wages
were typically paid monthly rather than weekly. Nevertheless, he cited a number of reports of plant
closures in 1873 as a consequence. In any case, they were relatively short-lived, however (as in 1893),
having essentially disappeared within a month, after mid-October. ‘‘Their place was taken by far more
serious causes of trouble, which were not, however, of a banking nature’’ (pp. 72–74).12 These instruments were employed in 1893 as well, although not to the extent as in 1907. Warner
claimed, for example, that clearinghouse certificates were frequently issued, particularly in the Southeast,
in towns where no clearinghouse existed—‘‘a travesty on the paper after which they were named’’ (1896,
71). He estimated ‘‘vaguely’’ around $80 million of emergency currency put in circulation (p. 73), or
about 8 % of the stock of high-powered money (Friedman and Schwartz 1970, 6).
108 J. A. James et al.
123
compared with a currency stock of $1,810 million in 1907 IV (Friedman and
Schwartz 1970, 65). ‘‘For 2 months or more these devices furnished the principal
means of payment for the greater part of the country, passing almost as freely as
greenbacks or bank-notes from hand to hand’’ (1908b, 515).13
During periods of cash restriction when the convertibility of bank balances into
currency at par could have been limited, a free market developed in New York
(most importantly) and other cities (such as Boston, Philadelphia, and Chicago)
where currency could be purchased with deposits (at a premium, needless to say).14
‘‘No civilized nation has ever experienced such a currency famine [in 1893]’’
(Warner 1896, 71). Demanders of cash included merchants, firms in order to meet
payrolls, and interior banks that needed to meet the withdrawal demands of their
country correspondents.15, 16 The daily currency premia in New York during the
suspensions of 1873, 1893, and 1907 against the number of days elapsed since the
onset of suspension are pictured in Fig. 1. The range between the high and low for
the day is shown for 1873 and 1907, while for 1893 it is the range between the
average daily buy and sell values.17 Although the premium reached 5 % in 1873, in
later suspensions it never rose above 4 %.18 One factor contributing to the day-to-
day variation in the premium was the timing of gold shipments from Europe.19 One
13 The largest component of this total, however, remained the traditional large-value clearinghouse loan
certificates ($238 million as compared with an issue of $69.1 million in 1893) (see Table 1).14 The currency premium apparently applied to all forms of currency, not just gold. The Commercial andFinancial Chronicle (5 August 1893, 196) noted the premium on small note currency, while a week later
it observed that ‘‘all kinds of currency’’ were in demand ‘‘even standard silver dollars’’ (12 August 1893,
232).15 In 1893, the early large purchasers were interior banks. ‘‘Then the premium declined, owing to a
smaller demand which was confined largely to those requiring money for pay-roll purposes,’’ with the
peak on August 19 resulting from the renewal of purchases by banks (Sprague 1910, 188). Andrew
(1908a, 291) similarly attributed the initial currency premium in New York at the end of October 1907 to
Western banks that paid ‘‘a bonus for large blocks of money.’’ These demands were supplemented by
‘‘orders from large mill-owners and manufacturers who were supplied with sufficient money to meet pay-
roll demands.’’16 Hoarding was stimulated by the incipient and actual currency premium. During the panic week in
October 1907 culminating in the declaration of suspension on Saturday, about six times the normal rate of
safe deposit boxes were rented. Similar surges in demand also occurred in Boston, Chicago, St. Louis, and
San Francisco. Interior country banks as well as individuals were enthusiastic hoarders. Their gain in cash
reserves was almost equal to the loss of cash by central reserve city banks over the same period. As
Andrew (1908a, 298) observed, ‘‘The most extraordinary aspect of the matter is that the hoarding of
reserves occurred for the greater part in places where cash payments were suspended by the banks and at a
time when the banks were inundating their respective communities with illegal money substitutes.’’17 Possibly, the reported low and high figures in 1873 and 1907 might have been prices paid by brokers
and retail prices paid by customers, respectively (as in 1893), so the daily bands measure the spread rather
than the range of intraday fluctuations (Sprague 1910, 57).18 However, Andrew (1908a, 290–291) noted that ‘‘even in the dire vicissitudes’’ in France in its war
with Prussia ‘‘the premium paid for coin rose only once to the level of 4 %, a surplus price which was
frequently paid for currency in New York during November, 1907.’’ On the other hand, Noyes (1894, 28)
attributed the fact that the premium never rose ‘‘exorbitantly high’’ to the expectation that early
resumption of cash payments by banks was ‘‘universally expected.’’19 Indeed during the month of suspension, August 1893, over $40 million in gold was imported from
Europe to New York, a sum greater than the cash sent from New York to the interior. At the same time,
European gold was also shipped directly to some other cities, such as Chicago (Sprague 1910, 190–195).
Wall Street and Main Street 109
123
guess put the volume of transactions in 1893 at around $15 million (Noyes 1894,
29).
The issue of currency substitutes may well have mitigated the problems in making
local payments in many cities, but there was no such substitute available to offset the
dislocations of the non-local, interregional payments system.20 Network or coordi-
nation problems among intercity banks interfered with normal settlement procedures.
01
23
45
high
/low
high
/low
0 10 20 30
dayselapsed
perc
ent
perc
ent
perc
ent
1873
01
23
45
sell/
buy
0 10 20 30dayselapsed
1893
01
23
45
0 20 40 60
dayselapsed
1907
Fig. 1 Currency premiums in New York. Sources: Andrew (1908a), Sprague (1910, pp. 57, 187,280–282)
20 A resolution of the Merchants’ Association of New York passed on November 21, 1907, read in part:
‘‘Checks payable ‘through clearing-house only’ are useful for local settlements, but do not pay non-local
debts. The business of all large manufacturing and mercantile concerns is chiefly non-local, and cannot go
on if local funds are everywhere tied up. Interstate exchange is essential to the conduct of interstate
business, and this constitutes the greater part of our domestic exchanges. Provision for the settlement of
local indebtedness is helpful, but provision for the settlement of non-local indebtedness is essential, and,
therefore, still more helpful’’ (Bankers’ Magazine, December 1907, 970).
110 J. A. James et al.
123
With the timeliness and predictability of intercity payments disrupted, the only
alternative was in many cases shipping currency (if available). Most basically, some
banks simply refused to accept or collect out-of-town checks. The New Orleans agent
of Bankers’ Magazine reported in 1873, for instance, the ‘‘refusal of the western and
other banks to receive checks on New York, as is the regular course in the settlement of
collections made here for their account’’ (November 1873, 385).21 Access to New
York balances, the fundamental interregional means of payment and settlement
medium, was similarly disrupted. The Comptroller of the Currency (1907, 70) noted
that ‘‘all domestic exchanges were at once thrown into disorder and the means of
remittance and collection were almost entirely suspended… This [derangement of the
machinery for making collections and remittances] has interfered with every kind and
class of business and led to great curtailment of business operations of every kind.’’22
Domestic exchange rates on New York, which in normal times were bounded by
the cost of shipping currency, rose or fell to dramatic levels, and on a number of
days, the market for New York exchange in interior cities simply dried up
altogether. In 1893, New York funds often sold at extraordinary discounts in mid-
western cities; in Chicago, for example, discounts reached $30 per $1,000 New
York as compared with normal levels of 50¢ or less.23 In 1907, these large discounts
gave way to substantial premiums, peaking at $2.50 per $1000 of New York funds
in the Chicago market. Moreover, the Chicago Tribune reported transactions on
only 1 day over the nearly two-week period from October 23 to November 4, 1907.
