Bank Fragility, “Money under the Mattress,” and Long-Run Growth
U.S. Evidence from the “Perfect” Panic of 1893
Carlos D. Ramírez *
Department of Economics
George Mason University
Fairfax, VA 22030-4444
[email protected]
Visiting Fellow
Center for Financial Research
FDIC
Revised: May 2008
JEL Classification Codes: N21, O16
Keywords:
Bank Failures, Panic of 1893, Convergence, Finance-Growth Nexus, Nebraska, West
Virginia, Deposits, “Money Hidden”
* I would like to thank, without implicating, Sean Campbell, Bob DeYoung, Paul Kupiec, Dan Nuxoll, Haluk
Unal, and seminar participants at the FDIC’s Division of Insurance and Research Workshop, as well as seminar
participants at the 2008 WAFA conference for helpful comments and suggestions. I would also like to thank
the FDIC’s Center for Financial Research for financial and logistical support. The views expressed in this paper
do not necessarily reflect those of the Center for Financial Research or the FDIC.
Bank Fragility, “Money under the Mattress,” and Long-Run Growth
U.S. Evidence from the “Perfect” Panic of 1893
Abstract
An issue that has received an increasing amount of attention has been the effect of financial
sector conditions (and of banking in particular) on long-run economic growth. In this paper,
I examine how the remnants of the U.S. financial crisis of 1893 (the “Perfect” Panic),
manifested through the incidence of bank failures, affected state output growth between
1900 and 1930. Using standard growth convergence regressions I find that the “elasticity” of
bank instability with respect to output growth is approximately 5 percent. This result
survives the inclusion of the usual factors known to influence long-term growth, including
initial measures of financial depth, and regulatory measures such as branch banking. In order
to show how financial disintermediation affects long-run growth, this paper compares two
extreme cases: Nebraska, with one of the highest bank failure rates, and West Virginia, which
did not experience a single bank failure during this panic. This comparison reveals that the
avenue by which disintermediation affected growth was through a long-lasting decline in the
state’s depository base (people simply stop trusting banks). Time series evidence from
newspaper articles printed between 1860 and 1970 confirms this result. I find that articles
with the words “money hidden” significantly increase during and after banking crises, and
die off very slowly over time. Taken together, the results imply that banking crises appear to
be particularly costly not just because they may amplify a business cycle downturn, but
because they, in the absence of institutions that restore confidence in the banking system,
may compromise long-run growth as well.
JEL Classification Codes: N21, O16
Keywords: Bank Failures, Panic of 1893, Convergence, Finance-Growth Nexus, Nebraska,
West Virginia, Deposits, “Money Hidden”
2
“[Mr.] Syndowsky had once lost some money through the failure of a savings
bank and was consequently possessed of a rooted abhorrence of all banks.
So, despite the urging of his wife and children, he refused to put his wealth
[$6,500] in a bank. Instead, he kept it in a little safe in his apartment at night
and in the day time he carried it about in his wallet.” [Syndowsky’s Savings,
Banker’s Magazine, November 1912, p. 536] 1
1. Introduction
Over the last two decades a substantial amount of research has investigated the
effect of financial development on long-term growth. 2 This literature generally finds that a
country’s long-run growth rate appears to be an increasing function of the country’s level of
financial development, leading to the conclusion that stimulating financial development can
be helpful for promoting growth. Although the finance-growth nexus literature is
voluminous, numerous other papers have questioned the empirical validity of such
relationship. Well-known economists Joan Robinson and Robert Lucas are frequently cited
in the literature as expressing skepticism over this relationship. 3 More recently, many other
economists have expressed doubt about the robustness of such relationship (Watchel 2003;
Manning, 2003). In fact, the disbelief has permeated the literature enough that Levine, a
leading proponent of the finance-growth relationship, mentions in his very comprehensive
survey: “We are far from definitive answers to the questions: Does finance cause growth,
and if it does, how?” (Levine, 2005, p.3).
This paper offers a modest contribution to this literature by studying the experience
of the U.S. at the turn of the 19th century. In particular, it focuses on the Panic of 1893 to
According to Williamson (2008), $6,500 in 1912 is equivalent to $143,375 in 2007 using the CPI.
Since King and Levine (1993)’s well known study, the empirical literature on this issue has essentially
ballooned. Hence, it is not practical to even attempt to provide a list here. For a comprehensive survey see
Levine (2005).
3 See, for example, Levine (2005) and Rousseau and Watchel (2005), who cite Robinson (1952) and Lucas
(1988), among others.
1
2
3
investigate the long-run growth consequences of financial disintermediation. The logic for
studying periods of financial disintermediation follows from the observation that if financial
development enhances growth, a country that experiences financial disintermediation should
experience, besides a temporary downturn in its business cycle, a decline in long-term
growth as well. Thus, one could evaluate the finance-growth association by studying the
long-run growth patterns of countries that endure episodes of financial disintermediation at
some point in their past. Episodes of financial or banking crises are clear candidates as
during these crises financial markets are disrupted and many financial institutions disappear.
The focus on historical panics to investigate the finance-growth nexus should be
straightforward—clearly enough time has elapsed in order to ascertain with enough statistical
precision it long term growth consequences. This is not to say that the numerous episodes of
banking crises around the world over the last three decades have not attracted the attention
of policymakers and academic researchers. They, in fact, have. 4 Indeed, a stylized fact that
emerges out of this research is that countries that experience a financial or banking crisis
endure a significant decline in economic activity in the years immediately following the crisis
(Dell’ Ariccia, Detragiache, and Rajan, 2008; Goldstein and Turner, 1996; Kaminsky and
Reinhart, 1999; Kroszner, Klingebiel, and Laeven, 2007). But this research focuses more on
the short-term macroeconomic consequences of banking crises and not on its long-term
growth consequences. 5 To study the long-term effects, one must rely on historical episodes
of banking crises.
The Panic of 1893 is particularly useful for studying this issue because, although the
panic itself was relatively short-lived, the incidence of bank failures it triggered continued
well into the 1890s, engulfing most of the states in the US by 1896. To show how this
4
5
For a comprehensive study see Demirguc-Kunt and Detragiache (2005).
Edwards (2007) is an exception.
4
adverse financial shock affects growth, I focus on the growth experience of the different
states in the U.S. between 1900 and 1930. 6 In particular, I estimate standard growth
convergence equations augmented by the inclusion of a variable that measures the aggregate
amount of failed bank liabilities relative to total bank deposits. The main results indicate that
a one percent increase in the incidence of bank failures reduced growth by about 5 percent
between 1900 and 1930. The magnitude of this elasticity is as large as that of eliminating
geographical restrictions on branch banking. In addition, the estimated elasticity is robust to
the inclusion of the standard set of controls, such as initial income per capita, measures of
education, and even initial levels of financial development. The results imply that the cost of
banking instability go beyond the short-term macroeconomic consequences the literature has
highlighted. They may also help to explain why it takes so long to restore growth among
countries that experience banking crises.
Convergence regressions may be suggestive, but, on their own, may not be fully
persuasive. 7 To complement the evidence, this paper provides an explanation, based on
theoretical underpinnings, of why financial disintermediation adversely affects long-run
growth. Two empirical tests, one based on a case study, and the other, on time series
evidence from newspaper articles, are provided in support of the explanation.
The intuition behind the explanation of why financial disintermediation affects
growth is straightforward. In the absence of deposit insurance or any other institutional
arrangement that induces confidence on the banking system, depositors who experience
losses or whose money becomes illiquid, even temporarily, may become reluctant to keep
their money in the banking system. They simply stop “trusting” banks. This lack of trust may
affect all depositors, including those that did not experience losses. With a high enough
6
7
The choice of this time period is explained in the next section of the paper.
The main drawbacks of standard growth convergence regressions are summarized further below.
