Clearinghouses as Credit Regulators Before the Fed?

Transcription

1 Colgate University Libraries Digital Colgate Economics Faculty Working Papers Economics Spring 2014 Clearinghouses as Credit Regulators Before the Fed? Matthew Jaremski Colgate University, mjaremski@colgate.edu Follow this and additional works at: Part of the Economic History Commons, Finance Commons, and the Macroeconomics Commons Recommended Citation Jaremski, Matthew, "Clearinghouses as Credit Regulators Before the Fed?" (2014). Economics Faculty Working Papers. Paper This Working Paper is brought to you for free and open access by the Economics at Digital Colgate. It has been accepted for inclusion in Economics Faculty Working Papers by an authorized administrator of Digital Colgate. For more information, please contact skeen@colgate.edu.

2 Clearinghouses as Credit Regulators Before the Fed? Matthew Jaremski * Colgate University, NBER, OFR Abstract: Clearinghouses were private organizations that not only had the power to audit member banks balance sheets and levy fines, but also provided emergency liquidity during large-scale financial panics. This paper studies how clearinghouses affected bank composition and solvency during stable periods as well as panics. An annual database of all national bank balance sheets from 1865 to 1914 indicates that national banks grew larger after the creation of a clearinghouse. Relative to the rise in assets, banks reduced their cash reserves and individual deposits and increased their loans, circulation, and interbank deposits. The analysis also shows that while clearinghouse members were less likely to fail during panics, they were more likely to fail in other periods, particularly those in non-financial centers. In this way, clearinghouses seem to have freed up additional resources during stable periods and delayed bank failures until the potential for contagion was removed. JEL: (G21, G32, N21) Keywords: Bank Regulation, Financial Panics, Clearinghouses, Bank Failure, National Banking Era * Dept. of Economics, Colgate University, 13 Oak Drive, Hamilton, NY 13346, USA. mjaremski@colgate.edu. The author would like to acknowledge the helpful comments of the participants of the "Instead of the Fed: Past & Present Alternatives to the Federal Reserve System" Conference at the Mercatus Center of George Washington University, particularly those from George Selgin, David Wheelock, and Ellis Tallman.

3 1. Introduction Clearinghouse associations--private, independent, and local organizations--filled a regulatory role from the inception of the first clearinghouse in 1853 until their virtual replacement by the Federal Reserve in Created by groups of state and national banks, the associations were established to facilitate the clearing of checks and bank notes, but quickly evolved to provide emergency liquidity during financial panics. Gorton (1985, p. 277) states that the "powers and functions that clearinghouses developed most resembled those of a central bank" and Cannon (1910, p. 96) calls them "one of the finest examples the country has ever seen of the ability of the people when left to themselves to devise impromptu measures for their own relief". The relatively small number of national bank failures during the late nineteenth century suggests that clearinghouses provided some level of protection amid the highly cyclical economy. While clearinghouses were not without their faults, questions remain as to how they affected surrounding banks and to what extent they were a viable private alternative to the Fed. Exploring the relative strengths and weaknesses of the clearinghouse system, though, has proved difficult because of a lack of bank-level data. Those studies that do exist have been too broad or too specific. The broad studies (e.g., Timberlake 1984, Gorton 1985, and Gorton and Mullineaux 1987) use anecdotal evidence from many cities and panics to discuss the structure and actions of clearinghouses. Specific studies (e.g., Moen and Tallman 1994 and 2000, and Hoag 2011) use detailed data on member and non-member banks in New York and Chicago during the Panics of 1893 and 1907 to examine whether clearinghouses mitigated member bank losses. Both types of studies often highlight the positive actions taken by clearinghouses, but neither test how clearinghouses affected bank composition during stable periods or whether those clearinghouses outside of financial centers also had an effect on banks. This paper uses a 1

4 comprehensive database of national bank balance sheets from 1865 to 1914 to provide a fuller account of the effects of clearinghouses. There is no consensus as to whether clearinghouses should have contributed to bank stability during normal periods. On the one hand, members had to satisfy a capital test, provide balance sheet information, and submit to examinations. Members also had an incentive to monitor each other because loan certificate losses were shared amongst the membership. At the same time, clearinghouses might have given banks the opportunity to take on additional risk. Weber (2012) argues that members lacked sufficiently strong incentives to monitor one another because the costs were high and losses were generally low. Gorton and Mullineau (1987) also point out that the costs of cheating were low because membership sent a public signal of confidence to depositors. As long as people believed that clearinghouses were monitoring members and preventing banks from taking on too much risk, they had little incentive to actively monitor, and members had little incentive to hold excess cash reserves. Before testing the effect of clearinghouses on individual banks, a data decision must be made: either examine all national banks from 1865 to 1914 or examine both state and national banks for a handful of states and years. 1 I have chosen to focus on national banks for several reasons. First, the uniform regulation of national banks reduces the possibility of omitted policy variables though, admittedly, their high requirements would likely reduce the effect of clearinghouses. Second, nearly all national banks were members of their local clearinghouse, whereas many state banks chose not to join or were excluded from becoming members. Third, I am able to separate national banks that were declared insolvent from those that voluntarily liquidated, changed name, or merged with another bank. Finally, the longer time series enables 1 Mitchener and Jaremski (2013) illustrate that few states reported bank balance sheets before