The reasons behind this dramatic shift from large discounts to premiums during
the 1893 and 1907 crises are discussed more fully in James et al. (2011). For our
purposes here, we emphasize two salient factors. On the demand side, banks after
the 1893 crisis depended increasingly on their New York correspondents and
exchange for settling customers’ long-distance check payments, but also for vital
securities and foreign exchange transactions (James and Weiman 2010). Their
responses to the 1907 crisis, individually and collectively through the clearinghouse,
disrupted the main sources of supply. Unlike under normal conditions, banks could
not rely on the regular channels of clearing and settlement to replenish their New
York balances. In local clearinghouses, deficit banks were more likely to settle their
accounts in loan certificates than in cash or with a New York draft.24 Without
21 Similarly, in 1893, Louisville banks were reported to have declined to receive country checks ‘‘even
for collection’’ and preferred not to ‘‘handle’’ New York exchange (Bradstreet’s, August 12, 1893, 511).
Or else, it became very expensive to use such payments instruments. In 1893, Sprague (1910, 206)
claimed that ‘‘drafts were often of little or no utility to the holder because the banks refused to take them
except at a ruinous discount or for collection.’’22 Just as was suggested with regard to payroll difficulties, Sprague (1910, 75–77) argued that the
‘‘dislocation’’ of the domestic exchanges had less serious consequences in 1873 than later because
markets had been less integrated geographically. Moreover, the exchanges then while ‘‘deranged for a
time… were never completely blocked [except in Chicago].’’23 The large relative premia of local funds relative to those in New York were not simply the result of the
interior origins of the Panic of 1893. In the Panic of 1873, which did originate in New York rather than in
the interior, domestic exchange rates in Chicago also fell dramatically, reaching levels of -$35.24 As striking evidence of this trend, the Independent Treasury withdrew from the Boston clearinghouse
on 30 October, precisely because it could not accept bank liabilities in the form of loan certificates in
settling its accounts (Boston Globe, 31 October 1907, 7).
Wall Street and Main Street 111
123
recourse to an emergency currency in long-distance payments, banks resorted
to even more dire measures; they effectively ignored collection notices and
either delayed payment or just refused to settle in New York funds (Sprague
1910, 296–298).25 On the flip side, if banks in, for example, Chicago could not
expect distant counterparties to settle their payments obligations promptly, then
they would likely respond in kind and transmit the liquidity shortage to other
centers.
Facing these supply uncertainties, banks were reluctant to sell their excess New
York funds in local markets even at a premium and to run the risk of future costly
shortfalls. Their hoarding of New York exchange, of course, only reinforced this
classic coordination failure by diminishing the liquidity of domestic exchange
markets—the marginal source of New York funds in commercial centers. While
Sprague (1910, 294–296) blamed banks for their short-sighted reactions, he did not
fully appreciate their dilemma. Without more systemic interventions in this case by
the U.S. Treasury, they had little recourse other than ‘‘to strengthen themselves
regardless of the consequences.’’26
Such interferences with long-distance payments and hence with the smooth
functioning of the payments system should have had an adverse effect on internal
trade. Bradstreet’s in 1893 noted ‘‘the clog to trade shown by prohibitive rates for
New York exchange at centers east, west, and northwest’’ (August 5, 1893, 495).
The Commercial and Financial Chronicle (September 16, 1893, 446) seconded that
observation:
the derangement of our financial machinery, which made it almost impossible
to obtain loans or sell domestic exchange, and which put money to a premium
over checks, had the effect of stopping the wheels of industry and of
contracting production and consumption within the narrowest limits, so that
our internal trade was reduced to very small proportions—in fact was brought
almost to a standstill—and hundreds of thousands of men thrown out of
employment.
Similarly, the Wall Street Journal observed in 1907 the ‘‘disorganization of
domestic exchanges … prevents the free movement of commodities for export’’
(December 2, 1907, 8).
25 According to Sprague (1910, 297–298), banks also refused to grant customers immediate credit for the
deposit of ‘‘foreign,’’ that is non-local, checks and instead made them wait for the funds until the items
were cleared and settled.26 An agent reported to Bradstreet’s from Philadelphia: ‘‘The scarcity and high rate of exchange on New
York has no doubt militated against the customary prompt settlements with that city, the banks for the
reason named being unwilling to part with their currency’’ (29 July 1893, 480). Similarly, Noyes (1894,
26) observed that in 1893 even before formal cash restrictions ‘‘country banks were charged with refusing
to remit their cash collections… The express companies did a very large business, during the panic, in
presenting out-of-town checks at the banks on which they were drawn, and bringing the money to the city
bank whence the check was remitted. The out-of-town banks frequently resisted this by paying in silver
dollars or fractional coin. Domestic exchange between two great Eastern cities was at one time fixed by
the express charges for transporting silver dollars.’’ And as well ‘‘banks in some larger cities were next
accused of withholding similar remittances.’’
112 J. A. James et al.
123
The grain trade seemed to have been particularly affected. ‘‘Naturally, there has
been some dislocation of the nation’s business, notably in domestic exchanges,
which has reacted on the collecting and forwarding forces by a time stopping the
buying of wheat in the Northwest and of cotton at the South’’ (Bradstreet’s,
November 2, 1907, 698). By mid-month, ‘‘one especially hopeful sign has been
renewal of grain purchases in the Northwest, exchange checks on larger interior
markets being the medium of exchange, thus allowing of the resumption of grain
forwarding…’’ (November 16, 1907, 730).27 In the flour trade, ‘‘shipments [were]
falling off by reason of difficulty in financing drafts, and the wheat price is now
secondary to the question of finance’’ (Bradstreet’s, November 23, 1907, 747).
5 Graphical analysis
It is, of course, impossible to isolate precisely the real impact of payments
disruptions from the concurrent effects of the credit contraction during panic and
cash restriction periods.28 But we shall try. In this section, we present some
graphical evidence; in the next two, more systematic econometric results at the
national and regional levels. In 1893 panic, payments disruptions preceded general
cash restrictions in several interior cities; in 1907 panic, cash restrictions were only
days apart (October 21 and 26, respectively). That said, the economy did seem to go
into a tailspin during periods of restrictions of cash payments. These episodes are
quite clear in the data.
Figure 2 presents monthly figures for three measures of trade and economic
activity before, during, and after the 1893 and 1907 crises—railroad freight ton-
miles, pig iron production, and the Babson (physical quantity) index of business
activity (U.S. Bureau of the Census 1949, 332–334; Moore 1961, 130). The Miron–
Romer index of industrial production (1990) is shown in Fig. 3. The NBER-dated
business cycle peaks came in January 1893 and May 1907 (Burns and Mitchell
1946, 78). Clearly the graphs show little indication of a serious downturn in either
case until the onset of the financial panics in May and particularly June 1893 and
October 1907, after which time the indicators fall sharply. The months bracketing
the restriction periods—July and September in 1893, October and January (1908)
for 1907—are marked by vertical lines.