5
degree of risk aversion and a high enough probability of a bank run or failure depositors may
be induced to reshuffle their liquid asset portfolio away from the banking system. 8 To the
extent that the panic induces a portfolio change in asset holdings away from the banking
system and into more rudimentary forms of savings, such as keeping the money “under the
mattress,” financial intermediation, and thus, growth are adversely affected. 9
Studying the cases of Nebraska and West Virginia is particularly illuminating in this
respect because it contrasts the experience of a state that did not suffer a single bank failure
during the panic (West Virginia) with that of a state that experienced one of the highest bank
failure rates during that period (Nebraska). Despite the fact that Nebraska was a much
wealthier state than West Virginia in the 1890s, it suffered a very drastic decline in its level of
deposits, and it did not recuperate from it until the 1920s, after it introduced state deposit
insurance. Thus, relative to West Virginia, Nebraska experienced a long-term decline in its
deposits, a finding consistent with the hypothesis that the panic induced a change in the
portfolio allocation of liquid assets.
The second piece of evidence supporting the explanation comes from newspaper
articles. In particular, I constructed a yearly index of newspaper articles with the words
“money hidden” in them to gauge the incidence of general distrust in the banking system.
The newspaper index starts in 1860 and ends in 1970, spanning all major historical banking
crises in the U.S. The vast majority of these articles tell tales of the extent people went
through to hide their savings in cash, often because they simply did not trust banks. The
time series evidence suggests three things: 1. That the incidence of “money hidden” articles
spikes during banking crises years; 2. The magnitude of the spike is substantial (about 80
percent increase within two years of the crisis); and 3. That the effect remains large in the
8
9
The quote at the beginning of the paper attests to the significance of this intuitive explanation.
Section 3 below provides a more detailed discussion and the relevant references.
6
long-run and appears to die off after the U.S. introduced deposit insurance at the national
level, clearly aiming at restoring confidence in the banking system.
The rest of the paper is organized as follows. Section 2 presents a brief overview of
the Panic of 1893, which explains in more detail why is it the case that focusing on this panic
is particularly useful for investigating the hypothesis that financial disintermediation reduces
long-term growth. Section 3 explains in more detail the theoretical underpinnings of how
financial disintermediation can adversely affect growth in the short-run and the long-run.
Section 4 presents the data and the empirical methodology behind the convergence
regressions, while section 5 discusses the growth regression results. Section 6 provides the
details of the case study (Nebraska versus West Virginia). Section 7 discusses the time series
evidence from newspaper articles, and lastly, section 8 offers some concluding remarks.
2. The “Perfect” Panic of 1893
As indicated in the introduction, this paper uses the Panic of 1893 to study the longterm growth consequences of financial disintermediation. I call it the “perfect” panic (in
allusion to the “The Perfect Storm”) because its unique characteristics make it particularly
useful for investigating this issue. There are several reasons for this. First, as Hoffman (1956)
points out, this depression was amongst the most devastating crises in U.S. history,
regardless of the criteria. Although it was not as damaging as the Depression of the 1930s, it
was large enough to cause serious economic disruption. For example, using Romer’s (1986)
conservative estimates, unemployment increased from about 4 percent in 1890 - 1892 to
over 11 percent between 1893 and 1898. The increase in bank suspensions and failures was
even more dramatic. Between 1890 and 1892, less than 1 percent of banks failed or
temporarily suspended operations. In 1893, over 5 percent suspended or failed and this
7
proportion did not return to pre-1893 levels until after 1897. 10 In addition, Hoffman (1956)
reports that business failures increased by as much as 50 percent between 1890 and 1893,
and by 1894, some 156 railroad companies, comprising approximately 30,000 miles, were
bankrupt. Thus, there is little doubt that the Panic of 1893 was disruptive enough to be
considered a financial shock that resulted in a significant amount of disintermediation,
making it a prime candidate for evaluating the hypothesis set out in the introduction.
Second, bank failures were widespread, affecting all but a handful of states. The first
column of Table 2 reports, by state, the number of banks that failed or temporarily
suspended operations between 1894 and 1896. The second column reports the aggregate
liabilities (from 1894 to 1896) of those institutions relative to total state deposits in 1896.
The figures reveal that the incidence of failures was indeed quite disperse, ranging from 0 in
six states to 52 in Nebraska. In terms the liabilities, the dispersion was equally wide—from 0
to nearly 35 percent in Washington. The fact that it was so diffused is in some sense “good”
because it allows for the identification of the effect of financial disintermediation on growth.
In addition, it suggests that it was not connected to problems in any specific sector of the
economy. For example, and unlike the banking failures that took place during the early
1920s, the banking failures of the 1890s were not directly linked to problems in the
agricultural sector.
This observation is bolstered by the fact that there appears to be no specific cause
for the Panic. Contemporary observers speculated that it was prompted by an attack on the
exchange rate (Noyes 1909, Stevens 1894). As concerns over the ability of the government
to maintain gold parity grew, investors rushed to convert their deposits into gold, much like
modern-day market participants do when they attack an exchange rate regime in response to
Statistics on bank failures and suspensions were computed using figures from the U.S. Bureau of the Census
(1975), Series X741, p. 1038.
10
8
anticipated currency depreciations. Other observers, however, put more emphasis on
domestic market conditions as the primary cause for the Panic (Sprague, 1910; Friedman and
Schwartz, 1963). More specifically, they argue that problems in the real sector, such as a
slowdown in railroad investment, the failure of a few large railroad companies, and turmoil
in the stock market in May of 1893, precipitated the failure of a few large banks, which
triggered the panic in June. However, there is no general consensus about this among
economic historians. Wicker (2000), for example, argues that the banking panic that took
place between June and August of 1893 was independent of the uncertainty surrounding the
stock market in May. In addition, he notes that the panic appears to have started with
smaller institutions in the interior of the country, spreading elsewhere a short time after that.
In fact, the identification of the specific causes of the Panic continues to be an area of active
research in economic history. For example, Carlson (2005) uses the 1893 panic episode to
test two competing theories of bank panics—the Diamond and Dybvig (1983) self-fulfilling
theory of bank panics and the real shock, asymmetric information theory of bank panics
advanced by Calomiris and Gorton (1991). He finds that actually both theories carry some
empirical support. Dupont (2007) finds that the bank panic that affected Kansas in 1893
took place despite the fact that newspapers provided ample information about the financial
quality of local banks. Thus, undoubtedly, the Panic of 1893 can be attributed self-fulfilling
runs as well as to real and financial markets disruptions domestically and maybe even abroad.
What is crucial for the purposes of this study is that, regardless of the underlying causes of
the panic, once having started, all states were vulnerable to (and indeed most of them
suffered from) it.
Third, the federal government did not react to it by implementing reforms aiming at
alleviating the social inequities caused by the economic calamity of the period. This stands in
9
sharp contrast to what happened after the Great Depression of the 1930s. The regulatory
infrastructure and the far-reaching social programs of the 1930s, such as social security and
federal deposit insurance, were implemented in part as a reaction to the conditions of that
period. Hence, the economic consequences of the financial shock of the 1890s can be
analyzed without the “contamination” of social welfare programs imposed at the federal
level and can be viewed as a “clean” experiment on the long-term economic consequences
of financial disintermediation.
Fourth, unfortunately it may not be possible to focus on other panics to test the
hypothesis that financial disintermediation reduces long-term growth despite the fact that
there were many others (Calomiris and Gorton, 1991; Wicker, 2000). To begin with, there is
no annual data on the incidence and liabilities of bank failures broken down by state before
1894. The main source of data used here, Dun’s Review, started publishing bank failure figures
consistently in that year. Bradstreet’s, the source that Wicker (2000) consults, does not break
down data at the state level and does not go back to the 1880s and 1870s either. 11 Calomiris
and Gorton (1991) consult Hunt’s Merchant magazine to obtain bank failure data for some
panics in the anti-bellum period. However, this publication lasted from 1839 to 1861. Hence,
it is not possible to focus on the relatively large banking panics of the 1870s and 1880s.