5 the examination of bank behavior before and after the creation of a clearinghouse, allowing for the control of previous trends and patterns of behavior that might have continued after entry. I use Jaremski's national bank database (2013) which contains the annual balance sheet of every national bank in operation from 1865 through I then link the bank data to the dates of clearinghouse entry as reported by the Merchants and Bankers Directory ( ), Rand McNally Bankers Directory ( ), and Polks Bankers Encyclopedia ( ). The data indicate that national banks expanded their balance sheets and shifted their portfolios in response to the creation of a clearinghouse. After controlling for the age of the bank, population, urbanization, and fixed effects, the creation of a clearinghouse increased a bank s proportions of loans, circulation, and interbank deposits but decreased its proportions of individual deposits and cash reserves. These changes might have provided for a more effective use of resources by freeing up excess reserves for use in the economy. The analysis also shows that while banks in clearinghouse cities were less likely to fail during financial panics, they were more likely to fail in other periods. The instability, however, is largest for clearinghouses outside of reserve cities where resources might not have been sufficient to prop up the system and calm nervous depositors. In this way, clearinghouses likely prevented additional losses due to contagion and fire sales during panics, but allowed idiosyncratic bank failures when spillover effects were negligible. The Federal Reserve Act of 1913 thus might have been necessary to provide stability to the entire banking system over time-- something that the reactionary approach of the clearinghouses could not provide. 3

6 2. The Development of Clearinghouses in the United States New York City established the first U.S. clearinghouse association in This organization was created to lower the costs of clearing checks and bank notes. 2 Before its creation, banks had to send a representative to every other bank in the area in order to redeem their debt. Costs were low when there were few banks, but they quickly rose with the number of banks as more individual visits were needed. Clearinghouses reduced these costs by providing a central location and time to clear debt every day. Each member bank would then only have had to send a representative to the clearinghouse. The association also reduced the amount of currency that members had to pay out each day, as they were only required to make up the difference between other banks debt that they returned and their debt that other banks returned. This reduced the need for excess cash reserves and could have led to a more efficient allocation of assets. The clearinghouse cleared nearly $6 billion in debt at a cost of roughly 0.02 percent of member deposits in its first year (Gibbons 1859, Comptroller of the Currency 1897 Table 90). Membership in a clearinghouse came with stipulations and requirements. Banks generally had to pay a fee and pass an admissions test. The test typically involved an examination of the bank's balance sheet and certification of its capital stock. After becoming a member, banks had to continue to submit financial statements and allow periodic examinations when requested. Members who did not follow these guidelines were subject to fines and could be expelled from the association. As an expulsion signaled bad behavior and could attract a bank run, members had some incentive to adhere to the rules. Information releases by members helped the clearinghouse to ensure that member banks followed the association s rules and also facilitated clearing. The clearinghouse's ability to audit 2 Timberlake (1984, p. 2) put it best: "clearinghouses are no more mysterious nor complicated than grocery stores". 4

7 member bank's books gave it the unique ability to judge each institution's credit-worthiness (Gorton and Mullineaux 1987). The receiver of a check could worry less about whether the issuer and the issuer s bank had enough money to pay the check. Because of the number of parties and amount of private information involved, it was difficult for individual banks or brokers to solve these problems. 2.1 Clearinghouses as Lenders of Last Resort The clearinghouse system was designed to clear bank debt, but over time it evolved to do much more. As the Panic of 1857 approached, depositors and country banks began to withdraw their deposits, and the reserves of New York City banks began to thin. After a few prominent bank defaults, the policy committee of the clearinghouse decided to take action and authorized the issue of clearinghouse loan certificates to qualified member banks. 3 The certificates were backed by collateral from the receiving bank, but they were drawn on the clearinghouse, making them a joint-liability of the members. When a member defaulted, the remaining members bore the loss in proportion to their capital stock. Member banks thus could conserve their limited currency reserves by settling their clearinghouse obligations with certificates rather than currency, and liquid banks that did not need reserves gained interest on the certificates. 4 The certificates were designed with large denominations to prevent them from circulating outside of the banking system and violating the restrictions on private currency. The clearinghouse, therefore, was not concerned about injecting liquidity into the economy, but it rather focused on protecting the banking system. (Gibbons 1859, Calomiris and Schweikart 1991) 3 Given that interest rates were fixed for all borrowers and everyone was on the hook for losses, the loan committee likely took great care in choosing borrowers and assessing the collateral. 4 It is important to note that clearinghouse members were required to accept loan certificates. 5

