The dawn of an ‘age of deposits’ in the United States

Accepted Manuscript

The Dawn of an ‘Age of Deposits’ in the United States Matthew Jaremski , Peter L. Rousseau PII: DOI: Reference:

S0378-4266(17)30257-1 10.1016/j.jbankfin.2017.10.010 JBF 5237

To appear in:

Journal of Banking and Finance

Received date: Revised date: Accepted date:

17 January 2017 10 October 2017 13 October 2017

Please cite this article as: Matthew Jaremski , Peter L. Rousseau , The Dawn of an ‘Age of Deposits’ in the United States , Journal of Banking and Finance (2017), doi: 10.1016/j.jbankfin.2017.10.010

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

The Dawn of an ‘Age of Deposits’ in the United States Peter L. Rousseau

Colgate University and NBER

Vanderbilt University

CR IP T

Matthew Jaremski

Abstract

Individual deposits in the United States grew from 5% to 23% of GDP between 1863 and 1913. A comprehensive database shows bank entry underlying this trend while historical events,

AN US

including the National Banking Acts, resumption in 1879, and the election of 1896, influenced deposits at the bank-level. The nation‟s embrace of deposits was thus driven by stability of the monetary system and confidence in the safety and utility of established and well-capitalized banks. Bank-level and county-level regressions confirm these patterns for national banks over

ED

M

the entire postbellum period and for a sample of Midwest state and national banks after 1888.

JEL codes: (G21, N21, E21)

PT

Keywords: Bank deposits; financial development; National Banking System; bank notes

CE

Jaremski: Associate Professor, Department of Economics, Colgate University, Hamilton, NY

AC

13346 and Faculty Research Fellow, National Bureau of Economic Research. Phone 1-315-2287524, Fax 1-315-228-7033, E-mail [email protected].

Corresponding Author: Rousseau: Professor, Department of Economics, Vanderbilt University, Box 1819 Station B, Nashville, TN 37235. Phone 1-615-343-2466, Fax 1-615-343-8495, E-mail [email protected].

ACCEPTED MANUSCRIPT

1. Introduction Despite the pervasiveness of deposits in banking systems today, personal banking accounts were not always so popular. Before the Civil War, the majority of bank liabilities in the United States took the form of dollar-denominated notes issued by individual banks. This

CR IP T

changed over the fifty years of the National Banking System (1863‒1913), when individual deposits expanded from about 5% of gross domestic product (GDP) to more than 23%. The rise in relation to bank capital is just as extraordinary, with deposits rising from 96% of capital to 592%. What caused the transition from a focus on bank notes to an “Age of Deposits” that

AN US

persists in the United States to the present day? General economic growth and changes in

financial regulation can account for part of the rise, but the introduction of clearinghouses and checking accounts along with greater public confidence in banks were also at play. Econometric

M

analysis of the relative contributions of these factors, however, has always been limited by a lack of the digitized bank-level data. We conduct such an analysis by collecting data on over 14,000

ED

individual banks in existence before 1913, and relate factors associated with economic activity, financial innovations, and depositor confidence to the growth and distribution of individual

PT

deposits across banks and over time.1

CE

Friedman and Schwartz (1963) and James (1978) describe the time path of aggregate deposits over the National Banking period and offer several hypotheses for their rise such as the

AC

utility of checking accounts for making payments, increases in national income, and the spread of the market economy through urbanization, but could not test these hypotheses empirically. 1

Because we are interested in consumer preferences over bank liability types, we focus on individual deposits rather than interbank or government deposits. We therefore refer to “individual deposits” and “deposits” interchangeably, considering interbank balances as a distinct quantity when applicable. Individual deposits include both personal and commercial deposit accounts because they were not separated in financial reports of the period.

ACCEPTED MANUSCRIPT

Bordo and Jonung (1987), in their study of the income velocity of money, show that a general rise in broad money with respect to GDP is shared by a variety of other countries over much of this period, including Canada, Norway, Sweden, Denmark, France, Germany, Italy, and Japan. Available data on actual bank deposits collected by Mitchell (2007) show similar upward trend

CR IP T

growth for Canada, Germany, and the United Kingdom. The time series from these sources, however, do not share the same apparent breaks in deposit activity over time that we observe for the United States, suggesting that country-specific events may have influenced the patterns in these trend breaks.2

AN US

A number of studies use disaggregated data to highlight particular factors in the rise, but data limitations prevent them from testing the implications fully. Using Nebraska and West Virginia as polar cases, Ramirez (2009) shows that bank failures during the Panic of 1893

M

preceded a long-run decline in deposits as individuals lost confidence in banks. James and Weiman (2011) and Rousseau (1998, 2011) examine deposit growth in New York City, while

ED

Krueger (1933) and Keene (1974) examine deposit growth in Wisconsin. All find specific reasons for local increases in deposits but do not test whether they apply to the nation as a whole.

PT

For instance, James and Weiman (2011) describe the conversion of banks in New York City

CE

from state to national charters at the end of the Civil War as instrumental to initiating the rise in

AC

deposits, whereas Rousseau (2011) emphasizes the effects of published bank stock prices on the

2

For example, Mitchell (2007) reports that Canada, which had the deposit pattern closest to that of the United States, shares a rise in deposits around 1896, but also has similar or larger percentage increases in 1884, 1890, and 1905 that do not occur in the United States. Canada also shows a small increase in deposits around 1879, but there are 14 other years that saw larger percentage increases from 1870 to 1913. These facts do not preclude the possibility that factors affecting deposits in the United States in 1879 and 1896 had spillover effects in Canada, perhaps due to the two nations‟ proximity and financial ties, but they do indicate that deposits in the two countries experienced quite different time series patterns.

ACCEPTED MANUSCRIPT

distribution of deposits across New York banks through 1897. Fulford and Schwartzman (2015) argue that the election of 1896 and the related reduction in inflation uncertainty were important for allowing banks to add leverage to their balance sheets. We gather two annual bank-level databases for our broader analysis. The first contains

CR IP T

balance sheet information on each national bank from 1865 through 1913, and allows us to study a geographically-diverse set of banks over a long time period. The second contains balance sheet information on both state and national banks in six Midwestern states (Indiana, Michigan,

Minnesota, Ohio, Pennsylvania, and Wisconsin) from 1888 through 1913, which permits us to

AN US

examine whether state banks operated differently than national banks and whether the aggregate growth that we observe was concentrated in certain types of banks. To test the various hypotheses for the rapid ascent of individual deposits, we link deposits to a county‟s population,

M

the availability of clearing facilities, U.S. bond prices, and measures of confidence in individual banks such as capital levels and the number of years in business.

ED

The first sharp rise in deposits relative to GDP occurred after the passage of the National Banking Acts in 1863 and 1864. The ratio also rose rapidly after the resumption of specie

PT

payments on government-issued notes in 1879 and following the election of 1896. Remarkably,

CE

the ratio of deposits to bank capital doubled within two years of each event, never to return to its earlier levels. The increases occurred across all banks – whether old, new, state-chartered, or

AC

national – suggesting that the rise in deposits was caused by system-wide factors rather than idiosyncratic or location-specific ones. Looking at the deposits of individual banks, no single factor seems to be driving their

growth. We do find, however, that factors plausibly related to a depositor‟s confidence in a particular bank were important, and that deposits would have followed a very different trajectory

ACCEPTED MANUSCRIPT

without them. It turns out that deposits accumulated most rapidly in older and more highly capitalized banks. Clearinghouses and the avoidance of bank closures also explain a significant part of the longest sustained episode of financial deepening in the nation‟s history. Even after controlling for these factors, however, other macroeconomic forces apparently

CR IP T

remain important. Not only are the fixed effects in our bank-level regressions large and

statistically significant, but they also display jumps in the same years we observe in the

aggregate data. Moreover, when we allow the effects of the microeconomic variables to vary over time, they explain nearly half of the variation in average bank deposits, and nearly all of the

AN US

variation in the average ratio of deposits to capital. These data suggest that while microeconomic factors clearly mattered, macroeconomic factors that boosted confidence in the banking system also played a significant role in the growth of deposits.

M

2. Aggregate individual deposits, 1800-1913

ED

Before looking at the microeconomic dynamics at the bank-level, it is useful to consider aggregates for the nation. Fig. 1 presents aggregates for the number of state and national banks

PT

(panel a), the ratios of deposits and circulation to GDP (panel b), and the ratios of deposits and circulation to capital (panel c) from 1800 through 1913. The number of banks in operation is a

CE

general measure of financial development that reflects extensive growth in banking. Given that GDP was rising more rapidly than population, increases in the ratio of deposits to GDP indicate

AC

that individuals as a whole were also placing more of their income in deposit accounts. The fresh deposits could appear on the extensive margin from new accounts in recently-formed and existing banks, or intensively as additional deposits in existing accounts. Increases in the ratio of deposits to capital, holding other liabilities constant, reflect compositional shifts on the liability side of bank balance sheets.

ACCEPTED MANUSCRIPT

There was relatively little deposit growth before 1860. The ratio of deposits to GDP saw only a small rise before 1820, while the ratio of deposits to capital was relatively steady across the entire period. The lack of growth is not surprising given that deposits were not the dominant bank liability before the Civil War. Rather, banks issued paper notes that circulated among the

CR IP T

public and were payable on demand in specie at the issuing bank. Banks preferred notes as they could issue large amounts while keeping only a fraction of their value in the vault as cash. The public may have even preferred notes because they were useful in transactions and deposits were not subject to required reserves. Therefore, while circulation declined over the period, banks in

AN US

1860 still had as much value outstanding in the form of notes as they attracted in deposits.