The pronounced declines in the series during the cash restrictions period in 1907
are evident. In 1893, the sharp decline begins in June but accelerates with cash
restrictions after July, although this is less evident to the untrained eye. Moreover,
rather dramatically in virtually every case, the decline stops and/or the series turn up
with the resumption of cash payments in September 1893 and January 1908. Sprague
(1910, 201–202) observed for the 1893 crisis: ‘‘Much of the decline in August, with
27 Disruptions in the foreign exchange market had similar effects. In 1873, the Chicago Tribune reported
that ‘‘the shipping movement was partially paralyzed by the news from New York that sterling exchange
was unnegotiable.’’ Shipments of wheat to Atlantic ports were interrupted and grain elevators became
‘‘crowded to their utmost capacity’’ (September 20 and September 25, 1873).28 Between the May 4 and October 4 call dates in 1893, loans of national banks fell by almost 15 %
(Sprague 1910, 208).
Wall Street and Main Street 113
123
freighttonmiles billions ofton-miles, seasonally adjusted
Fre
ight
ton-
mile
s, 1
893
date
01/9
305
/93
07/9
308
/93
09/9
312
/93
7
7.58
8.5
pigironproductiondaily average in thousands of gross
Pig
iron
pro
duct
ion,
189
3da
te01
/93
05/9
307
/93
08/9
309
/93
12/9
3
10152025
babsonindex1923-27=100
babsonindex1923-27=100
Bab
son
inde
x of
bus
ines
s ac
tivity
, 189
3da
te
01/9
305
/93
07/9
308
/93
09/9
312
/93
202530
freighttonmiles billions ofton-miles, seasonally adjusted
Fre
ight
ton-
mile
s, 1
907
date
01/0
705
/07
10/0
701
/08
16182022
Pig
iron
pro
duct
ion,
190
7da
te01
/07
05/0
710
/07
01/0
8
304050607080
Bab
son
inde
x of
bus
ines
s ac
tivity
, 190
7da
te01
/07
05/0
710
/07
01/0
8
405060
tons, seasonally adjusted
pigironproductiondaily average in thousands of gross
tons,seasonally adjuste
Fig
.2
Mo
nth
lyin
dic
ato
rso
fec
on
om
icac
tiv
ity
in1
89
3an
d1
90
7.
So
urc
es:
See
tex
t
114 J. A. James et al.
123
the subsequent partial recovery, can only be ascribed to the trade paralysis produced
by the financial situation at that time.’’ Of course, the economy only stabilized rather
than bouncing back after resumption, but ‘‘after the beginning of September the
course of the crisis of 1893 was no longer a banking affair’’ (p. 209).
For most of the data series, the declines over the periods of cash restrictions in
1893 and 1907 are quite comparable. The Miron–Romer industrial production
index, for example, fell by 16.7 % between July and September in 1893 and by
21.7 % between October 1907 and January 1908.29 The period of cash restriction
was twice as long in 1907 as in 1893, so if one thinks of the effects of restriction as
being a continuing process—the difficulty of obtaining cash to meet local payrolls
or of making payments at a distance—then the per month real effects of restriction
in 1907 were less severe than in 1893, even though the cash restrictions in 1907
have been characterized as both more widespread and more severe than in 1893
miro
nrom
erip
inde
x19
09 =
100
miro
nrom
erip
inde
x19
09 =
100
Miron-Romer Industrial Production lndex, 1893
Miron-Romer Industrial Production lndex,1907
date01/93 07/93 08/93 09/93 12/93
40
45
50
55
60
date01/07 05/07 10/07 01/08
70
80
90
100
110
Fig. 3 Miron–Romer index of industrial production in 1893 and 1907. Source: Miron and Romer (1990)
29 Freight ton-miles fell by 10.8 % between July and September 1893 and by 10.4 % between October
1907 and January 1908. The declines in pig iron production were 42.2 and 52.8 %, respectively, while in
the Babson index, they were 12 and 18 %.
Wall Street and Main Street 115
123
(Andrew 1908b, 497). There are a couple of reasons why this may have been so.
First, of course, as has been noted, the issue and use of local cash substitutes was
much more extensive in 1907. Second, by around the turn of the century, checks had
become standard payments instruments in interregional transactions (Kinley 1910).
To the extent that recipients were willing simply to deposit checks received in their
accounts, rather than attempting to cash them, and in turn use the proceeds to issue
checks of their own, there would not have been a total collapse of the payments
system, although collections may have been slower.
6 Econometric analysis I: monthly aggregate data
Consider a standard aggregate demand function (in Eq. 1) in which real output
growth is a function of lagged real output growth (two lags), growth in real high-
powered money (lagged once), and bank failures. The variables y, p, and h are
logarithms of the output measure, the price level (measured by the Warren–Pearson
wholesale price index), and the stock of high-powered money (total currency
outside the Treasury—NBER series 14135).30 Bank failures might have affected
aggregate demand through several channels. In addition to effects on money
demand and supply, there may have been non-monetary effects through increased
costs of credit intermediation (Bernanke 1983). We measure bank failures as the
number of national banks placed in the hands of receivers (e.g., U. S. Comptroller of
the Currency 1894, 266–285). To be sure, this is an imperfect measure. For one
thing, it might include a few banks that were originally placed in receivership but
later restored to solvency. It does not include, however, banks that only temporarily
suspended cash payments. More importantly, from early on in the postbellum
period, the number of non-national bank failures substantially exceeded those of
national banks, but those figures are not available at a monthly frequency. However,
the correlation between annual national bank failures and those of state and private
banks is 0.958 (U.S. Comptroller of the Currency 1931, 3–8). Since national and
other bank failures moved so closely together at an annual frequency, monthly
national bank failures would seem a reasonable proxy for total bank failures.31
Finally, we follow Grossman (1993, 309–310) in allowing for a nonlinear influence
of bank failures. The square of bank failures is included as well as the level to
capture a ‘‘panic effect’’ in which ‘‘some critical mass of failures must be attained
before producing monetary effects.’’
Dyt ¼ a1Dyt�1 þ a2Dyt�2 þ a3ðDht � DptÞ þ a4ðDht�1 � Dpt�1Þ þ a5 failurest
þ a6 failures2t þ a0 þ et ð1Þ
30 All cited NBER series were retrieved from the MacroHistory database at http://data.
nber.org/databases/macrohistory/contents.31 Wicker (2000, 3) distinguished between bank suspensions, a closure that might have been temporary
or permanent, and failure that was permanent. Here we focus on the latter.