8 While the issue of loan certificates was important, the clearinghouse further protected its members from negative information releases during the panic. This included replacing the weekly publication of individual bank statements with the publication of a single aggregate clearinghouse statement, limiting predatory interest rate practices, and guarding the names of banks that received loan certificates (Cannon 1910, p. 90). These steps were important to prevent spillover effects. If a handful of members were identified as insolvent, then the public might question the solvency of all members. Even if depositors knew only one bank was insolvent and ran on it, its losses and sudden liquidation would have chipped away at the security of stable banks. 5 The system thus had an incentive to string along failing banks until after a panic had subsided. Through his examination of the minutes of the New York Clearinghouse, Gorton (1985, p. 282) finds that banks were not allowed to fail during the period of suspension of convertibility, but were expelled from clearinghouse membership for failure to repay loan certificates after the period of suspension had ended. 6 Each successive panic pushed clearinghouses to innovate. During the Panic of 1860, the New York Clearinghouse pooled their members' reserves to create a fund for mutual aid for the first time. Reserves were pooled again in 1861 and 1873 but never thereafter. The clearest reason for its discontinued use seems to be the difficulty of forcing member banks' compliance, as some members attempted to hide reserves (Wicker 2005). The Panic of 1873 ushered in the use of certified checks. These checks differed from loan certificates because receiving banks did not have to give up collateral. The lack of collateral made them easier for banks to issue and let banks keep safe assets on their balance sheets. To 5 Gorton (1985, p. 280) comes to a similar conclusion: "Banks had an incentive to avoid other members' failures because of the information externalities. 6 Gorton references the Clearinghouse Committee Minutes on January of

9 counter this increased usability, clearinghouses kept denominations of these checks high, like loan certificates, to continue to prevent their wide circulation. The Panic of 1893 brought the use of low denomination clearinghouse notes. Unlike clearinghouse loan certificates and certified checks, these notes could be issued to anyone and operated as a makeshift currency. Cannon (1910) relates that Atlanta's clearinghouse issued notes as low as $10 and others issued denominations as low as $0.25 to help make change during the Panic. The use of small denomination notes saw even greater use during the Panic of 1907 when clearinghouses issued three and a half times more notes than issued in 1893 (Andrew 1908). Cannon relates that during the Panic of 1907 some clearinghouses even printed blank checks made out to "bearer" so that banks could quickly fill in the needed amount. Hepburn (1924, p. 352) states: This temporary currency performed so valuable a service...in moving crops and keeping business machinery in motion, that the government, after due deliberation, wisely forbore to prosecute [the issue of illegal currency]." 7 It should be noted that not every clearinghouse chose to issue notes and certificates. For instance, New York City s clearinghouse only ever issued clearinghouse loan certificates, focusing on the stability of their banks rather than the community at large. By the creation of the Federal Reserve in 1913, clearinghouse associations were issuing large amounts of emergency funds during panics. The public also took notice of the actions of clearinghouses. Moen and Tallman (2000, p. 152) reports that the Chicago Daily Tribune reported early on in the panic [of 1907] about the importance of the Chicago Clearinghouse in reducing the riskiness of the Chicago banking system and in moderating the effects of the panic. The clearinghouse thus had a variety of effects during panics. The loan certificates and certified 7 Timberlake (1984) contains a more detailed discussion of these topics and sources. 7

10 checks impacted those inside of the banking system, the clearinghouse notes helped those outside of the banking system, and the actions provided a public signal of stability. Despite the measures taken during panics, clearinghouses ultimately sought to protect their members from financial panics but not necessarily protect the rest of the banking system or to prevent crises from occurring. 8 This is evident in several respects. First, clearinghouses had limited memberships. As shown by Moen and Tallman (2000), state banks that were not allowed to join Chicago s clearinghouse and trust companies that were largely excluded from the New York City s clearinghouse experienced greater losses of deposits in Banks in locations without clearinghouses were also generally excluded from receiving emergency liquidity. 10 Second, clearinghouses only provided emergency liquidity during financial panics. A New York Clearinghouse report states: "It must always be remembered that the absence of any important central institutions, such as exists in other commercial nations, the associated banks are the last resort in this country, in times of financial extremity" (1858, p.50; cited by Wicker 2005, p. 20). Indeed, clearinghouses only issued loan certificates in 1857, 1860, 1861, 1863, 1873, 1884, 1890, 1893, and 1907 (Sprague 1910, pp ). Finally, clearinghouses responded to crises and did not seem concerned with preventing them. They generally issued clearinghouse certificates only to prevent the closure of member banks from hurting the stability of others, and thus clearinghouses did not spring into action until the panic was clearly visible and banks were in trouble. For example, the New York 8 Moen and Tallman (2013) for instance show that the size of clearinghouse loan issues in New York City varied based on the degree that local banks were exposed to panic rather than how large the panic itself was. 9 New York City's Clearinghouse explicitly excluded trust companies until However, even after 1902, no trust companies became members due to the clearinghouse s reserve requirements. 10 Cannon (1910, pp ) describes that clearing-house certificates were taken out in the Southeast in cities were no clearing-house existed" suggesting that others took advantage of emergency liquidity. However, given the potential for federal prosecution, these occurrences were likely few and far between. 8

11 Clearinghouse waited nearly a month after the start of the Panic of 1893 to issue clearinghouse loan certificates as the panic started in the west and only later moved to the east. The historical period during which clearinghouses operated saw relatively few national bank failures. Figure 1 illustrates national bank failures from 1870 through Just over 500 of the almost 11,000 national banks were declared insolvent over the 50 year period, amounting to a little over 1 percent per year. The figures show that banks in clearinghouse cities often had a higher failure rate than banks in other cities. However, an interesting pattern emerges around several of the panics: the fraction of clearinghouse bank failures is about the same (or less) than the fraction for other banks during the panic year but is significantly higher in the years after. This is true during the Panic of 1893 when the fraction of failed banks was around 1.5% for both types, but this rose to 2% in 1896 for clearinghouse cities and fell to 0.7% for others. A starker pattern is visible during the Panic of 1885 when clearinghouse cities had a lower failure rate in 1885 but nearly double the rate of others in This suggests that clearinghouses kept some banks from failing during panics in order to avoid contagion, and let them fail when losses would not have affected the rest of the membership The fact that clearinghouse failures were still relatively high during panics might be driven by their not issuing loan certificates until after a bank had failed. Because of the small number of clearinghouse banks in the denominator, even a single failure would cause a large spike in the fraction of bank failures. For instance, only four clearinghouse banks became insolvent in 1908 yet the fraction of insolvent banks was 0.5%. 9

