The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Federal Reserve Bank of Chicago
Bank Panics, Government Guarantees,
and the Long-Run Size of the Financial
Sector: Evidence from Free-Banking
America
Benjamin Chabot and Charles C. Moul
Winter 2013
WP 2013-03
Bank Panics, Government Guarantees, and the Long-Run Size of the Financial Sector:
Evidence from Free-Banking America
Benjamin Chabot♠
Charles C. Moul♥
Federal Reserve Bank of Chicago
Miami University
Winter 2013
ABSTRACT: Governments often attempt to increase the confidence of financial
market participants by making implicit or explicit guarantees of uncertain
credibility. Confidence in these guarantees presumably alters the size of the
financial sector, but observing the long-run consequences of failed guarantees is
difficult in the modern era. We look to America’s free-banking era and compare
the consequences of a broken guarantee during the Indiana-centered Panic of
1854 to the Panic of 1857 in which guarantees were honored. Our estimates of a
model of endogenous market structure indicate substantial negative long-run
consequences to financial depth when panics cast doubt upon a government’s
ability to honor its guarantees.
JEL: D53, G21, L11, L13, N21
Keywords: banking panics, government guarantees, leverage cycles, endogenous
market structure, economic history
♠Corresponding author. Email: Ben.Chabot@chi.frb.org. The views expressed in this article are those of
the author and do not necessarily reflect those of the Federal Reserve Bank of Chicago or Federal Reserve
system.
♥Email: moulcc@miamioh.edu
1
Bank Panics, Government Guarantees, and the Long-Run Size of the Financial Sector:
Evidence from Free-Banking America
“Indiana presents the anomaly of having been a State organizing
the most admirable system of banking of any State in the Union
and also of having a banking system at one time so vicious that
under it the banks bankrupted nearly the whole people.”
–John Jay Knox, A History of Banking in the United States (1900)
1. Introduction
A bank run is the canonical example of a self-fulfilling prophecy. Bank liabilities are
fixed in value while most assets are illiquid and fluctuate in value. If creditors lose confidence in
the banking system’s ability to convert assets into legal tender on command, even solvent banks
may fall victim to a run. Governments understand the danger of self-fulfilling bank runs and
often make guarantees in order to increase the confidence of financial market participants. The
recent financial crisis has illustrated that government guarantees may not be credible when the
size of the required financial sector bailout dwarfs the sovereign’s ability to raise funds. The
long-term consequences of a broken government guarantee, however, remain unexplored. Do
sovereigns who fail to live up to the market’s ex ante expectations of support quickly regain the
trust of financial markets after the crisis has passed, or does lost reputation doom a sovereign’s
territory to a reduced steady state level of financial development?
Measuring the long-run consequences of broken promises is difficult in part because
government insurance programs and implicit guarantees only fail during panics and a banking
panic itself necessarily leads to changes in the equilibrium size of the financial sector. During
panics banks exit from (and then perhaps enter into) the financial sector until a new steady state
is achieved. Unfortunately, one cannot simply observe a sample of modern panics and compare
the change in steady states across markets with and without broken guarantees. Because modern
banks operate across many jurisdictions, cross-sectional analysis is confounded by international
linkages and the harmonized regulatory response to modern crises. Financial integration has
essentially created one global market, robbing analysts of the cross-sectional variation necessary
2
to identify the causal impact of broken guarantees. Furthermore, it is difficult to disentangle the
long-run impact of panics on government creditability from the regulatory reforms that typically
follow modern banking crises. To properly measure the effect of guarantees on market structure,
one would ideally observe many similar markets that differ only in the ex post credibility of their
guarantees and measure their long-run responses to a common panic. While such an ideal design
does not exist with contemporary globalized banking, the variation in state guarantees during
antebellum America’s free banking era comes remarkably close.
During the free-banking era (1837-63), a still sparsely populated America had many
geographically isolated towns that constituted localized banking markets. Furthermore, the era
saw bank runs, panics, and closures occur with some regularity, but these events tended to be
more localized than they are today. The existence of such markets and banking events can
provide the cross-sectional variation lacking in today’s global capital market. Regulation was
also fragmented. Regulation was a state rather than federal power, and banks were limited to a
single state. Each state adopted regulations designed to assure bank creditors that banknotes
issued by banks within that state’s borders would always be redeemed at par. While no state
explicitly guaranteed redemption at par, the collateral and monitoring regulations amounted to an
implicit guarantee that notes would be redeemed at par as long as the state’s regulatory
authorities fulfilled their statutory requirements and the size of the bank panic did not impair the
state’s sovereign credit. These implicit guarantees were only credible in practice, though, if the
bond market was willing to hold state debt at low yields. Historical circumstances periodically
called these guarantees into question and provide us with examples of broken guarantees.
Finally, regulations were rarely revised in the aftermath of panics, so there is little trouble in
distinguishing the impact of the panic from that of reform. The period thus had many features
ideal for the study of the interaction of government guarantees and the size and structure of the
financial system.
We investigate the consequences of a broken guarantee by comparing the Indianacentered Panic of 1854 with the Panic of 1857 during which regulatory guarantees were honored.
In the latter panic of 1857, many markets across the country endured bank closures, but banknote
holders were made whole after banks were unwound. Despite the system working, it is possible
that confidence in, and therefore demand for, banking decreased. This lack of confidence may
then generate less banking in markets where closures occurred. In the earlier panic of 1854,
3
though, largely idiosyncratic forces revealed flaws in Indiana’s collateral system which led to
instances of banknotes being redeemed at less than par, despite the state’s implicit guarantee.
Indiana’s regulatory reputation was damaged and subsequent potential bank investors may have
lowered their demand for banknotes issued by Indiana banks when deciding where to allocate
resources. This broken guarantee would then compound any reduced banking confidence that
resulted from closures during the panic, spilling over into lower demand for banks located in
Indiana markets that saw no closures but were nonetheless covered by the flawed regulatory
regime.
To measure the impacts of these historical events, we make use of the isolated nature of
banking markets during this period and examine what determines market structure. Our major
results come from a model of banking in which banks enter where it is profitable to do so and
fixed costs limit the number of entrants. Employing Abraham, Gaynor, and Vogt’s (2007)
extension of Bresnahan and Reiss (1990, 1991), we relate the number of banks in a market to
market characteristics. 1 Specifically, we infer the determinants of latent (unobserved) banking
profits by employing an ordered probit to link market characteristics and panic history with the
consequent equilibrium market structure. Such models are now standard in the industrial
organization literature but, for the aforementioned reasons, are difficult if not impossible to apply
meaningfully to the question of panics in modern banking markets. Our formal model enables us
to control for market-level variables and draw conclusions about profitability even though many
components of bank profits cannot be ascertained. 2 Essentially, we recognize that the era’s
potential bankers wished to locate in the most profitable locations and let the observed entry and
exit decisions of antebellum bankers illuminate their implied profits.
Our data come from isolated markets in Indiana and surrounding states in mid-1854
(before that year’s panic) and late 1860 (three years after the ’57 panic). The Indiana-centered
panic in 1854 provides us with an exogenous shock to government credibility, and closures from
the national panic of 1857 offer cases when panics occurred but guarantees were not in question.
Our results suggest that in mid-1854 Indiana was an exceptionally banking-friendly state, in that
Indiana markets had far more banks than the underlying population and other market
1
Cohen and Mazzeo (2007) use related techniques to exploit localized conditions and consider the causes and
impacts of endogenous bank-branching decisions.
2
Measurement issues of profitability have been a recurring problem in past studies of the free banking era. See
Bodenhorn (1995) for an excellent contribution and review of the literature.
4
characteristics would suggest. By late 1860, though, Indiana’s implicit guarantee had been
questioned for six years, and the state was underbanked relative to both its population and other
free-banking states. The difference in banking activity between these six years is both
statistically and economically significant and provides us with a stark example of the interaction
between sovereign credibility and the size of the financial sector.
We explore two mechanisms to explain Indiana’s decline, distinguishing them by
considering how much of the Indiana impact is concentrated in the local markets that
experienced actual bank closures. The first – which we call the Kindleberger hypothesis, after
Charles Kindleberger’s observations of French banking before the Revolution 3 – revolves around
the local depression and loss of confidence that individual creditors experience as a consequence
of personal losses due to bank failures. This explanation applies equally to any market that is
struck by bank failure, regardless of regulatory regime. Psychology and behavioral economic
models of reinforced learning emphasize the importance of personal experience in decision
making. 4 If a town’s bank closed during a panic, that town may be a less profitable location in
steady state if its citizens are subsequently less likely to trust (and hold notes of) local banks than
citizens of towns that did not suffer panics. Our second mechanism focuses on information about
the state’s ability to fulfill its guarantee. Unlike the first explanation, the loss of confidence in the
government guarantee should affect each market within the state rather than only markets that
suffered failures.
When guarantees are honored, we find little evidence that bank failures have long lasting
implications for the steady state level of banking. Markets outside of Indiana that suffered bank
failures in 1857 have insignificantly lower implied steady state profits in 1860. The same cannot
be said for Indiana markets where a closure of a bank during the panic lowers the steady state
level of banking. We attribute these results to Indiana suffering a state-wide reputational cost
from its failure to fulfill its implicit guarantee during the earlier panic of 1854.
The structure of this paper is as follows. We begin with discussions of the regulation of
antebellum banking and of the practice of free banking in Section 2. We then discuss the curious
3
Kindleberger, A Financial History of Western Europe, 1984.
Reinforced learning occurs when personal experiences are given more weight than other sources of information in
the formation of expectations about risk and return (Choi et al., 2009). Osili and Paulson (2009) show that
immigrants to the United States who personally experienced banking panic in their home country are less likely to
trust U.S. banks. See their literature review and cites therein for an excellent summary of the empirical evidence
that individuals over-extrapolate from personal risk and return experience when making banking decisions.
4
5
details of Indiana’s Panic of 1854 and the national Panic of 1857 in Section 3. Section 4 details
our basic model of endogenous market structure. Sections 5 and 6 outline the data with some
basic analysis and then consider the model’s estimates, and we conclude in Section 7.
2. State Regulation, Trust and Profits during the “Free Banking” Era
After the Second Bank of the United States’ charter expired in 1836, banks were
regulated entirely at the individual state level. States had previously granted banking charters on
a bank-by-bank basis via acts of the legislature or discretionary approval by a state appointed
regulator. Beginning with Michigan in 1837, many states adopted laws that permitted individuals
who abided by general banking laws to form limited liability banks with no discretionary
approval. These so called “Free-Banking” statutes allowed for free entry of banks subject to each
state’s regulatory requirements.
The required capital, reserves, and type of security accepted as collateral varied from
state to state. Antebellum regulation was based on the desire to create safe banknotes that could
be used as a transaction medium and store of value and was predicated on the belief that bank
runs could be prevented if noteholders were confident that all losses in bankruptcy would be
borne by equity holders. To that end, banks were required to post collateral securities backing
each banknote in sufficient quantity that no monitoring of individual banks was required by
noteholders. 5
States wished to make their banknote circulations safe enough that they could serve as a
medium of exchange both within and outside the state. After all, a merchant in Cincinnati or
Chicago could not be expected to know the riskiness of each bank whose notes passed through
his business. Regulations therefore required notes to be backed by sufficient transparent
collateral that no due-diligence was necessary on the part of noteholders. Banknotes that were
sufficiently collateralized became what Gorton (2010, 2012) has described as “secretless” or
“information-insensitive” debt – debt with value that was known with such certainty that it was
readily accepted as a medium of exchange.
5
Ng (1988), Bodenhorn (1990), Economopolous and O’Neill (1995), and Dwyer (1996) each survey antebellum
capital and entry requirements.
6
Contemporary commentators referred to the collateral backing note issues as the
“ultimate security” but realized that, while this backing may eliminate the need to monitor
individual bank assets, the need to convert these securities into specie during a panic meant that
banks could be subject to what we refer to today as “wrong-way risk” – collateral that was likely
to lose value in exactly the state of the world in which it was needed.
“This ‘ultimate security’ is, it may be admitted, better than no security at all. The mischief is that it is least
available when most wanted. The very causes which prevent the banks from redeeming their issues
promptly, produce a fall in the value of the stocks and mortgages, on the ‘ultimate security’ of which their
notes have been issued” - The Bankers Magazine (Sept.1857 p.169)
A bank that wished to circulate banknotes first had to use specie to purchase an
acceptable federal or state government bond. Next, the bank deposited the bond with the state
auditor who returned endorsed banknotes that could be legally circulated. The bank was required
to redeem the banknotes in specie on demand, and noteholders were protected against default by
the posted collateral. Should a bank fail to redeem a banknote, the noteholder could protest to the
state auditor who was empowered to sell the posted collateral and redeem the note in gold. As
long as the bond collateral could be sold in the open market for at least the face value of
banknotes, the noteholders were assured redemption at par regardless of the value of the bank’s
other illiquid assets and therefore had no reason to “run” on banks at the first hint of trouble.