116 J. A. James et al.
123
The aggregate supply function (in Eq. 2) is a standard Phillips curve relation in terms
of rates of change. Such a specification is consistent with an expectations-augmented
Phillips curve of either the old or new Keynesian flavor when expected inflation is
usually stable as under a gold-standard regime (Hanes and Rhode 2011, 21; Hanes
1999). A dummy variable for the periods of restrictions of cash payments in New York
is also included to capture the negative supply shock effects as per the discussion in the
introduction. Even though decisions of whether to and when to restrict and resume cash
payments were generally made independently by local clearinghouses, in the great
majority of cases (of which we know) they followed the timing of the New York
Clearing House. Similarly, just as there was significant variation in the geographical
extent and timing of cash restrictions, the scope of restrictions and the issue of cash
substitutes also differed markedly across cities. Such variation in the extent, scope, and
timing of cash restrictions across cities make a binary variable necessarily an imperfect
measure of their overall nationwide implementation. But New York banks clearly
called the tune, and a focus on the timing of events there seems most reasonable. We
assign a value of one to the suspension dummy variable if there is a period of suspension
in New York within the given month.
Dpt � Dpt�1 ¼ b1Dyt þ b2 suspensiont þ b0 þ gt ð2Þ
Solving for the change in output, we may write:
Dyt ¼ c1Dyt�1 þ c2Dyt�2 þ c3 failurest þ c4 failures2t þ c5 suspensiont þ c6Dht
þ c7Dht�1 þ c8Dpt þ c9Dpt�1 þ c0 þ mt
ð3Þ
In the econometric specification, we also include eleven monthly dummies (the
estimated coefficients of which we do not present here for compactness) and also a
variable, month of down turn, which measures the number of elapsed months since
the NBER peak to capture any systematic influence on output growth of the stage of
the downturn. Adding additional lags, say to failures, or changes in high-powered
money or prices, resulted in no notable changes to the tenor of results reported
below. The estimated coefficients of the additional lagged variables were not sta-
tistically significant themselves and had little effect on the estimated magnitude and
significance of the suspension variable, so we do not report them here. Similarly
adding real federal government expenditures as an exogenous variable had little
effect, and so those results are not presented here either.
The equation is estimated by instrumental variables. The basic instruments are
two lags of the following variables: Dy, Dp, suspension, failures, failures2, monthly
dummies, month of down turn. From January 1, 1879, the United States was on a de
facto gold-standard regime (but not de jure until 1900). So in column 1 of Tables 2,
3, and 4, Dh is taken as endogenous. Additional instruments used in those cases
include two lags of Dh, contemporaneous values and two lags of changes in British
price level (Craighead 2010; Klovland 1993; NBER series m4053) and of the
British open-market discount rate (NBER series 13016). Alternatively, following
Grossman (1993), in columns 2 we take Dh as exogenous. Instruments there would
also include contemporaneous and two lagged values of Dh.
Wall Street and Main Street 117
123
We use three alternative proxies for the change in real output. The first is the
Miron–Romer index of industrial production (1990), which begins in 1884. Those
results are shown in Table 2. Second is the series on railroad freight ton-miles,
which starts in August 1866 and is the closest direct measure of the effects of
suspensions on internal trade (U.S. Bureau of the Census 1949, 334).32 Those results
Table 2 Instrumental variables regressions, Miron–Romer IP index (p values in brackets)
Right-side variables (1) (2)
h endogenous h exogenous
1884 II–1914 XII 1884 II–1914 XII
Dyt-1 -0.34779
[0.000]
-0.35163
[0.000]
Dyt-2 -0.05797
[0.313]
-0.06435
[0.299]
Failures—national banks 0.00937
[0.424]
0.00914
[0.515]
Failures2 -0.00065
[0.529]
-0.00045
[0.714]
Suspension -0.20488
[0.046]
-0.22851
[0.087]
Month of downturn -0.00008
[0.899]
-0.00002
[0.982]
Dpt -0.41791
[0.535]
-0.20335
[0.917]
Dpt-1 -0.02507
[0.936]
-0.02134
[0.942]
Dht 1.10204
[0.517]
0.54803
[0.492]
Dht-1 -1.28317
[0.038]
-1.14609
[0.088]
Constant -0.02448
[0.243]
-0.02655
[0.290]
R2 0.2438 0.2428
Wald v2 97.43
[0.000]
99.14
[0.000]
NOBS 303 303
Instruments: (1) Dyt-1, Dyt-2, failurest-1, failurest-2, failurest-12 , failurest-2
2 , suspensiont-1, suspen-
siont-2, rgbt, rgbt-1, rgbt-2, Dpt-1, Dpt-2, Dpgbt, Dpgbt-1, Dpgbt-2, Dht-1, Dht-2, month of downturnt,
month1t - month11t
(2) Dyt-1, Dyt-2, failurest-1, failurest-2, failurest-12 , failurest-2
2 , suspensiont-1, suspensiont-2, Dpt-1,
Dpt-2, Dht, Dht-1, Dht-2, month of downturnt, month1t - month11t
32 The series is constructed from data on monthly gross real earnings. From 1879 to 1890, it is based on
24–27 railroads accounting for about 50 % of total traffic drawn from Poor’s Manual of Railroads. After
1890, similar techniques were applied to ICC data for all railroads. The data are seasonally adjusted. A
detailed description may be found in U.S. Bureau of the Census (1949, 323–324).
118 J. A. James et al.
123
appear in Table 3. Third is pig iron production beginning in 1877 (U.S. Bureau of
the Census 1949, 332–333),33 used by Grossman (1993) and by Calomiris and
Hubbard (1989). Results are presented in Table 4. The frequency of the data in the
regressions is monthly. Given that periods of cash restrictions were matters of just
Table 3 Instrumental variables regressions, railroad freight ton-miles (p values in brackets)
Right-side variables (1) (2)
h endogenous h exogenous
1879 II–1914 XII 1879 II–1914 XII
Dyt-1 0.28235
[0.000]
0.28868
[0.000]
Dyt-2 -0.30375
[0.000]
-0.29993
[0.000]
Failures—national banks -0.001761
[0.515]
-0.00212
[0.412]
Failures2 0.00019
[0.438]
0.00022
[0.342]
Suspension -0.08219
[0.006]
-0.07148
[0.001]
Month of down turn -0.00018
[0.118]
-0.00017
[0.133]
Dpt 0.17852
[0.304]
0.22905
[0.109]
Dpt-1 -0.08207
[0.300]
-0.06253
[0.366]
Dht 0.27546
[0.641]
-0.02555
[0.861]
Dht-1 0.02209
[0.890]
0.07410
[0.549]
Constant 0.00609
[0.218]
0.00715
[0.278]
R2 0.1712 0.1903
Wald v2 104.13
[0.000]
107.02
[0.000]
NOBS 370 370
Instruments: (1) Dyt-1, Dyt-2, failurest-1, failurest-2, failurest-12 , failurest-2
2 , suspensiont-1, suspen-
siont-2, rgbt, rgbt-1, rgbt-2, Dpt-1, Dpt-2, Dpgbt, Dpgbt-1, Dpgbt-2, Dht-1, Dht-2, month of downturnt,
month1t - month11t
(2) Dyt-1, Dyt-2, failurest-1, failurest-2, failurest-12 , failurest-2
2 , suspensiont-1, suspensiont-2, Dpt-1,
Dpt-2, Dht, Dht-1, Dht-2, month of down turnt, month1t - month11t
33 These data are daily averages in gross tons constructed by dividing figures based on the weekly
capacity of furnaces in blast by the number of calendar days in the month. They are not seasonally
adjusted (U.S. Bureau of the Census 1949, 323). The original data are found in Macaulay (1938, A252–
A270).