12 Figure 1: Bank Outcomes ( ) 50 Number of Insolvent Banks No CH CH 2.0% Fraction of Insolvent Banks 1.5% 1.0% 0.5% 0.0% No CH CH Notes: Insolvency is taken from Comptroller of the Currency's Annual Report. "No CH" denotes counties without a clearinghouse in that year, and "CH" denotes those with a clearinghouse. 2.2 The Rise of Clearinghouses The success of the clearinghouse model in lowering clearing costs depended upon establishing intimate relationships with member banks and keeping operations relatively local. Banks in other large financial centers, therefore, formed their own associations rather than

13 clearing debt through New York City. Most associations adopted a structure similar to New York City's but installed some of their own rules and regulations. For instance, Buffalo and Sioux City limited interest rates on deposits (Cannon 1910), whereas Chicago s clearinghouse allowed trust companies but not state banks to become members (Moen and Tallman 2000). As a result, over 200 clearinghouses were spread across every state but Arizona by Figure 2 provides a geographic view of clearinghouses across time. The early expansion seems to have been limited to large financial centers, particularly those labeled as central reserve clearinghouses between 1853 and The first five clearinghouses (New York, Boston, Baltimore, Cleveland, and Philadelphia) were among the largest cities both in terms of population and banks. The timing of entry indicates that debt clearing and not emergency liquidity drove the early growth of clearinghouses. These cities had the largest number of bank and thus the largest cost of clearing debt each day as representatives had to visit each bank individually. Alternatively, only a few clearinghouses entered right after the Panics of 1857 or 1873, yet 12 entered in-between. The Panic of 1884 represents a turning point in clearinghouse entry. While a relatively minor panic, it was centered on the collapse of the Marine National Bank of New York City. The New York Clearinghouse averted a major panic by injecting liquidity, but it was clear that Wall Street was vulnerable and panics could easily spread to the interior. After the panic, more clearinghouses began to be established outside of financial centers, often directly after financial crises. Seen in Figure 3, 5 clearinghouses were established within three years after 1884, 25 were established within three years after 1893, and 34 were established within three years after Despite the potential that these clearinghouses were different from the rest, their inclusion in the sample does not alter the results in any significant way. 11

14 Figure 2: Clearinghouses By County ( ) Notes: Figures display the active clearinghouses in each year. Locations and dates of operation are from the bank directories described in the data section. County boundaries obtained from Minnesota Population Center (2004). The desire for a lender of last resort and building distrust of New York City s ability to contain crises thus seems to have incentivized smaller cities to establish clearinghouses. 3. Data I use Jaremski's database of national bank balance sheets (2013) to analyze the effects of clearinghouses on bank behavior. Culled from the Comptroller of the Currency's Annual Report, 12

15 25 Figure 3: Number of New Clearinghouses By Year ( ) Notes: Figure presents the number of new clearinghouses that entered in each year. the data contain the annual balance sheet of every national bank in operation during the third quarter of each year. 13 Data are missing for 1885 when balance sheets were not reported and for 1905 when certain balance sheet items are combined. Using the tables provided by the Comptroller's each year, I separate national banks that were declared insolvent and placed in the hands of a receiver from those that voluntarily liquidated, changed their name, or merged with another bank. 14 I also drop any balance sheets published during the year of a bank s closure in order to avoid changes made in anticipation of closure. The resulting database contains 174,124 observations from 10,843 banks. 13 While there is some possibility of window dressing early in the data, the Comptroller effectively randomized the call date after 1869 in order to gain a more accurate view of balance sheets. 14 Banks that were declared insolvent but later returned to solvency are not treated as insolvent. 13

16 The dates of clearinghouse entry come from the Merchants and Bankers Directory ( ), Rand McNally Bankers Directory ( ), and the Polks Bankers Encyclopedia ( ). Similar to modern telephone books, the bank directories provide a list of banks and trust companies in each year, but they also listed clearinghouses under their corresponding city. I designate entry as the year before the clearinghouse first appeared in the directories and verify dates with Cannon (1900) and the Comptroller s Annual Report whenever possible Empirical Analysis Using the micro-level database on banks and clearinghouses, I examine three questions. First, what types of locations created clearinghouses and when were they created? Second, did the entry of a clearinghouse cause incumbent banks to change their composition? Third, were clearinghouse members more likely to become insolvent? 4.1 Entry of Clearinghouses In order to determine whether clearinghouses affected bank behavior, we have to know something about what led to the creation of a clearinghouse. If the creation was endogenous to the underlying institutions or banking system, then we must control for those factors before examining the effect of clearinghouses on banks. Based on the historical narrative, earlier clearinghouses were created in locations that had reached some level of urban or financial development, whereas later clearinghouses were created in response to financial crises. While crises can be controlled for through year fixed effects, urbanization might have affected subsequent bank behavior. I thus begin by testing for them. 15 Dating entry one year prior to its appearance best matches the other sources, but the results are similar when using the contemporaneous year. 14