With banknotes backed by posted collateral, potential noteholders did not have to monitor
individual banks. Instead, noteholders only had to trust the bank’s home state regulator to
properly monitor the specie reserves and collateral value of posted bonds.
Table 1 summarizes the regulatory differences across select states in our study. The key
differences were the amount, quality and diversity of collateral required to back note issues.
Some states adopted remarkably careful collateral requirements that resulted in well diversified
collateral pools backing their note issue. They did so in large part by granting less than full value
(i.e., applying haircuts) to less than ideal collateral.
Indiana, on the other hand, used its collateral requirements to encourage banks within its
borders to hold domestic state bonds as collateral. This provided Indiana’s government with
cheap funding but also exposed the banks in the state to the danger of undiversified portfolios. 6
6
The implications of collateral value on equilibrium security prices have been examined by a number of authors
including Geanakaplos (2003, 2010), Brunnermeir and Pedersen (2009), and Garleanu and Pedersen (2011).
7
By construction, these portfolios were vulnerable to the “wrong way risk” of collateral that was
likely to default in exactly the state of the world in which it was needed.
Indiana’s collateral requirements were lenient with respect to haircuts and designed in
such a way that they inevitably encouraged the concentration of Indiana bonds in the collateral
portfolios of her free banks. Before 1851 Indiana had issued bonds with coupons of 5% and 2.5%
while virtually every other state bond had coupons of 6% or more. Due to their low coupon rate,
these Indiana bonds were ineligible as collateral in most states and subject to extremely large
haircuts in others. Indiana’s legislature perceived that the low collateral value of Indiana’s
existing bonds was depressing their market price and raising the state’s funding costs. Indiana’s
free banking legislation therefore attempted to provide the state with cheap funding by accepting
Indiana bonds as collateral.
“Thus it will be perceived that this banking law was a shrewd financial measure on behalf of the State [of
Indiana] to enhance their bonds, which only bore five percent interest” - The Bankers Magazine (Sept.1857
p.165-166)
Cross-state differences in regulation resulted in Indiana bonds having a higher collateral value to
Indiana banks than to other potential holders. Indiana banks consequently purchased large
quantities of Indiana bonds as note collateral for their growing circulation, and Indiana bonds
became concentrated in Indiana bank collateral portfolios.
Figure 1 reports the diversification of bank portfolios in three states that published
sufficiently detailed collateral reports before the panics of 1854 and 1857. Indiana banks had far
more concentrated holdings than Illinois or Wisconsin. To make matters worse, Indiana 5%
bonds traded close to par so the value of banknotes in circulation was roughly equivalent to the
market value of posted collateral. This concentration and lack of over-collateralization would
prove disastrous in the events that follow.
2.A. Trust and Profits
The issuance and lending of banknotes would seem a highly lucrative enterprise, but
competitive pressure constrained bank leverage and profits. Banks could only maintain a
circulation of notes by convincing noteholders that their currency was sound. Bank notes traded
in major trade centers, and the price of a bank’s notes served as both a public signal of its
8
soundness and a measure of its potential leverage and profits. Notes could obviously not trade
below par in a city of redemption without creating opportunities for arbitrageurs to buy
discounted notes and immediately redeem them at par. Notes could, however, trade at a discount
outside of redemption cities, provided the discount was smaller than the transaction costs of
returning notes to the bank for redemption. 7 If a bank’s notes were trusted and traded near par in
far away cities, note holders would not find it profitable to pay the transport costs of returning
notes for redemption. As long as a bank was considered sound, its owners could profit from the
leverage created by circulating more notes. 8
Banks leveraged their initial capital of gold by buying an acceptable bond to post as
collateral for state auditor-endorsed banknotes. If the bank’s notes circulated at prices near par,
the bank could use its notes to acquire more bonds and post those bonds as collateral for still
more notes. Each time this process was repeated, the bank earned the coupons on the bonds
posted as collateral. As long as the market trusted the bank and was willing to accept its notes
near par, the owners of a bank could continue to repeat this cycle and leverage their initial
investment. With each purchase of bonds, though, the number of banknotes in circulation
increased, and the bank’s owners risked that the price of notes would fall below the threshold at
which noteholders would find it profitable to return their notes for redemption in gold. Thus the
market price of the notes put a limit on a bank’s leverage. 9 A bank made profits by guarding its
reputation so that its notes would continue to circulate and its owners would earn the coupons on
the collateral bonds backing note circulation.
Given this setting, banks could increase their profits by circulating notes far from home.
A bank with notes accepted far from home had more scope for profitable leverage. Notes
circulated locally were more likely to be redeemed for purely idiosyncratic reasons and required
banks to hold more non-earning specie reserves than notes circulated in distant cities. By
locating in states with reputations for sound banking, banks found their notes in demand in
distant cities, while banks located in states with poor reputations or costly regulations found it
more profitable to borrow notes from banks in sound states than to circulate their own notes. The
7
Excellent descriptions of note circulation and redemption practices during the free banking era can be found in
Rockoff (1975, 1985) and Gorton (1996).
8
Banks could also increase leverage by accepting uncollateralized deposits on which they paid interest, but the
issuance of banknotes was the lower cost option.
9
Our data and the relevant literature suggest that the era’s banks were able to achieve leverage ratios of between two
and three times their capital stock.
9
most notorious example of the second case was Michigan, which earned a reputation for wildcat
banking and a lax regulatory regime in the late 1830s. Its banks’ notes faced larger discounts
well into the 1850s, and many of its banks found it impossible to maintain a circulation far from
home. Indiana, on the other hand, earned a reputation for sound banking during the panics of
1837 and 1841. As a result, Indiana notes traded closer to par on the East Coast before the panic
of 1854, and many Indiana banks were able to maintain a profitable circulation on the East
Coast. 10
Ohio had a reputation for sound regulation as well, but the state taxed local banknote
circulation and adopted unprofitable collateral and specie reserve requirements. As a result, Ohio
banks found it more profitable to borrow and circulate notes from Indiana banks than to maintain
their own circulation. 11 When the auditor of the State of Ohio questioned why an Ohio bank did
not keep up its own circulation, he was told, “It could if it were not for the high rates of taxation.
If taxes were low, a circulation would be a profit.”12
A regulatory arbitrage developed in which banks that were located in states with poor
reputations or costly regulations purchased bonds and loaned them to banks in low-cost, goodreputation states. These banks then re-hypothecated the bonds to their state regulator in return for
banknotes which they loaned to banks in states where note issue was unprofitable. For example,
from a bank examiner’s report we know that on November 2nd, 1853, the Union Bank of
Sandusky Ohio borrowed $50,000 of banknotes from the Steuben County Bank of Indiana and
delivered $33,000 worth of Indiana state bonds and coin for collateral. The Steuben County Bank
promptly re-hypothecated the collateral with the Indiana regulator to replenish its banknote
supply. 13 There is good reason to believe these interstate loans were a common means of
concentrating note issue in regulatory friendly states. Writing about the pre-1854 arbitrage
opportunities, the 1858 Encyclopedia of Commerce noted that “the oppressive tax law of Ohio
10
Eighty-eight Indiana banks were able to maintain a circulation in Philadelphia during the first eight months of
1854. Indiana banknotes traded at an average discount of 1.58% and a range of 1.13-1.75% between Jan. 1 and Aug.
1, 1854. At the same time, only ten Michigan banks issued notes that were accepted at a low enough discount to
maintain a circulation in Philadelphia. These Michigan notes traded at an average discount of 1.82% with a range
between 1-5%. Source: Gorton & Webber
(http://www.minneapolisfed.org/research/economists/wewproj.cfm#discounts)
11
Writing of Ohio banks in 1856, H.F. Baker cites a businessman’s complaint to the Cincinnati city council that in
six years of business he had received but four banknotes circulated by Cincinnati banks. (Huntington, p. 222,
footnote 44)
12
Reemelin (1855), Appendix to annual report of Auditor of State: Series of reports made on the condition of the
Ohio stock banks, as ascertained by Charles Reemelin, Esq., p.14
13
Ibid p. 117.
10
has driven banking capital out of the state, it was at once invested in the free banks of
Indiana”. 14 The 1857 Bankers Magazine further noted that the ease of establishing a bank in
Indiana and the widespread acceptance of Indiana notes “was enough to encourage brokers and
bankers from other States to avail themselves of its facilities for supplying railroads and porkpackers with a convenient currency and receive in payment therefore their bills of exchange on
New York”. 15 Banks with trustworthy notes were only too happy to lend far from home in
exchange for a promise that their notes would be given a “good circulation” and “scattered” in a
way that decreased the probability that the notes would find their way home and thus limit the
need to hold costly specie reserves. 16 And before the panic of 1854, no banks had a better
reputation than banks regulated by Indiana. As a result Indiana’s free and State banks found their
notes accepted across the nation, and those banks employed every means available to increase
their circulation and profits.
“Every possible avenue of circulation was surcharged – every opportunity for substitution and exchange for
other notes improved - brokers were employed to scatter them on rail-roads and steamboats, until the
traveling public found nothing but Indiana notes in their pockets” - The Bankers Magazine (Sept.1857
p.170)
In this setting, a bank panic could alter the long run leverage, profitability and equilibrium
number of banks if it altered the perceived safety of the banknotes in circulation. While each
note was over-collateralized by the collateral on deposit, specie reserves, and other assets, the
fact that banks were leveraged ensured that the aggregate number of notes in circulation
exceeded the value of gold in the economy. When a single bank failed, the regulator could sell
the bond collateral and redeem all notes at par with little trouble, but, if many banks were forced
to convert their bond collateral into gold at once, the equilibrium price of bonds would decline.
The possibility of such an event created the risk that not all noteholders could simultaneously be
made whole. 17 A state’s potential for systemic risk was thus dictated by the extent to which that
14
Homans’ (1858 p. 137)
Bankers Magazine Sept.1857 p.166
16
See quotes from letters examined by Reemelin (1855), p. 117-118.
17
The fire sale declines in state bond collateral during the panics of the free-banking era provide some of the best
empirical examples of what Allen and Gale (2005) and Acharya and Yorulmazer (2008) call “Cash-in-the-Market”
pricing.
15
11
state’s regulations permitted or encouraged the concentration of note collateral in a single
collateral bond. 18
3. Bank Panics and Regulator Reputation
State-wide panics revealed information about the stability of the regulatory system,
specifically the elasticity of demand for required note collateral, the excess specie held in bank
vaults, the liquidity of banks’ loan portfolios, and the quality of capital buffers. If these
safeguards proved adequate, banks could expect their notes to circulate near par and to enjoy all
the leverage and profits that implied. If, however, a panic revealed inadequacies in the regulatory
regime, noteholders could respond by demanding a larger discount to hold notes issued by banks
located in the territory of sovereigns whose guarantees were revealed to be lacking.
3.A. The Panics of 1837 & 1841
Indiana earned a reputation for sound banking during the panics of 1837 and 1841 when
“of all the banks then possessing government deposits, the Indiana Bank was the only one that
offered or paid any specie”. 19 Previous to the panic of 1837, Indiana had established a system of
chartered banks known as the branches of the State Bank of Indiana. Despite its name, the State
Bank of Indiana system was not branches of a single bank but rather a collection of banks with
independent stock holders and management. Each branch raised its own capital, made its own
lending and investment decisions and retained all profits and losses outside of bankruptcy.
Should one of the branches fail, however, the remaining branches were mutually liable for the
losses incurred by the bankrupt branch. This mutual insurance scheme encouraged monitoring
and collective action to discourage excessive risk taking. 20
When eastern banks suspended in May 1837, the Indiana State Bank branches held
approximately $1.2M specie and feared runs from both specie-short eastern bankers who held
approximately $1M of bank’s $2.5M notes in circulation and the Federal Government whose
18
A number of authors have documented the importance of collateral regulation in determining the cause of free
bank panics. See Rolnick and Weber (1983, 1984), Hasan and Dwyer (1994), and Jaremski (2010).
19
Esarey (1912 p.259)
20
Calomiris (1990) surveys pre-FDIC bank insurance systems and concludes the mutual insurance system adopted
by Indiana to be one of the most effective.