The first substantive expansion of deposits began in 1863, and panel (b) of Fig. 1 shows the ratio of deposits to GDP rising by more than over 50% over the next five years. This

M

represents the largest five-year increase in deposits achieved at any time before 1870, but unlike previous episodes the expansion had only just begun. Indeed, it was not until the Panic of 1907

ED

that the ratio ever declined over more than a single year. The period‟s other large scale panics in 1837, 1839, 1873, 1884, and 1893 also correspond to slight and temporary dips in deposits.

PT

Despite these, by 1913 the ratio of individual deposits to GDP stood at 23% and the ratio of

CE

deposits to capital had reached nearly 600%. The ratio of deposits to banking capital closely resembles a step function with three large

AC

jumps. The ratio grew from 71% to 143% between 1862 and 1868, from 136% to 226% between 1878 and 1883, and from 258% to 500% between 1896 and 1901. As a result, by 1913 the ratio of deposits to bank capital had reached 592% ‒ ten times its value in 1860. Although the ratio of deposits to GDP does not display the same clear step function, it did grow from 5.1% to 7.7% between 1862 and 1868, from 9.0% to 11.1% between 1878 and 1883, and from 14.7% to 20.2%

ACCEPTED MANUSCRIPT

between 1896 and 1901. These smaller jumps are partially due to the robust pace of GDP growth over the post-bellum period. For example, the jump observed in the ratio of deposits to capital in 1879 is not as obvious because GDP had grown rapidly from just over $8 billion in 1878 to $12 billion in 1883. Results reported in Appendix B indicate that the same breaks occurred across all

CR IP T

regions of the country.

To confirm these jumps econometrically, Table 1 presents estimates for the dates of structural breaks using two different methods. The Perron and Vogelsang (1992) approach

selects the single time-series observation in a sample with the highest probability of representing

AN US

a structural break, while the approach of Clemente et al. (1998) selects the two observations with the highest probabilities. In the upper panel we report the years of structural breaks estimated over the full sample period, whereas in the bottom panel we report the dates of structural breaks

M

for the postbellum only. Regardless of the time period or variable used (i.e., the logarithm of deposits in constant 1870 dollars or the ratios of deposits to GDP or bank capital), the selected

ED

structural breaks match the years described above as coinciding with stabilization and greater higher confidence in the nation‟s monetary and banking systems.3

PT

The start of the sustained growth in deposits roughly corresponds to the National Banking

CE

Acts of 1863 and 1864. Although these were designed primarily to raise revenue for the Union during the Civil War, the Acts also lowered default rates by making the requirements for starting

AC

a new bank more stringent. This was accomplished through the creation of the Office of the Comptroller of the Currency, a nationwide requirement that national banks back their note issues 3

These three events were the only national-level events that would have significantly increased public confidence in banks. The Sherman Silver Purchase Act of 1890, for example, introduced additional uncertainty about the real value of current bank deposits. National bank legislation was left largely unchanged or was weakened through 1913. And while state bank legislation was changing over time, most changes involved lowering entry barriers and the timing varied greatly across states.

ACCEPTED MANUSCRIPT

with holdings of U.S. bonds, and increases in minimum capital and reserve requirements required for operating a national bank. The reserve requirements on deposits in particular increased the security of deposits and encouraged more individuals to take advantage of deposit accounts as vehicles for saving. To raise additional revenue and gain control over note issues

CR IP T

across the system, Congress followed the Acts with a 10% tax on outstanding notes of state banks, which went into effect in August 1866. The Acts also installed a $300 million ceiling on the aggregate notes of national banks, which was reached in 1866 and not eliminated until 1875 (Cagan, 1963, p. 33; Sylla, 1969, p. 663).Taken together, these changes led to larger and

AN US

generally safer banks, and the substitution of deposits for circulation.

The growth in deposits during the 1860s was not driven solely by new bank formation. Existing banks that adopted national bank charters saw large increases in individual deposits.

M

While individual bank data were not reported during the Civil War, Fig. 2 shows average balance sheet quantities from 1850 to 1860 and from 1865 to 1875 for state banks that converted to a

ED

national charter before 1868. Each measure includes a linear time trend extrapolated from the 1850-60 data which indicates whether state banks altered their liability structures after switching

PT

to a national charter. The data indicate that the average ratio of deposits to capital was nearly

CE

double what would have been expected had the ratio followed its pre-1860 trend.4 While the ratio of circulation to capital did not substantially increase for these banks, there is still a rise in the

AC

ratio above its expected trend. Thus, the jump in the ratio of deposits to capital was driven

4

Between 1860 and 1865, capital levels for the sample of converting banks shown in Fig. 2 continued on the upward trend of the previous decade. Out of the 808 converting banks, the average change in bank capital across the Civil War was 13.3 percent, with 400 banks maintaining the same level of capital, 298 banks increasing their capital, and 110 banks decreasing it. Looking at the level of real deposits or circulation over the period rather than their ratios to capital would thus lead to more pronounced increases than those displayed in Fig. 2.

ACCEPTED MANUSCRIPT

largely by increased deposits at those institutions that were becoming less risky by virtue of their national charters. The resumption of specie payments in 1879 after an 18-year suspension marks the second jump in deposits, and represents a restoration of both domestic and international confidence in

CR IP T

the nation‟s financial system. Congress took a gamble when it passed the Resumption Act in 1875 which mandated a return to specie payments for Greenbacks at par on January 1, 1879, and with the exception of the creation of a specie reserve, took no significant direct actions to assist in its implementation. Mitchell (1908) argues that the success of the legislation was far from

AN US

certain at the time, and Kindahl (1961) demonstrates that parity was reached only in late 1878. The plan‟s success helped restore confidence by removing the immediate threat of specie drains, given that depositors had good reason to fear that a return to the gold standard would lead to a

M

suspension similar to that experienced in the 1860s. Moreover, Resumption marked the end of the sharp deflation that had been required to absorb the Greenbacks at par, and also brought forth

ED

capital from abroad and a modest rise in U.S. prices relative to Britain, all of which were conducive to economic growth.

PT

The remarkable growth in deposits after 1896 corresponds to the end of the free silver

CE

movement. While the nation had largely been under a de facto Gold Standard since 1873, the populist movement advocated for the return of silver money. The movement was concentrated in

AC

agricultural regions, but was large enough to affect expectations more generally. Friedman and Schwartz (1982, p. 629) write that “before 1896, the political strength of the free silver movement made it entirely reasonable for farsighted investors to anticipate inflation,” while Calomiris (1993) describes a growing apprehension among the public that the movement might lead to a cessation of gold convertibility. The movement came to an abrupt end with the

ACCEPTED MANUSCRIPT

presidential election of 1896, not only because the Democratic Party and William Jennings Bryan‟s free silver platform were soundly defeated, but also because gold became more readily available due to the discovery of new mines and mining technology around this time. The clear support for a single monetary basis eliminated the potential for suspension or inflation due to an

CR IP T

excessive monetization of silver, and the Gold Standard Act of 1900 then formally committed the nation to this course.5 Using quarterly data, recent work by Fulford and Schwartzman (2015) indicates that the increase in the ratio of deposits to capital occurred immediately after the

election rather than as a general trend, which is consistent with our view that commitment to the

AN US

Gold Standard and the removal of inflation uncertainty was particularly important for drawing in new deposits.

As dramatic as the increases were around these three events, the lack of any large or

M

sustained decline in deposits around the financial crises of the period is important to note. The ratios of deposits to GDP fell by only -0.4%, -0.2%, -0.8%, and -0.5% respectively in the four

ED

crisis years 1873, 1884, 1893, and 1907, whereas the ratios of deposits to capital fell by -1.3%, -21.1%, -30.1%, and -12.3%. These declines were much smaller than the substantial increases in

PT

the deposit ratios described above, and with the exception of 1907, deposit growth in the year

CE

following each crisis more than compensated for the decline. The lack of large depositor responses in the crisis years is likely due to the role of illiquidity rather than insolvency in these

AC

disturbances (Calomiris and Gorton, 1991). Once the money markets unfroze, it became clear that most banks were solvent and that there would be few depositor losses. Moreover, since banks outside of New York City were generally not responsible for precipitating these panics, depositors might have still believed that their local deposits were secure. 5

Bordo and Rockoff (1996) illustrate the benefits of adherence to the Gold Standard for the cost of capital in international markets.

ACCEPTED MANUSCRIPT

3. Bank level data We use data collected from the Comptroller of the Currency‟s Annual Report to examine individual national banks from 1865 to 1913. The data contain the annual balance sheet of every national bank in operation during the third quarter of each year, but no information on income or

CR IP T

interest rates. Over time the Comptroller began to publish less detailed data on specific items and instead aggregated items into larger categories (i.e., total loans, individual deposits, etc.). Data are missing for 1885 when balance sheets were not reported and for 1905 when individual

deposits and interbank deposits were combined. While there is some possibility of window

AN US

dressing the early years, the Comptroller effectively randomized the call date after 1869 to minimize this and thereby gain a more accurate view of bank condition.