17 I model clearinghouse creation using a linear probability model where the dependent variable is a dummy variable that takes the value of "1" if a clearinghouse was in operation in the specified year. 16 While the bank data exist for every year, clearinghouses were often only created in large cities and thus data on the population are also needed. Unfortunately, even the best source of this data, the county-level Census database assembled by Haines (2004), only contains data each decade. 17 This data limitation is restrictive because many clearinghouses were not created at the end of a decade, meaning that the model would often pick up changes occurring after the clearinghouse entered. To obtain better estimates of the causes of clearinghouse creation, I assume that the county-level census variables grew linearly over time to estimate county-level aggregates in-between each decennial observation. 18 The resulting sample consists of county-level observations in 1870, 1875, 1880, 1885, 1890, 1895, 1900, 1905, and As the creation of a clearinghouse was likely to have quickly influenced the banking system in the county but not economic activity, the banking measures are lagged but the census variables are not. Moreover, observations after a clearinghouse was created are dropped to avoid changes that occurred after creation, and counties without banks are dropped to avoid locations that would not need a clearinghouse. The full specification is:,,,,, 1 where, is a vector of county-level variables including the logarithm of population, the logarithm of farm value per person, and the fraction of the population living in an area of 25, The selection of the model was chosen for easily translatable coefficients and the ease of including fixed effects. The choice, however, does not affect the conclusions as other models (e.g. hazard, probit, etc) yield similar results. 17 I drop any counties that changed their area by more than 5% between 1870 and 1910 in order to create a more consistent panel over time. County areas are taken from the Minnesota Population Center (2004). 18 The results are similar when only using the decennial observations. 15

18 Table 1: County-Level Determinants of Clearinghouse Entry ( ) Probability of Gaining a Clearinghouse (1) (2) (3) (4) (5) (6) Ln(Population) 0.012*** 0.082*** 0.012*** 0.086*** 0.012*** 0.087*** [0.004] [0.020] [0.004] [0.021] [0.004] [0.021] Fraction Living in Area 0.264*** 0.387*** 0.264*** 0.380*** 0.264*** 0.379*** of 25,000 or More [0.035] [0.064] [0.035] [0.064] [0.035] [0.064] Ln(Farm Value Per Person) [0.003] [0.005] [0.003] [0.005] [0.003] [0.005] L.Number of 0.003* 0.006** 0.003* 0.007*** 0.003* 0.007*** National Banks [0.002] [0.002] [0.002] [0.003] [0.002] [0.003] L.Ln(Assets) *** [0.001] [0.001] L.Ln(Circulation) [0.001] [0.001] L.Ln(Deposits) [0.001] [0.002] L.Ln(Capital) [0.002] [0.002] Fixed Effect State County State County State County Observations 7,149 7,149 7,149 7,149 7,149 7,149 R-squared Notes: Table presents the results of an OLS regression. The dependent variable is whether or not a county gained a clearinghouse in that particular year. Each observation is a county during a 5-year period. Counties that changed their boundaries more than 5% between 1870 and 1910, had a clearinghouse before 1870, or did not have a bank are dropped. Period fixed effects are included in all specifications. Counties are also dropped from the sample after a clearinghouse was established. Dollar values are deflated to 1870 using Officer (2008). Robust standard errors are provided in brackets. * denotes significance at 10%; ** at 5% level and *** at 1% level. people or more. 19, is number of national banks in the county at the beginning of the 5- year period., is a measure of national banking activity (i.e., the logarithms of assets, circulation, deposits, or capital) that captures whether any specific aspects of bank balance sheets attracted clearinghouses. The model includes fixed effects for each five-year 19 All dollar values are deflated to an 1870 basis using Officer (2008). While a formal measure of urbanization (i.e., a place of more than 2,500) could be used, only a handful of clearinghouses were established in a city with fewer than 25,000 people and thus the included variable should better capture the determinants of entry. 16

19 period ( ) and either for state or county ( ). The inclusion of county-fixed effects is an attempt to identify whether the variables explain the creation of a clearinghouse within a county as well as across counties. Table 1 shows that banks in counties with large population and urban centers were the most likely to create a clearinghouse. The coefficients on population and the fraction of the population living in an area of 25,000 or more are always statistically significant and positive. A county that had 10 percent more population had a 0.1 to 0.8 percentage point higher probability of creating a clearinghouse, whereas a county that had 10 percentage points more people living in a large city had a 2.6 to 3.8 percentage point higher probability. Even after dropping counties without a bank, the number of national banks is important to the creation of a clearinghouse. Every additional bank in a county increased the probability of establishing a clearinghouse by between 0.3 and 0.7 percentage points. The effect is statistically significant, but is small relative to the effect of population and urbanization. The small coefficient is due to dropping counties without banks. When these counties are included, the coefficient on the number of national banks increases substantially. The coefficients on the other banking variables display an interesting pattern. The value of assets does not matter when comparing across counties; however, within a county, clearinghouse creation is negatively correlated with bank assets. While very small, the negative coefficient is likely picking up the effect of clearinghouses being established after financial crises. The fact that the number of banks is important but not the size of their liabilities is unsurprising given the costs of clearing debt. Without a clearinghouse, banks had to send a representative to each bank regardless of the how much debt had accumulated. Therefore, the primary cost of clearing would have been the number of banks rather than their size. 17