After the 1890s, only two serious panic episodes merit consideration as potential
candidates: the Panic of 1907 and the panics of the Great Depression. Focusing on the
panics of the Great Depression is not practical because, as already discussed, the
intervention of the federal government “contaminates” the test. This leaves only the Panic
of 1907 as a candidate. Unfortunately (or, rather, fortunately), while this panic was very
Calomiris and Gorton (1991) consult Hunt’s Merchants’ Magazine, which reports the incidence of failures
and suspensions for various states. Unfortunately, this publication began in 1839 and stopped in 1861.
11
10
serious, it was mostly concentrated to trust companies in New York City and had limited
fallout effects elsewhere (Moen and Tallman, 1992; Wicker, 2000). Hence, it was not
“widespread” enough to render the hypothesis testable.
3. Consequences of Bank Failures and Financial Disintermediation: Theory
The banking literature emphasizes several ways through which banking crises affect
macroeconomic performance. These can be roughly divided into those that emphasize
short-term effects, and those that highlight long-term consequences.
A. Short-Term Effects
Most papers that estimate the effect of banking crises underscore the “credit crunch”
hypothesis as the primary avenue through which banking distress affect the real side of the
economy. This hypothesis, which is directly or indirectly based on the work of Bernanke
(1983), refers to the notion that an adverse shock to the banking sector permeates into the
real sector via a reduction in bank lending to customers who are heavily dependent on bank
loans. For example, a decline in commercial lending affects the financial health of bankdependent firms, and may even force some of them into bankruptcy. In addition, a decline
in consumer lending tends to aggravate a slump in aggregate demand as banks tighten
consumer credit.
Evidently, these effects tend to operate in the short-run, and work by amplifying a
business cycle downturn. But as soon as banking conditions are normalized, credit is
restored, and, in theory, the economy should return to its pre-banking distress level of
activity. Although bank failures tend to delay the speed at which the economy recovers from
its slump, the effects, at least according to these theories, are temporary.
B. Long-Run
11
Banking crises and failures can have long term consequences as well. A substantial
literature on finance, growth, and development has established that financial intermediaries
can help foster economic growth (Cameron, 1967; Goldsmith, 1969; McKinnon, 1973;
Shaw, 1973; Levine, 2005). Some papers have even established this relationship from a
theoretical standpoint (e.g. Bencivenga and Smith, 1991). Naturally, this implies that financial
disintermediation will tend to retard growth. But how do banking crises can translate into a
long-lasting financial disintermediation problem? In the absence of deposit insurance, or
another institutional mechanism that promotes or restore financial sector confidence,
depositors who experience the adverse consequence of a crisis (such as a partial or
temporary loss of deposits) will tend to be very reluctant to keep their money in the banking
system, when they finally retrieve it (if they ever do). To the extent that crises induce a
portfolio change in savings away from banks and more into rudimentary forms of savings
such as money “under the mattress,” financial intermediation, and hence long-term growth is
compromised. Under such a scenario, even the mere probability of having a bank run can
affect long-run growth. Ennis and Keister (2003) use a theoretical model to show how this
can happen. They set up a stylized endogenous growth model in which bank runs (a la
Diamond and Dybvig, 1983) can take place in equilibrium. Their model demonstrates that
the probability of bank runs can affect the stock of capital and growth. Below I provide
some evidence consistent with this explanation.
4. Empirical Methodology
One approach followed in the finance-growth nexus literature is to estimate growth
convergence regressions where the growth rate of income per capita is regressed on several
control variables such as: initial income per capita, measures of initial levels of education,
12
measures of initial levels of financial depth, and, occasionally, regulatory variables
hypothesized to influence growth. 12 Studies that investigate the robustness of the financegrowth nexus using cross-country data must also control for a large host of conditions such
as political, legal, regulatory, and even institutional differences across countries in order to
make appropriate inferences about the role of finance in economic growth. 13 The statistical
finding that finance precedes long-run growth without controlling for these differences may
be misleading as it may just be the result of an omitted variable bias. But adding more
controls to correct for this bias necessarily reduces the degrees of freedom because the
number of countries (observations) is limited. Moreover, the controls may be correlated with
each other, making it more difficult to ascertain with statistical precision which is the true
determinant. In the end, researchers typically settle for the inclusion of a “large enough” set
of controls that does not comprise the efficiency of the estimates by consuming degrees of
freedom. 14 These econometric challenges are mitigated when the research is performed in
homogeneous regions, such as the different states in the U.S., a strategy that motivates the
research work that looks at regions within a country, and motivates this paper as well. 15 For
example, focusing on homogeneous regions reduces the necessity to control for differences
in political and legal aspects, saving on the number of degrees of freedom consumed, and
ameliorating the multicollinearity problem.
Thus, to test the effect of financial disintermediation on growth, I fit a regression of
the growth rate of income per capita on the standard controls, and augment it by including
The original paper that investigates the effect of financial development on growth using growth convergence
regressions is King and Levine (1993). Levine (2005) offers a comprehensive survey.
13 For a comprehensive list of controls included see Durlauf, Johnson, and Temple (2005) as well as Sala-iMartin, Doppelhofer, and Miller (2004).
14 These econometric problems are summarized in Acemoglu (2008), Chapter 2.
15 See, for example, Barro and Sala-i-Martin (1991) and Mitchener and McLean (1999), and more recently,
Higgins, Levy, and Young (2006).
12
13
the liabilities of failed banks relative to total deposits, which I use as a measure of banking
instability. Formally, the regression takes the following form:
yˆ i ,1900 −1930 = α 0 + α 1 y i ,1900 + α 2 illri ,1900 + α 3 bf i ,1894 −96 + α 4 bri ,1900 + α 5 bd i ,1900 + α 6 fd i ,1900 + (1
)
Where yˆ i ,1900 −1930 is the growth rate of income per capita between 1900 and 1930 in state i;
y i ,1900 is the log of the 1900 level of income per capita in state i; illri ,1900 is the ratio of
illiterate persons over 21 years of age divided by total population over 21 years of age;
bf i ,1894 −96 represents bank instability and it is defined as the sum of the liabilities of bank
failures for 1894, 1895, and 1896, all divided by total deposits in 1896 in each state; bri ,1900 is
an indicator variable equal to 1 if the state permitted branching in 1900, and equal to 0
otherwise; bd i ,1900 represents bank density and is defined as the total number of banks in the
state in 1900 divided by the state’s area in square miles; fd i ,1900 represents financial depth
and is defined as the 1900 level of deposits per capita divided by the 1900 level of income
per capita; and ε i represents the error term. 16
The first independent variable included, the log of the 1900 level of income per
capita, captures the degree of convergence across states. This variable is expected to be
negative, indicating that, all else constant, poorer states should grow faster than richer ones.
Given that the 48 states can be viewed as 48 countries with full labor and capital mobility, I
expect this coefficient to be larger than the one estimated using cross-section data. The
second variable, illiteracy rate, is a proxy for education, and it is included to control for the
influence of human capital on growth. Both of these variables are among the most prevalent
ones in standard growth regressions (Sala-i-Martin, Doppelhofer, and Miller, 2004; Darlauf,
16
A more detailed description of the data, including sources, is presented in the Data Appendix.
14
Johnson, and Temple, 2005). The remaining independent variables included capture the
influence of financial market conditions.
Banking instability (represented as bf in equation 1) is the variable of interest in this
paper. It is hypothesized to have a negative effect on growth. As discussed in the
introduction, recent research has documented the fact that banking crises are typically
followed by a dramatic slowdown in economic activity, especially in the years around the
banking crises. The inclusion of banking instability in equation 1 tests whether the decline in
economic activity persists in the long run.