Because most states did not report balance sheet data until relatively late in the National

M

Banking period, we cannot achieve the same coverage for state banks. Nevertheless, to compare national and state banks at the bank-level, we collected balance sheet data for individual state

ED

banks from the available state reports of Indiana, Michigan, Minnesota, Ohio, Pennsylvania, and Wisconsin from 1888 to 1913.6 We use data from these states for a variety of reasons. First,

PT

banking data for these states are available for a large number of years. Second, these states had 6

AC

CE

Data for Indiana are from the Annual Report of the Auditor of State of the State of Indiana. Data for Michigan are from the Annual Report of the State Treasurer of the State of Michigan before 1889 and the Annual Report of the Commissioner of the Banking Department of the State of Michigan thereafter. Data for Minnesota are from the Report of the Public Examiner of the State of Minnesota to the Governor before 1910 and the Annual Report of the Banking Department of the State of Minnesota thereafter. Data for Ohio are from Annual Report of the Auditor of State to the Governor of Ohio. Data for Pennsylvania are from the Reports of the Several Banks and Savings Institutions and Banks Organized under the Free Banking Law of Pennsylvania before 1895 and the Annual Report of the Commissioner of Banking thereafter. Data for Wisconsin are from the Semi-Annual Statement on the Condition of State and Private Banks of Wisconsin before 1895, the Annual Report of the Bank Examiner from 1896 to 1902, and the Annual Report of the Commissioner of Banking thereafter. State bank balance sheets are missing for Ohio in 1893, and we use a linear trend to fill in balance sheet observations for that year.

ACCEPTED MANUSCRIPT

only a few pre-existing financial centers that might affect our comparison of bank types. Finally, because banks in these states operated in a wide range of locations including urban and rural areas, we argue that they provide a representative sample of the nation as a whole. The Northeast was the only other region with data available for the period, but given the region‟s large number

CR IP T

of existing banks in 1860 and high concentration of savings banks and trust companies, it is not well suited for our study.

In all, the data represent the largest collection of postbellum balance sheets in existence. For both the national and state bank databases, the only adjustments we make are to aggregate

AN US

unique items that are broken down only in some publications. For instance, time deposits,

checking deposits, demand deposits, savings deposits, and certificates of deposits are merged into a single individual deposits measure. We then supplement the balance sheet data by

M

obtaining banks‟ starting dates from the Merchants and Bankers Directory and the Rand McNally Bankers Directory, and the original entry year for national banks that were formerly

ED

state banks using the bank conversion data from Jaremski (2013) and the entry dates of antebellum banks from Weber (2005). The national bank database starts in 1865 and contains

PT

166,616 observations from 10,642 banks, whereas the Midwest database starts in 1888 and

CE

contains 70,597 observations for 2,479 national banks and 3,954 state banks. It is important to note that the data do not contain any bank branches and very few banks

AC

with branches. National banks were prohibited from branching, and only a few states allowed state banks to branch outside the city of the headquarters bank. Where branches were present, state officials generally did not provide separate balance sheet data on the branches or list out those banks with branches. According to Banking and Monetary Statistics, in 1910, Ohio had 18 branches outside the city of its headquarters bank, Pennsylvania had four, and Wisconsin had

ACCEPTED MANUSCRIPT

two. In 1900, the number of branches was even smaller, with only Ohio having eight branches and Pennsylvania having two. The Rand McNally’s Bankers Directory also indicates that some of these “branching” banks were non-chartered private banks not included in the state reports.

CR IP T

4. Determinants of deposits There are few studies devoted to the determinants of deposit growth, but many offer insights into their pattern. In this section, we highlight and discuss these factors before exploring them empirically at the bank and county levels.

AN US

4.1. Checking accounts

The growth of deposits is often associated with the increasing use of checking accounts, which allowed customers to access funds easily and served as a substitute for bank notes. While

M

most states did not separate checking from deposits accounts in their records, the Comptroller‟s reports include disaggregated data on checking accounts for a few years. These data indicate that

ED

96.7% of bank receipts in reserve cities and 81.7% outside of them were already in the form of checks and drafts by 1881. There is also little change in these shares between 1881 and 1890, a

CE

the Civil War.”

PT

fact that led James (1974, p. 25) to assert that “checks were already widely in use by the end of

The expansion of clearing facilities also likely encouraged individuals and banks to use

AC

deposit accounts more intensively. Clearinghouses were large institutions that sorted and processed checks, thereby lowering a bank‟s cost of accepting them, and were particularly attracted to cities with large populations. Although the entry of a clearinghouse might also be viewed as an endogenous response to increases in banking activity, Jaremski (2015) shows that entry of a clearinghouse was driven primarily by population and the number of national banks in

ACCEPTED MANUSCRIPT

the vicinity rather than the level or change in deposits or banking capital. This renders clearinghouse entry at least a plausibly exogenous factor in our analysis. Early clearinghouses were concentrated primarily in the Northeast, but moved into the Midwest during the 1880s. Areas in the South and West eventually established a few clearinghouses but they never covered

deposit growth but not the rise across all banks.

CR IP T

a large portion of these regions. We might therefore expect clearinghouses to explain local

As advanced by Friedman and Schwartz (1963) and others, the expansion of the banking system itself might also have increased the demand for checking, with drafts between banks

AN US

maintaining correspondent relationships more likely to be accepted. As the utility of checks depends crucially on their acceptability, a large number of banks in a location would increase consumer demand for such accounts. Competition among local banks could also encourage

ED

4.2. The return of state banks

M

higher interest rates on deposits to attract funds.

Another important hypothesis for the rise in deposits is the rapid resurgence of state

PT

banking after 1880. Because of the 10% tax levied on state bank notes in 1866, state banks could operate only if they had sufficient deposits, and this provided a strong incentive to pursue

CE

depositors aggressively and to implement new technologies. The lower entry requirements of state banks also allowed them to open in rural areas where deposit accounts had been previously

AC

unavailable. In this way, demand for deposit accounts in rural areas may have led to the sudden influx of state banks during the 1880s and 1890s. At the same time, existing state banks might have encouraged additional deposits in order to expand their businesses. Fig. 3 indicates that state banks were not solely responsible for the rise in deposits. First, national banks held nearly twice the aggregate deposits as state banks from 1864 to 1913 despite

ACCEPTED MANUSCRIPT

their exclusive ability to issue notes. Second, deposits in state and national banks as well as their ratios of deposits to capital saw comparable growth patterns, with the only important deviation across the two bank types coming after the Panic of 1907 when state bank deposits began to

CR IP T

decline and national bank deposits continued to grow. 4.3. Reputation

Studying deposit growth in New York City over the second half of the 19th century, Rousseau (2011) argues that confidence and reputation of banks were important determinants to

AN US

fluctuations in deposits. While we do not have a precise measure of how banks were viewed by the public, we can look at an indirect measure of reputation: bank age. Older banks were often better capitalized and could establish a reputation over a number of years, and these factors could increase confidence among depositors and lead them to prefer established banks. Banks that

M

survived the Civil War and successfully converted to national charters were in the most

ED

advantageous positions. In this sense, a reputation built upon longevity and a sufficiently high level of capital could allow deposits to expand without commensurate increases in capital,

PT

allowing reputation itself to substitute for it. Indeed, bank advertisements often highlighted the

CE

name, establishment date, and capital level above any other characteristics. 4.4. Bank note circulation

AC

Among macro-level factors, the price of U.S. bonds should affect the amount of notes in circulation, and by extension the quantity of deposits (James, 1976; Champ et al.,1992). Fig. 4 shows that the aggregate ratio of circulation to bank capital fell when bond prices were high and rose when prices were low, as might be expected given that national banks were required to back their notes fully with U.S. bonds and high bond prices made this collateral expensive. On the

ACCEPTED MANUSCRIPT

other hand, bond prices did not have a consistently positive relation with deposits. The rise of deposits after 1896, for example, occurs in a period of declining bond prices. Further, the rise in deposits continued after 1900 even though the Gold Standard Act made note issue more

CR IP T

attractive. 4.5. Urbanization

The size of and economic activity in a bank‟s location should also influence the quantity of its deposits. For example, Cagan (1963) argues that urbanization can explain the rise observed

AN US

in the aggregate. Indeed, the expanding railroad network and the rapid development of integrated national markets coincided with a wave of urbanization after the Civil War (Atack et al., 2010). Because individuals in densely populated areas were more likely to hold wealth in the form of

population alone.

ED

4.6. Bank risk and financial stability

M

deposits than in real estate, urbanization could have affected deposits more than increases in

Modern banking theory and the related empirics also offer guidance about factors that

PT

may influence deposit growth. In money markets, such as that for deposits, debt holders not only

CE

price risk but demand a very low level of default risk as well. This intolerance for risk in uninsured deposit markets in turn disciplines banks to target a low level of default risk on

AC

deposits (Gorton and Pennacchi, 1990; Martinez Peria and Schmukler, 2001; Calomiris and Powell, 2001; Calomiris and Wilson, 2004), and the risk premium required by depositors ends up being higher than the actuarially fair one. Since banks unwilling to pay the required premium lose deposits, the discipline of the market forces them to meet the premium constraint by maintaining a low-risk loan portfolio and/or a sufficiently high ratio of equity to assets. In

ACCEPTED MANUSCRIPT

particular, Calomiris and Wilson (2004) show that U.S. banks during the interwar and Great Depression, in an effort to retain deposits, were forced to reduce loans as a substitute for raising scarce capital. Our explanation of the rise in deposits around 1866, 1879, and 1896 focuses on general reductions in banking risk due to events that resolved uncertainty about the real value of

CR IP T

deposits rather than direct reduction in loans, but the mechanism is similar in that banks with more capital and longer-standing reputations likely had lower perceived risk and thereby

benefited most from a system-wide rise in depositor confidence. While our historical bank data do not include information on loan quality or interest rates, we can and do include the ratio of

AN US

loans to assets as well as the level of capital in our regression models to assess whether

idiosyncratic portfolio choices by banks are driving the rise and distribution of deposits. Of course, the observed growth in deposits could also be simply a result of the general

M

stability of the post-bellum period. For example, Jalil (2015) finds evidence of major panics only in 1873, 1893, and 1907, and argues that, with the exception of 1907, these panics generally led

ED

to fewer losses than those sustained in the antebellum period. The high capital and collateral requirements put into place by the National Banking Acts also protected depositors better from

PT

losses, allowing national banks to usually lose fewer deposits during panics. This was clearly

CE

seen in the Panic of 1907. After 1907, aggregate deposits at state banks declined despite growth in the number of state banks, while deposits at national banks continued to grow despite a

AC

reduction in the rate of entry. As discussed by Moen and Tallman (1992), the panic affected state banks more than national ones, with deposits continuing to grow for national banks while declining for state banks. The pattern suggests that deposit growth corresponded with stability and confidence in the banking sector.