Wall Street and Main Street 119
123
weeks or a couple of months at most, the use of high-frequency data in order to
capture their effects is essential. The downside, however, is that this limits the range
of observations because monthly high-powered money data begin only in 1879.
Consequently, the time period covered for the railroad freight ton-mile regressions
(Table 3) and pig iron production (Table 4) runs from February 1879 to the end of
1914. The Miron–Romer regressions (Table 2) run from February 1884 to the end
of 1914. In all of the alternative specifications, suspension is taken as an endogenous
variable.
Table 4 Instrumental variables regressions, pig iron production (p values in brackets)
Right-side variables (1) (2)
h endogenous h exogenous
1879 II–1914 XII 1879 II–1914 XII
Dyt-1 -0.57801
[0.000]
-0.57545
[0.000]
Dyt-2 -0.26107
[0.000]
-0.26388
[0.000]
Failures—national banks -0.06691
[0.125]
-0.06990
[0.095]
Failures2 0.00573
[0.152]
0.00601
[0.118]
Suspension -0.76515
[0.105]
-0.68616
[0.051]
Month of down turn -0.00169
[0.365]
-0.00159
[0.381]
Dpt -2.10334
[0.456]
-1.70560
[0.465]
Dpt-1 -0.36044
[0.780]
-0.20390
[0.857]
Dht 2.18010
[0.817]
-0.12553
[0.958]
Dht-1 -1.39483
[0.589]
-0.99294
[0.624]
Constant 0.06607
[0.425]
0.07565
[0.303]
R2 0.1993 0.2037
Wald v2 136.93
[0.000]
137.68
[0.000]
NOBS 370 370
Instruments: (1) Dyt-1, Dyt-2, failurest-1, failurest-2, failurest-12 , failurest-2
2 , suspensiont-1, suspen-
siont-2, rgbt, rgbt-1, rgbt-2, Dpt-1, Dpt-2, Dpgbt, Dpgbt-1, Dpgbt-2, Dht-1, Dht-2, month of downturnt,
month1t - month11t
(2) Dyt-1, Dyt-2, failurest-1, failurest-2, failurest-12 , failurest-2
2 , suspensiont-1, suspensiont-2, Dpt-1,
Dpt-2, Dht, Dht-1, Dht-2, month of downturnt, month1t - month11t
120 J. A. James et al.
123
In the Tables 2, 3, and 4 results, we see first of all that whether the change in
high-powered money is taken as endogenous (column 1) or as exogenous (column
2) has little effect on the magnitudes or significance levels of other estimated
coefficients. There is a general consistency as well in the pattern of results across
choice of output variable. The lagged output variable is always highly significant,
although the signs differ—negative for the Miron–Romer IP index (Table 2) and pig
iron production (Table 4) but positive for railroad freight ton-miles (Table 3). The
effects of bank failures on the contemporaneous change in output is not significant
for the Miron–Romer IP index and railroad freight ton-miles, but is somewhat so for
pig iron production.34
More importantly here, regardless of the output measure and of the endogeneity/
exogeneity specification of high-powered money, the estimated suspension coef-
ficients are substantial in magnitude and highly significant. Other things equal,
Miron–Romer industrial production is estimated to have fallen about a third during
suspension months; freight ton-miles, a bit less than 10 %; and pig iron production,
a remarkable two-thirds to three-quarters. The real effects of suspension show up
strongly and consistently.
These equations encompass two periods of suspension—1893 and 1907. It
would be nice to go back a bit further, to include 1873 as well, but, as noted,
there are no monthly data on high-powered money available for the earlier
period. In view of the general insensitivity of the other coefficient estimates to
the specification of the high-powered money variable and the low significance
levels of the high-powered money coefficients themselves, we may take a leap of
faith and do without the Dh variable. Table 5 shows the estimation results again
using instrumental variables with railroad freight ton-miles as the dependent
variable over the period from September 1866 to December 1914 (adding twelve
more years to the range). Here, contemporaneous and lagged values of British
price change and interest rates (see above) are explicitly included in the equation
in lieu of Dh. Unlike in Table 3, the bank failures coefficient is statistically
significant and of the expected sign. But similar to those estimates in Table 3,
the estimated coefficient of suspension is of comparable magnitude and strongly
statistically significant.35 Expanding the sample does not weaken the earlier
results.
34 Also using pig iron production as the output measure over this period, Grossman (1993) found aucontraire a strong and statistically significant influence of bank failures. Perhaps this might have been due
to the fact that result is based on quarterly data that might conflate the effects of bank failures and
suspensions (suspensions do not appear explicitly in his specification).
In any case, the message in these results accord with Wicker (2000, 2) who argued that bank failures in
the interior (although qualified by excepting 1893) ‘‘were generally few in number, region-specific, and
too localized geographically to have exerted economy-wide effects… It was the suspension of cash
payments and not bank runs nor bank failures through which the public in the rest of the country
experienced the effects of banking panics.’’35 And this notwithstanding the real effects of suspension in 1873 may have been weaker than in 1893
and 1907. Indeed, Sprague (1910, 77) argued that it ‘‘had relatively little influence on the course of
trade.’’
Wall Street and Main Street 121
123
7 Econometric analysis II: daily regional data
To corroborate the results in the preceding section, we analyze here the real impacts of
New York banks’ convertibility restrictions at the regional level during the 1893 and
1907 crises. To measure real activity, we follow an earlier literature and track the trends
in ‘‘outside’’ bank clearings, that is, the value of checks and related payment
instruments exchanged through clearinghouses in regional centers.36 Because
individual checks and bank drafts were the predominant payments instruments in
wholesale trade during these episodes, their clearing and settlement reflect the current
Table 5 Instrumental variables
regression, railroad freight ton-
miles (p values in brackets)
Right-side variables 1866 IX–1914 XII
Dyt-1 0.16413
[0.006]
Dyt-2 -0.35821
[0.000]
Failures—national banks -0.00667
[0.101]
Failures2 0.00052
[0.165]
Suspension -0.10083
[0.011]
Month of downturn -0.00009
[0.563]
Dpt -0.87155
[0.306]
Dpt-1 0.00685
[0.938]
Dpgbt 0.64490
[0.124]
Dpgbt-1 0.20751
[0.473]
Drgbt 0.00235
[[0.343]
Drgbt-1 -0.00225
[0.344]
Constant 0.01497
[0.088]
R2
Wald v2 78.3
[0.000]
NOBS 516
36 See Persons (1919), Frickey (1925, 1930), and Garvy (1959). They exclude New York bank clearings,
which are skewed by the large value of financial (or ‘‘speculative’’) transactions.
122 J. A. James et al.
123
flows of ‘‘business activity’’ (U.S. Comptroller of the Currency 1894, 1896; Kinley
1895, 1897). Moreover, unlike the aggregate indices, bank clearings were reported at
daily and weekly frequencies, and so can detect more precisely the impacts of the onset
and end of banks’ cash payments restrictions.