12
deposits accounted for approximately $1.5M of the bank’s $1.9M deposit liability. 21 The
branches quickly dispatched a representative to Washington, D.C., with $80,000 in specie to give
to the Treasury in hopes of forestalling a run. The Treasury Secretary, who suddenly found his
deposits throughout the nation tied up in suspended banks, gratefully accepted the money and
agreed not to run on the Indiana bank if the bank would pay out the government’s specie when
drawn in the regular course of business. 22
The State Bank responded to the risk of a run on its notes in the traditional manner – by
suspending convertibility of State Bank notes into specie. Due to a number of mitigating
circumstances, however, the suspension enhanced rather than harmed the State Bank’s
reputation. The most important factor was the fact that virtually every other bank in the nation
also suspended note convertibility during the panic. Additionally, the State Bank adopted a
number of measures that maintained the market value of its notes until the full resumption of
specie payment in 1841. First, the bank accepted its own notes at par in payment of its more than
$2M in loans outstanding. Secondly, the bank convinced the local merchants that the suspension
was one of choice rather than necessity, and the merchants of Indianapolis collectively agreed to
accept the State Bank’s notes at par in their business transactions. These two actions created
sufficient demand for the State Bank’s notes in Indiana to maintain a par market value in the
state and discounts in New Orleans and eastern markets that were relatively small compared to
other banks in the west. At a time when many Western and Southern banks were ruined, the
State Bank of Indiana used its freedom from specie payments to increase its circulation and
convert the privileged price its notes received in eastern exchange into large profits. 23
The State of Indiana did not fare as well as the State Bank during the panic of 1837 and
the depression that followed. Falling land values and the disappointing prospects for bondfunded public works projects forced the State into default on its bonds in 1841. 24 This default
sparked a new panic as Indiana’s defaulted 5% bonds were a popular collateral backing the
banknote issuance of New York free banks. Although Indiana’s bond default resulted in
widespread bank failures in New York and another suspension of convertibility in Philadelphia,
21
The balance sheet data and specific liabilities during the panic of 1837 are estimates of Esarey (1912 p.259-260)
and The Banker’s Magazine Sep. 1857 p. 162-164.
22
Esarey (1912 p.259) and Lainer (1877, p.16-17)
23
Esarey (1912 p.260)
24
See Wallis (2003) for an excellent account of the interaction of land values, tax revenues and bond prices in
Indiana.
13
the State Bank of Indiana was relatively unscathed. The State Bank had yet to resume specie
payments from 1837 and had used the past years’ profits to build a fortress balance sheet that
allowed the bank to weather the panic despite the default of its home sovereign.
Taken together, the panics of 1837 and 1841 left the capital market with the impression
that the banking regulations in Indiana produced the most prudent and safe banks in the nation.
This confidence manifested itself in a willingness to hold Indiana banknotes in foreign markets
that the State Bank branches and later Indiana’s free banks translated into profits that were
unavailable to banks located in less trusted regulatory environs.
3.B. Free Banks Enter Indiana
Indiana’s free banking law took effect on July 1, 1852. New entrants flocked to Indiana
to capture the profits from issuing notes in a state with such a sterling reputation for sound
banking. The free banks notes were implicitly guaranteed by Indiana’s regulators promise to
maintain sufficient collateral to back all notes, but the free banks lacked the mutual liability of
the State Bank branches. Despite the differences between free and State Bank governance, the
government’s implicit guarantee served its purpose and transformed banknotes issued by free
banks and the State Bank branches into information-insensitive equivalent securities in the eyes
of the market. Of the 738 unique monthly Indiana free bank banknote quotations that appeared
on the Philadelphia market between October 1852 and pre-panic October 1854, 694 (94%) were
identical to the discount on the State Bank notes 25 and every free bank note was quoted at the
same price as State Bank notes for all 582 observations between January 1854 and the panic of
1854. 26
3.C. The Panics of 1854 & 1857
Due to the sterling reputation and regulatory advantages they had, Indiana banks were the
preferred counterparties of Ohio and Illinois banks that preferred to borrow out-of-state notes
rather than maintain their own circulation. By the summer of 1854, Indiana banks stood at the
center of a vast interconnected network of interbank lending in which foreign banknotes made up
25
Free banks were quoted at a higher discount 42 times and a lower discount 2 times.
Source: Authors’ calculations from data provided in Weber, Warren E. Quoted Discounts on State Bank Notes in
New York, Cincinnati, and Cleveland, selected dates, 1827-1858. Research Department, Federal Reserve Bank of
Minneapolis. http://minneapolisfed.org/research/economists/wewproj.html
26
14
the majority of the circulating currency of Ohio and Illinois. 27 Table 2 reports the circulation per
capita, circulation per dollar of capital and specie reserves per dollar of circulation for banks in
select Midwestern states. Before the panic of 1854, Indiana’s reputation allowed Indiana banks
to hold fewer specie reserves and circulate more notes per dollar of capital than other states. That
is, Indiana’s funding advantage manifested itself in increased leverage and a relatively high
circulation per capita among Indiana banks.
Indiana’s Panic of 1854 began in Ohio when, in an attempt to increase domestic
circulation and tax revenue, Ohio banking authorities enacted a law making it illegal for any
resident after November 1, 1854, to possess small denomination banknotes issued by an out-ofstate bank. 28 In response, Ohio banks stopped discounting out-of-state banknotes and returned
their banknote inventories to Indiana for redemption in gold. The inflow of notes for redemption
caused Indiana banks to exhaust their gold on hand, forcing the state regulator to sell the state
bond collateral to raise specie. The simultaneous selling of so many banks’ Indiana state bonds
pushed the market price of Indiana collateral down to 22-25% below face value. 29 Noteholders
throughout the state realized that their notes were no longer fully collateralized, and a statewide
panic ensued as noteholders attempted to convert their notes into specie before the banks were
inevitably forced to suspend redemption.
The panic quickly spread to Illinois, but Illinois regulations had encouraged a more
diversified portfolio of collateral by applying haircuts to Illinois bonds. The Illinois collateral
regulations accepted bonds with 6% coupons at a minimum of market or face value, but bonds
with coupons below 6% faced 50% haircuts and bonds issued by the state of Illinois were haircut
20% regardless of coupon. In early 1854, 6% bonds traded above par, and roughly 30% of
Illinois’ banknote circulation was secured by Illinois or California 5% bonds with large haircuts.
As a result, Illinois’ regulator in May 1854 was in the enviable position of holding a diversified
portfolio of collateral bonds with a market value of $1.16 for every $1 in banknotes
outstanding. 30 Fear of counterparty exposure to Indiana banks and the decline of heavily used
27
Huntington (1915) claimed banknotes from Indiana, Illinois and other states supplied Ohio’s domestic currency to
the exclusion of Ohio banknotes (p. 222). The Illinois Bank Commission report of May 1854 estimated that foreign
banknotes accounted for 70% of the paper in circulation in Illinois.
28
This account relies primarily on Hansen and Dwyer (1994), p. 276-7.
29
The price of Indiana 5% bonds fell to 77.75% of par in New York (Banker’s Magazine, Dec. 1854) and 75% of
par in Boston (Martin (1871), p. 90).
30
Authors’ calculation from the market value and circulation figures reported in the Bank Commissioner’s Report of
May 1854.
15
Missouri and Virginia bonds to respectively 93% and 95% of par triggered a run that closed eight
Illinois banks. These banks, however, were sufficiently over-collateralized that noteholders
suffered no losses.
The panic was over by December of 1854. The majority of banks throughout the Midwest
were forced to suspend redemption, but most reopened quickly after the panic had subsided.
Outside of Indiana, noteholders of every failed bank received 100 cents on the dollar from the
sale of note collateral. In Indiana, though, the panic exposed the critical flaw in the regulatory
framework, namely the systemic and wrong-way risk that arose from encouraging
collateralization in primarily Indiana state bonds. Even with these deficiencies, banknotes issued
by many Indiana banks were redeemed at par by the auditor. 31 The notes of the remaining banks,
however, were redeemed at 80-97 cents on the dollar. Thus, while the average redemption was
quite high, the panic introduced doubt regarding the system’s infallibility.
The panic of 1854 revealed information about the relative stability of each state’s
regulatory regime. Despite the state’s promises, Indiana’s regulator was unable to redeem all
banknotes at par. While many banks failed in Illinois as well, the bonds and capital backing
Illinois circulation proved more robust, and the state auditor was able to make noteholders whole
in a timely fashion. The Indiana state attorneys concluded that the state was not obligated to
make noteholders whole, and no reforms were made to limit the use of Indiana bonds as
collateral or reduce the leverage of Indiana banks. The legislator did increase the haircut on
bonds posted as collateral requiring the state regulator to issue only $100 in notes for every $110
of collateral posted. 32 This change did make Indiana notes safer in the event of a bankruptcy and
should, all else equal, have made Indiana notes more appealing to noteholders after 1854.
The national Panic in October of 1857 provides a useful counterpoint, in that it revealed
no new structural flaws in the various regulatory regimes. That panic’s causes are less
transparent than Indiana’s panic of 1854, but, regardless of its cause, the panic of 1857 led to
widespread closures across many states with Indiana again being hit especially hard. 33 Note
31
At least fifteen failed Indiana banks were redeemed at Par. Source: Authors’ calculations from various issues of
The Annual Report of the Auditor of the State of Indiana and the Letter from the Secretary of the Treasury May 13,
1856 House Ex Doc no.102 34th Congress 1st Session.
32
This reform reduced the statutory maximum leverage, but, as the market-imposed limits on leverage were much
lower, these statutory limits were binding neither before nor after the reform.
33
Calomiris and Schweikart (1991) review the possible causes and consequences of this panic. Candidate causes
include lower crop and land prices following the Crimean War, the sinking of the S.S. Central America with a loss
of 30,000 pounds of gold bound for the East Coast, and the Dred Scott decision in March 1857 which lessened the
16
holders of banks that closed were uniformly paid off in full, though after several months and
presumable uncertainty. If the Kindleberger thesis of first-hand uncertainty decreasing banking
demand holds, we would expect to see that markets that suffered closures during the Panic of
1857 should be especially underbanked by 1860.
What were these panics’ long term effects on the market structure of banking? Did note
holders reward banks in those states where the regulatory regime proved robust to bank runs and
failures? Did creditors raise the cost of capital to punish banks in the state of Indiana where
systemic risk was revealed and not corrected? Or were noteholders uniformly scarred by bank
failures and any regulatory concerns were swamped? To answer these questions, we now take
advantage of the (endogenous) entry and exit decisions of banks observed in mid-1854 and late
1860 to estimate a model of bank profits across geographic space.
4. An empirical model of endogenous market structure
We employ techniques similar to those developed by Bresnahan and Reiss (1990, 1991)
and Berry (1992) and extended by Abraham, Gaynor, and Vogt (2007, henceforth AGV) for
examining the determinants of market structure. In particular, we employ a discrete dependent
variable model relating the number of banks in a market to characteristics of that market.
Although variables such as profits, prices, and costs are not observed, this approach permits the
estimation of the impact of market characteristics, most notably market size, on the number of
banks in a market and in turn allows the derivation of threshold conditions on bank entry. In
other words, even without data on a number of economically relevant variables, one can still
examine the determination of market structure through observation of actual decisions by banks.
Thus, actual realizations of market structure reveal features of underlying profitability even
without directly observing profits.
A bank’s entry decision depends upon the profit it expects to earn following entry given
the entry decisions of other potential banks in the market, the nature of post-entry competition,
and both demand and cost factors. We assume that the long-run profits of a bank in a market can
be expressed as a function of the number of active bank in the market and characteristics of the
prospect of lucrative westward railroad expansion. They also note and emphasize the disproportionate impact the
’57 Panic had on Indiana’s banking sector.
17
market. In other words, the profits of bank i in market k are given by
Π𝑖 (𝑁𝑘 , 𝑦𝑘 , 𝑥𝑘 )
(1)
in which Nk is the number of banks in the market, yk is a scalar such as the total population of
consumers describing the size of market k, and xk is a vector of variables that can affect both
costs and demand. This expression should be interpreted as a reduced form discounted long-run
profit function reflecting the outcome of competition between the banks in market k. Thus, if
Nk = 1, (1) describes the equilibrium profits of a monopolist in market k, while, if Nk = 2, (1)
describes the per-firm equilibrium profit outcome for duopolists.
Following earlier empirical studies of entry, we view banks’ equilibrium post-entry
profits as unobserved random variables. As our data suggest that allowing for heterogeneous
banks is unimportant, we instead impose very strong restrictions on profits for banks in a given
market. 34 In particular, we assume that the equilibrium profits for bank i in market k are given by
Π𝑖 (𝑁𝑘 , 𝑦𝑘 , 𝑥𝑘 ) = 𝜋(𝑁𝑘 , 𝑦𝑘 , 𝑥𝑘 ) + 𝜀𝑘
(2)
This formulation says that equilibrium profits are composed of expected profits, 𝜋(𝑁𝑘 , 𝑦𝑘 , 𝑥𝑘 ),
and a random term εk, in which both components depend only on the market k. That is, there are
no observed or unobserved variables unique to bank i in market k.