ACCEPTED MANUSCRIPT

5. Explaining deposit growth Accounting for the rise of deposits requires distinguishing the contributions of growth in population, income, regulation, and the number of banks themselves from less easily quantified yet still fundamental shifts in confidence, individual preferences, and business practices. In

CR IP T

Section 2 we established that deposits per bank rose sharply after the National Banking Acts, resumption of convertibility in 1879, and the election of 1896, and in Section 4, we discussed the role of bank and local characteristics in effecting the rise. We now explore econometrically the determinants of deposit taking among all national banks after 1865, the relative behavior of

AN US

national and state banks in the Midwest sample after 1888, and the growth of deposits in national banks using county-level data.

5.1. Explaining national bank deposits at the bank level, 1865-1913

M

We begin the econometric analysis by exploring the determinants of deposits and circulation from 1865 to 1913 using our database of national bank balance sheets. Each

ED

observation is a bank year. The dependent variables that we consider (

are the logarithms of

PT

deposits and circulation as well as the ratio of deposits to capital. To account for inflation over time, we adjust all dollar values used in the analysis to an 1870 basis using the GDP deflator

is a vector of county-level variables including the number of other national banks in

AC

where

CE

from Officer (2008). The full specification is

year t, the value of assets of national banks that closed or merged over the previous ten years, and a binary variable for whether the county had a clearinghouse within its borders.

is a

vector of bank-level variables including the ratio of loans to assets and the logarithms of capital and age in years. We also include fixed effects for years ( ), states (

), and their interactions.

ACCEPTED MANUSCRIPT

Table 2 presents the results. The number of banks in the county relates positively and significantly to both the real value of deposits and the ratio of deposits to capital, yet relates negatively and significantly to real value of bank notes in circulation. The value of assets in closed banks in a county over the previous ten years relates positively with a bank‟s deposits but

CR IP T

has mixed effects on the other two outcome variables. The coefficients on the clearinghouse variable indicate that national banks in areas with a clearinghouse, despite an array of other

control variables, could expect to attract 11-14.3% more deposits, decrease their circulation by 5.2-9.5%, and increase their ratio of deposit to capital 18.8-20.1% than those in areas without a

AN US

clearinghouse. The real value of the capital stock of each bank is positively and significantly correlated with both the real values of deposits and circulation. In the specifications with bank age, a 10% increase in capital is predicted to increase deposits by 6.3-7.5% and circulation by

M

8.9-10.4%. Because capital is as much a scaling variable in these models that controls for bank size as a measure of stability and reputation, these findings are as expected.

ED

Most striking, however, are the second columns of each panel, which include the log of each bank‟s age. With capital already in the specification and time controlled with year fixed

PT

effects, the positive and significant coefficients on bank age strongly suggest that banks in

CE

operation longer secured more real deposits, issued more notes, and maintained higher ratios of deposits to capital than banks entering later in the period. Specifically, the model suggests that an

AC

average-aged national bank of 11.7 years would have about 69.4% more deposits, 76.7% more circulation, and a ratio of deposits to capital that was 169 percentage points higher than that of a newly created bank. If longevity and bank size engender confidence, older and larger banks that had survived multiple banking panics were likely perceived as models of stability and sound

ACCEPTED MANUSCRIPT

management and bank depositors seemed particularly attracted to them. 7 Unlike the other variables, the coefficient on the ratio of loans to assets is sensitive to the inclusion of the log of bank age. When age is not included in the model, the ratio is positively correlated with deposits, negatively correlated with circulation, and uncorrelated with the ratio of

CR IP T

deposits to assets. When log of age is included, however, the loan-to-asset ratio is uncorrelated with deposits, is still negatively related with circulation, and is negatively correlated with the deposit-to-asset ratio. The latter set of results matches most cleanly with to the theoretical

literature: depositors prefer banks with lower default risk all things equal. We take the pattern to

AN US

suggest that depositors tolerate higher loan-to-asset ratios among older banks because they take these banks‟ established reputations and stability into account.

While not reported in Table 2 due to space constraints, the time fixed effects are also

M

particularly important. As discussed in Section 2, much of the growth in aggregate deposits occurred around a few key dates, but this observation does not shed light on whether the

ED

observed jumps in the data are driven by micro-level determinants or by the specific bank and location factors included in our specifications. To address this question, in Fig. 5 we plot the

PT

coefficients associated with the fixed effects in Table 2 in Fig. 5. The upper panel corresponds to

7

CE

the year fixed effects in a specification that does not include their interaction with each bank‟s

AC

The log of bank capital and age have a correlation coefficient of 0.45 across banks, and to explore the effects of this we repeated the regressions dividing the sample into nine groups based on low, medium, and high age and capital levels. The coefficients on the log of capital remained positive and statistically significant for deposits across all nine combinations, but the log of bank age was positive and significant only for banks of young and medium age. This is not surprising: when a bank becomes established (and therefore in the “old” age category), age variation within this category would likely not matter as much as the level of capital, which appears sufficient for measuring relative soundness. We also computed variance inflation factors for the coefficients on capital and age in Table 2, and found that their standard errors could be understated by as much as 53 and 28 percent, respectively. Given that the t-statistics for these variables are more than 60, however, we conclude that multi-collinearity between bank capital and age is not driving our interpretation of the regression results.

ACCEPTED MANUSCRIPT

respective state, while the lower panel plots the coefficients for the year fixed effects associated directly with the specifications in columns (2) and (6) of Table 2. Even after controlling for the available bank-level and local factors, as well as the host of state-by-year fixed effects, the main events in the sample period (resumption in 1879 and the election of 1896) still coincide with

CR IP T

sharp increases in the real value of deposits. For instance, the fixed effects between 1896 and 1901 in the lower panel of Fig. 5 indicate that bank-level deposits would have risen by over 50 percent and the ratio of deposits to capital would have risen by 175 percentage points. Fig. 5 also shows that the Panics of 1873, 1884, 1890, 1893 and 1907 in general led to temporary declines in

AN US

our measures of deposits.

5.2. National and state bank deposits in the Midwest, 1888-1913

We now examine individual state and national banks together in our sample of Midwest

M

states (Indiana, Michigan, Minnesota, Ohio, Pennsylvania, and Wisconsin) from 1888 to 1913,

ED

where bank-year is again the observational unit. The dependent variables are the logarithms of the real value of deposits and the ratio of deposits to capital. We cannot examine circulation

PT

because state banks ceased issuing notes after the 10% tax on their notes took effect in 1866. The

is a vector of county-level or nationwide variables and

is a vector of bank-level

AC

where

CE

full specification is:

variables. Because the Midwest sample contains observations for both national and state banks, we examine the effects of each bank type on deposits using a dummy variable set to unity if bank i is a national bank, and the differential effects of each variable on the bank types by interacting the X and Z vectors with the national bank dummy variable. We also include fixed effects for years ( ), and states (

), and their interactions.

ACCEPTED MANUSCRIPT

Table 3 presents the findings. The real value of a given bank‟s deposits and its ratio of deposits to capital relate positively to the number of state banks in its county. Specifically, banks in a location where the number of state banks doubled would expect to have 0.7-1.1% more deposits and a ratio of deposits to capital that was 5.1-10.0 percentage points higher. At the same

CR IP T

time, the coefficient on the number of national banks is not statistically significant for deposits or their ratio to capital. The value of assets of closed banks seems to have increased deposits in surviving banks but generally lowered their ratios of deposits to capital. The result is likely due to surviving banks gaining some of the deposits removed from closed banks, but the general

AN US

public also requiring them to hold more capital going forward as additional security.

When included, the dummy variable for whether bank i is a national bank has a positive and statistically significant coefficient for real deposits but not for the ratio of deposits to capital.

M

The results in columns (2) and (5) imply that national banks on average had 17.5% more deposits and a ratio of deposits to capital that was 168 percentage points lower than state banks. As

ED

before, the real value of capital and the age of the bank are always positively and significantly correlated with deposits and the ratio of deposits to capital. In this case, the effects are somewhat

PT

larger. Based on columns (2) and (5), a bank with 11.7 years in operation would be expected to

CE

have 105% more deposits and a ratio of deposits to capital ratio that was 322 percentage points higher than that of a newly created bank.