At the same time we recognize the limitations in using clearings as a barometer
of real activity. Perhaps most important, they omit on-us check and cash
transactions that bypassed the clearinghouse entirely. Conversely, they could
partially double-count the value of check transactions, as banks often used drafts to
settle their clearinghouse balances.37 These factors would only bias our results if
their share of total clearinghouse transactions changed dramatically over the time
periods covered here. The short durations—3 months for the 1893 and 6 months for
the 1907 case—make this outcome unlikely.
A more serious bias is the possible negative impacts of convertibility
restrictions on the use of bank payments instruments in wholesale trade and
hence on total bank clearings. For example, local clearings, but not real activity,
would drop precipitously if banks’ deposit customers substituted cash or cash
‘‘substitutes’’ for checks in conducting routine business activities during crises
(Andrew 1908b). Businesses might resort to cash on the margins but not for their
large-value transactions that dominate the total value of bank clearings. Not only
would they incur significant transactions and information costs in shifting into a
cash economy, but they also would deplete their cash reserves at the very moment
of a ‘‘currency famine.’’ If anything, bank customers would tend to hoard their
cash withdrawals, which would reinforce the decline caused by banks’ hoarding of
currency reserves (Warner 1896; Andrew 1908a; Sprague 1910). Additionally, as
has been noted, most workers could not afford a deposit account and so were paid
on a cash basis.
Finally, we observe that businesses actually had an incentive to shift their
demand for means of payment from currency to deposits. After all, convertibility
restrictions had no impact on, for example, check transactions that were settled via
ledger entries to banks’ clearinghouse and their customers’ deposit accounts. In fact,
for these payments, banks could use clearinghouse loan certificates to economize on
their scarce currency reserves. Moreover, the proliferation of cash substitutes tended
to funnel payments through clearinghouses. For example, bank customers were
urged to conduct local transactions in conveniently denominated certified checks or
script issues but only if noted ‘‘payable… through the clearing house’’ (quoted in
Andrew 1908b, 501). During restriction periods, in other words, we would expect to
see clearings biased in the opposite direction, that is, against our hypothesis, as
banks and their customers routed normal currency transactions through the banking
system.
Our empirical perspective is decidedly partial, focused on the brief period
embracing the restriction of cash payments. For both 1893 and 1907, there are fairly
consistent daily reports of bank clearings in the largest regional centers—Baltimore,
37 These are the main factors that differentiate bank clearings from the more economically pertinent
variable, debits to deposit accounts (Garvy 1959, 46–55). Still, as Garvy (pp. 69–70) showed, during the
postbellum period, annual ‘‘outside’’ bank clearings and GNP were highly correlated.
Wall Street and Main Street 123
123
Boston, Chicago, New Orleans, Philadelphia, and St. Louis.38 In each case, we
collected data over the restriction episode and equal duration (roughly 1 month in
1893 and 2 months in 1907) prior to and following it. Such a snapshot enables us to
capture the overall trend in clearings during the crisis and the impact, if any, of banks’
decision to restrict deposit convertibility. We also collected daily clearing data for the
same time period in 1892 and 1906, periods of relative normalcy, to test for any
seasonal impacts that could routinely deflate clearings during the restriction months.
The graphs in Figs. 4 and 5 trace out daily clearings in Boston, Chicago, and St.
Louis during the 1893 and 1907 crises. The vertical lines in the middle of each
graph demarcate the restriction period in New York. In both instances, daily
clearings follow a similar trend, declining heading into the restriction period and
then recovering afterward. The recovery is more pronounced in 1893, as shown by
the U-shaped overall trend for all three cities. In the 1907 crisis, by contrast, the
macroeconomy did not fully recover until April 1908, as evidenced by overall
downward trend in clearings from the end of August 1907 to the end of February
1908. Despite differences in the overall trends, clearings in both periods dropped
sharply during the restriction months. To be more precise, we characterize the
impact of the convertibility restrictions in terms of a downward parallel shift in the
overall trend line, that is, a decrease in the vertical intercept.
Fig. 4 Daily bank clearings in Boston, Chicago, and St. Louis, June 29 to September 28, 1893. Sources:See text
38 The figures for 1892 and 1893 come from the local money-securities markets reports in the ChicagoDaily Tribune. For the later years, the data are taken from the ‘‘Bank Clearings’’ section of the Wall StreetJournal and exclude New Orleans.
124 J. A. James et al.
123
To test this generalization, we estimated simple linear regression equations on the
panel of daily bank clearings for the six cities in 1893 and five in 1907.39 The
dependent variable is the logarithm of bank clearings, and the independent variables
specify the overall trend (whether quadratic or linear) and a dummy variable for the
convertibility restriction period. To capture the large differences in clearings across
cities, we also included dummy variables for each location and so effectively
estimated a fixed-effect model (for the quadratic version):
ln(clearingsitÞ ¼ d1 timeþ d2 time2 þ d3 suspensiont þ d4i cityi þ eit ð4ÞBecause clearings were subject both to common national and also more
idiosyncratic regional shocks, we tested and corrected for heteroskedasticity and
cross-equation correlation in the error terms.40
The results presented in Table 6 support our empirical generalization as well as the
results from the aggregate analyses in the preceding section. In 1893, clearings closely
follow the flat U-shaped trend evident in Fig. 4, but drop markedly by an estimated
12.7 % during the suspension period. The regression estimates for 1907 yield similar
results, although the magnitude of the suspension effect depends on the specification
of the overall trend. To bias the case against our hypothesis, we focus on the results
derived from linear trend line specification (equation 1907.2), which downplay the
Fig. 5 Daily bank clearings in Boston, Chicago, and St. Louis, August 31, 1907, to February 28, 1908.Sources: See text
39 The panels for each year cover the same time period, but are unbalanced because of inconsistencies in
data reporting and collection. For this reason, we applied the panel-corrected standard error and not
generalized least squares estimation procedure.40 We also tested for but found no evidence of serial correlation in the error terms for each city.
Wall Street and Main Street 125
123
impact of the surge in clearings immediately preceding the suspension. As is evident
from the graph, clearings tended to fall (by 4.8 % per month) over the entire period, but
dropped by an additional 9.3 % during the suspension period.41
The relevant coefficients in the estimated equations are statistically significant at
the 1 % confidence level. More importantly, they are economically relevant, vividly
capturing the suspension impact on economic activity. To show this, each figure
also includes a graph of the Chicago trend line excluding the restriction effect
(which is set to zero). Ironing out the seasonal and idiosyncratic fluctuations in
clearings, the trend line shows the overall ups and downs of the regional economy
during the crisis, but also the substantial gap in activity caused by the diminished
liquidity of bank deposits. To affirm our interpretation of the suspension coefficient,
we also estimated regressions on the same panel of cities over the same time frame
in 1892 and 1906.42 During those relatively normal times, there was no notable
Table 6 Panel regressions of daily bank clearings during the 1893 and 1907 crises (p values in brackets)
Right-side variables 1893 1907.1 1907.2
June 29 to September 28, 1893 August 31, 1907, to February 28, 1908
Time -0.0139
[0.000]
0.0030
[0.003]
-0.0016
[0.000]
Time2 0.000137
[0.002]
-0.000029
[0.000]
Suspension -0.1355
[0.000]
-0.1710
[0.000]
-0.0978
[0.000]
Baltimore 7.9741
[0.000]
8.4822
[0.000]
8.5882
[0.000]
Boston 9.7246
[0.000]
10.1251
[0.000]
10.2236
[0.000]
Chicago 9.7479
[0.000]
10.5879
[0.000]
10.5796
[0.000]
New Orleans 7.12567
[0.000]
Philadelphia 9.4633
[0.000]
10.0304
[0.000]
10.0232
[0.000]
St. Louis 8.2764
[0.000]
9.2959
[0.000]
9.2887
[0.000]
R2 0.9705 0.9561 0.9539
Wald v2 31844.68
[0.000]
22972.77
[0.000]
22500.91
[0.000]
NOBS 427 683 683
41 Regression estimates based on the quadratic specification yield an inverted-U trend line and imply a
much larger suspension impact of 15.7 %. Strikingly, the range parallels the results from the aggregate
analysis in the preceding section.42 The results are available from the authors upon request.