20 Table 2: Effect of Previous County-Level Changes on Clearinghouse Entry ( ) Probability of Gaining a Clearinghouse (1) (2) (3) (4) (5) (6) Ln(Population) 0.014*** 0.090*** 0.013*** 0.094*** 0.012*** 0.095*** [0.005] [0.022] [0.004] [0.023] [0.004] [0.023] Fraction Living in Area 0.281*** 0.430*** 0.278*** 0.419*** 0.275*** 0.419*** of 25,000 or More [0.038] [0.076] [0.037] [0.076] [0.037] [0.076] Ln(Farm Value Per Person) [0.003] [0.005] [0.003] [0.006] [0.003] [0.006] L.Number of ** ** ** National Banks [0.002] [0.003] [0.002] [0.003] [0.002] [0.003] L.Change in Number of Banks 0.007** *** *** [0.003] [0.003] [0.004] [0.004] [0.004] [0.004] L.Ln(Assets) *** [0.001] [0.001] L.Change in Ln(Assets) *** [0.001] [0.001] L.Ln(Circulation) [0.001] [0.002] L.Change in Ln(Circulation) [0.001] [0.001] L.Ln(Deposits) [0.002] [0.004] L.Change in Ln(Deposits) [0.002] [0.002] L.Ln(Capital) [0.003] [0.005] L.Change in Ln(Capital) [0.002] [0.002] Fixed Effect State County State County State County Observations 6,660 6,660 6,660 6,660 6,660 6,660 R-squared Notes: Table presents the results of an OLS regression. The dependent variable is whether or not a county gained a clearinghouse in that particular year. Each observation is a county during a 5-year period. Counties that changed their boundaries more than 5% between 1870 and 1910, had a clearinghouse before 1870, or did not have a bank are dropped. Period fixed effects are included in all specifications. Counties are also dropped from the sample after a clearinghouse was established. Dollar values are deflated to 1870 using Officer (2008). Robust standard errors are provided in brackets. * denotes significance at 10%; ** at 5% level and *** at 1% level. In Table 2, I test whether changes in the structure of the financial system pushed banks to establish a clearinghouse by including the previous growth as well as the initial lag. The results are quite similar. Population and urbanization still seem to be the primary determinants of clearinghouse entry. The coefficients on the lagged number of banks and value of assets and the 18

21 change trade off significance depending on the inclusion of fixed effects. The other balance sheet variables remain statistically insignificant. As expected, clearinghouses thus were likely installed in densely populated counties with a large number of banks in order to reduce clearing costs (i.e., the number of individual bank trips an employee would have to make). 4.2 Effect of Clearinghouses on Individual Balance Sheets While I could examine the effect of clearinghouses on any balance sheet variable, I have chosen several key variables based on historical studies of banking and the CAMELS measures used by modern bank regulators. Ln(Assets) measures bank size. The ratio of Capital to Assets measures capital adequacy. Cash to Assets is an index of bank reserves and liquidity, measuring the bank s capacity to meet bank runs. 20 The ratios of Individual Deposits to Assets and Circulation to Assets measure liability diversity, whereas the ratio of Loans to Assets measure asset diversity. The ratios of Due From Banks to Assets and Due to Banks to Assets measure the size of interbank relationships and the bank's exposure to market participants. I also separately regress the logarithm of each individual item to capture their extensive changes. I take several steps to isolate the effect of clearinghouses on banks. First, I include the number of other national banks in the city to control for any endogeneity between the variable and the creation of a clearinghouse. Second, I drop banks that entered a county after a clearinghouse was established because their composition might have been endogeneous to the existing clearinghouse. Third, I include the logarithm of the bank s age and year fixed effects to capture any trends that existed over the lifetime of the bank and across the financial system. The model takes the form of: 20 Cash includes any items accepted as money and thus includes cash, cash items, and specie. 19

22 ,,,,, 2 where the, consists of the balance sheet variables,, is the clearinghouse dummy, and the other variables retain their definitions. Similar to before, I include bank-fixed effects in some specifications to examine whether banks significantly changed their previous behavior in response to a clearinghouse rather than changed their behavior relative to banks in other locations. Table 3 shows that the size of existing banks increased after a clearinghouse was established. The largest effects seem to have been on Assets, Loans, and Due from Banks, but the value of nearly every balance sheet increased between 11 to 35 percentage points. While every other item increased relative to Assets, Deposits and Cash significantly decreased relative to Assets. The ratio of Deposits to Assets decreased by 3.7 percentage points and the ratio of Due to Banks to Assets increased by about 2.5 percentage points, but most of the ratios change by just under 1 percentage point after a clearinghouse was established. 21 While the results are statistically and economically significant, the annual regression specification cannot rule out the possibility that banks altered their behavior due to other countylevel changes. I address this criticism by adding the five-year Census variables and dropping bank data for years in-between. 22 The model is:,,,,,, 3 21 As reported in Table A1, the results are similar when aggregating the banking data to the county-level. 22 There is a potential geographic mismatch, but as most clearinghouses were in the largest if not only city in the county, the county-level aggregates and specifically urbanization should be highly correlated with city-level aggregates. Similar to the previous specifications, I drop any counties with large boundary changes. 20