A statistical relationship between banking instability and growth, by itself, is not
necessarily indicative that a true association exists. It is possible that such a relationship may
be driven by other factors, such as regulatory differences across states. For example,
Calomiris and Gorton (1991) show that banking panics and failures were more likely to have
taken place in unit-banking states. Moreover, Dehejia and Lleras-Muney (2007) find that
states that permitted bank branching also enjoyed faster growth rates between 1919 and
1940. These two set of findings would imply that the relationship between banking instability
and growth may be driven by state-level differences in branch banking restrictions. For this
reason, equation (1) includes a bank branching indicator variable. 17
The remaining two variables included in equation (1), bank density and financial
depth, control for the effect of the level of financial development on long-term growth.
Their inclusion is motivated by the empirical cross-country growth regressions literature
Another potential regulatory variable of interest is state-sponsored deposit insurance. Before the
establishment of federal deposit insurance in 1934, 8 states experimented with their own deposit insurance
schemes. The regressions do not test for the influence of these schemes as they were implemented between
1908 and 1929. Since the dependent variable is the growth rate between 1900 and 1930, its inclusion in the
regressions would introduce an obvious endogeneity problem. Dehejia and Lleras-Muney (2007) test for the
effect of deposit insurance on growth between 1919 and 1930 and find that it reduced the growth rate of state
income.
17
15
which King and Levine (1993) started, which finds that financial development variables
explain a great deal of the variation in long run growth.
5. Results
The main results are present in Tables 3 and 4. Table 3 shows the regression results
for various versions of equation (1), while Table 4 presents the elasticities implied by the
regressions. Each table presents three different sets of regressions—ordinary least squares
(OLS), weighted least squares (WLS) with the 1900 population levels being used as weights,
and weighted least squares with the 1900 level of income being used as weights instead. Each
set of regressions introduces the branching regulation and level of financial development
variables in a nested form in order to evaluate the robustness of the banking instability
variable.
Overall, the results are not out of line with those typically reported in the literature.
Most of the coefficients have the expected signs and are statistically significant. For example,
in every specification the initial level of income enters negatively and it is statistically
different from zero in virtually all of them, a result consistent with previous research (Barro
and Sala-i-Martin, 2003; Darlauf, Johnson, and Temple, 2005). Moreover the implied
coefficient of convergence is estimated to be between 0.012 and 0.028 (standard error 0.004
to 0.008), depending on the specification. This range of values corresponds nicely with the
one reported by Barro and Sala-i-Martin (1991). 18
18
The coefficient of convergence, or
⎛ 1 − α1T ⎞
− log⎜
⎟ , where
⎝ T ⎠
1
-convergence as is known in the growth literature, is given by
is the regression coefficient from equation (1) and T is 30. This coefficient
measures the speed at which the states incomes converge over time. For more details see Barro and Sala-iMartin (1991) and Darlauf, Johnson, and Temple (2005).
16
Illiteracy rate, a proxy for the level of education, also enters negatively (and
significant) in the regressions.
In virtually every specification the effect of banking instability is estimated to be
negative. In addition, it is so precisely estimated that it is statistically different from zero in 8
out of the 9 regressions. Thus, the negative effect of banking instability on growth is quite
robust. Another robust result is the effect of branching regulation. In most regressions its
coefficient is positive and statistically significant. This implies that states that permitted bank
branching enjoyed faster growth rates. This result is consistent with the findings of Dehejia
and Lleras-Muney (2007) and Jayaratne and Strahan (1996) who find evidence suggesting
that the elimination of bank branching restrictions accelerated real per capita income growth
in the U.S. during the 1980s.
The estimated effect of bank density is statistically zero in all of the regressions.
Thus, there is no evidence that initial bank density had an effect on long-term growth.
However, it is important to point out that its inclusion was not to test directly whether bank
density affects growth, but rather to test the robustness of the bank instability coefficient. As
the regression results indicate, its inclusion had an almost negligible effect on bank
instability.
Table 3 shows that the effect of financial depth is positive and statistically significant.
Thus, the level of initial financial depth appears to have had a strong influence on growth
between 1900 and 1930.
Although the results in Table 3 are useful for identifying variables that are statistically
relevant in explaining the growth rate, they are not as useful for ascertaining the variables’
economic effects. To enable us to recognize these effects, Table 4 presents the implied
elasticities of each of the variables included in the regressions presented in Tables 3. The
17
implied elasticities measure the percentage increase in a state’s growth rate in response to a
one percent increase in the independent variables. For branching regulation, a discrete
variable, the implied elasticity measure the percentage gain in growth rate for allowing
branching in the state.
What stands out the most out of these two tables is the fact that the implied elasticity
for banking instability—2 to 5 percent, depending on the specification considered—is
approximately equal (in absolute terms) to the impact of allowing branch banking in the
state.
6. Nebraska versus West Virginia
As indicated in the introduction, comparing the experience of these two states is
useful for learning the extent to which the 1893-96 panic affected long run growth. Table 5
presents a set of basic statistics for both states. It is clear that even in the 1890s West
Virginia was a much poorer state than Nebraska. Using Easterlin (1960)’s income per capita
figures, in 1900 West Virginia’s income level was about 60 percent of Nebraska’s. Illiteracy
rate figures also reveal a stark discrepancy between these two states—1.18 percent for
Nebraska, while nearly 12 percent for West Virginia. In addition, the deposit per capita
figures show a sizable difference in wealth between these two states. For example, in 1890,
the average deposit per capita in West Virginia was 23 percent of the level of Nebraska. The
other banking statistics indicate that both states were unit banking states, and both states
required double liability. But the level of capital requirements for state banks was very
different. The minimum capital required to charter a bank in West Virginia was $25,000,
while it was only $10,000 in Nebraska. This difference suggests that it should have been
relatively less costly to open a state bank in Nebraska than to open one in West Virginia.
18
And, indeed, the data reveals this—banks in West Virginia tended to be larger institutions
than banks in Nebraska. Using deposits per bank as an indicator of size, banks in West
Virginia were nearly twice the size of banks in Nebraska.
The table also reveals one of the most important differences in banking performance
outcomes: West Virginia did not experience a single bank failure during the 1890s, while
Nebraska did. In fact, the roughly 8 percent of total deposits compromised by failures
(whether temporary or permanent) put Nebraska at one of the worst in the nation (see Table
1). Perhaps the fact that they were smaller, more fragile institutions can explain part of the
reason these two states faired so differently during the 1893-96 panic in terms of failures.
But the basic set of statistics suggest that Nebraska’s banking system should have been
relatively more resilient—Nebraska was far wealthier, and the cost of chartering banks was
much lower than it was in West Virginia.
Despite this, the evidence suggests that the loss of deposits Nebraska suffered in the
mid 1890s was very long-lasting. Figures 1 and 2 display of the log of deposits per capita
indices for both states from 1890 to 1930. (Both series are scaled to start at 1 in 1890.) The
series for state banks is presented in Figure 1, while the series for national banks in presented
in Figure 2. Both figures reveal an overall similar pattern, although more pronounced for
state banks than for national banks. The loss of deposits in Nebraska, relative to West
Virginia, is striking. Despite the fact that Nebraska continued to charter many state banks
after the panic of the 1890s, its level of deposits (on a per capita basis) did not catch up to
West Virginia’s until the early 1920s, when Nebraska saw an acceleration in the growth rate
of deposits a few years after it introduced deposit insurance. This finding is consistent with
the proposition that the panic led to a loss in financial intermediation due to a long-lasting
loss in the depository base. Figure 1 suggests that the introduction of deposit insurance in
19
Nebraska may have been pivotal in attracting deposits back into the banking sector—the
growth rate of bank deposits per capita accelerated a few years after its introduction. It must
be pointed out, however, that this acceleration was temporary at best as Nebraska’s deposit
insurance system became insolvent in 1930. 19
7. Time Series Evidence from Newspaper Articles
As indicated earlier, theoretical models show that bank runs (and banking crises in
general) can affect the savings portfolio composition away from bank deposits and towards
other forms of savings. When financial markets are underdeveloped and lack of trust on the
financial system abounds, the only effective alternative are rudimentary forms of savings
such as keeping the money “under the mattress.”
