The Panic of 1907: JP Morgan, Trust Companies, and the Impact of the Financial Crisis,

The Panic of 1907:  JP Morgan, Trust Companies, and the Impact of the Financial Crisis,
by
Carola Frydman, Boston University and NBER,
Eric Hilt, Wellesley College and NBER,
Lily Y. Zhou, Federal Reserve Bank of New York,

Abstract:  The outbreak of the Panic of 1907 occurred following a series of scandalous revelations about the investments of some prominent New York financiers, which triggered widespread runs on trust companies throughout New York City.  The connections between the trust companies that came under severe strain during the crisis, and their client firms, may have transmitted the financial crisis to nonfinancial companies.  Using newly collected data, this paper investigates whether corporations with close ties to trust companies were differentially affected during the panic.   The results indicate that firms connected to trust companies that faced severe runs performed worse in the years following 1907.  The data also suggest that many of the rescue efforts organized by J.P. Morgan may have been motivated by self-interest.

1. Introduction

In 1907 the United States experienced one of its most severe financial crises prior to the Great Depression.  A panic was triggered by a series of bank runs in New York, and quickly spread throughout the financial system.  Over the following year, real GNP declined by 11 percent, industrial production contracted by 16 percent, and the unemployment rate almost doubled (Balke and Gordon, 1986; Davis 2004; Romer, 1983). Although the causes of the Panic of 1907 have been the subject of considerable research, the micro-level consequences of the panic have never been analyzed. In particular, little is known about the channels through which the contraction of financial intermediation may have been transmitted to the real economy, or why particular firms or sectors were differentially affected.  Given the extensive debates on the consequences of financial crises, much of which has focused on the Great Depression, this gap in the literature is significant.
Research on the Panic of 1907 has also taken on renewed importance because of its many parallels to the financial crisis of 2007-08. The Panic of 1907 originated with runs on a type of financial intermediary that was mostly outside the payments system, trust companies.  Similar to modern “shadow” banks, trust companies grew rapidly and became important financial intermediaries in the years prior to the crisis.1

  Less regulated than commercial banks, trust companies were highly levered, held low cash reserve balances, and issued uninsured liabilities. In addition, they did not have direct access to a lender of last resort because they did not belong to the private clearinghouse association that facilitated partial co-insurance of commercial banks at that time. There are of course many important differences between these two crises. Most significantly the Fed, which was created in 1913, injected an enormous amount of liquidity into financial markets in the recent crisis, whereas the Panic of 1907 was at its core a liquidity crisis, resolved only through
a halting series of privately organized rescues and suspensions.  In addition, the solvency of many modern financial intermediaries was threatened by the Panic of 2007-08, whereas all of New York’s trust companies were revealed to be solvent in the 1907 crisis. These differences, however, highlight the importance of learning from historical events, in order to understand how markets function within different institutional contexts.

This paper analyzes the consequences of the Panic of 1907 at the firm level, by studying the effect of relationships with the New York trust companies that came under strain during the crisis on the outcomes of non-financial firms.  The paucity of extant research on the impact of the panic is due largely to the lack of data on individual firms or bank lending patterns. An important contribution of this paper is to construct a firm-level panel dataset with detailed financial information on all NYSEtraded industrials and railroads for the years 1900-1911, and establish their connections to major financial institutions. One of the unique characteristics of bank-firm relationships in the early twentieth century was that banks and trust companies would often place their own directors on their clients’ boards.  By collecting a comprehensive dataset of directors of trust companies, as well as directors of NYSE-traded non-financial corporations, we can identify ties between trust companies and their clients through board interlocks. Using this measure of connections, we investigate whether ties to the trust companies that were most severely affected by the crisis, in the sense that they lost the most deposits, had negative consequences on the performance of non-financial firms. We posit two main channels through which a connection to a trust company that came under acute pressure may have had negative consequences for non-financial firms.  First, ties through the board of directors may have reflected a lending or underwriting relationship, or the provision of other financial services.  In this case, non-financial firms may have experienced a negative shock to the supply of external financing or other financial services.2

  A second possibility is that a relationship with a troubled financial institution may have made other lenders, suppliers, or customers of the firm uneasy about the quality of the firm’s own assets or operations.  This
mechanism may have been important for the Panic of 1907 since the runs on trust companies were partly triggered by associations with individuals involved in a financial scandal.  Regardless of the channel, the negative shock to trust companies may have been transmitted to non-financial firms because of the financial frictions they most likely faced.  In an environment with relatively little financial disclosure and many new industries and enterprises emerging, asymmetries of information were likely significant, and building new relationships with alternative financial institutions would have taken time.  Thus, both mechanisms suggest that the effects should have been worse for smaller and less “established” firms, with assets whose value was more difficult to ascertain for use as collateral.

Our empirical analysis proceeds in three steps.  First, we establish that much of the variation in deposit losses among the New York trust companies at the center of the panic was due to their associations with a handful of men involved in a financial scandal.  Since this scandal did not impact any of the trust companies directly, but instead raised fears among households that their deposits may have been threatened, this characteristic of the panic helps rule out the possibility of “reverse causation”—that concerns regarding the non-financial client firms of the trusts led to runs.  Second, we present an event study of the stock market’s reaction to the onset of the runs, and show that the non-financial companies with ties to at least one of the trust companies most severely affected were discounted more heavily (by about 6.5 percent relative to other firms).  Thus, investors already perceived these connections to negatively affect non-financial firms when the runs started.  Finally, we analyze whether shocks to trust companies had a differential effect on the performance of firms
in the years following the panic.  The results indicate that the firms’ profitability and dividends each fell by an amount equivalent to around 10 percent of a standard deviation.  Moreover, the average interest rates paid by these firms, measured by their interest expense as a fraction of outstanding debt, rose substantially. Consistent with the notion that credit intermediation suffered following the panic, these effects were largest for smaller firms and for industrials, whose collateral was more difficult to value than that of railroads.