23 Table 3: Annual Effect of Clearinghouses on Bank Balance Sheets ( ) Ln(Assets) Ln(Loans) Loans/Assets Ln(Cash) Cash/Assets Clearinghouse 0.255*** 0.233*** 0.276*** 0.249*** * 0.173*** 0.113*** *** *** in City [0.007] [0.023] [0.008] [0.024] [0.002] [0.004] [0.010] [0.027] [0.001] [0.001] Ln(Bank 0.311*** 0.291*** 0.393*** 0.377*** 0.020*** 0.031*** 0.273*** 0.218*** *** *** Age) [0.002] [0.005] [0.004] [0.007] [0.001] [0.001] [0.003] [0.006] [0.001] [0.001] Number of 0.046*** 0.036*** 0.059*** 0.045*** 0.003*** 0.004*** 0.037*** 0.012** ** *** Nat Banks [0.001] [0.005] [0.002] [0.006] [0.001] [0.001] [0.002] [0.006] [0.001] [0.001] Fixed Effect State Bank State Bank State Bank State Bank State Bank Observations 160, , , , , , , , , ,692 R-squared Ln(Due from Banks) Due from Banks/Assets Ln(Capital) Capital/Assets Ln(Indiv. Deposits) Clearinghouse 0.348*** 0.270*** 0.007*** 0.006*** 0.239*** 0.214*** 0.008*** 0.007** 0.183*** 0.128*** in City [0.012] [0.033] [0.001] [0.002] [0.006] [0.019] [0.001] [0.003] [0.011] [0.026] Ln(Bank 0.286*** 0.225*** *** *** 0.100*** 0.072*** *** *** 0.381*** 0.346*** Age) [0.004] [0.008] [0.001] [0.001] [0.002] [0.004] [0.001] [0.001] [0.004] [0.007] Number of 0.049*** 0.015** *** *** 0.045*** 0.035*** 0.002*** 0.003*** 0.033*** Nat Banks [0.002] [0.006] [0.001] [0.001] [0.001] [0.004] [0.001] [0.001] [0.002] [0.006] Fixed Effect State Bank State Bank State Bank State Bank State Bank Observations 160, , , , , , , , , ,692 R-squared Indiv. Deposits/Assets Ln(Circulation) Circulation/Assets Ln(Due to Banks) Due to Banks/Assets Clearinghouse *** *** 0.224*** 0.201*** 0.006*** 0.009*** 0.350*** *** 0.024*** in City [0.002] [0.005] [0.017] [0.035] [0.001] [0.002] [0.037] [0.097] [0.001] [0.003] Ln(Bank 0.022*** 0.028*** 0.550*** 0.541*** *** 0.002*** 0.933*** 0.764*** 0.006*** 0.006*** Age) [0.001] [0.001] [0.010] [0.014] [0.001] [0.001] [0.016] [0.035] [0.001] [0.001] Number of *** *** 0.045*** 0.031*** 0.001*** 0.002*** 0.203*** 0.092*** 0.004*** 0.004*** Nat Banks [0.001] [0.001] [0.003] [0.008] [0.001] [0.001] [0.007] [0.018] [0.001] [0.001] Fixed Effect State Bank State Bank State Bank State Bank State Bank Observations 160, , , , , , , , , ,692 R-squared Notes: Table presents the results of an OLS regression. The dependent variable is described in the column heading. Each observation is a bank-year, and year fixed effects are included in all specifications. Banks that entered after a clearinghouse was established are dropped. Dollar values are deflated to 1870 using Officer (2008). Robust standard are provided in brackets. * denotes significance at 10%; ** at 5% level and *** at 1% level.

24 where the variables keep their definitions, and t represents each five-year period. Note that the timing mismatch between clearinghouse creation and balance sheet observations should bias the coefficient on, towards zero, and the low number of observations per bank reduces the usefulness of the bank-fixed effects regressions. The results in Table 4 are similar to those in Table 3. A bank's Assets, Loans, Due from Banks, Capital, and Deposits grew larger after the entry of a clearinghouse. However, the levels of Cash and Due to Banks now have a negative and significant coefficient when including bankfixed effects and the coefficient on circulation is no longer statistically significant. When bankfixed effects are present, only the ratios of Loans, Due from Banks, Circulation, and Due to Banks to Assets grew larger after a clearinghouse entered, whereas the ratios of Deposits and Cash grew smaller. The data indicate that the creation of a clearinghouse increased the amount of national banking activity and allowed banks to make better use of their cash reserves. Moreover, clearinghouses seemed to create greater connections between banks. The reduction in cash holdings might have exposed banks to greater scrutiny. However, as long as clearinghouses reliably monitored its members and provided emergency liquidity during panics, the changes should not have been an issue. Indeed, they might have been more efficient as they freed up resources that would have previously gone unused. 4.3 Effect of Clearinghouses on Bank Failure I model the probability of insolvency over the following year using a linear probability model. Clearinghouses could have affected a bank s probability of failure by altering its composition, penalizing its risky actions, and providing emergency liquidity. I measure the first