But finding compelling evidence supporting this hypothesis is difficult. For obvious
reasons, people do not divulge how much money the store away under mattresses, carpets,
trunks, etc. Thus, surveys, even those conducted in recent years, are not very reliable, let
alone surveys done at the turn of the 20th century, if there are any at all. Despite their
statistical defects, they are useful for at least assessing the extent of the amount of
disintermediation.
Some recent surveys conducted in several Eastern European and South American
countries indicate that a sizable proportion of the population do not trust banks or, worse,
don’t even have a bank account. The countries in which the surveys were conducted,
incidentally, also experienced significant disruption in financial markets, including financial
crises. 20
For more details see Federal Deposit Insurance Corporation (1956).
For example, a survey done in March 2003 in the Czech Republic reports that 75% of population do not
trust banks. Source: Ceska Tiskova Kancelar, April 22., 2003. A survey done in Argentina in 2004 revealed that
19
20
20
Unfortunately, there are no similar surveys done in the U.S. at the turn of the 20th
century (or at least, I am not aware of any). However, despite this, we can get an idea of how
bad the savings portfolio composition may have been by relying on evidence from
newspaper articles. There are at least two ways in which evidence from newspapers can be
exploited. The first, most obvious one, is to simply read the content of the articles. But
although newspapers stories of individual experiences can be helpful and revealing, it is of
limited statistical value, as it is impossible to generalize those individual stories to the entire
population.
The second way in which newspaper evidence can be exploited is by studying the
incidence of relevant articles over time, and test whether this time series of newspaper
articles correlates with episodes of financial distress. This is exactly what is done in this
section. Specifically, I collected data on the dates of newspaper articles that included the
words “hidden money” in it. The newspapers searched were: The New York Times, The
Chicago Tribune, Boston Globe, Atlanta Constitution, Los Angeles Times, Washington
Post, The Wall Street Journal, and the Christian Science Monitor from 1860 to 1970. 21 The
search produced 641 articles, the vast majority of them telling tales, some of them sad, but
many of them humorous, of people losing large quantities of cash due to robberies, fires,
accidents, deaths, etc, and how they kept their holding hidden in various locations because
they did not trust banks.
Next, I constructed a yearly index of “hidden money” news defined as log (1 +
number of “money hidden” articles in year t). The log form is used as it facilitates the
interpretation of regression coefficients as an elasticity. This news index is then regressed on
86% population do not trust banks. Source: The Russian Business Monitor, June 2, 2004. On July 28, 2004
The Moscow Times reported that according to a polling done by the All Russian Public Opinion Survey
approximately 2/3 of population in Russia do not have bank accounts.
21 The choice of newspapers was dictated by availability.
21
a trend variable (“year”), and an indicator variable equal to 1 if there was a banking crisis in
that particular year, 0 otherwise. 22 I also included another indicator variable, labeled “postFDIC,” which equals 1 for the post 1935 period, after the FDIC was created. The inclusion
of this variable controls for the possibility that confidence in the banking system was
enhanced with the introduction of deposit insurance at the national level.
Tables 6 and 7 report the regression results for various lag structures of the banking
crisis variable, and for different dynamic specifications. Table 6 reports the results with the
lagged dependent variable is excluded, while Table 7 reports the results with the lagged
dependent variable included. The results are indeed very supportive of the hypothesis that
people lose trust on the banking system after a banking crisis. Newspaper articles with the
words “hidden money” in them increase anywhere from 60 to about 80 percent within two
years of the panic. After 10 years, the elasticity increases anywhere from 2.64 (Table 7) to
3.67 (Table 6), implying that the number of articles trebles. Moreover, the results appear to
be large even in the long-run—the implied long-term elasticities (computed as the
cumulative banking crisis coefficient divided by one minus the lagged dependent variable
coefficient) are estimated to be anywhere between 1.1 and 3.7, and virtually all of them are
statistically significant that the 5 percent level.
The post-FDIC indicator variable is included to control for the possibility that
distrust on the banking system declined after national deposit insurance was introduced. The
results are consistent with this interpretation. The post-FDIC indicator variable is negative
and statistically significant. Its magnitude indicates that the number of “money hidden” news
was cut in half after national deposit insurance was introduced.
The banking crisis dummy equals 1 for the following years: 1873, 1884, 1893, 1894, 1895, 1907, 1921, 1930,
1931, and 1932. These are the years that correspond to a major financial crisis in the U.S.
22
22
Taken together, these set of results point clearly to the hypothesis that distrust in the
banking system increases phenomenally after banking crises, remaining high for an extended
period of time, and that introduction of institutions such as the FDIC may have helped by
restoring trust in the banking system.
8. Concluding Remarks
This paper investigates the effect of bank distress on long run growth. In particular,
it examines how the remnants of the U.S. financial crisis of 1893, manifested through the
incidence of bank failures, affected state output growth between 1900 and 1930. The main
results indicate that bank instability caused by the panic appears to have adversely affected
output growth by 2 to 5 percent. This effect is similar in magnitude to the effect of allowing
branch banking in the state. Given the importance the literature has attached to the
elimination of geographical restrictions in banking for promoting growth (Jayaratne and
Strahan, 1996; Dehejia and Lleras-Muney, 2007), this results implies that the costs of
financial crises for growth are not negligible, and clearly go beyond a temporary decline in
real output. To the extent that long-term growth is compromised the real costs of bank
distress may have to be revised upwards.
This paper also argues that the mechanism through which bank instability affects
long-run growth most likely happen through a reduction in the depository base, stemming
from a loss of confidence in the banking system. Depositors that lose access to their money,
even if temporarily, tend to become more apprehensive about keeping their money in the
banking system. To the extent that they adjust their liquid portfolio away from banks and
into more rudimentary forms of savings, such as hiding the money “under the mattress,”
bank deposits, and hence, financial intermediation is compromised in the long run.
23
As evidence in favor of this argument, the paper looks at the pattern of deposit
growth in Nebraska (one of the most severely affected states during the panic) and contrasts
it with the pattern of deposit growth rate in West Virginia (which did not see a single bank
failure during the panic). The comparison show that, relative to West Virginia, Nebraska
suffered a severe contraction on deposits per capita, and did not recuperate from it until the
early 1920s, a few years after it had introduced deposit insurance.
Further evidence supporting the lack of trust explanation comes from newspaper
articles. In particular, I use a yearly index of newspaper articles with the words “money
hidden” in them to gauge the incidence of general distrust in the banking system. The
regression results three important conclusions: 1. that the incidence of “money hidden”
articles spikes during banking crises years; 2. the magnitude of the spike is substantial (about
80 percent increase within two years of the crisis); and 3. that the effect remains large in the
long-run and decline after the U.S. introduced deposit insurance at the national level, clearly
aiming at restoring confidence in the banking system.
Taken together, the results imply that banking crises appear to be particularly costly
not just because they may amplify a business cycle downturn, but because they, in the
absence of institutions that restore confidence in the banking system, may compromise longrun growth as well.
24
Data Appendix
The data were collected from several sources, which are described here:
1. Real income per capita figures. 1900 figures are from Easterlin (1960) and from
Mitchener and McLean (2003). 1930 figures are from the Bureau of Economic
Analysis. 1940 figures are from Mitchener and McLean (2003). Easterlin (1960)
figures as well as those of the Bureau of Economic Analysis are in nominal terms.