These assumptions have several useful implications. Most importantly, if expected profits
decline in Nk, the assumption that profits involve only market-specific factors implies that the
Nash equilibrium of many standard entry games will involve banks entering a market until entry
by additional banks would be unprofitable. Put differently, the equilibrium number of active
banks in a market is the maximum number that the market can profitably sustain. Thus, this
assumption implies simple threshold conditions on profits that characterize the equilibrium
number of firms in a market. In particular, the equilibrium number of banks in market k, denoted
𝑁𝑘∗ , is characterized by
𝑁𝑘∗ = 0 𝑖𝑓 𝜋𝑘1 + 𝜀𝑘 < 0
𝑁𝑘∗ = 𝑁 𝑖𝑓 𝜋𝑘𝑁 + 𝜀𝑘 ≥ 0 𝑎𝑛𝑑 𝜋𝑘𝑁+1 + 𝜀𝑘 < 0
in which 𝜋𝑘𝑁 = 𝜋(𝑁𝑘 = 𝑁, 𝑦𝑘 , 𝑥𝑘 ). That is, there will be N banks in market k if it is profitable for
each of those banks to enter given conditions in that market but unprofitable for additional banks
to enter. These simple conditions contrast with those involving unobserved bank-specific
34
Unreported regressions indicate that incumbents that were state banks (rather than free banks) or banks with
higher-than-typical assets had no significantly different impact on market structure.
18
heterogeneity, as those conditions would need to account for different combinations of entrants
and the differences in their unobserved profitability. 35 With the distributional assumption that
the random component is i.i.d. normal across markets, the probabilities observing N banks in
market k are
𝑃(𝑁𝑘∗ = 0) = 1 − Φ(𝜋𝑘1 )
𝑃(𝑁𝑘∗ = 1) = Φ(𝜋𝑘1 ) − Φ(𝜋𝑘2 )
(3)
𝑃(𝑁𝑘∗ = 2) = Φ(𝜋𝑘2 ) − Φ(𝜋𝑘3 )
𝑃(𝑁𝑘∗ ≥ 3) = Φ(𝜋𝑘3 )
in which Φ(•) is the cdf of a standard normal random variable with the variance of the
disturbance term normalized to one. 36
Our restrictions on heterogeneity among banks are certainly not innocuous. They imply
that all variation in outcomes across markets is attributable solely to differences in characteristics
of those markets. This inference may be erroneous if heterogeneity among firms is in fact the
source of variation in market outcomes. Without additional information or assumptions about the
number of potential entrants, though, it is difficult to distinguish the effect of unobserved
heterogeneity from market characteristics. The lack of such information in the data that we use in
this study necessitates the imposition of these assumptions. The widespread ease of entry during
the free-banking era, of course, also presumably makes this lack of heterogeneity less onerous. 37
Turning to functional form assumptions, we assume that expected profits can be
decomposed into variable profits and fixed costs. 38 Unlike in Bresnahan and Reiss, market size
here does not play a central role in identification. We follow the endogenous market structure
literature and assume the slope of profits in market size may vary with market structure. We
additionally assume that fixed costs may vary over market structures because of endogenously
35
Ciliberto and Tamer (2009) provide a solution to the problem of unobserved heterogeneity among firms in their
analysis of airlines; however, their method requires knowledge of the characteristics of each potential entrant and is
therefore not applicable to banking markets in an era of unit banks.
36
N+1
It is straightforward to extend the analysis to more than three entrants, as 𝑃(𝑁𝑘∗ = 𝑁) = Φ�πN
�. Our
k � − Φ�πk
data set, however, includes very few observations of markets containing more than three banks, and we therefore
consider all markets with three or more banks as a single outcome. Regardless, the last probability in (3) provides a
necessary condition for observing at least three banks.
37
Manuszak (2002) likewise exploits a truly free-entry situation of minimal regulation in his analysis of late 19th
century American brewing.
38
Because they have quantity data, AGV find value in breaking variable profits into market size, per-capita quantity,
and average variable profit. Without such data, we gain little from such a breakdown.
19
determined barriers to entry (represented as higher costs to all firms). Formally, expected profits
for a bank in a market with N banks are
1
𝜋𝑘𝑁 = 𝑁 𝜐(𝑁, 𝑦𝑘 , 𝑥𝑘 ) − 𝐹(𝑁, 𝑥𝑘 )
(4)
in which 𝑣(•) denotes the industry’s variable profits (revenues less variable costs) and 𝐹(•)
denotes fixed costs. We follow AGV and further assume that industry variable profits and fixed
costs can be approximated by exponential functions of N and variables affecting demand and
costs. We specifically assume that
𝜐(𝑁, 𝑦𝑘 , 𝑥𝑘 ) = exp(𝑦𝑘 𝜆 + 𝑥𝑘 𝛿𝑋 + 𝛿𝑁 + 𝜀𝑘𝑣 )
(5)
𝐹(𝑁, 𝑥𝑘 ) = exp(𝑥𝑘 𝛾𝑋 + 𝛾𝑁 + 𝜀𝑘𝐹 )
The parameters δN and γN are coefficients on dummy variables for market structure. They capture
differences in industry variable profits and fixed costs between markets with one bank and
markets with N banks. The idiosyncratic disturbances εν and εF are assumed to be normally
distributed with mean zero and constant variance. As discussed in AGV, this specification
makes explicit the identification troubles that arise when one tries to distinguish the impacts of
variable profits and fixed costs on market structure without quantity data. 39 Like AGV, we use
the log of population, ln(Pop), as our measure of y, so our specification for industry variable
profits can be rewritten as
𝜐(𝑁, 𝑦𝑘 , 𝑥𝑘 ) = 𝑃𝑜𝑝𝑘𝜆 exp(𝑥𝑘 𝛿𝑋 + 𝛿𝑁 + 𝜀𝑘𝑣 )
(5a)
Substituting these specifications into the prior inequalities that determine market
structures, we see that the Nth bank enters when its share of industry variable profits exceeds its
fixed costs. That is,
1
exp(𝑦𝑘 𝜆 + 𝑥𝑘 𝛿𝑋 + 𝛿𝑁 + 𝜀𝑘𝑣 ) − exp(𝑥𝑘 𝛾𝑋 + 𝛾𝑁 + 𝜀𝑘𝐹 ) > 0
𝑁
Rearranging and taking the log of both sides yields
𝑦𝑘 𝜆 + 𝑥𝑘 𝛿𝑋 + 𝛿𝑁 + 𝜀𝑘𝑣 − ln 𝑁 > 𝑥𝑘 𝛾𝑋 + 𝛾𝑁 + 𝜀𝑘𝐹
Denote µX = δX – γX and µN = γN – δN, and further denote εΠ = ευ – εF, which is distributed
39
While the signs of our critical parameters are robust to the use of Bresnahan and Reiss’s usual model that
identifies variable profit margin and fixed costs from the linear functional form, parameters were imprecise unless
we considered only one or the other.
20
N(0, σ2). We then see that
0
1
𝑁=
⎨ 2
⎪
⎩ 3+
⎧
⎪
𝑖𝑓
𝑖𝑓
𝑖𝑓
𝑖𝑓
𝑦𝑘 𝜆 + 𝑥𝑘 𝜇𝑋 + 𝜀𝑘Π < 𝜇1
𝜇1 < 𝑦𝑘 𝜆 + 𝑥𝑘 𝜇𝑋 + 𝜀𝑘Π < 𝜇2 + ln 2
𝜇2 + ln 2 < 𝑦𝑘 𝜆 + 𝑥𝑘 𝜇𝑋 + 𝜀𝑘Π < 𝜇3 + ln 3
𝜇3 + ln 3 < 𝑦𝑘 𝜆 + 𝑥𝑘 𝜇𝑋 + 𝜀𝑘Π
The normality assumption on the error term εΠ yields an ordered probit (with threshold values of
µN) that can be estimated via maximum likelihood techniques in which the dependent variable in
the ordered probit is the number of banks in a market. All parameters are rescaled by σ, which
we normalize to σ = 1. Note that, in this specification and unlike many applications of
Bresnahan and Reiss, coefficients do not accumulate. For example, the relevant test of whether
entry of the third firm affects per-firm profits will be µ2 = µ3 rather than µ3 = 0.
The above framework applies nicely for our pure cross-sectional analyses, but we are also
interested in the change across cross-sections, that is, whether Indiana’s banking parameter
decreased from 1854 to 1860. To implement this test, we offer a slight extension of AGV in
which we pool our two cross-sections, imposing that the primitive parameters are identical for
the two years but allowing the scalars representing the distribution’s variance (σ) to differ. That
is, we will normalize the scalar for 1854 (σ54=1) but estimate the 1860 scalar σ60 as a free
parameter. Our explicit control for market population makes the restriction that parameters are
fixed across the two cross-sections somewhat plausible; we will let the reader judge how onerous
those restrictions are by comparing the estimates of the separate cross-sections to the pooled
estimates.
This framework also implicitly assumes that all markets are subject to free entry, an
assumption at odds with the presence of several states that required charters from the state
legislature. This problem, however, is readily resolved by revising the above entry thresholds so
that, for state-charter-only states, they reduce to the probit case in which the market structure
possibilities are either zero banks or at least one bank (i.e., N=0 or N≥1). As we observe few
markets with more than a single bank in states that require charters, this agglomeration leads to
little information being lost.
Our emphasis in this study is to consider the impact of various market characteristics
(specifically whether or not a market is in Indiana) on firm profitability, and thus our interest is
primarily upon the relevant parameters µX. We would be remiss, however, if we did not trace the
21
(difficult to interpret) estimates back to their real-world implications. Estimation of the profit
function provides information about the market sizes required to support entry. Specifically, the
market sizes yM and yD at which monopoly and duopoly profits are respectively zero can be
formed from the estimates of the profit function parameters. For example, the exact expressions
for yM and yD are given by
𝜋 𝑀 = 0 ↔ 𝑦 𝑀 = exp �
𝜋 𝐷 = 0 ↔ 𝑦 𝐷 = exp �
𝜇1 −𝑥𝜇𝑋
𝜆
�
𝜇1 +𝜇2 −𝑥𝜇𝑋
𝜆
(7)
�
in which the parameters are replaced by maximum likelihood estimates in practice. The
thresholds yM and yD therefore provide estimates of the sizes of markets required to support the
long-run equilibrium of one and two firms. 40 How these entry thresholds differ over time and
between Indiana and other free-banking states thus offers an extra check of the estimates’
plausibility.
5. Data review and analysis
We begin by compiling cross-sections of markets from 1854 and 1860 located in
(largely) Midwestern states. Potential markets consist of every enumerated place in the InterUniversity Consortium for Political and Social Research (ICPSR) machine-readable 1850 and
1860 Censuses of the United States. A Census enumerated place consists of a city, town or
village, so observations at the township level were excluded. These ICPSR data are not
complete. The aggregate population of ICPSR places does not approach the aggregate population
of many states (e.g., Tennessee, Arkansas, Georgia), and we consequently focus on states for
which coverage appeared complete. As our eventual goal is to construct a sample of
geographically isolated markets, we exclude most of the densely populated Eastern Seaboard.
We also exclude Michigan because of its highly erratic banking history and long-standing
reputation for wildcat banking. These restrictions leave us to consider seven states: Missouri,
Iowa, Illinois, Wisconsin, Indiana, Ohio, and western Pennsylvania. In preliminary work, we
allowed observations from western New York, but results indicated that idiosyncratic features of
40
To facilitate estimation and interpretation, we demean most of our variables x. The thresholds in this case reduce
𝜇 + 𝜇2�
𝜇1 + 𝜇3�
𝜇
to 𝑀 = exp� 1�𝜆�, 𝐷 = exp � 1
𝜆�, and 𝑇 = exp �
𝜆�
22
New York made it too dissimilar to include. 41 We gather latitude and longitude coordinate data
for the center of each of these Census observations, and we use these coordinate data to construct
distances between observations’ center-points. Given these restrictions, we begin with a sample
of 1352 not necessarily isolated observations.