25 Table 4: Five-Year Effect of Clearinghouses on Bank Balance Sheet Ratios With County Controls ( ) Ln(Assets) Ln(Loans) Loans/Assets Ln(Cash) Cash/Assets Clearinghouse 0.189*** 0.123*** 0.205*** 0.145*** * 0.074*** ** ** *** in City [0.019] [0.024] [0.021] [0.026] [0.004] [0.006] [0.027] [0.031] [0.001] [0.001] Ln(Bank 0.359*** 0.279*** 0.425*** 0.384*** 0.010*** 0.031*** 0.369*** 0.215*** *** *** Age) [0.005] [0.008] [0.008] [0.015] [0.001] [0.003] [0.007] [0.013] [0.001] [0.001] Number of 0.110*** 0.024*** 0.138*** 0.029*** 0.006*** *** *** ** Nat Banks [0.004] [0.007] [0.005] [0.008] [0.001] [0.002] [0.006] [0.010] [0.001] [0.001] Fixed Effect State Bank State Bank State Bank State Bank State Bank Observations 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 R-squared Due from Banks/Assets Ln(Capital) Capital/Assets Ln(Indiv. Deposits) Ln(Due from Banks) Clearinghouse 0.229*** 0.145*** 0.008*** 0.008*** 0.175*** 0.104*** 0.009*** *** 0.062* in City [0.041] [0.046] [0.002] [0.003] [0.017] [0.019] [0.003] [0.004] [0.028] [0.034] Ln(Bank 0.214*** 0.072*** *** *** 0.175*** 0.059*** *** *** 0.419*** 0.342*** Age) [0.007] [0.016] [0.001] [0.002] [0.003] [0.006] [0.001] [0.002] [0.008] [0.014] Number of 0.168*** 0.029** *** 0.116*** 0.038*** *** 0.116*** ** Nat Banks [0.006] [0.014] [0.001] [0.001] [0.004] [0.006] [0.001] [0.001] [0.007] [0.011] Fixed Effect State Bank State Bank State Bank State Bank State Bank Observations 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 R-squared Indiv. Deposits/Assets Ln(Circulation) Circulation/Assets Ln(Due to Banks) Due to Banks/Assets Clearinghouse *** ** ** *** 0.014*** 0.010*** in City [0.005] [0.006] [0.034] [0.047] [0.002] [0.003] [0.129] [0.154] [0.003] [0.004] Ln(Bank 0.016*** 0.032*** 0.626*** 0.921*** *** *** *** Age) [0.001] [0.002] [0.020] [0.049] [0.001] [0.001] [0.029] [0.070] [0.001] [0.001] Number of *** *** 0.138*** 0.043*** *** 0.004*** 0.552*** 0.102** 0.006*** 0.002** Nat Banks [0.001] [0.002] [0.007] [0.015] [0.001] [0.001] [0.023] [0.049] [0.001] [0.001] Fixed Effect State Bank State Bank State Bank State Bank State Bank Observations 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 23,565 R-squared Notes: Table presents the results of an OLS regression. The dependent variable is described in the column heading. Each observation is a bank every five years, and time fixed effects are included. Banks in counties that changed their boundaries more than 5% between 1870 and 1910 are dropped. County controls for ln(population), ln(farm value per person) and fraction living in area with more than 25,000 are included in all specifications. Dollar values are deflated to 1870 using Officer (2008). Robust standard are provided in brackets. * denotes significance at 10%; ** at 5% level and *** at 1% level.

26 effect by including the balance sheet items discussed in the previous section 23 and the second effect using the clearinghouse dummy. Because clearinghouses only issued loan certificates in a handful of panic periods, the clearinghouse dummy will not clearly capture the importance of emergency liquidity, and I therefore isolate the clearinghouses' importance as a lender of last resort later in this section. The model is:,,,,,, 4 where the variables retain their previous definitions. In this case, I only include state fixed effects as I am interested in both the cross-sectional variation and time series variation. To determine whether clearinghouses in large financial centers (e.g. New York City, Chicago, etc.) were better able to mitigate shocks, I also estimate the regressions when separating, into three categories: (1) Central Reserve City Clearinghouses, (2) Reserve City Clearinghouses, and (3) Other Clearinghouses. 24 The first half of Table 5 presents the results when estimated for the entire set of national banks, whereas the second half presents the results when only including those banks that existed before a clearinghouse entered. This approach attempts to understand if clearinghouses attracted risky banks wanting to take advantage of the public signal. This would still be part of the effect of clearinghouses, but it is helpful to know whether it was the only reason for the instability. Clearinghouses increased a bank's probability of insolvency by 0.3 to 0.4 percentage points. The effect might seem small, but given that only about 5 percent of national banks 23 I drop Circulation, Due from Banks, and Due to Banks to avoid multicollinearity. 24 Similar to the separation of cities, I have estimated regressions where the clearinghouse dummy was split by whether the clearinghouse entered immediately after the Panics of 1893 and 1907 (defined either as 2 or 3 years) or at another time to determine whether they biased the results in any particular way. The positive coefficient on clearinghouses is slightly larger for locations that entered around panics (around 0.005), but the coefficient on other clearinghouses retains a positive and significant coefficient (around 0.003). Given that most early clearinghouses were created in reserve cities, the effects are not too surprising and thus are not reported.