AC

Columns (3) and (6) add interactions between the national bank dummy and other explanatory variables in the model. In Column (3), we see that the real value of deposits in national banks has a small positive correlation with both the presence of other national banks in bank i‟s county (i.e., the negative coefficient of -0.004 on the level is smaller than the positive coefficient of 0.007 on the interaction) and other state banks in the county (i.e., the positive

ACCEPTED MANUSCRIPT

coefficient of 0.011 on level is larger than the negative coefficient of -0.007 on the interaction). In Column (6), however, the ratio of deposits to capital at national banks has a much larger positive correlation with other national banks in the county and a much larger negative correlation with other state banks in the county. Combined with the opposite results for state

CR IP T

banks, we conclude that conditional on bank size, national and state banks seem to have

competed for depositors with the opposite bank-type rather than with its own bank-type.

As before, it is important to examine the time-fixed effects to determine whether the included variables can explain the discrete jumps in aggregate data or whether they occurred

AN US

across all banks. Fig. 6 plots the coefficients on the year fixed effects for a specification without state-year interactions in the upper panel, and the year fixed effects directly related to the specifications from columns (3) and (6) of Table 3 in the lower panel. Even after controlling for

M

bank-level and location factors as well as state-by-year fixed effects, the plots show a steep rise in the real value of deposits and the ratio of deposits to capital starting around 1896, and there is

ED

a large downturn during the panics of 1890 and 1893 while only a slight downturn during 1907. In this case, the macroeconomic effects are somewhat smaller than in Fig. 5, but still imply an

CE

1896 and 1901.

PT

80% growth in deposits and a 200 percentage point rise in the ratio of deposits to capital between

5.3. National bank deposits at the county level, 1870-1913

AC

The bank-level data allow us to control for key historical events, yet their high frequency

precludes the inclusion of local factors such as population and urbanization that are observed only with the decadal censuses. To account for these factors and focus on the intensity of banking activity, we aggregate the data on individual national banks throughout the United States to the county-level and match them to decennial Census data collected by Haines (2004),

ACCEPTED MANUSCRIPT

dropping out any county with significant boundary changes between 1870 and 1913 and any county without a bank. Aggregation at the county level allows us to average out idiosyncratic components in individual bank balance sheets. The baseline specification is: (3)

CR IP T

where the dependent variables are the logarithms of deposits, circulation, and bank capital, all measured in the Census year (i.e., 1870, 1880, etc.) on a per capita basis and deflated to reflect 1870 dollars.

is the number of people living in the county,

population living in a location with more than 25,000 persons, and

is the per capita

is a vector of county-level banking variables including the

AN US

value of farm land in the county.

is the percent of the

asset value of national banks that closed over the previous ten years and indicators for whether there was a clearinghouse in operation or whether the county had a bank that converted from a

M

state to national charter in the 1860s.

The regressions reported in Table 4 indicate that more populated and urbanized counties

ED

had a higher real value of deposits and bank capital per person though the coefficient on urbanization for deposits is not statistically significant when we account for counties with

PT

converted state banks. In the models without the converted bank dummy, a county with a 10

CE

percentage point higher urbanization share would be expected to have 1.9% more deposits, an insignificant 1.4% higher circulation, and 3.4% more capital, whereas counties with a 10%

AC

higher population would be expected to have 1% more deposits, an insignificant 0.1% more circulation, and an insignificant 0.4% lower capital. The coefficients on the clearinghouse indicator are also positive and statistically significant in all specifications with economically meaningful effects of at least 13% on deposits, 19.3% on circulation, and 15.5% on capital.

ACCEPTED MANUSCRIPT

Neither agricultural counties, as measured by the value of farms per thousand persons, nor bank closures significantly relate to any of the outcome variables. Again, we find a positive and highly significant effect of having an early bank in the county on later banking intensity therein. A county that started with a converted state bank is

CR IP T

expected to have 58.1% more deposits, 53.4% higher circulation, and 59.4% more capital. One might think that having an early bank is just an indicator of economic potential, but our controls for wealth, population, and urbanization make it more likely to reflect confidence in banks or at least familiarity with the benefits of banking services. Banks that had converted from state to

AN US

national charters in the 1860s and remained in existence were ones that county residents had recognized for some time, and their survival may have served as a testimonial to the security afforded by a well-managed bank. Being more acclimated to banks and banking services than

M

counties that saw their first bank only later in the century, residents of such counties were most

ED

likely to use deposits and accept bank notes in transactions as the age of deposits took hold. 6. Microeconomic effects and macroeconomic events

PT

In our bank-level specifications, we control for macroeconomic factors affecting deposits using time-fixed effects. This approach, however, potentially misses how relationships between

CE

the explanatory variables and deposits change over time. For example, the confidence-building events we emphasize may have reduced the effects of bank-specific factors such as capital and

AC

age over time. In this section, we decompose the real value of deposits to determine how much their average over time can be explained by changes in the average characteristics of the banks relative to how those characteristics change deposits over time. We conduct a similar examination of the ratio of deposits to capital.

ACCEPTED MANUSCRIPT

To carry out this analysis, we estimate a model similar to those reported in columns (2) and (6) of Table 2 and columns (3) and (6) of Table 3), but replace the time-fixed effects with interactions between the time fixed effects and each explanatory variable. The expanded model allows the coefficients on each determinant to vary over time as there is a separate coefficient

CR IP T

estimated for each variable in each year. We then use the model to predict real deposits and the ratio of deposits to capital in two ways. First, we evaluate the importance of time effects by applying the estimated time varying coefficients of each variable to the time series average of the respective observed variables, themselves averaged across all banks in each year (i.e., “Holding

AN US

Variables Constant”). Second, we evaluate the importance of the microeconomic factors by applying the average of each variable‟s estimated time varying coefficients to the average of the respective observed variables across all banks in each year in the sample (i.e., “Holding

M

Coefficients Constant”).

Fig. 7 compares the actual averages of deposits and the ratio of deposits to capital in each

ED

year to the two predicted series for the national bank sample. Fig. 8 provides the same but for the Midwest bank sample. The two sets of figures are quite similar. The average growth in real

PT

deposits was driven almost equally by the changing loadings on the variables and the changing

CE

means of the variables. Alternatively, most of the rise in the ratio of deposits to capital is driven by the time-varying loadings on the microeconomic factors. The predicted values of the deposits-

AC

capital ratio are relatively flat when holding the coefficients constant over time. The evidence thus supports our hypothesis that the jumps in deposits and the ratio of deposits to capital were both driven by macroeconomic factors, but that macroeconomic factors coinciding with our confidence-building events had a larger influence on the ratio than on the level.

ACCEPTED MANUSCRIPT

7. Conclusion Deposits grew from 5% to 23% of GDP between 1863 and 1913. A comprehensive bank database shows that the aggregate growth in deposits was mainly caused by bank entry while the growth of deposits at the bank-level was mainly caused by historical events. Specifically, bank

CR IP T

deposits across all banks rose sharply after the National Banking Acts, the resumption of 1879, and the presidential election of 1896. The nation‟s embrace of deposit banking thus seems driven by the stability of the monetary system and the confidence that individuals held for the safety and utility of established and well capitalized banks. A series of bank-level and county-level

AN US

regression models confirm these patterns in disaggregated data for all national banks over the entire postbellum period and for a comprehensive sample of state and national banks for a selection of Midwest states after 1888. At the same time, we also find support for traditional

widespread clearing facilities.

M

explanations for the rise in deposits such as increasing incomes, urbanization, and more

ED

Economic historians well before even Friedman and Schwartz suspected that confidence among the public was central to the rise of deposit banking, and the lack of such confidence

PT

seems alive and well in many of today‟s emerging and not-so-emerging markets. Given that most

CE

bank liabilities were in the form of redeemable bank-issued notes before the Civil War, we investigate how subsequent financial innovations may have contributed to a shift in U.S.

AC

preferences, but also find that measures of confidence provide strong explanatory power. Perhaps it is natural that they should, given that in order for demandable debt contracts to gain acceptance, there must be confidence that they are indeed “demandable.” We consider our contribution a step toward assembling the hard evidence required to raise the status of

ACCEPTED MANUSCRIPT

“confidence” from a reasonable hypothesis for a gain in momentum over the “Age of Deposits”

AC

CE

PT

ED

M

AN US

CR IP T

to an empirical construct that can be plausibly measured and assessed.

ACCEPTED MANUSCRIPT

References Atack, J., Bateman, F., Haines, M., Margo, R. A., 2010. Did railroads induce or follow economic growth? Urbanization and population growth in the American midwest, 1850–1860. Social Science History 34, 171-197.

CR IP T

Bordo, M. D., Jonung, L., 1987. The Long-Run Behavior of the Circulation of Money: The International Evidence. Cambridge University Press, New York.

Bordo, M. D., Rockoff, H., 1996. The gold standard as a „good housekeeping seal of approval.‟ Journal of Economic History 56, 389-428.

AN US

Bodenhorn, H., 2006. Bank chartering and political corruption in antebellum New York: free banking as reform. In: Glaeser, E., Goldin, G. (Eds.), Corruption and Reform: Lessons from America‟s History. University of Chicago Press, Chicago, IL.

M

Cagan, P., 1963. The first fifty years of the National Banking System: a historical appraisal. In: Carson, D. (Ed.), Banking and Monetary Studies. Richard D. Irwin, Homewood, IL.

ED

Calomiris, C. W., 1993. Greenback resumption and silver risk: the economics and politics of monetary regime change in the United States, 1862-1900. In: Bordo, M. D., Capie, F.

PT

(Eds.), Monetary Regime Transformations. Cambridge University Press, New York.