We consider Census enumerated observations in these states to be geographically isolated
(and thus included in our 1860 sample) if they satisfy either of two criteria in 1860. First, we
include any observation that is at least ten miles from its nearest neighboring observation. Given
the speed of transportation at the time and an absence of communication options 42, ten miles
seems a reasonable distance beyond which there would be substantial pressure to create a more
local bank. Second, we consolidate observation-pairs that are near one another but far from any
other markets. To satisfy this criterion, two observations must be within five miles of each other
and at least ten miles away from the next closest observation. In these latter cases, we create a
new observation that aggregates all relevant features of the two observations. We use similar
conditions in constructing the 1854 sample, with the additional requirement that we see the 1850
population for that observation. Because of their relative size and easy transportation access, we
also exclude St. Louis and Cincinnati even though they otherwise survive our isolation criteria.
Applying these criteria reduces our original sample of Census observations to 254 and
362 isolated observations in 1854 and 1860 respectively. 43 Maps of our 1854 and 1860 markets
can be found in Figures 2 and 3. We will henceforth refer to these observations as markets. A
reasonable robustness check would be to consider even more stringent requirements for isolation.
Unfortunately, about half of the markets of these 362 markets from the 1860 cross-section that
survive our initial ten-mile-minimum criterion have a neighboring observation within 12.5 miles.
Fewer than 100 observations are at least 15 miles from the nearest observation. We therefore
trade off more compelling degrees of isolation in exchange for a substantially larger sample size.
Market definition is of course only the first step in the analysis. We next determine how
many banks existed in each market and link each market with population, weather history and
access to railroads. Populations and demographic characteristics are taken from the 1850, 1860,
41
We define western Pennsylvania and New York by constructing a line that connects Albany, NY, and Allentown,
PA. Pennsylvania and New York Census enumerated places to the west of this line are eligible for inclusion in the
sample.
42
While Western Union would offer transcontinental telegraph service in 1861, our isolated small markets generally
lacked telegraph access in 1860.
43
Iowa had banned banks prior to its free-banking act of 1858, so we exclude it from our 1854 sample.
23
and 1870 Censuses 44 and include all persons. 45 Bank number and closure information come from
Warren Weber’s (2006) impressive census of antebellum banking available at the Federal
Reserve Bank of Minneapolis. 46 We deviate from Weber only when state auditor reports indicate
a bank was circulating notes before Weber’s entry date or newspaper articles indicate the bank
was suspended never to reopen before Weber’s exit date.
Economists have long been interested in the link between weather, agricultural output
and financial distress. 47 Many of our sample markets suffered drought conditions between 1854
and 1860. We control for these weather shocks with an annual rainfall index derived from treering records. This weather information consists of the annual time series of the Palmer Drought
Severity Index (PDSI) inferred by taking a weighted average of the nearest grid points’ values
from Zhang et al. (2004). After experimentation, we combine this drought information into a
single variable as shown in the appendix. This measure looks back five years, ignores wetterthan-average years, places more weight on more recent events, and exhibits a convex relation
with PDSI.
We include controls for access to railroads. Contemporaries complained that banks
located in hard to reach locales to raise the cost of returning their note issues, and Gorton (1990)
documents a link between transportation costs and banknote discounts in Philadelphia. We
therefore construct a railroad dummy equal to 1 if a railroad was operating in the bank’s county
in the year in question.
Table 3 provides variable definitions, and Table 4 summarizes population, demographic
features, bank presence, and other variables for these markets. Most notable is the number of
markets that have no reported banks. In the 1854 cross-section, about 80% of markets in freebanking states (including Indiana) lack banks, while close to 95% of markets in state-charteronly states have no bank. These percentages are lower in the 1860 cross-section for all states.
The number of markets with more than one bank is not large in either cross-section, and that with
more than two banks is quite small. This limitation of the data guides our later caution when we
estimate explicit triopoly-plus specifications in addition to the more easily justifiable duopolyplus specifications. Populations of markets with different numbers of banks are generally
44
The 1870 Census included population totals for the 1850, 1860, and 1870 Censuses, and we use those figures for
those years.
45
Missouri, the only slave state in our sample, thus includes both freemen and slaves in its population count.
46
Weber’s bank census is available http://www.minneapolisfed.org/research/economists/wewproj.cfm#censusstate
47
See for instance Landon-Lane, Rockoff, and Steckel (2011) and cites therein.
24
intuitive. Markets with one bank have higher populations than markets with no banks, and
markets with two banks have higher population than markets with one bank. The fact that
average population for markets with more than two banks is less than that for markets with two
banks for free-banking states in 1860 is somewhat troubling and highlights the need for caution
when interpreting the estimates from the triopoly-plus specification.
Drought data indicate that the primary meteorological event of the 1850s was a western
drought in 1856 that lingered into 1857 in Illinois and Missouri. About 44% of our markets
qualified as being hit by drought (PDSI < -2) in 1856. As would be expected, there was a great
deal of regional variation in these figures. In 1856, 70% of our Indiana markets had drought
conditions, as did over 80% of our Illinois markets, 30% of our Iowa markets, and all of our
Missouri markets. As only 5% of our Ohio markets and no Wisconsin or Pennsylvania markets
suffered drought conditions in 1856, we expect that we have sufficient variation in the data to
separate out impacts of banking regulation from the primary competing shock of drought.
Because we expect drought to affect rural agriculture-centered markets most directly, we include
an interaction of our drought measure and market population in the summary statistics and later
empirics.
The last variables of Table 4 highlight how we will handle the Panics of 1854 and 1857.
For a market to suffer a panic-closure, it must have first had at least one bank and therefore
probably had exceptional banking demand. Examining panic-closures without controlling for the
original level of banking will likely lead to results erroneously indicating that panic-closures
increase the steady-state level of banking. We see that the Panic of 1854 caused the closure of
almost half the banks that were in operation in August immediately prior to the panic. Observing
the September 1857 level of banking, we can see that banking in Indiana had recovered
somewhat from the 1854 panic (though far from its previous peak), but the robustness of those
new banks is uncertain. Within our isolated-market sample, the closures in the Panic of 1857
disproportionately struck Indiana, as noted in Calomiris and Schweikart (1991). These closures
moved the Indiana average number of banks closer to the free-banking state average. Our
analysis now illuminates the extent of additional decline between that panic and the 1860
observations.
25
6. Estimation results
Table 5 presents the maximum likelihood estimates of the expected profit functions for
the separately estimated cross-sections. To reiterate, the estimation involves a discrete dependent
variable model in which the dependent variable is the number of banks in a market, and the
econometric goal is to uncover features of underlying profitability. Specifically, the estimated
equations related to the latent profit functions are
𝜐(𝑁, 𝑦𝑘 , 𝑥𝑘 ) = 𝑃𝑜𝑝𝑘𝜆 exp(𝑥𝑘 𝛿𝑋 + 𝛿𝑁 + 𝜀𝑘𝑣 )
(6)
𝐹(𝑁, 𝑥𝑘 ) = exp(𝑥𝑘 𝛾𝑋 + 𝛾𝑁 + 𝜀𝑘𝐹 )
in which the estimated parameters are µX = δX – γX and µN = γN – δN. We represent the crosssection-specific standard deviations of the normally distributed εΠ=εν-εF as σt for year t. In
addition to separate estimates by cross-section, we also display estimates using the duopoly-plus
and triopoly-plus specifications. Entry thresholds beneath the estimates themselves indicate the
estimated number of persons required to support a market structure in a given regime.
The estimates from the 1854 cross-section are all of the expected sign, and many
coefficient estimates are significant at conventional levels. Banking is clearly increasing with
population, and markets in charter-only states are at a sizable disadvantage. To support at least
three banks, markets must be significantly larger than triple the size needed to support a
monopoly bank, but no similar conclusions can be drawn with respect to duopoly and monopoly
or triopoly and duopoly. Our drought measure is negatively related to profits and its impact is
decreasing with market population, consistent with expectations. Of most importance, Indiana
significantly outperforms its free-banking peers in both the duopoly-plus and triopoly-plus
specifications.
The implied entry thresholds are plausible, if not especially precise, and consistent with
our story of Indiana being a bank-friendly state in 1854. Depending on specification, at least
4305 persons were needed to support a monopoly bank in non-Indiana free-banking states, while
only 1729 were required in Indiana. Both contrast favorably to the monopoly entry threshold in
charter-only states (at least 9,518 persons needed). The exponential specification of AGV
ensures that other entry thresholds are proportional (e.g., Indiana’s monopoly and duopoly
thresholds are both 40% of those for non-Indiana free banking states).
Indiana’s position reverses when we consider the estimates from the 1860 cross section.
The Indiana indicator’s estimated coefficient is now significantly negative in both the duopoly26
plus and triopoly-plus specifications (respective p-values of 0.015 and 0.018). Signs of other
estimated coefficients are the same as from the 1854 estimation, though the impact of additional
entrants on profits is now statistically significant 48 and the impacts of drought and being in a
charter-only state are now insignificant. The implied entry thresholds obviously reflect these
estimates. While approximately 1729 persons were needed to support a monopoly bank in
Indiana in 1854, close to 6,831 are required by 1860. This implication contrasts with the nonIndiana free-banking states in which the threshold declined by at least 20% across specifications
and the charter-only states in which the threshold fell dramatically (50-75% depending on
specification). It seems reasonable to conclude that whatever happened to Indiana between 1854
and 1860 was not a shock with uniform impact across our sample. Furthermore, the loss of
Indiana’s paramount position in the regulatory arbitrage network coincided not only with a
diminution of its banking sector but a reduction to a level below that of its peers that had never
experienced such financial success.
Table 6 displays the results when we pool our cross-sections, restricting the parameters to
be identical across years but allowing the standard deviation of the normal distribution to vary.
These are strong assumptions that may generate a great deal of insight and additional precision if
they are not especially onerous. While certain restrictions (e.g., the impact of state-charter-only
banking being the same in 1854 and 1860) show strain, the general implications that are relevant
to our paper are robust between the cross-sections and pooled regressions. We will therefore
take the pooled estimates at face value.
Aside from the coefficients of the drought variables (which are still the expected sign for
duopoly-plus), the key parameters are estimated precisely. Markets with greater population have
more banks, charter-only states are under-banked relative to their free-banking peers, and the
entry of additional banks lowers profits, both when moving from monopoly to duopoly and from
duopoly to triopoly. Most importantly for our purposes, we can now explicitly reject the
hypothesis that Indiana did not undergo a change between 1854 and 1860. Entry thresholds are
quite similar to those implied by the separate cross-sectional estimates and confirm the
magnitude of the change that Indiana banking underwent.
48
Using the test statistics that are standard in the IO literature, we can reject with high confidence (p=0.028 in the
duopoly-plus case and p=0.026 in the triopoly-plus case) that a doubling of the monopoly threshold will support a
second entrant. Given the institutional details of the era’s banking sector, this is presumably because of the increased
competition limiting leverage and reducing interest rates on loans rather than higher fixed costs, though our
estimates give us no way to test this story.
27
These results strongly indicate something happened to Indiana banking between mid1854 and 1860, and Indiana’s panic of 1854 is the likeliest candidate. From a policy perspective,
however, the mechanism through which the panic had its impact is the critical issue. Did Indiana
become a less profitable location to locate a bank because of the loss of faith in its regulatory
guarantees, or did so many Indianans personally suffer loses and shun future banks that much of
Indiana was unprofitable for future banking? If panics reveal no regulatory shortcomings but
negatively affect long-run banking demand (à la Kindleberger), then there may be sizable gains
to early intervention. If, however, panics make their impact by revealing regulatory flaws and not
by affecting consumer demand, then early intervention without remedies for the underlying flaw
are likely to fail. We now turn to an extended examination of our 1860 cross-section in which we
consider the impacts of closures during the two panics.
We use the panic of 1857 to investigate the cause of Indiana’s banking decline. The panic
of 1857 resulted in a number of bank failures across our sample states. Unlike the panic of 1854,
banknote guarantees were honored. The panic of 1857 therefore provides us with an opportunity
to examine the effect of bank panics on individual markets without the confounding revelation
that a state guarantee will not be honored. If the Kindleberger hypothesis is true, we should
observe a decrease in implied banking profits across all markets that suffered bank failures in
1857. Furthermore, if this hypothesis is the entire story, the impact of such closures should draw
away the explanatory power of the Indiana indicator. If loss of faith in government guarantees is
also a culprit, though, we should still witness a decline in bank profits in Indiana markets even
controlling for closures.
Table 7 displays the results of maximum likelihood estimation when we extend our
model to include a market’s number of prior closures. 49 As column (2) indicates that the
triopoly-plus threshold is significantly greater (p=0.055) than the duopoly threshold, we employ
the triopoly-plus specification going forward; results (not shown) are similar under the duopolyplus specification. We begin our extensions by incorporating the number of banks present in a
market in September 1857 (immediately prior to the October panic) and then add variables
capturing bank closures from the Panic of 1857. Given the endogeneity of a market’s banking
level, the coefficient on the pre-panic number of banks can be most usefully interpreted as
absorbing market-specific characteristics that are constant over time. We stress that, unlike our
49
Because Iowa had no banks during the Panic of 1857, we omit its markets from this sample.