Both series were converted into real terms using the WPI index, Series E23, from the
Historical Statistics of the United States. Mitchener and McLean (2003) figures are in real
terms, so there was no need for further adjustment.
2. Bank failures: Dun’s Review Jan 1895, Jan 1896, and Jan 1897 issues.
3. Illiteracy rate figures were computed using data from the 1900 Census, available
online from the Historical Census Browser at the University of Virginia.
(http://fisher.lib.virginia.edu/collections/stats/histcensus/)
4. Population figures are from the US Census, various years. These are available online
from the Historical Census Browser at the University of Virginia.
(http://fisher.lib.virginia.edu/collections/stats/histcensus/)
5. Bank density is defined as the total number of banks in the state divided by the
state’s area. Number of banks figures is from Flood (1998).
6. Financial depth is defined as the 1900 level of deposits per capita divided by the
1900 level of income per capita. Total deposits figures are from Flood (1998).
7. Branching is an indicator variable equal to 1 if the state permitted bank branching in
1900, 0 otherwise. Source: Dehejia and Lleras-Muney (2007).
25
References
Acemoglu, Daron (2008) Introduction to Modern Economic Growth. Book manuscript.
Department of Economics, MIT. http://econ-www.mit.edu/faculty/acemoglu/books
Barro, Robert J. and Xavier Sala-i-Martin (2003) Economic Growth, 2nd Edition.
Cambridge, MA: MIT Press
Barro, Robert J. and Xavier Sala-i-Martin (1991) “Convergence Across States and
Regions,” Brookings Papers on Economic Activity, No. 1, pp. 107-182.
Bencivenga, Valerie R. and Bruce D. Smith (1991) “Financial Intermediation and
Edogenous Growth,” Review of Economic Studies, Vol. 58, pp. 195-209.
Bernanke, Ben S, (1983) "Nonmonetary Effects of the Financial Crisis in Propagation of
the Great Depression," American Economic Review, vol. 73(3), pages 257-76.
Calomiris, Charles W., and Gary Gorton (1991) “The Origins of Banking Panics.” In R.
Glenn Hubbard (ed.) Financial Markets and Financial Crises, pp. 107–143. Chicago:
University of Chicago Press.
Cameron, Rondo (1967) Banking in the Early Stages of Industrialization, New York: Oxford
University Press.
Carlson, Mark (2005) “Causes of Bank Suspensions in the Panic of 1893,” Explorations in
Economic History, vol. 42, pp. 56-80.
Durlauf, Steven N. & Quah, Danny T. (1999) "The new empirics of economic growth,"
In J. B. Taylor & M. Woodford (ed.), Handbook of Macroeconomics, volume 1, chapter 4,
pages 235-308, Elsevier.
Darlauf, Steven N., Paul A. Johnson, and Jonathan R.W. Temple (2005) “Growth
Econometrics,” In Philip Aghion and Steven N. Darlauf (eds.), Handbook of Economic
Growth, Volume 1, Chapter 8, pp. 555-677.
Dehejia, Rajeev and Adriana Lleras-Muney (2007) “Institutions, Financial Development,
and Pathways of Growth: The United States from 1900 to 1940,” Journal of Law and
Economics, Vol. 50, No. 2 (May), pp. 239-272.
Dell’ Ariccia, Giovanni, Enrica Detragiache, and Raghuram Rajan (2008) “The Real
Effect of Banking Crises,” Journal of Financial Intermediation, Vol. 17 (January), pp. 89-112.
Demirguc-Kunt, Asli and Enrica Detragiache (2005) Cross-Country Empirical Studies of
Systemic Bank Distress: A Survey. IMF Working Paper No. 05/96.
Diamond, Douglas W. and Dybvig, Philip H. (1983) “Bank Runs, Deposit Insurance and
Liquidity,” Journal of Political Economy, Vol. 91, No. 3, pp. 401-419
26
Dupont, Brandon (2007) “Bank Runs, Information and Contagion in the Panic of 1893”
Explorations in Economic History, Vol. 44, No. 3 (July), pp.
Easterlin, Richard A. (1960) “Interregional Difference in Per Capita Income, Population,
and Total Income, 1840-1950.” In William Parker (ed.) Trends in the American Economy in
the Nineteenth Century, Studies in Income and Wealth Vol. 24, pp. 73-140. Princeton, NJ:
Princeton University Press.
Edwards, Sebastian (2007) “Crises and Growth: A Latin American Perspective,” NBER
Working Paper No. 13019.
Ennis, Humberto M. and Todd Keister (2003) “Economic Growth, Liquidity, and Bank
Runs,” Journal of Economic Theory, Vol. 109, pp. 220-245.
Federal Deposit Insurance Corporation (1956). Annual Report. Washington, DC: Federal
Deposit Insurance Corporation.
Flood, Mark D. (1998). United States Historical Data on Bank Market Structure, 18961955 (Computer File). ICPSR Version. Ann Arbor, MI: Inter-university Consortium for
Political and Social Research (distributor).
Friedman, Milton, and Anna J. Schwartz. (1963) A Monetary History of the United States,
1867–1960. Princeton: Princeton University Press.
Goldsmith, Raymond W. (1969) Financial Structure and Development. New Haven, CT: Yale
University Press.
Goldstein, Morris and Philip Turner (1996). Banking Crises in Emerging Economies: Origins
and Policy Options. Basel: Bank for International Settlements.
Higgins, Matthew J., Daniel Levy, and Andrew T. Young (2006) “Growth and
Convergence Across the United States: Evidence from County-Level Data,” Review of
Economics and Statistics, Vol. 88, no. 4, pp. 671-681.
Hoffman, Charles (1956). The Depression of the Nineties. Journal of Economic History, Vol.
16, No. 2, June, pp. 137-164.
Hoggarth, Glenn, Ricardo Reis, and Victoria Saporta (2002) “Costs of Banking System
Instability: Some Empirical Evidence,” Journal of Banking and Finance, Vol. 26, pp. 825855.
Jayaratne, Jith, and Philip Strahan (1996) “The Finance-Growth Nexus: Evidence from
Bank Branch Deregulation,” Quarterly Journal of Economics, Vol. 111, pp. 639-670.
Kaminsky, Graciela and Carmen Reinhart (1999) “The Twin Crises: Causes of Banking
and Balance-of-Payments Problems,” American Economic Review, Vol. 89, No. 3, pp.473500.
27
Kroszner, Randall S., Daniela Klingebiel, and Luc Laeven (2007) “Financial Crises,
Financial Dependence, and Growth,” Journal of Financial Economics, Vol. 84, No. 1, pp.
187-228.
King, R.G. and Ross Levine (1993) “Finance and Growth: Schumpeter Might Be Right,”
Quarterly Journal of Economics, Vol. 108, pp. 717-738.
Levine, Ross (2005) “Finance and Growth: Theory and Evidence,” In P. Aghion and S.
Durlaff, eds., Handbook of Economic Growth. Netherlands: Elsevier Science.
Lindgren, Carl-Johan, Gillian Garcia, and Matthew Saal (1996) Bank Soundness and
Macroeconomic Policy, Washington, DC: International Monetary Fund.
Lucas, Robert E. (1988) “On the Mechanics of Economic Development,” Journal of
Monetary Economics, Vol. 22, pp. 3-42.
Manning, Mark J. (2003) “Finance Causes Growth: Can We Be So Sure?” Contributions to
Macroeconomics, Vol. 3, Issue 1, Article 12.
McKinnon, R. I. (1973) Money and Capital in Economic Development. Washington, DC:
Brookings Institution.
Mitchener, Kris James and Ian McLean (1999) “U.S. Regional Growth and Convergence,
1880 – 1980.” Journal of Economic History, Vol. 59, no. 4, pp. 1016-1042.
Mitchener, Kris James and Ian McLean (2003) “The Productivity of U.S. States since
1880,” Journal of Economic Growth, Vol. 8, no. 1, pp. 73-114.