CE

Calomiris, C. W., Gorton, G., 1991. The origins of banking panics: models, facts, and bank regulation. In: Hubbard, R. G. (Ed.), Financial Markets and Financial Crises. University

AC

of Chicago Press, Chicago, IL. Calomiris, C. W., Powell, A., 2001. Can emerging market bank regulators establish credible discipline? The case of Argentina, 1992-99. In: Mishkin, F. (Ed.), Prudential Supervision: What Works and What Doesn‟t. University of Chicago Press, Chicago, IL.

ACCEPTED MANUSCRIPT

Calomiris, C. W., Wilson, B., 2004. Bank capital and portfolio management: the 1930s „capital crunch‟ and the scramble to shed risk. Journal of Business 77, 421-456. Cannon, J., 1900. Clearing Houses‒Their History, Methods, and Administration. D. Appleton and Company, New York.

CR IP T

Champ, B., Wallace, N., Weber, W. E., 1992. Resolving the National Bank note paradox. Federal Reserve Bank of Minneapolis Quarterly Review, 13–21.

Clemente, J., Montanes, A., Reyes, M., 1998. Testing for a unit root in variables with a double change in the mean. Economics Letters 59, 175-182.

AN US

Davis, A. M., 1911. Origin of the National Banking System. Government Printing Office, Washington, DC.

Davis, L. E., 1965. The investment market, 1870-1914: the evolution of a national market.

M

Journal of Economic History 25, 355-399.

Dwyer, G. P. Jr., Hasan, I., 2007. Suspension of payments, bank failures, and the nonbank

ED

public‟s losses. Journal of Monetary Economics 54, 565-580. Friedman, M., Schwartz, A. J., 1963. A Monetary History of the United States, 1867–1960.

PT

Princeton University Press, Princeton, NJ.

CE

Friedman, M., Schwartz, A. J., 1982. The role of money. In: Friedman, M., Schwartz, A. J. (Eds.), Monetary Trends in the United States and United Kingdom: Their Relation to

AC

Income, Prices, and Interest Rates, 1867-1975. University of Chicago Press, Chicago, IL. Fulford, S. L., Schwartzman, F. F., 2015. The benefits of commitment to a currency peg: lessons from the National Banking System. Federal Reserve Bank of Richmond Working Paper.

Gorton, G., Pennacchi, G., 1990. Financial intermediaries and liquidity creation. Journal of Finance 45, 49-71.

ACCEPTED MANUSCRIPT

Haines, M. R., 2004. Historical, demographic, economic, and social data: the United States, 1790-2000. ICPSR Study 2896. Inter-university Consortium for Political and Social Research, Ann Arbor, MI. Haxby, J. A., 2009. Standard Catalog of United States Obsolete Bank Notes 1782-1866. Krause

CR IP T

Publications, Inc., Iola, WI.

Jalil, A., 2015. A new history of banking panics in the United States, 1825-1929: construction and implications. American Economic Journal: Macroeconomics 7, 295-330.

James, J. A., 1976. The conundrum of the low issue of National Bank notes. Journal of Political

AN US

Economy 84, 359-368.

James, J. A., 1978. Money and Capital Markets in Postbellum America. Princeton University Press, Princeton, NJ.

M

James, J. A., Weiman, D. F., 2011. The National Banking Acts and the transformation of New York City banking during the Civil War era. Journal of Economic History 71, 338-362.

ED

Jaremski, M., 2010. Free bank failures: risky bonds vs. undiversified portfolios. Journal of Money, Credit, and Banking 42, 1565-1587.

PT

Jaremski, M., 2013. State banks and the National Banking Acts: measuring the response to

CE

increased financial regulation, 1860-1868. Journal of Money, Credit, and Banking 45, 379-399.

AC

Jaremski, M., 2015. Clearinghouses as credit regulators before the Fed? Journal of Financial Stability 17, 10-21.

Johnston, L., Williamson, S. H., 2013. What was the U.S. GDP then? MeasuringWorth. http://www.measuringworth.org/consumer/

ACCEPTED MANUSCRIPT

Keehn, R., 1974. Federal bank policy, bank market structure, and bank performance: Wisconsin 1863-1914. Business History Review 48, 1-21. Kindahl, J. K., 1961. Economic factors in specie resumption in the United States, 1865-79. Journal of Political Economy 69, 30-48.

CR IP T

Krueger, L., 1933. History of Commercial Banking in Wisconsin. University of Wisconsin Press, Madison, WI.

Martinez-Peria, M. S., Schmukler, S. L., 2001. Do depositors punish banks for bad behavior? Market discipline, deposit insurance, and banking crises. Journal of Finance 56, 1029-

AN US

1051.

Merchants and Bankers‟ Directory (also called Merchants and Bankers‟ Almanac). 1861-1890. Bankers‟ Magazine Office, New York.

M

Minnesota Population Center, 2004. National Historical Geographic Information System: Prerelease Version 0.1. University of Minnesota, Minneapolis, MN. http://www.nhgis.org.

ED

Mitchell, B. R., 2007. International Historical Statistics. Europe, 1750-2005. Palgrave Macmillan, New York.

PT

Mitchell, W. C., 1908. Gold, Prices, and Wages under the Greenback Standard. University of

CE

California Publications in Economics, Vol. 1. The University Press, Berkeley, CA. Mitchener, K. J., Jaremski, M., 2015. The evolution of bank supervision: evidence from U.S.

AC

states. Journal of Economic History 75, 819-859. Moen, J., Tallman, E., 1992. The bank panic of 1907: the role of trust companies. Journal of Economic History 52, 611-630.

Officer, L. H., 2008. What was the value of the U.S. consumer bundle then? MeasuringWorth. http://www.measuringworth.org/consumer/

ACCEPTED MANUSCRIPT

Perron, P., Vogelsang. T. J., 1992. Nonstationarity and level shifts with an application to purchasing power parity. Journal of Business and Economic Statistics 10, 301-320. Rand McNally Bankers‟ Directory. 1890-1914. Rand McNally, Chicago, IL. Ramirez, C. D., 2009. Bank fragility, „money under the mattress,‟ and long-run growth: U.S.

CR IP T

evidence from the „perfect‟ panic of 1893. Journal of Banking and Finance 33, 21852198.

Rousseau, P. L., 1998. The permanent effects of innovation on financial depth: theory and U. S. historical evidence from unobservable components models. Journal of Monetary

AN US

Economics 42, 387-425.

Rousseau, P. L., 2011. The market for bank stocks and the rise of deposit banking in New York City, 1866–1897. Journal of Economic History 71, 976-1005.

M

Sylla, R., 1969. Federal policy, banking market structure, and capital mobilization in the United States, 1863-1913. Journal of Economic History 29, 657-686.

ED

Weber, W. E., 2005. Listing of all State Banks with Beginning and Ending Dates. Research Department, Federal Reserve Bank of Minneapolis.

PT

http://research.mpls.frb.fed.us/research/economists/wewproj.html.

CE

Weber, W. E., 2008. Balance Sheets for U.S. Antebellum State Banks. Research Department, Federal Reserve Bank of Minneapolis.

AC

White, E. N., 1983. The Regulation and Reform of the American Banking System, 1900-1929. Princeton University Press, Princeton, NJ.

ACCEPTED MANUSCRIPT

Appendix A. Construction of data on aggregate deposits The aggregate data draw from a variety of sources. Antebellum data are from two databases collected by Weber (2005, 2008). The first is a census of banks that provides the location and dates of operation for each bank ever in existence before 1861. The second contains

CR IP T

items from their annual balance sheets including information on size and portfolio composition. While the most complete source of bank data for the period, balance sheet data are missing in some years for several banks, and a few banks have no balance sheet data available at all. The gaps occurred most frequently for years prior to 1830 and for banks that operated for less than a

AN US

year. When building aggregates we linearly interpolate across missing observations in a single bank‟s time series. As nearly all the gaps are for single years and other banks display no large jumps, the interpolation should not bias our aggregates. For banks missing balance sheet data

M

altogether, we fill in observations using the average values of the lower quartile of the bank‟s closest geographic neighbors. The process begins by matching banks in the same county. Banks

ED

without an immediate match are then matched with banks in the same state. Because banks in a given area often had similar balance sheet compositions, the matching process should lead to

PT

reasonable aggregates for examining general trends in liability structures.

CE

The decline of state banks for some time after 1861 and their resurgence after 1880 creates challenges for building a comprehensive bank-level dataset for the postbellum era. But

AC

we do have full coverage for national banks from the Annual Report of the Comptroller of the Currency. To facilitate building an aggregate series of banking variables that includes state banks, we obtain the total number and aggregate balance sheet composition of state banks for 1861-1863 and 1873-1913 as reported by the various issues of the Annual Report of the Comptroller of the Currency. To fill in the missing years, we assume that state banks adjusted

ACCEPTED MANUSCRIPT

their balance sheets linearly from the average per bank in 1863 to the average in 1873 and apply the interpolated average to the number of state banks. This approach produces smooth aggregate statistics with estimates that merge seamlessly into the observed series and match the national bank averages that display no sharp jumps during the period. Note also that the number of state

CR IP T

banks was small compared to national banks between 1865 and 1872. For example, there were 349 state banks and 1,512 national banks in 1865 and still only 566 state banks and 1,918

national banks in 1872, and the average state bank was less than half the size of the average

aggregate values. Appendix B. Regional deposit growth

AN US

national bank. In these ways, our assumptions for state banks have relatively small effects on the

Davis (1965), Sylla (1969), and James (1978) emphasize the large regional differences

M

that existed before 1900. They argue that the lack of banking in rural areas allowed some banks

ED

to act as monopolists and charge high interest rates. In Fig. A1, we categorize banks into five groups: reserve city national banks and country banks located in each of the four geographic

PT

regions (i.e. Northeast, Midwest, South, and West). The data show large regional differences in the levels of deposits relative to GDP as a result of their different levels of population and

CE

economic activity. Likewise, all regions displayed relatively similar growth patterns but at different rates. When deflating deposits by capital, the regions are more similar. Each saw sharp

AC

increases in the ratio after 1879 and 1897. The evidence further indicates that the growth in deposits around the three key dates was a response to nation-wide events rather than locationspecific ones.