28
other variables and their coefficients, we make no attempt to build on this estimate; its inclusion
is necessary so that the closure variables do not inadvertently capture those same market-specific
and time-invariant characteristics.
Our first closure variable is the number of banks that closed in a market during the 1857
panic, and we also include an interaction of that variable with the Indiana indicator (columns (4)
and (5)). The first of these specifications offers some support for the Kindleberger hypothesis, in
that inferred profits are significantly decreasing in the number of closures that a market
experienced in the ’57 panic. While the interaction of (5) is not estimated precisely, there is little
evidence that closures in Indiana had greater negative impact than closures in other states.
Despite the inclusion of market-specific closure variables, the Indiana indicator is always
significantly negative, as one would expect if the impact of Indiana’s Panic of 1854 was a loss of
confidence in the regulatory structure. We tentatively conclude that this loss of confidence in the
regulatory regime is a primary reason for Indiana’s lessened banking activity and that diminished
demand arising from direct experience with bank closures does not fully explain the data.
We further explore the Kindleberger hypothesis in Table 8, in which we consider
statewide rather than market-specific bank closures. Dropping Iowa because of its banking-ban
prior to 1858, we construct cumulative statewide failure rates as the ratio of bank failures in the
state in the ’54 and ’57 panics to the cumulative number of pre-panic banks. These failures and
bank-counts are thus not limited to our sample of isolated markets. These failure rates are PA,
0.96%; OH, 5.51%; IN, 40.15%; IL, 19.75%; WI, 8.43%; and MO, 0%. If banking demand falls
because potential noteholders fear losses on account of bank failures anywhere in their state,
these failure rates should be negatively correlated with levels of banking.
Column (1) of Table 8 confirms that Indiana is still significantly under-banked for this
smaller sample. After dropping the Indiana indicator (column (2)), we see that there is a negative
but insignificant relation between failure rates and bank profits in column (3). This result,
however, disappears when we introduce an interaction of the Indiana indicator and the failure
rate. These estimates indicate that outside of Indiana the states with higher historical panic
failure rates (e.g., Illinois and Wisconsin) are associated with higher demand for banking in
1860. Indiana, however, shows the opposite relationship. The estimates are consistent with a
story in which potential banknote holders contrast the resiliency of the Illinois and Wisconsin
regulatory regimes with that of Indiana and choose to allocate their money accordingly.
29
7. Conclusion
The panics of 1854 and 1857 present us with a near ideal setting to examine the
consequences of financial and regulatory fragility. The many isolated markets and state-level
regulation of the antebellum era provide cross-sectional variation in both bank failures and
regulatory regime, variation that is unmatched in modern times. We apply a model of
endogenous market structure in order to examine the impact of panics on the steady state level of
banking.
Prior to the panic of 1854, Indiana enjoyed a reputation for sound regulation that elevated
Indiana banks to a status of preferred counterparties to other banks throughout the Midwest.
Indiana’s banks were able to translate this trust advantage into profits by holding lower reserves
and to access cheap funding by circulating more banknotes than their out-of-state competitors.
New entrants flocked to Indiana. Indeed, our model of endogenous market structure implies that
Indiana towns in mid-1854 could support an initial bank with less than half the population
necessary in other free-banking states.
The panic of 1854, though, exposed a critical flaw in the Indiana regulatory regime.
Indiana’s collateral requirements encouraged the holding of domestic state debt to an extent
unmatched in other states. As a result, Indiana’s noteholders were exposed to wrong-way risk as
the majority of collateral backing their notes declined in value at exactly the time it was most
needed. The panic of 1854 resulted in suspensions and failures throughout the Midwest, but
regulators in states with more diversified portfolios were able to make good on their implicit
promise to convert collateral backing notes of failed banks into sufficient gold to redeem all
notes at par. Indiana’s regulation resulted in a collateral portfolio disproportionately backed by
Indiana state bonds that could not be liquidated at a price sufficient to make noteholders whole.
There is good reason to believe that the loss of trust in Indiana’s regulatory regime after
the panic of 1854 was the primary cause of the decrease in its steady state level of banking.
While markets outside of Indiana saw insignificantly reduced levels of banking after
experiencing closures in the Panic of ’57, Indiana’s markets with no direct experience in
panicked closures also saw reduced banking activity. This statewide effect is not explained by
Indiana’s high level of closures in the ’54 and ’57 panics. Other states with high numbers of
closures but with more robust regulatory regimes saw banking activity increase appreciably after
30
the Panic of ’57. These results are consistent with a statewide decline in profitability rather than
a decline in only markets where citizens had firsthand experience of bank failures.
Our results confirm that trust is an important trait in a well-functioning financial system.
Unaddressed regulatory failure can leave long-lasting impacts on the expectations of consumers
and counterparties and result in a long-lasting decline in the level of financial intermediation.
31
Appendix: Construction of Drought Variable
The Palmer Drought Severity Index (PDSI) “uses temperature and rainfall information to
determine dryness”. (NOAA website, http://www.drought.noaa.gov/palmer.html, accessed
7/5/2012). The measure is standardized to local climate and normalized to 0. Negative numbers
indicate drought, with -2/-3/-4 denoting moderate/severe/extreme drought. We use the
reconstructed measures of the Palmer Drought Severity Index provided by Zhang, Mann, and
Cook (2004). This study employs tree-ring data over the coterminous United States to estimate
PDSI measures for a system of latitude-longitude grid points back to the year 1700. Our marketappropriate PDSI measures then use market latitude-longitude coordinates to construct a
weighted average of the provided gridpoints.
The above measure indicates a severe and widespread drought in 1856. The average
PDSI in 1856 for our 362 markets observed in 1860 was -1.8689 (i.e., almost the moderate
drought threshold of -2). The drought continued in many but not all places in 1857 (average
PDSI in 1857 was -1.4291). Furthermore, roughly 44% of our 362 markets from 1860
experienced at least one drought-year of at least moderate severity between 1850 and 1859. The
most recent drought year (compared to 1860) was 1857 for 11% of the markets and 1856 for
33% of the markets. All other markets experienced no drought for the time period.
Our market-level measure is constructed for 1854 and 1860. Given the cumulative effects
of yearly precipitation, it looks back five years with recent years more heavily weighted than
distant years. The measure also builds on the intuition that droughts are worse for agriculture
than overly wet seasons (so asymmetric) and that the agricultural costs of drought are non-linear
(so convex). Each market-year’s drought measure is
�𝑃𝐷𝑆𝐼𝑗,𝑡−𝑘 |𝑃𝐷𝑆𝐼𝑗,𝑡−𝑘 < 0�
2
so that a market-year’s drought measure is positive if there was any degree of drought and zero
otherwise. Given a look-back of LB years, the weight applied to k years ago follows the formula
𝜔𝑘 = 1 −
𝑘−1
𝐿𝐵
So for our proposed look-back of LB=5, the immediately prior year is given full weight (ω=1)
and preceding years are given weights 0.8, 0.6, 0.4, and 0.2. Our full drought measure is then
𝐿𝐵
2
𝐷𝑟𝑜𝑗𝑡 = � 𝜔𝑘 ��𝑃𝐷𝑆𝐼𝑗,𝑡−𝑘 |𝑃𝐷𝑆𝐼𝑗,𝑡−𝑘 < 0� �
𝑘=1
32
While different specifications yielded much different estimates of the impact of drought
on equilibrium levels of banking, our paper’s core results are robust to the use of many drought
measure (e.g., varying number of years looked back, linear rather than convex, symmetric rather
than asymmetric, etc.).
33
References
Abraham, J. M., M. Gaynor, and W. B. Vogt, “Entry and Competition in Local Hospital
Markets,” Journal of Industrial Economics, 60(2): 265-88, 2007.
Acharya, Viral, and Tanju Yorulmazer “Cash-in-the-Market Pricing and Optimal Resolution of
Bank Failures,” Review of Financial Studies, 21(6): 2705-42, 2008.
Allen, Franklin, and Douglas Gale “From Cash-in-the-Market Pricing to Financial Fragility,”
Journal of the European Economic Association, 3: 535-46, 2005.
Berry, Steven T., “Estimation of a model of entry in the airline industry,” Econometrica,
60(4): 889–917, 1992.
Bodenhorn, Howard, “Entry, Rivalry, and Free Banking in Antebellum America,” Review of
Economics and Statistics, 72(4): 682-86, 1990.
Bodenhorn, Howard, “Was There a Note Issue Conundrum in the Free Banking Era?” Journal of
Money, Credit, and Banking, 27(3): 702-12, 1995.
Bresnahan, Timothy F., and Peter C. Reiss, “Entry in monopoly markets,” The Review of
Economic Studies, 57(4): 531-53, 1990.
Bresnahan, Timothy F., and Peter C. Reiss, “Entry and Competition in Concentrated Markets,”
Journal of Political Economy, 95(5): 977-1009, 1991.
Brunnermeir, Marcus. K. and Lasse H. Pedersen, “Market liquidity and funding liquidity”
Review of Financial Studies, 22 (6): 2201-2238, 2009
Calomiris, Charles W., “Is Deposit Insurance Necessary? A Historical Perspective,” Journal of
Economic History 50, 283-95. 1990
Calomiris, Charles W., and Larry Schweikart, “The Panic of 1857: Origins, Transmission, and
Containment,” Journal of Economic History, 51(4): 807-34, 1991.
Choi, James J., Laibson, David, Madrian, Brigitte C. and Andrew Metrick, “Reinforcement
Learning and Savings Behavior,” The Journal of Finance, 64(6): 2515-2534, 2009.
Ciliberto, Federico, and Elie Tamer. “Market Structure and Multiple Equilibria in the Airline
Industry,” Econometrica, 77(6): 1791-828, 2009.
Cohen, Andrew, and Michael Mazzeo, “Market Structure and Competition among Retail
Depository Institutions,” Review of Economics and Statistics, 89(1): 60-74, 2007.
Dwyer, Gerald P. Jr., “Wildcat Banking, Banking Panics, and Free Banking in the United
States,” Federal Reserve Bank of Atlanta Economic Review, 1996.
34
Economopolous, Andrew, and Heather O’Neill, “Bank Entry during the Antebellum Period,”
Journal of Money, Credit, and Banking, 27(4): 1071-85, 1995.
Esarey, Logan, State Banking in Indiana, 1814-1873, Indiana University Studies, 1912
Garleanu, Nicolae and Pedersen Lasse, “Margin-based Asset Pricing and Deviations from the
Law of One Price”, Review of Financial Studies, 24 (6): 1980-2022. (2011)
Geanakaplos, John, “Liquidity, default and crashes: Endogenous contracts in general
equilibrium” In M. Dewatripont, L. P. Hansen, and S. J. Turnovsky (Eds.), Advances in
Economics and Econometrics: Theory and Applications, Eighth World Conference, Volume 2,
Chapter 5, pp. 170–205. Econometric Society Monographs. 2003
Geanakaplos, John , “The Leverage Cycle,” in D. Acemoglu, K. Rogoff, and M. Woodford, eds.,
NBER Macroeconomics Annual 2009, Vol. 24, University of Chicago Press, Chicago (2010), pp.
1–65.
Gorton, Gary, “Reputation Formation in Early Bank Note Markets,” Journal of Political
Economy, 104(2): 346-97, 1996.
Gorton, Gary, Slapped by the Invisible Hand: The Panic of 2007, Oxford University Press, 2010
Gorton, Gary Misunderstanding Financial Crises: Why We Don't See Them Coming, Oxford
University Press, 2012
Hasan, Iftekhar, and Gerald P. Dwyer, Jr., “Bank Runs in the Free Banking Period,” Journal of
Money, Credit, and Banking, 26(2): 271-88, 1994.
Huntington, Charles Clifford, A history of banking and currency in Ohio before the Civil War, F.
J. Heer Printing, 1915. Print.
Homans, Isaac, A Cyclopedia of Commerce and Commercial Navigation, Harper, 1858
Jaremski, Matt, “Free Bank Failures: Risky Bonds versus Undiversified Portfolios,” Journal of
Money, Credit and Banking, 42: 1565-87, 2010.
Kindleberger, Charles P., A Financial History of Western Europe, Routledge: London, 1984.
Knox, John Jay, A History of Banking in the United States, B. Rhodes & Co.: New York, 1900.