Moen, Jon and Ellis Tallman (1992) “The Bank Panic of 1907: The Role of Trust
Companies,” Journal of Economic History, Vol. 52, No. 3, pp. 611-630.
Noyes, Alexandre (1909) Fourty Years of American Finance, New York: G.P. Putnam’s
Sons.
Robinson, Joan (1952) “The Generalization of the General Theory,” In The Rate of Interest
and Other Essays. London: Macmillan.
Romer, Christina (1986) “Spurious Volatility in Historical Unemployment Data,” Journal
of Political Economy, Vol. 94, February, pp. 1-37.
Rousseau, Peter L. and Watchel, Paul (2005) “Economic Growth and Financial Depth:
Is the Relationship Extinct Already?” Manuscript prepared for UNU/WIDER
conference on Financial Sector Development for Growth and Poverty Reduction, Helsinki,
Finland.
Sala-i-Martin, Xavier, Gernot Doppelhofer, and Ronald I. Miller (2004) “Determinants
of Long-Term Growth: A Bayesian Averaging of Classical Estimates (BACE)
Approach,” American Economic Review, Vol. 94, No. 4, pp. 813-835.
28
Shaw, E. S. (1973) Financial Deepening in Economic Development. New York: Oxford
University Press.
Sprague, Oliver M. W. (1910) History of Crises under the National Banking System.
Washington, DC: National Monetary Commission.
Stevens, Albert C. (1894) “Analysis of the Phenomena of the Panic in the United States,”
Quarterly Journal of Economics, Vol. 8, No. 2, January, pp. 117-148.
U.S. Bureau of the Census (1975) Historical Statistics of the United States, Colonial Times to
1970, Part 2. Washington, DC: U.S. Government Printing Office.
Wachtel, Paul (2003) “How Much Do We Really Know About Growth and Finance?”
Federal Reserve Bank of Atlanta Economic Review, Vol. 88, pp. 33-47.
Wicker, Elmus (2000) Banking Panics of the Gilded Age, New York, NY: Cambridge
University Press.
Williamson, Samuel H. (2008) “Five Ways to Compute the Relative Value of a U.S.
Dollar Amount, 1790 - 2007,” http://www.MeasuringWorth.com
29
Table 1
Summary Statistics
Variable
1900-1930 growth
Initial income
Illiteracy rate
Initial bank instability
Branching allowed?
Bank density
Financial depth
Median
Standard
Dev
Mean
0.037
417
0.027
0.013
0.000
0.004
263
0.035
431
0.095
0.043
0.354
0.010
451
30
0.011
184
0.113
0.066
0.483
0.016
452
Minimum
0.006
186
0.007
0.000
0.000
0.000
44
Maximum
0.052
981
0.356
0.348
1.000
0.092
1892
Table 2:
Incidence of Bank Failures across States
State
Bank Failures
Liabilities
Alabama
1
0.17
Arizona
1
1.84
Arkansas
3
3.01
California
8
0.74
Colorado
8
10.83
Connecticut
3
0.39
Delaware
0
0.00
Florida
4
7.15
Georgia
7
3.29
Idaho
3
9.04
Illinois
32
5.51
Indiana
5
0.50
Iowa
30
5.28
Kansas
31
23.10
Kentucky
6
1.30
Louisiana
6
12.33
Maine
2
0.17
Maryland
1
0.00
Massachusetts
1
0.01
Michigan
11
0.90
Minnesota
36
9.72
Mississippi
1
1.36
Missouri
34
2.43
Montana
3
18.50
Nebraska
52
8.25
Nevada
0
0.00
New Hampshire
0
0.00
New Jersey
2
0.08
New Mexico
0
0.00
New York
23
0.59
North Carolina
1
0.04
North Dakota
2
6.14
Ohio
9
0.42
Oklahoma
11
4.81
Oregon
7
6.02
Pennsylvania
17
0.52
Rhode Island
2
1.20
South Carolina
2
1.30
South Dakota
5
5.63
Tennessee
6
1.25
Texas
8
7.46
Utah
2
3.82
Vermont
1
0.30
Virginia
5
4.02
Washington
42
34.81
West Virginia
0
0.00
Wisconsin
14
2.26
Wyoming
0
0.00
This table presents the number of banks that failed or temporarily suspended operations between 1894 and 1896 (Bank Failures), and the
liabilities of failed banks as a percent of total deposits in the state (Liabilities). Source: Dun’s Review.
31
Table 3
1900-1930 Growth Regressions
Variable
Initial income
Illiteracy rate
Banking instability
Branching allowed?
OLS
WLS: Population 1900
-0.021
-0.021
-0.021
-0.006
-0.010
-0.016
-0.013
-0.015
-0.020
(0.004)
(0.004)
(0.004)
(0.005)
(0.006)
(0.005)
(0.006)
(0.005)
(0.004)
0.000
0.000
0.000
0.277
0.108
0.004
0.035
0.005
0.000
-0.060
-0.052
-0.041
-0.036
-0.038
-0.043
-0.053
-0.053
-0.052
(0.014)
(0.013)
(0.014)
(0.017)
(0.017)
(0.016)
(0.019)
(0.017)
(0.016)
0.000
0.000
0.006
0.047
0.025
0.011
0.008
0.003
0.002
-0.043
-0.031
-0.018
-0.053
-0.042
-0.024
-0.059
-0.046
-0.029
(0.015)
(0.013)
(0.013)
(0.012)
(0.012)
(0.011)
(0.013)
(0.012)
(0.012)
0.006
0.023
0.196
0.000
0.002
0.046
0.000
0.000
0.024
0.006
0.006
0.005
0.006
0.006
0.002
0.006
0.007
0.002
(0.002)
(0.002)
(0.002)
(0.003)
(0.003)
(0.002)
(0.003)
(0.003)
(0.002)
0.012
0.017
0.037
0.033
0.026
0.295
0.035
0.024
0.248
Bank density
0.165
-0.009
0.124
-0.039
0.129
-0.038
(0.103)
(0.098)
(0.080)
(0.074)
(0.076)
(0.066)
0.118
0.924
0.131
0.600
0.098
0.578
Financial depth
0.100
0.090
0.085
(0.032)
(0.018)
(0.017)
0.002
0.000
0.000
Adjusted R-Squared
0.466
0.526
0.604
F statistic
10.130
10.800
12.290
Prob>F
0.000
0.000
0.000
45
45
45
45
Num of Observations
WLS: Income 1900
0.351
0.384
0.524
0.425
0.467
0.596
6.320
5.710
15.610
0.001
0.001
0.000
6.630
7.130
15.920
0.000
0.000
45
45
45
0.000
45
45
Dependent variable: Real per capita growth from 1900 to 1930. Independent variables: “Initial income” is the log of real
income per capita in 1900. “Initial Bank Failures” is the log of the liabilities of failed banks in 1894, 1895, and 1896 divided
by total deposits in 1896. “Illiteracy rate” is the ratio of illiterate persons 21 years or older divided by total adult population
(21 years or older). “Deposit Insurance” is an indicator variable equal to 1 if the state had some form of deposit insurance
between 1900 and 1930, 0 otherwise. “Branch Banking” is an indicator variable equal to 1 if the state permitted bank
branching in 1900, 0 otherwise. “Bank density” is the total number of banks in the state divided by the state’s area in
squared miles. “Financial depth” is total deposits in 1896 divided by the state income in 1900. Robust (White, 1980corrected) standard errors are in parenthesis under the estimated coefficients. Significance levels are presented in italics
under the standard errors.
32
Table 4
1900-1930 Growth Regressions: Implied Elasticities
Variable
Initial income
Illiteracy rate
Banking instability
Branching allowed?