ACCEPTED MANUSCRIPT

Table 1 Tests for structural breaks in deposits series.a Estimated Structural Breaks, 1800-1913 Log of deposits Ratio deposits to capital Ratio deposits to GDP

Clemente et al. (1998) - 2 Breaks

1896

1895

1878

1878, 1897

1861, 1896

1863, 1896

CR IP T

Perron and Vogelsang (1992) - 1 Break

Estimated Structural Breaks, 1865-1913 Log of deposits Ratio deposits to capital Ratio deposits to GDP Perron and Vogelsang (1992) - 1 Break

1895

1878, 1897

1878, 1897

AN US

Clemente et al. (1998) - 2 Breaks

1897

a

1896

1878, 1897

AC

CE

PT

ED

M

The table presents results from two different tests for structural breaks in the aggregate data on deposits. The top panel contains all 114 years in the sample, and the bottom panel contains years after 1865. Deposits are deflated to constant 1870 dollars using the GDP deflator from Officer (2008) before taking their logarithms. The column headings list the variables upon which each test is based.

ACCEPTED MANUSCRIPT

Table 2 Bank-level determinants of national bank deposits, 1865-1913.a Log of deposits (1) (2)

Log of circulation (3) (4)

Ratio deposits to capital (5) (6)

0.004***

0.003***

-0.006***

-0.006***

0.008

0.002

[0.001]

[0.001]

[0.001]

[0.002]

[0.006]

[0.006]

Log assets of closed national banks in county over previous ten years

0.002***

0.003***

-0.002**

0.000

0.004*

0.006**

[0.001]

[0.001]

[0.001]

[0.001]

[0.002]

[0.002]

Clearinghouse in county

0.110***

0.143***

-0.095***

-0.052**

0.201**

0.188**

[0.018]

[0.018]

[0.022]

[0.022]

[0.094]

[0.093]

0.751***

0.629***

1.042***

0.892***

[0.010]

[0.010]

[0.021]

[0.025]

Log of capital stock Log of bank age

0.282***

Ratio loans to assets

a

0.688***

[0.011]

[0.019]

0.086*

0.054

-0.075**

-0.107***

-0.021

-0.118**

[0.050]

[0.043]

[0.034]

[0.041]

[0.046]

[0.058]

Yes Yes Yes 166,616 0.588

Yes Yes Yes 166,616 0.632

Yes Yes Yes 166,616 0.435

Yes Yes Yes 166,616 0.471

Yes Yes Yes 166,616 0.303

Yes Yes Yes 166,616 0.341

M

State fixed effects Year fixed effects State x Year fixed effects Observations R2

0.312***

AN US

[0.005]

CR IP T

Number of other national banks in county

AC

CE

PT

ED

The table presents results of ordinary least squares regressions. Each observation is a national bank-year, and the dependent variable is listed in the column headings. All money values are deflated to 1870 dollars using Officer (2008). Standard errors appear in brackets beneath the coefficients and are clustered by state. *, **, and *** denote statistical significance at 10%, 5%, and 1% levels, respectively.

ACCEPTED MANUSCRIPT

Table 3 Midwest bank-level determinants of deposits, 1888-1913.a Log of deposits (1) (2) (3) Number of other national 0.001 -0.001 -0.004 banks in county [0.002] [0.002] [0.004]

Ratio deposits to capital (4) (5) (6) 0.016 0.037 0.118 [0.028] [0.031] [0.078]

0.007*** [0.002]

0.007*** [0.002]

0.011*** [0.003]

0.096*** [0.013]

0.051*** [0.013]

0.100*** [0.019]

Log assets of closed banks in county over previous ten years

0.002** [0.001]

0.002*** [0.001]

0.002*** [0.001]

-0.016*** [0.005]

-0.008* [0.005]

-0.009** [0.005]

Clearinghouse in county

0.053** [0.024]

0.065*** [0.023]

0.029 [0.049]

-0.233 [0.183]

-0.182 [0.177]

-0.420 [0.388]

Log of capital stock

0.684*** [0.034]

0.516*** [0.034]

0.511*** [0.047]

National bank dummy

0.412*** [0.043]

0.175*** [0.040]

-0.512*** [0.160]

-1.688*** [0.178]

Ratio loans to assets

0.312*** [0.084]

National*Other national banks

National*Log assets of closed banks

National*US bond price

PT

National*Capital

0.215*** [0.067]

0.318*** [0.110]

ED

National*Clearinghouse

0.423*** [0.009]

M

National*Other state banks

0.426*** [0.009]

AN US

Log of bank age

CE

AC

Observations R2 a

0.022 [0.194]

1.311*** [0.049]

1.250*** [0.050]

-0.189 [0.194]

-0.342 [0.310]

0.007* [0.004]

-0.093 [0.074]

-0.007** [0.003]

-0.121*** [0.022]

-0.000 [0.000]

-0.000* [0.000]

0.048 [0.053]

0.283 [0.440]

0.000 [0.003]

-0.000 [0.004]

0.026 [0.034]

National*Ratio loans to assets

State fixed effects Year fixed effects State x Year fixed effects

CR IP T

Number of other state banks in county

-0.225* [0.120] Yes Yes Yes 70,597 0.471

Yes Yes Yes 70,597 0.537

Yes Yes Yes 70,597 0.536

0.261 [0.321] Yes Yes Yes 70,597 0.118

Yes Yes Yes 70,597 0.166

Yes Yes Yes 70,597 0.163

The table presents results from ordinary least squares regressions. Each observation is a bank-year, and the dependent variables are listed in the column headings. The sample only contains observations for Indiana, Michigan, Pennsylvania, and Wisconsin. All money values are deflated to 1870 dollars using Officer (2008). Standard errors appear in brackets beneath the coefficients and are clustered by state. *, **, and *** denote statistical significance at ten percent, five percent, and one percent levels, respectively.

ACCEPTED MANUSCRIPT

Table 4 County-level determinants of national banking by decade.a Log of deposits per person (1) (2) 0.102*** 0.061* [0.035] [0.035]

Log of population

Counties with a bank Log of circulation per person (3) (4) 0.017 -0.022 [0.030] [0.029]

Log of capital per person (5) (6) -0.040 -0.083*** [0.029] [0.028]

0.194* [0.101]

0.149 [0.101]

0.141 [0.088]

0.092 [0.088]

0.343*** [0.088]

0.300*** [0.087]

Log of farm value per person

0.037 [0.029]

0.030 [0.029]

0.043 [0.028]

0.037 [0.027]

0.039 [0.027]

0.032 [0.026]

Log assets of closed national banks in county over previous ten years

0.001 [0.002]

0.001 [0.001]

0.002 [0.001]

0.001 [0.001]

-0.001 [0.001]

-0.001 [0.001]

Clearinghouse in county

0.130** [0.057]

0.135** [0.057]

0.193*** [0.053]

0.199*** [0.053]

0.155*** [0.053]

0.162*** [0.052]

0.581*** [0.060]

Decade fixed effects State fixed effects State x Year fixed effects Observations R2

Yes Yes Yes 4,763 0.441

a

AN US

Had converted state bank

CR IP T

Fraction of population living in area of +25,001

Yes Yes Yes 4,763 0.458

Yes Yes Yes 4,763 0.358

0.534*** [0.050] Yes Yes Yes 4,763 0.389

Yes Yes Yes 4,763 0.308

0.594*** [0.052] Yes Yes Yes 4,763 0.349

AC

CE

PT

ED

M

The table presents results from ordinary least squares regressions. Each observation is a county-decade. The bank sample only contains counties that had a national bank, and did not experience large boundary changes. The binary variable “Had a converted state bank” denotes counties that had at least one state bank during the 1860s convert to a national bank charter. Dollar values are deflated to 1870 using Officer (2008). Non-bank variables are from Haines (2004). Robust standard are provided in brackets beneath the coefficient estimates. * denotes significance at 10%; ** at 5% level and *** at 1% level.

ACCEPTED MANUSCRIPT

Number of Banks

14,000 12,000 10,000 8,000 6,000

CR IP T

4,000 2,000

1800 1804 1808 1812 1816 1820 1824 1828 1832 1836 1840 1844 1848 1852 1856 1860 1864 1868 1872 1876 1880 1884 1888 1892 1896 1900 1904 1908 1912

0 State Bank

National Banks

Aggregate Values (% of GDP)

AN US

25.0% 20.0% 15.0% 10.0%

M

5.0%

ED

1800 1804 1808 1812 1816 1820 1824 1828 1832 1836 1840 1844 1848 1852 1856 1860 1864* 1868* 1872* 1876 1880 1884 1888 1892 1896 1900 1904 1908 1912

0.0%

Deposits

Circulation

Relative to Bank Capital

PT

600% 500%

CE

400% 300% 200%

AC

100%

1800 1804 1808 1812 1816 1820 1824 1828 1832 1836 1840 1844 1848 1852 1856 1860 1864* 1868* 1872* 1876 1880 1884 1888 1892 1896 1900 1904 1908 1912

0%

Dep/Cap

Circ/Cap

Fig. 1. Banking statistics before 1914. Notes: See Appendix A for sources of the aggregate data. Estimates of annual GDP are from Johnston and Williamson (2013).