Lanier, James and Lanier, Sidney, Sketch of the life of J. F. D. Lanier, 1877
Landon-Lane, John; Hugh Rockoff and Richard Steckel, “Droughts, Floods, and Financial
Distress in the United States” in The Economics of Climate Change edited by Gary Libecap and
Richard Steckel, University of Chicago Press, 2011
35
Manuszak, Mark D., “Endogenous Market Structure and Competition in the 19th Century
American Brewing Industry,” International Journal of Industrial Organization, 20(5): 673-92,
2002.
Martin, Joseph, Seventy-Three Years’ History of the Boston Stock Market, From January 1, 1798
to January 1, 1871. Published by the Author: Boston, 1871.
Ng, Kenneth, “Free Banking Laws and Barriers to Entry in Banking, 1838-1860,” Journal of
Economic History, 48(4): 877-89, 1988.
Osili, Una Okonkwo, and Anna Paulson, “Banking Crises and Investor Confidence: An
Empirical Investigation,” Federal Reserve Bank of Chicago Financial Markets Group Policy
Discussion Paper Series, 2009.
Reemelin, Charles, Appendix to annual report of Auditor of State. Series of reports made on the
condition of the Ohio stock banks, as ascertained by Charles Reemelin, Esq., acting as special
examiner under the appointment of the Auditor and Secretary of State. Statesman Steam Press:
Columbus, 1855.
Rockoff, Hugh, The Free Banking Era: A Reexamination. Arno Press: New York, 1975.
Rockoff, Hugh, “New Evidence on Free Banking in the United States,” American Economic
Review, 75(4): 886-89, 1985.
Rolnick, Arthur J., and Warren E. Weber, “New Evidence on the Free Banking Era.” American
Economic Review, 73(5): 1080-91, 1983.
Rolnick, Arthur J., and Warren E. Weber, “The Causes of Free Bank Failures.” Journal of
Monetary Economics, 14(3): 269-91, 1984.
Rolnick, Arthur J., and Warren E. Weber. “Banking Instability and Regulation in the U.S. Free
Banking Era.” Federal Reserve Bank of Minneapolis Quarterly Review 9 (Summer 1985): 2-9.
Weber, Warren E., “Early State Banks in the United States: How Many Were There and When
Did They Exist?,” Journal of Economic History, 433–55, June 2006.
Weber, Warren E., “Quoted Discounts on State Bank Notes in New York, Cincinnati, and
Cleveland, selected dates, 1827-1858.” Research Department, Federal Reserve Bank of
Minneapolis. http://minneapolisfed.org/research/economists/wewproj.html
Zhang, Zhihua, Michael E. Mann and Edward R. Cook, “Alternative methods of proxy-based
climate field reconstruction: application to summer drought over the conterminous United States
back to AD 1700 from tree-ring data,” The Holocene, 14(4): 502-516 2004.
36
Figure 1: Collateral Portfolios Held by State Regulators
1854
WI
Other
(4)
1857
Other
(10)
WI *
VA
MO
CA
WI *
NC
MO
IN
LA
MO
MI
TN
Est.
Circulation
Market
Value of
Collateral
Other
(5)
Est.
Circulation
Market
Value of
Collateral
Other
(6)
VA
VA
VA
LA
IN *
IN *
MO
Est.
Circulation
Market
Value of
Collateral
IL
Est.
Circulation
Market
Value of
Collateral
Other
(6)
Other
(5)
IL *
VA
IL *
TN
LA
VA
MO
Est.
Circulation
Market
Value of
Collateral
MO
Est.
Circulation
Market
Value of
Collateral
* Indicates home state bonds. Source: Author’s calculations from relevant state auditor reports and
New York and Boston stock market quotations.
Figure
2:
1854
Markets
f-r-71:
"'
i) "'------l£1'~.!~·~~
-/r :~ }""'",~
.-r---·~--- ~
~
/~
-~I
~
II
1
\
~ -~
·Y~
;I\
r.J
II/
•
"'
I
?~
~
~)
) '
rJ!r·\
.
.
"\\
l ~-J.
··"·· ( I
••• 't ~- .r' ··~
)
tilii"
·~
\ -·--·-----~
..........
.......
.
.
•
••
••; • r *.:. .,. ~
--
)
"
J•
I
(
'h
\
-
0
•
f..·
I
•• ••
--- r-\--·
.. ---·
•
•
I • :. • • ·\• :
~~·.
. ."j.. . .;>
·~
~
~(~-......s----LJ
,f
•\
• N• •
0
oO
I•
• ••
0
\
I
A , .__._..--,.-----.
• .,• •
-I '
-"!' • •. . . . ..
0 0
•
(
--' ...--
••
•
·- - - . . \ _{
I
.. ••
1}
0 0
IJI
,
0
l •
• .. '
•
•
•
•
0
...
•
•
e:·
•
. r -:
•• 0
\.!..-,.
,-
-:
0
~t.~
~ -9
'
~J=-ot-
).,_//
I\
• ,.,....
c.------7-------------i.
£
~q12.C.AU'l....-.,~ zeroBa~k~umberof8ank.s
/"
_-.
0
--
lEGEND
State
Townsb
*• 0
Two-+
Bank
40
~ eo
•
.,...
One 6ank
>20
t6o
Figure 3: 1860 Markets
• ••
~ ,.1
•• *•
.
*
I • •
l•
• •
•• •.....-•. ~. ~-.!'---....
•
••
••
• • •
••• •
' .
.-.
CJ
Slate by Humbe< of
Banlc$
Towns
One
a.nt
....
~
*o
Two• <10
-
ao
120
160
Table 1: Free Bank Regulations
Ohio
Capital
Min: $25,000 Max:
$500,000
Required 30% (in specie +
Reserves current notes)
US or Ohio State
Note
Bonds marked at
Collateral
min of market or
Rules
par value
Indiana
Illinois
Wisconsin
Min: $50,000
Min: $50,000
Min: $25,000 Max:
$500,000
12.5%
None
None
US or State Bonds
paying 6% marked at
min of Par or Market
Any current state
bond marked at min Value. Illinois state
securities marked at
of market or par.
After 1855: marked 80% of market value.
Bonds paying less
at 91% of min of
market or par value than 6% coupon
marked at 50% of
market value.
US or State Bonds
paying 6% marked
at min of market
or par. Instate RR
7% bonds marked
at 80% of NYSE
market value
Table 2: Bank balance sheet info, by state
Circulation
Deposits
Population Circ per cap Circ/Capital Spec/Notes
1854
1856
1858
1860
8,074,132
9,153,629
8,040,304
8,143,611
5,450,556
6,543,420
4,389,831
4,046,811
2,116,862
2,188,618
2,262,807
2,339,511
3.81
4.18
3.55
3.48
0.87
0.90
0.75
1.11
0.22
0.18
0.30
0.21
Indiana
1854
1856
1858
1860
8,165,856
4,731,705
5,379,936
5,755,201
2,289,605
1,852,742
1,723,840
1,841,051
1,119,831
1,191,952
1,268,718
1,350,428
7.29
3.97
4.24
4.26
1.02
0.70
0.75
0.82
0.15
0.22
0.19
0.15
Illinois
1854 2,283,526
1856 5,534,945
1858 5,707,045
1860 11,010,837
1,286,102
1,002,399
640,058
807,763
1,125,896
1,294,680
1,488,768
1,711,951
2.03
4.28
3.83
6.43
0.96
0.94
1.43
1.71
0.23
0.11
0.05
0.02
Wisconsin
1854
1856
1858
1860
1,482,053
3,365,562
3,022,384
4,083,131
443,437
534,342
643,884
775,881
1.67
3.19
7.29
5.56
0.63
0.58
0.59
0.58
0.31
0.32
0.15
0.09
Ohio
740,764
1,702,570
4,695,170
4,310,175
Source: Statement Exhibiting the Number of Banks, with Circulation, Deposits, and Amount Due to Other Banks, By States, for the Nine Years
Prior to January 1863. P.49-51 in "Report of the Comptroller of the Currency Dec 7, 1896" 54th Congress 2d Session, House Doc. No.10 Part1.
Table 3: Definition of Variables
Variable name
N
IN
Definition
Number of state registered banks in market
Binary variable indicating that market in state of Indiana
POP
Total (free and slave) population from US Census: linear interpolation for 1854,
as observed for 1860
Measure of cumulative drought severity, constucted from Palmer Drought
Severity Index (see Appendix)
Binary variable indicating that market in county with railroad line
Binary variable indicating that market in state that requires banks to have
charter from state legislature
Number of banks in market on August 1, 1854
Number of market's bank closures from Panic of 1854
Number of banks in market on September 1, 1857
Number of market's bank closures from Panic of 1857
Binary variable indicating that county had state bank branch (OH & IN only)
DROUGHT
RR
SCO
54BANKS
CLOS54
57BANKS
CLOS57
SB
Table 4: Summary statistics
#Mkts
8/1854 (254 mkts)
IN = 1
IN = 0
SCO = 0
SCO = 0
SCO = 1
49
140
65
12/1860 (362 markets)
IN = 1
IN = 0
SCO = 0
SCO = 0
SCO = 1
59
214
89
#Mkts|N=0
31
119
60
46
158
71
#Mkts|N=1
12
12
5
13
38
15
#Mkts|N=2
3
6
0
0
13
3
#Mkts|N>2
3
3
0
0
5
0
AvgPop
1876
1702
2072
2083
1634
1806
AvgPop|N=0
1427
1108
1650
1489
1113
1249
AvgPop|N=1
1658
2656
7141
4188
2903
2291
AvgPop|N=2
2146
7718
---
---
4075
12569
AvgPop|N>2
7125
9340
---
---
2110
---
Drought
0.56
0.39
0.49
1.78
2.53
2.78
Dro*Pop
1.13
0.65
0.90
3.83
3.51
4.18
Railroad
0.71
0.58
0.48
0.81
0.57
0.54
8/54 Banks
---
---
---
0.559
0.122
0.034
Clos54
---
---
---
0.288
0.005
0.000
9/57 Banks
---
---
---
0.373
0.196
0.034
Clos57
---
---
---
0.119
0.014
0.000
Notes: Counts MO & PA as SCO60=1, omits IA in 1854
Table 5: Separate cross sections (all states)
*Maximum likelihood estimates of ordered probit
Date
n
µ1/σt
8/1854
254
1
b
7.31
µ2/σt
0.41
µ3/σt
---
Pop
0.83
λ
/σt
Drought
µ
/σt
s.e.
2.49
8/1854
254
2
b
9.77
s.e.
2.09
12/1860
362
3
b
5.19
0.21
0.38
0.20
0.31
0.93
0.33
---
0.34
1.17
0.28
0.64
s.e.
1.06
12/1860
362
4
b
5.18
s.e.
1.04
0.13
0.32
0.13
0.54
0.20
0.15
0.64
0.15
-2.45
0.94
-1.63
0.72
-0.15
0.07
-0.14
0.07
Dro*Pop
0.73
0.36
0.26
0.18
0.02
0.02
0.01
0.02
IN
µ /σt
1.01
0.29
1.07
0.28
-0.49
0.22
-0.47
0.22
SCO
/σt
-0.98
0.38
-0.93
0.37
-0.29
0.20
-0.29
0.20
RR?
/σt
-0.18
0.26
-0.17
0.26
-0.50
0.18
-0.53
0.18
µ
/σt
µ
µ
/log-lik/
p: µ2=µ3
87.06
---
Implied entry thresholds
yM|IN=0, SCO=0
6555
M
M
D
D
T
T
y |IN=0, SCO=1
y |IN=1
y |IN=0, SCO=0
y |IN=1
y |IN=0, SCO=0
y |IN=1
97.24
0.06
201.57
---
213.70
0.18
4204
4305
1205
3267
898
3276
892
21334
23160
9518
4157
5143
2085
5120
2047
1936
536
1729
314
6991
3390
6831
3244
24631
28555
10802
5132
15736
9717
15872
9655
7274
5144
4337
1405
33669
27268
33095
26243
---
24399
16591
---
42667
36340
---
9797
5096
---
88966
91413
Notes: Only MO and PA treated as SCO=1 states in 1860; IA included only in 1860 sample
Table 6: Pooled cross sections (all states)
*Maximum likelihood estimates of ordered probit on pooled cross-sections
* 616 markets (254 in August 1854, 362 in December 1860)
1
2
b
s.e.
b
s.e.