126 J. A. James et al.
123
seasonal decline in clearings during the suspension period; in other words, we could
not reject the hypothesis that the suspension coefficient was equal to zero.
8 Conclusion
Most contemporary writers (e.g., Noyes 1894; Sprague 1910; and also Walter
Bagehot in Wicker 2000, 130–131), Colwell excepted, thought restrictions of cash
payments were economic disasters. Taking a more nuanced view in his analysis of
the 1893 crisis, Sprague (1910, 200) recognized that suspension ‘‘may have enabled
some banks which might have failed or suspended completely to escape those
misfortunes’’ and that it was ‘‘also possible that more drastic loan contraction would
have been enforced by the banks if they had been exerting every effort to maintain
cash payments.’’ Still, he insisted that suspension was a ‘‘potent factor accentuating
the depression in trade which characterized the month of August’’ because of
difficulties in meeting pay rolls which led to the ‘‘temporary shutting down of many
factories’’ and the ‘‘derangement’’ of the domestic exchanges, which caused ‘‘a
slackening in the movement of commodities and needless delays in collections.’’
Moreover, he concluded, it ‘‘increased the general feeling of distrust which, as
always in a crisis, does so much to bring about greater inactivity than the actual
condition of affairs warrants.’’
While the more recent literature downplays or neglects the effects of these
restrictions, our evidence provides support for the original position. Monthly and
daily data on economic activity in 1893 and 1907 clearly suggest that contractions
intensified during suspension periods. More systematic regression analysis affirms
this impression and quantifies the output loss during restriction periods. We cannot
judge the macroeconomic costs of banks’ restriction on cash payments versus the
alternative, a rash of bank failures. Like Sprague and his contemporaries, though,
our results recommend a search for reforms to the banking system that would
minimize the risks of both.
In fact, the prevention of widespread and severe disruptions of the payments
system in the wake of financial crises was a, if not the, fundamental financial reform
issue to many or most contemporaries and led directly to the establishment of the
Federal Reserve System. A principal feature of the new central bank was the
nationalization of the interbank settlement network. Fed institutions such as the gold
settlement fund and Fedwire (for telegraphic transfers of reserves) replaced their
private analogues, New York correspondent balances, and the Clearing House
Association on the one hand and domestic exchange markets on the other.
The Fed’s takeover of the interbank settlement system was not peculiar to the
U.S. payments system. According to the most recent survey of payments systems by
the Bank for International Settlements (2012), central banks own and operate the
main large-value (interbank) payments network in virtually all developed econo-
mies, either outright or in a partnership arrangement. What distinguishes the U.S.
payments system from those of other countries and remains controversial to this day
was the Fed’s entry into the check clearing system and the relative efficiency of
public and private clearing systems (Stevens 1996; Lacker et al. 1999; Gilbert
Wall Street and Main Street 127
123
2000). However, even today in the face of the increased privatization of the
payments system spurred in large part by the Monetary Control Act of 1980, the
Federal Reserve still plays a potentially crucial role as the clearinghouse of last
resort in financial crises (Summers and Gilbert 1996; James and Weiman 2005).
Acknowledgments The views expressed in the paper are those of the authors and do not necessarily
represent the views of the Federal Reserve Bank of New York or of the Federal Reserve System. Earlier
versions of this paper were presented at the Institute for Monetary and Economic Studies, Bank of Japan,
London School of Economics, Rutgers University, 7th BETA Workshop in Historical Economics. We
thank the participants and Richard Grossman and Christopher Hanes for comments and suggestions but
alas must still accept responsibility for any errors. Weiman acknowledges the financial support of Barnard
College, and research assistance from Keren Baruch, Olivia Benjamin, and Nancy Greenewalt.
References
Andrew AP (1908a) Hoarding in the panic of 1907. Q J Econ 22:290–299
Andrew AP (1908b) Substitutes for cash in the panic of 1907. Q J Econ 22:497–516
Bank for International Settlements, Committee on Payment and Settlement Systems (2012) Statistics on
payment and settlement systems in the CPSS countries—figures for 2010. CPSS Publications No. 99
Bankers’ Magazine (1907) various issues
Barth M, Ramey V (2002) The cost channel of monetary transmission. In: NBER macroeconomics
annual, 2001, pp 199–239
Bernanke B (1983) Nonmonetary effects of the financial crisis in the propagation of the great depression.
Am Econ Rev 73:257–276
Bodenhorn H (2000) A history of banking in antebellum America. Cambridge University Press,
New York
Bordo MD, Kydland FE (1995) The gold standard as a rule: an essay in exploration. Explor Econ Hist
32:423–464
Bradstreet’s (1893, 1907) various issues
Burns AF, Mitchell WC (1946) Measuring business cycles. National Bureau of Economic Research,
New York
Calomiris C, Gorton G (1991) The origins of bank panics: models, facts, and bank regulation. In: Hubbard
RG (ed) Financial markets and financial crises. University of Chicago Press, Chicago, pp 109–173
Calomiris C, Hubbard RG (1989) Price flexibility, credit availability, and economic fluctuations: evidence
from the United States, 1894–1909. Q J Econ 104:429–452
Cannon JG (1910) Clearing houses. National Monetary Commission. Government Printing Office,
Washington, DC
Chandler AD (1977) The visible hand. Belknap Press, Cambridge
Chari VV (1989) Banking without deposit insurance or bank panics: lessons from a model of the U.S.
national banking system. Fed Reserve Bank Minneap Q Rev 13:3–19
Chicago Daily Tribune (1893, 1907) various issues
Chordia T, Sarkar A, Subrahmanyan A (2004) An empirical analysis of stock and bond market liquidity.
Rev Financ Stud 18:85–129
Christiano LJ, Eichenbaum M (1992) Liquidity effects and the monetary transmission mechanism.
Am Econ Rev 82:346–353
Christiano LJ, Eichenbaum M, Evans C (1997) Sticky price and limited participation models of money.