ACCEPTED MANUSCRIPT

1.6 1.4 1.2

M

Deposits/Capital

1875

1874

1873

1872

1868

1867

1866

1865

1864

1863

1862

1860

1859

1858

1857

1856

1855

1854

1853

1852

1851

1850

0

1861

0.2

AN US

0.4

1871

0.6

1870

0.8

1869

CR IP T

1

Circulation/Capital

Fig. 2. Average balance sheet ratios of state banks converting to national charters between 1863 and 1868.

AC

CE

PT

ED

Notes: The figure includes a linear trend line for each series based on the antebellum data. Data on bank conversions are from Jaremski (2013).

ACCEPTED MANUSCRIPT

Aggregate Values (% of GDP) 16.0% 14.0% 12.0% 10.0%

Nat Deposits

1913

1911

1909

1907

1899

1897

1895

1893

1891

1889

State Capital

Nat Capital

AN US

State Deposits

1887

1885

1883

1881

1879

1877

1875

0.0%

1905

2.0%

1903

4.0%

1901

6.0%

CR IP T

8.0%

Relative to Bank Capital

700% 600% 500% 400%

State Dep/Cap

1913

1911

1909

1907

1905

1903

1901

1899

1897

1895

1893

1889

1887

1885

1883

1879

CE

PT

1877

1875

0%

1881

100%

ED

200%

1891

M

300%

Nat Dep/Cap

Fig. 3. National and state bank deposits, 1875-1913.

AC

Notes: See Appendix A for sources of the aggregate data. Estimates of annual GDP are from Johnston and Williamson (2013).

ACCEPTED MANUSCRIPT

$120

500%

$115

400%

$110

CR IP T

600%

$105

$100

$95

1913

1911

$90 1909

1905

1903

1901

1899

1897

1895

1891

Nat Circ./Cap (Left Axis)

M

Nat Dep/Cap (Left Axis)

1889

1887

1885

1883

1881

1879

1877

1875

0%

1893

100%

AN US

200%

1907

300%

U.S. Bond Price (Right Axis)

Fig. 4. U.S. bond prices, 1875-1913.

AC

CE

PT

ED

Notes: See Appendix A for sources of the aggregate data. Estimates of annual GDP are from Johnston and Williamson (2013). Bond prices were obtained from the Comptroller of the Currency‟s Annual Report. The prices (relative to face value) were averaged from the high and low quotes from the New York City market in November of each year. The bond price series comes from various issues of the Comptroller‟s reports. Since no single bond issue spans the entire National Banking period, we compute averages of several of the longer maturity bonds as they were available. To do this, we use U.S. 6s of 1865 (18651878), U.S. 6s of 1881 (1865-1881), U.S. 5s of 1881 (1872-1881), U.S. 4.5s of 1891 (1876-1890), U.S. 4s of 1907 (1877-1907), U.S. 5s of 1904 (1894-1903), U.S. 4s of 1925 (1895-1913), U.S. 2s of 1930 (19001913), and U.S. 3s of 1908-1918 (1898-1913). All bonds are as quoted in New York and expressed relative to face value. The average of the available bonds for each year is computed as the average of the monthly high and low prices in November. This synchronizes the series for bond prices with our balance sheet data.

ACCEPTED MANUSCRIPT

Without state-by-year interactions 3

4 3.5

2.5

3 2.5

2

Ln(Deposits) - Column (2) (Left Axis)

1.5 1 0.5 0 -0.5 -1

1913

1911

1902

1900

1898

1896

1894

1892

1890

1888

1886

1883

1881

1879

1877

1875

1873

1871

1869

1867

0

AN US

0.5

1909

1

1907

1.5

1904

CR IP T

2

Deposits/Capital - Column (6) (Right Axis)

With state-by-year interactions

4

M

3

3.5 3 2.5

ED

2.5 2

1.5

PT

1.5

2

1 0.5 0 -0.5

Ln(Deposits) - Column (2) (Left Axis)

1913

1911

1909

1907

1904

1902

1900

1898

1896

1894

1892

1890

1888

1886

1883

1881

1879

-1

1877

1871

1869

1867

AC

0

1875

0.5

1873

CE

1

Deposits/Capital - Column (6) (Right Axis)

Fig. 5. Plot of year fixed effects from Table 2. Notes: The omitted reference year is 1865. The Column number refers to the regression specification. The other specifications yield nearly identical patterns. While not displayed standard errors are generally small and fit tightly around the coefficient estimates.

ACCEPTED MANUSCRIPT

Without state-by-year interactions 0.6

5

0.5 4 0.4 3

-0.3

Ln(Deposits) - Column (3) (Left Axis)

2

1

1913

1912

1911

1910

1909

1905

1904

1903

1902

1901

AN US

-0.2

1900

1899

1898

1897

1896

1895

1894

1893

1892

1891

1890

-0.1

1889

0

1908

0.1

1907

0.2

1906

CR IP T

0.3

0

-1

Deposits/Capital - Column (6) (Right Axis)

With state-by-year interactions

0.5

3 2

1913

1912

1911

1910

1909

1908

1907

1906

1905

1904

1903

1902

1901

1900

1899

1 1898

1896

1895

1894

1893

1891

AC

CE

1890

1889

0

1892

0.1

PT

0.2

1897

0.3

-0.2

4

ED

0.4

-0.1

5

M

0.6

0

-0.3

-1

Ln(Deposits) - Column (3) (Left Axis)

Deposits/Capital - Column (6) (Right Axis)

Fig. 6. Plot of year fixed effects from Table 3. Notes: The omitted reference year is 1888. The column number refers to the regression specification. The other specifications yield nearly identical patterns. While not displayed standard errors are generally small and fit tightly around the coefficient estimates.

ACCEPTED MANUSCRIPT

Log(Deposits) 14 12 10

Ln(Deposits)

Holding Variables Constant

1912

1910

1908

1899

1897

1895

1893

1891

1889

1887

1884

1882

1880

1878

1876

1874

1872

1870

1868

1866

0

AN US

2

1906

4

1903

6

1901

CR IP T

8

Holding Coefficients Constant

Deposits/Capital

M

7

ED

6 5

PT

4

CE

3 2

AC

1

Dep/Cap

Holding Variables Constant

Holding Coefficients Constant

Fig. 7. Separating time effects from mean effects in the national bank regressions. Notes: See the discussion in Section 6 of the text for a description of Fig. 7.

1912

1910

1908

1906

1903

1901

1899

1897

1895

1893

1891

1889

1887

1884

1882

1880

1878

1876

1874

1872

1870

1868

1866

0

ACCEPTED MANUSCRIPT

Log(Deposits) 14 12 10

Ln(Deposits)

Holding Variables Constant

1913

1912

1911

1910

1906

1905

1904

1903

1902

1901

1900

1899

1898

1897

1896

1895

1894

1893

1892

1891

1890

1889

1888

0

AN US

2

1909

4

1908

6

1907

CR IP T

8

Holding Coefficients Constant

Deposits/Capital

M

9 8

ED

7 6

PT

5 4

Deposits/Capital

Holding Variables Constant

Holding Coefficients Constant

Fig. 8. Separating time effects from mean effects in the midwest bank regressions. Notes: See the discussion in Section 6 of the text for a description of Fig. 8.

1913

1912

1911

1910

1909

1908

1907

1906

1905

1904

1903

1902

1901

1900

1899

1898

1897

1896

1895

1894

1890

1889

1888

AC

0

1893

1

1892

2

1891

CE

3

ACCEPTED MANUSCRIPT

Aggregate values (% of GDP) 0.06 0.05 0.04

Midwest

South

1913

1911

1909

1907

1904

1896

1894

1892

1890

1888

1886

1883

Northeast

West

AN US

Reserve Cities

1881

1879

1877

1875

1873

1871

1869

1867

1865

0

1902

0.01

1900

0.02

1898

CR IP T

0.03

Relative to bank capital 8 7 6 5

Reserve Cities

Northeast

Midwest

South

1913

1911

1909

1907

1904

1902

1900

1898

1896

1894

1892

1890

1883

1881

1879

1877

PT

1873

1871

1869

1867

1865

0

1875

1

ED

2

1888

3

1886

M

4

West

CE

Fig. A1. National bank deposits by region, 1865-1913.

AC

Notes: See Appendix A for sources of the aggregate data. Estimates of annual GDP are from Johnston and Williamson (2013). The separation of reserve city banks is necessary because they were created to obtain more interbank deposits. Reserve Cities: New York City, St. Louis, Chicago, Boston, Cincinnati, New Orleans, Philadelphia, Albany, Cleveland, Detroit, Leavenworth, Louisville, Milwaukee, Pittsburgh, San Francisco, and Washington D.C.. Northeast: Maine, Vermont, New Hampshire, Massachusetts, Rhode Island, Connecticut, New York, New Jersey, and Pennsylvania. South: Delaware, Maryland, Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Alabama, Mississippi, Louisiana, Texas, Arkansas, Kentucky, Tennessee, and Oklahoma. Midwest: Ohio, Indiana, Illinois, Michigan, Wisconsin, Minnesota, Iowa, Missouri, North Dakota, South Dakota, Nebraska, and Kansas. West: Washington, Oregon, California, Arizona, Nevada, Utah, Montana, Wyoming, Colorado, New Mexico, and Idaho.