M
10.91
1.43
10.88
1.40
D
0.72
0.18
0.68
0.17
T
---
1.17
0.26
µ
µ
µ
Pop
λ
σ60
1.36
1.84
0.19
0.22
1.35
1.76
0.19
0.20
µDro
-0.19
0.12
-0.15
0.12
Dro*Pop
0.01
0.03
-0.01
0.03
IN
0.85
0.24
0.88
0.23
IN*60
-1.80
0.49
-1.72
0.46
SCO
-0.80
0.26
-0.77
0.25
RR
-0.39
0.20
-0.42
0.19
loglik
300.32
322.69
---
0.0082
µ
µ
µ
µ
µ
D
T
p: µ =µ
Implied Entry Thresholds
M
3137
386
3106
367
M
5674
1093
5479
1010
M
1673
240
1625
229
M
6324
2086
5800
1780
D
8915
1986
8553
1776
D
4753
954
4476
865
D
17971
7239
15974
5956
T
---
16656
4631
T
---
8716
2209
T
---
31108
13346
y |IN=0, SCO=0
y |IN=0, SCO=1
y |IN54=1
y |IN60=1
y |IN=0, SCO=0
y |IN54=1
y |IN60=1
y |IN=0, SCO=0
y |IN54=1
y |IN60=1
Notes: σ54 normalized to 1; {MO PA} as SCO60=1; IA dropped from 1854 sample
Table 7: 1860 Cross-section with '57 panic closures (n = 301) … all states (except IA)
*Maximum likelihood estimates from ordered probit, scaled by σ60
1
3
4
5
s.e.
b
s.e.
b
s.e.
b
s.e.
b
s.e.
M
5.21
1.12
5.20
1.10
3.20
1.09
2.69
1.10
2.72
1.11
D
0.30
0.13
0.30
0.13
0.56
0.17
0.58
0.17
0.58
0.17
T
---
0.57
0.21
0.96
0.25
1.00
0.26
1.01
0.26
µ
µ
µ
λ
2
b
Pop
0.67
0.16
0.67
0.15
0.38
0.15
0.31
0.15
0.31
0.15
Dro
0.01
0.09
0.02
0.08
0.09
0.08
0.07
0.08
0.08
0.08
Dro*Pop
0.01
0.02
0.00
0.02
-0.02
0.02
-0.01
0.02
-0.01
0.02
RR?
-0.63
0.19
-0.66
0.19
-0.73
0.19
-0.74
0.19
-0.74
0.19
IN
-0.61
0.22
-0.59
0.22
-0.80
0.24
-0.69
0.25
-0.67
0.25
SCO
-0.59
0.22
-0.58
0.22
-0.36
0.22
-0.34
0.22
-0.34
0.22
0.17
1.21
0.19
1.21
0.19
0.40
-0.71
0.71
0.75
µ
µ
µ
µ
µ
PrePanicBanks
---
---
1.02
#57Clos
---
---
---
-0.88
#57Clos*IN
---
---
---
---
-0.22
182.27
194.04
172.96
170.40
170.36
µ
µ
µ
loglik
Notes: Treats {MO, PA} as SCO=1 states, Iowa excluded (no banks in 1854 or 1857)
Table 8: 1860 Cross-section with statewide panic failure rates (n = 301)
*Maximum likelihood estimates of ordered probit by individual cross-sections
1
2
3
4
µ /σ60
b
s.e.
5.2020 1.0958
b
5.1268
s.e.
1.0884
b
5.1441
s.e.
1.0927
b
5.7267
s.e.
1.1299
µD/σ60
0.2996 0.1327
0.2642
0.1278
0.2768
0.1296
0.3247
0.1359
µ /σ60
0.5724 0.2083
0.5066
0.2005
0.5310
0.2035
0.6137
0.2128
Pop
/σ60
0.6709 0.1518
0.6390
0.1503
0.6499
0.1511
0.6995
0.1546
Dro
µ /σ60
0.0195 0.0842
0.0162
0.0842
0.0343
0.0851
-0.0629
0.0896
Dro*Pop
0.0016 0.0220
0.0016
0.0219
0.0010
0.0219
0.0040
0.0222
IN
-0.5912 0.2199
---
-0.5845 0.2150
-0.4275
M
T
λ
µ
/σ60
µ /σ60
SCO
µ
/σ60
StFailRate
µ
/σ60
StFailRate*IN
µ
RR
/σ60
--0.2067
-0.6376
0.2504
0.0600
0.3171
0.6723
4.9185
1.7798
-5.0237
1.3888
-0.6249
0.1873
---
---
-0.9999
---
---
---
µ /σ60
-0.6578 0.1862 -0.6783
/log-lik/
194.04
0.1847
197.84
Notes: Treats {MO, PA} as SCO=1 states, Iowa excluded (no banks in 1854 or 1857)
---
-0.6756
196.72
0.1853
190.20
Working Paper Series
A series of research studies on regional economic issues relating to the Seventh Federal
Reserve District, and on financial and economic topics.
Comment on “Letting Different Views about Business Cycles Compete”
Jonas D.M. Fisher
WP-10-01
Macroeconomic Implications of Agglomeration
Morris A. Davis, Jonas D.M. Fisher and Toni M. Whited
WP-10-02
Accounting for non-annuitization
Svetlana Pashchenko
WP-10-03
Robustness and Macroeconomic Policy
Gadi Barlevy
WP-10-04
Benefits of Relationship Banking: Evidence from Consumer Credit Markets
Sumit Agarwal, Souphala Chomsisengphet, Chunlin Liu, and Nicholas S. Souleles
WP-10-05
The Effect of Sales Tax Holidays on Household Consumption Patterns
Nathan Marwell and Leslie McGranahan
WP-10-06
Gathering Insights on the Forest from the Trees: A New Metric for Financial Conditions
Scott Brave and R. Andrew Butters
WP-10-07
Identification of Models of the Labor Market
Eric French and Christopher Taber
WP-10-08
Public Pensions and Labor Supply Over the Life Cycle
Eric French and John Jones
WP-10-09
Explaining Asset Pricing Puzzles Associated with the 1987 Market Crash
Luca Benzoni, Pierre Collin-Dufresne, and Robert S. Goldstein
WP-10-10
Prenatal Sex Selection and Girls’ Well‐Being: Evidence from India
Luojia Hu and Analía Schlosser
WP-10-11
Mortgage Choices and Housing Speculation
Gadi Barlevy and Jonas D.M. Fisher
WP-10-12
Did Adhering to the Gold Standard Reduce the Cost of Capital?
Ron Alquist and Benjamin Chabot
WP-10-13
Introduction to the Macroeconomic Dynamics:
Special issues on money, credit, and liquidity
Ed Nosal, Christopher Waller, and Randall Wright
WP-10-14
Summer Workshop on Money, Banking, Payments and Finance: An Overview
Ed Nosal and Randall Wright
WP-10-15
Cognitive Abilities and Household Financial Decision Making
Sumit Agarwal and Bhashkar Mazumder
WP-10-16
1
Working Paper Series (continued)
Complex Mortgages
Gene Amromin, Jennifer Huang, Clemens Sialm, and Edward Zhong
WP-10-17
The Role of Housing in Labor Reallocation
Morris Davis, Jonas Fisher, and Marcelo Veracierto
WP-10-18
Why Do Banks Reward their Customers to Use their Credit Cards?
Sumit Agarwal, Sujit Chakravorti, and Anna Lunn
WP-10-19
The impact of the originate-to-distribute model on banks
before and during the financial crisis
Richard J. Rosen
WP-10-20
Simple Markov-Perfect Industry Dynamics
Jaap H. Abbring, Jeffrey R. Campbell, and Nan Yang
WP-10-21
Commodity Money with Frequent Search
Ezra Oberfield and Nicholas Trachter
WP-10-22
Corporate Average Fuel Economy Standards and the Market for New Vehicles
Thomas Klier and Joshua Linn
WP-11-01
The Role of Securitization in Mortgage Renegotiation
Sumit Agarwal, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet,
and Douglas D. Evanoff
WP-11-02
Market-Based Loss Mitigation Practices for Troubled Mortgages
Following the Financial Crisis
Sumit Agarwal, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet,
and Douglas D. Evanoff
WP-11-03
Federal Reserve Policies and Financial Market Conditions During the Crisis
Scott A. Brave and Hesna Genay
WP-11-04
The Financial Labor Supply Accelerator
Jeffrey R. Campbell and Zvi Hercowitz
WP-11-05
Survival and long-run dynamics with heterogeneous beliefs under recursive preferences
Jaroslav Borovička
WP-11-06
A Leverage-based Model of Speculative Bubbles (Revised)
Gadi Barlevy
WP-11-07
Estimation of Panel Data Regression Models with Two-Sided Censoring or Truncation
Sule Alan, Bo E. Honoré, Luojia Hu, and Søren Leth–Petersen
WP-11-08
Fertility Transitions Along the Extensive and Intensive Margins
Daniel Aaronson, Fabian Lange, and Bhashkar Mazumder
WP-11-09
Black-White Differences in Intergenerational Economic Mobility in the US
Bhashkar Mazumder
WP-11-10
2
Working Paper Series (continued)
Can Standard Preferences Explain the Prices of Out-of-the-Money S&P 500 Put Options?
Luca Benzoni, Pierre Collin-Dufresne, and Robert S. Goldstein
Business Networks, Production Chains, and Productivity:
A Theory of Input-Output Architecture
Ezra Oberfield
WP-11-11
WP-11-12
Equilibrium Bank Runs Revisited
Ed Nosal
WP-11-13
Are Covered Bonds a Substitute for Mortgage-Backed Securities?
Santiago Carbó-Valverde, Richard J. Rosen, and Francisco Rodríguez-Fernández
WP-11-14
The Cost of Banking Panics in an Age before “Too Big to Fail”
Benjamin Chabot
WP-11-15
Import Protection, Business Cycles, and Exchange Rates:
Evidence from the Great Recession
Chad P. Bown and Meredith A. Crowley
WP-11-16
Examining Macroeconomic Models through the Lens of Asset Pricing
Jaroslav Borovička and Lars Peter Hansen
WP-12-01
The Chicago Fed DSGE Model
Scott A. Brave, Jeffrey R. Campbell, Jonas D.M. Fisher, and Alejandro Justiniano
WP-12-02
Macroeconomic Effects of Federal Reserve Forward Guidance
Jeffrey R. Campbell, Charles L. Evans, Jonas D.M. Fisher, and Alejandro Justiniano
WP-12-03
Modeling Credit Contagion via the Updating of Fragile Beliefs
Luca Benzoni, Pierre Collin-Dufresne, Robert S. Goldstein, and Jean Helwege
WP-12-04
Signaling Effects of Monetary Policy
Leonardo Melosi
WP-12-05
Empirical Research on Sovereign Debt and Default
Michael Tomz and Mark L. J. Wright
WP-12-06
Credit Risk and Disaster Risk
François Gourio
WP-12-07
From the Horse’s Mouth: How do Investor Expectations of Risk and Return
Vary with Economic Conditions?
Gene Amromin and Steven A. Sharpe
WP-12-08
Using Vehicle Taxes To Reduce Carbon Dioxide Emissions Rates of
New Passenger Vehicles: Evidence from France, Germany, and Sweden
Thomas Klier and Joshua Linn
WP-12-09
Spending Responses to State Sales Tax Holidays
Sumit Agarwal and Leslie McGranahan
WP-12-10
3
Working Paper Series (continued)
Micro Data and Macro Technology
Ezra Oberfield and Devesh Raval
WP-12-11
The Effect of Disability Insurance Receipt on Labor Supply: A Dynamic Analysis
Eric French and Jae Song
WP-12-12
Medicaid Insurance in Old Age
Mariacristina De Nardi, Eric French, and John Bailey Jones
WP-12-13
Fetal Origins and Parental Responses
Douglas Almond and Bhashkar Mazumder
WP-12-14
Repos, Fire Sales, and Bankruptcy Policy
Gaetano Antinolfi, Francesca Carapella, Charles Kahn, Antoine Martin,
David Mills, and Ed Nosal
WP-12-15
Speculative Runs on Interest Rate Pegs
The Frictionless Case
Marco Bassetto and Christopher Phelan
WP-12-16
Institutions, the Cost of Capital, and Long-Run Economic Growth:
Evidence from the 19th Century Capital Market
Ron Alquist and Ben Chabot
WP-12-17
Emerging Economies, Trade Policy, and Macroeconomic Shocks
Chad P. Bown and Meredith A. Crowley
WP-12-18
The Urban Density Premium across Establishments
R. Jason Faberman and Matthew Freedman
WP-13-01
Why Do Borrowers Make Mortgage Refinancing Mistakes?
Sumit Agarwal, Richard J. Rosen, and Vincent Yao
WP-13-02
Bank Panics, Government Guarantees, and the Long-Run Size of the Financial Sector:
Evidence from Free-Banking America
Benjamin Chabot and Charles C. Moul
WP-13-03
4
@FedFRASER