Bank density
Financial depth
OLS
WLS: Population 1900
WLS: Income 1900
-0.584
-0.594
-0.587
-0.163
-0.245
-0.398
-0.314
-0.383
-0.488
0.114
0.100
0.108
0.148
0.149
0.130
0.145
0.128
0.118
0.000
0.000
0.000
0.274
0.102
0.002
0.030
0.003
0.000
-0.159
-0.140
-0.110
-0.083
-0.089
-0.099
-0.082
-0.083
-0.082
0.037
0.035
0.038
0.041
0.038
0.037
0.029
0.025
0.024
0.000
0.000
0.003
0.041
0.020
0.008
0.005
0.001
0.001
-0.053
-0.038
-0.022
-0.041
-0.033
-0.019
-0.048
-0.038
-0.023
0.018
0.016
0.016
0.009
0.009
0.009
0.010
0.010
0.010
0.004
0.019
0.185
0.000
0.001
0.041
0.000
0.000
0.020
0.063
0.060
0.048
0.046
0.049
0.016
0.048
0.053
0.015
0.024
0.023
0.022
0.020
0.020
0.014
0.021
0.023
0.013
0.009
0.012
0.030
0.025
0.019
0.291
0.026
0.016
0.245
0.046
-0.003
0.039
-0.012
0.048
-0.014
0.029
0.028
0.025
0.023
0.028
0.024
0.108
0.923
0.122
0.598
0.088
0.576
0.138
0.129
0.146
0.044
0.028
0.030
0.002
0.000
0.000
Dependent variable: Real per capita growth from 1900 to 1930. Weights: 1900 state population. Independent variables:
“Initial income” is the log of real income per capita in 1900. “Initial Bank Failures” is the log of the liabilities of failed banks
in 1894, 1895, and 1896 divided by total deposits in 1896. “Illiteracy rate” is the ratio of illiterate persons 21 years or older
divided by total adult population (21 years or older). “Deposit Insurance” is an indicator variable equal to 1 if the state had
some form of deposit insurance between 1900 and 1930, 0 otherwise. “Branch Banking” is an indicator variable equal to 1
if the state permitted bank branching in 1900, 0 otherwise. “Bank density” is the total number of banks in the state divided
by the state’s area in squared miles. “Financial depth” is total deposits in 1896 divided by the state income in 1900. Robust
(White, 1980-corrected) standard errors are in italics under the estimated coefficients. Significance levels are presented
under the standard errors.
33
Table 5
Comparing Nebraska and West Virginia
Number of Bank Failures in 1894-96
Failed Bank Liabilities (1894-96) /State Deposits
Deposits per capita, 1890 (1967 $)
Deposits per capita, 1900 (1967 $)
State population, 1900
Income per capita, 1900 (1967 $)
Income per capita, 1930
Growth rate, 1900- 1930
Illiteracy rate, 1900
Capital Requirements 1900, State banks
Double Liability?
Branching Allowed?
Deposits per bank, 1890 (1967$)
34
Nebraska
West Virginia
52
8.25%
$104
$163
1,066,300
$464
$1,168
3.1%
1.18%
$10,000
Yes
No
$330,972
0
0
$24
$63
958,800
$274
$915
4.01%
11.8%
$25,000
Yes
No
$809,048
Figure 1
Growth of Bank Deposits for State Banks: Nebraska versus West Virginia
3.500
3.000
NE introduces DI
2.500
2.000
1.500
Nebraska
West Virginia
1.000
0.500
0.000
1890 1892 1894 1896 1898 1900 1902 1904 1906 1908 1910 1912 1914 1916 1918 1920 1922 1924 1926 1928 1930
-0.500
-1.000
35
Figure 2
Growth of Bank Deposits for National Banks: Nebraska versus West Virginia
3.000
2.500
2.000
1.500
Nebraska
West Virginia
1.000
0.500
-1.000
36
8
6
4
2
0
8
6
4
0
19
3
19
2
19
2
19
2
19
2
19
2
19
1
19
1
2
0
8
6
4
2
0
8
6
4
2
-0.500
19
1
19
1
19
1
19
0
19
0
19
0
19
0
19
0
18
9
18
9
18
9
18
9
18
9
0
0.000
Table 6
Effects of Banking Crises on “Money Hidden” News, Version 1
Year
n=2
n=4
n=6
n=8
n = 10
0.033
(0.003)
0.000
0.032
(0.004)
0.000
0.033
(0.004)
0.000
0.033
(0.004)
0.000
0.033
(0.004)
0.000
n
∑ Crisis
t−i
i= 0
Post-FDIC
Num. Obs.
F-test
Prob>F
Adj. R Sqrd.
0.807
1.249
1.837
2.840
3.665
(0.331)
0.016
(0.452)
0.007
(0.536)
0.001
(0.635)
0.000
(0.693)
0.000
-1.663
(0.245)
0.000
109
22.64
0.000
0.501
-1.553
(0.259)
0.000
107
15.10
0.000
0.482
-1.553
(0.257)
0.000
105
11.93
0.000
0.486
-1.498
(0.254)
0.000
103
10.07
0.000
0.495
-1.496
(0.251)
0.000
101
8.71
0.000
0.501
Dependent variable: Log (1 + number of “hidden money” news during the year). Independent variables: “Year” is a trend variable; the sum
of “Crisis” variable is the sum of the coefficients from i = 0 to n of the banking crisis dummy variable. The banking crisis dummy variable
equals 1 if there was a banking crisis on that year, 0 otherwise. “Post-FDIC” is a dummy variable equal to 1 if year > 1934, 0 otherwise.
“Num. Obs.” stands for the number of observations. “Adj. R Sqrd.” is the adjusted R squared. Standard errors are included in parenthesis
under each coefficient. Below the standard errors, in italics, are the p-values.
Table 7
Effects of Banking Crises on “Money Hidden” News, Version 2
n=2
n=4
n=6
n=8
n = 10
Lagged dep. vbl.
0.411
(0.085)
0.000
0.406
(0.089)
0.000
0.398
(0.088)
0.000
0.347
(0.093)
0.000
0.290
(0.098)
0.004
Year
0.020
(0.004)
0.000
0.020
(0.004)
0.000
0.020
(0.004)
0.000
0.022
(0.005)
0.000
0.024
(0.005)
0.000
0.644
0.760
1.222
1.916
2.638
(0.302)
0.035
(0.427)
0.078
(0.508)
0.018
(0.645)
0.004
(0.750)
0.001
-0.985
(0.263)
0.000
1.092
(0.519)
0.038
109
26.77
0.000
0.589
-0.950
(0.271)
0.001
1.281
(0.696)
0.069
107
18.38
0.000
0.567
-0.970
(0.268)
0.000
2.029
((0.816)
0.015
105
14.90
0.000
0.572
-1.011
(0.272)
0.000
2.934
(0.912)
0.002
103
11.66
0.000
0.556
-1.090
(0.277)
0.000
3.718
(0.937)
0.000
101
9.43
0.000
0.541
n
∑ Crisis
t−i
i= 0
Post-FDIC
Implied LT elas.
Num. Obs.
F-test
Prob>F
Adj. R Sqrd.
Dependent variable: Log (1 + number of “hidden money” news during the year). Independent variables: “Lagged dep. vbl.” stands for
lagged dependent variable; “Year” is a trend variable; the sum of “Crisis” variable is the sum of the coefficients from i = 0 to n of the
banking crisis dummy variable. The banking crisis dummy variable equals 1 if there was a banking crisis on that year, 0 otherwise. “PostFDIC” is a dummy variable equal to 1 if year ≥ 1935, 0 otherwise. “Implied LT elas.” stands for implied long term elasticity. This variable
measures the long term effect on “hidden money” news of the banking crisis variable. “Num. Obs.” stands for the number of
observations. “Adj. R Sqrd.” is the adjusted R squared. Standard errors are included in parenthesis under each coefficient. Below the
standard errors, in italics, are the p-values.
37