Eur Econ Rev 41:1201–1249
Colwell S (1859) The ways and means of payment. J. B. Lippincott, Philadelphia
Commercial and Financial Chronicle (1893, 1907–1908) various issues
Craighead WD (2010) Across time and regimes: 212 years of the US–UK real exchange rate. Econ
Inquiry 48:951–964
Dewald, WG (1972) The national monetary commission: a look back. J Money Credit Banking 4:930-956
Diamond DW, Dybvig PH (1983) Bank runs, deposit insurance, and liquidity. J Polit Econ 91:401–419
Dunbar CF (1906) Chapters on the theory and history of banking, 2nd edn. G. P. Putnam’s Sons,
New York
128 J. A. James et al.
123
Dwyer GP Jr, Hassan I (2007) Suspension of payments, bank failures, and the nonbank public’s losses.
J Monet Econ 54:565–580
Frickey E (1925) Bank clearings outside New York City, 1875–1914. Rev Econ Stat 7:252–262
Frickey E (1930) A statistical study of bank clearings, 1875–1914. Rev Econ Stat 12:112–138
Friedman M, Schwartz A (1963) A monetary history of the United States, 1867–1960. Princeton
University Press, Princeton
Friedman M, Schwartz A (1970) Monetary statistics of the United States. National Bureau of Economic
Research, New York
Garvy G (1959) Debits and clearing statistics and their use. Board of Governors of the Federal Reserve
System, Washington, DC
Gilbert RA (2000) The advent of the Federal Reserve and the efficiency of the payments system: the
collection of checks, 1915–1930. Explor Econ Hist 37:121–148
Gorton G (1985a) Bank suspension of convertibility. J Monet Econ 15:177–193
Gorton G (1985b) Clearinghouses and the origin of central banking in the United States. J Econ Hist
45:277–283
Gorton G, Mullineaux D (1987) The joint production of confidence: endogenous regulation and
nineteenth century commercial-bank clearinghouses. J Money Credit Banking 19:457–468
Grossman RS (1993) The macroeconomic consequences of bank failures under the national banking
system. Explor Econ Hist 30:294–320
Hanes CL (1999) Degrees of processing and changes in the cyclical behavior of prices in the United
States, 1869–1900. J Money Credit Banking 31:35–53
Hanes CL, Rhode P (2011) Harvests and financial crises in gold-standard America. Unpublished
manuscript
James JA, Weiman DF (2005) Financial clearing systems. In: Nelson RR (ed) The limits of market
organization. Russell Sage Foundation, New York, pp 114–155
James JA, Weiman DF (2010) From drafts to checks: the evolution of correspondent banking networks
and the formation of the modern U.S. payments system, 1850–1914. J Money Credit Banking
42:237–265
James JA, McAndrews J, Weiman DF (2011) Panics and the disruption of private payments networks: the
United States in 1893 and 1907. Unpublished manuscript
Johnson JF (1905) Money and currency. Ginn and Co., Boston
Kemmerer EW (1911) American banks in times of crisis under the national banking system. Proc Acad
Polit Sci City N Y 1:233–253
Kinley D (1895) Credit instruments in retail trade. J Polit Econ 3:203–217
Kinley D (1897) Credit instruments in business transactions. J Polit Econ 5:157–174
Kinley D (1910) The use of credit instruments in payments in the United States. National Monetary
Commission. Government Printing Office, Washington, DC
Klovland JT (1993) Zooming in on Sauerbeck: monthly wholesale prices in Britain, 1845–1890. Explor
Econ Hist 30:195–228
Knodell J (1998) The demise of central banking and the domestic exchanges: evidence from antebellum
Ohio. J Econ Hist 58:714–730
Kocherlakota N (2000) Money is memory. J Econ Theory 81:232–251
Koeppl T, Monnet C, Temzelides T (2005) Mechanism design and payments. 2005 meeting papers 11,
Society for Economic Dynamics
Kroszner RL (2000) Lessons from financial crises: the role of clearinghouses. J Financ Serv Res
18:157–171
Lacker JM, Walker JD, Weinberg JA (1999) The Fed’s entry into check clearing reconsidered. Fed
Reserve Bank Richmond Econ Q 85:1–32
Macaulay FR (1938) The movements of interest rates, bond yields and stock prices in the United States
since 1856. National Bureau of Economic Research, New York
McAndrews J, Potter SM (2002) Liquidity effects and the events of September 11, 2001. Fed Reserve
Bank N Y Econ Rev 8:59–79
McAndrews J, Rajan S (2000) The timing and funding of Fedwire funds transfers. Fed Reserve Bank N Y
Econ Rev 6:17–32
McAndrews J, Roberds W (1999) Payment intermediation and the origins of banking. FRB of New York
Staff Report No. 85
Miron JA, Romer CD (1990) A new monthly index of industrial production, 1884–1940. J Econ Hist
50:321–337
Wall Street and Main Street 129
123
Mishkin FS (1996) The channels of monetary transmission: lessons for monetary policy. NBER Working
Paper 5464
Moore GH (ed) (1961) Business cycle indicators, vol II. Princeton University Press, Princeton
Myers MG (1931) The New York money market, volume 1: origins and development. Columbia
University Press, New York
Noyes AD (1894) The banks and the panic of 1893. Polit Sci Q 9:12–30
Persons WM (1919) An index of general business conditions: I. the index: a statement of results. Rev
Econ Stat 1:111–117
Roberds W (1995) Financial crises and the payments system: lessons from the national banking era. Fed
Reserve Bank Atlanta Econ Rev 80:15–31
Rockoff H (1993) The meaning of money in the great depression. NBER Historical Paper No. 52
Selgin G (1993) In defense of bank suspension. J Financ Serv Res 7:347–364
Sprague OMW (1910) A history of crises under the national banking system. National Monetary
Commission. Government Printing Office, Washington, DC
Stevens AC (1894) Analysis of the phenomena of the panic in the U.S. in 1893. Q J Econ 9:117–145
Stevens EJ (1996) The founders’ intentions: sources of the payments services franchise of the Federal
Reserve Banks. Federal Reserve Bank of Cleveland Working Paper Series, 03-96
Summers BJ, Gilbert RA (1996) Clearing and settlement of U.S. dollar payments: back to the future? Fed
Reserve Bank St. Louis Rev 78:3–27
Timberlake RH Jr (1984) The central banking role of clearinghouse associations. J Money Credit Banking
16:1–15
U.S. Bureau of the Census (1949) Historical statistics of the United States, 1789–1945. Government
Printing Office, Washington, DC
U.S. Comptroller of the Currency (1894, 1896, 1907, 1908, 1931) Annual report. Government Printing
Office, Washington, DC
U.S. Senate (1908) Refusal of national banks in New York City to furnish currency for needs of interior
banks. Senate document No. 435, 60th Congress, 1st Session
Wall Street Journal (1893, 1907) various issues
Wallace N (1990) A banking model in which partial suspension is best. Fed Reserve Bank Minneap Q
Rev 14:11–23
Warner JD (1896) The currency famine of 1893. Sound Curr 2:339–356
Weber WE (2003) Interbank payments relationships in antebellum Pennsylvania. J Monet Econ
50:455–474
Wicker E (2000) Banking panics of the gilded age. Cambridge University Press, Cambridge
130 J. A. James et al.
123