A potential source of concern is that our findings may reflect the selection of particular types of firms into relationships with trust companies.  The empirical framework includes firm fixed effects, and therefore controls for time-invariant unobserved characteristics such as firm ‘quality.’ However, selection would remain a problem if the differentially affected trusts were represented on the boards of firms most vulnerable to a shock or recession.  In order to address this possibility, we analyze the performance of firms with ties to the trust companies that came under severe strain in 1907 during the recession and financial panic of 1903-04.  We find that these firms did not experience worse outcomes during that earlier crisis, which is supporting evidence that our findings are not the result of a selection effect.

Our paper also sheds light on the private lending arrangements organized by J.P. Morgan that eventually halted the runs on trust companies.  Morgan decided against providing an emergency loan to the Knickerbocker Trust, the first trust company to face a deposit run.  On the day after Knickerbocker was forced to close its doors, Morgan began to arrange emergency loans to a similar institution experiencing a run, the Trust Company of America.  Although Morgan was hailed as the savior of the financial system, these particular decisions may have been motivated by self-interest.  Our board-interlock data reveal that The Trust Company of America had ties to many clients of J.P. Morgan & Company, whereas the Knickerbocker Trust did not.

The results of this paper contribute to the growing literature on the channels through which financial crises impact the real economy.  Following the work of Bernanke (1983), recent scholarship has emphasized the consequences of the breakdown of financial intermediation during financial crises as an important transmission mechanism independent of the monetary channel emphasized by Friedman and Schwartz (1963).  Recent contributions to this literature, in the context of the Great Depression, include Calomiris and Mason (1993), Ziebarth (2012) and Mladjan (2012) and in the context of more recent crises include, Kashyap, Lamont and Stein (1994), Khwaja and Mian (2008), Schnabl (2011) and Amiti and Weinstein (2009). Our findings are also closely related to Fernando, May, and Megginson (2012), who document a negative stock market reaction to the investment banking clients of Lehman Brothers when that firm went bankrupt in 2008. We also contribute to the growing literature on the Panic of 1907.  The causes and
macroeconomic context of this crisis have been the focus of a substantial body of research in the years immediately following the crisis (Sprague, 1910; Barnett, 1910) and more recently (Moen and
Tallman, 1992, 2000; Odell and Weidenmier, 2004; Hansen 2011; and Rogers and Wilson, 2011).3

This paper extends this literature by analyzing the microeconomic impact of the crisis, and the consequences of the disruption of the financial system for the real economy.  Finally, some of our findings relate to studies of the role of trust in financial markets (Guiso Sapienza and Zingales, 2008) and, in particular, of the effects of impaired reputations of corporate directors.  This literature mostly focuses on the consequences of a negative reputational shock on directors’ future careers (Agrawal, Jaffe and Karpoff, 1999; Fich and Shivdasani, 2007).  In contrast, our results may indicate that a firm may suffer losses when its directors are perceived to be associated with a scandal not directly connected to the firm.

2. Historical Background

The Panic of 1907 occurred following a series of economic shocks, which precipitated the onset of a recession.4 The San Francisco earthquake and fire of 1906 had had a profound monetary and financial impact, both domestically and internationally (Odell and Weidenmeir, 2004). Gold flowed into the United States as foreign insurers paid claims on their San Francisco policies; New York financial institutions also faced reduced gold reserves resulting from their own transfers to San Francisco.  In response, the Bank of England, followed by the German and French central banks, raised its discount rates in order to reverse the flow of gold. The Bank of England also acted to halt acceptances of American “finance bills,” which were used to finance gold imports into the United States.  This policy resulted in a significant fall in American securities markets, as the collateral for those bills was sold, and led to significant gold outflows from the United States (Sprague, 1910, p. 241).  A relatively weak cotton harvest in 1907 resulted in low export revenues, further aggravating the stress on the financial system (Hanes and Rhode, 2011).  The New York money market thus entered the fall of 1907 low on gold reserves and vulnerable to shocks.

At that time, New York’s banking system had also experienced an important structural change, in the form of the rapid growth of trust companies.  In the ten years ending in 1907, trust company assets in New York State had grown 244 percent (from $396.7 million to $1.364 billion) in comparison to a 97 percent growth (from $915.2 million to $1.8 billion) in the assets of national banks (Barnett, 1910, p. 235).   The impressive growth of these institutions can be explained by the advantages of the trust form. Originally created to serve as fiduciaries, trust companies enjoyed broad powers, including the ability to hold corporate equity and debt, and to underwrite and distribute securities (Smith, 1928; Neal 1971).  Although they were not permitted to issue bank notes, they could make loans, and competed with national banks for deposits.8 Incorporated under permissive state laws, trust companies were not subject to the strict regulations of the National
Banking Act, and often specialized in providing financing for corporate investments and acquisitions.  One observer noted that the industry’s profits were “derived largely from the skill of their officers in financing important combinations and aiding in the creation of new enterprises” (Conant 1904, p. 223).    By the onset of the panic, trust companies played major role in banking and financial markets: They provided lending and underwriting services, were major purchasers of securities, and acted as financial agents for corporations.11
Many prominent private bankers, as well as former U.S. Treasury Secretaries, were among the directors of these enterprises, which enhanced their reputations.12

The rapid proliferation of trust companies may have contributed to the vulnerability of the financial system to crises.  Whereas the national banks located in New York City were required to hold reserves equivalent to 25% of their deposits in specie, New York’s trust companies faced no minimum reserve requirement at all until 1906.13 In 1906, a 15% reserve requirement was imposed, but trust companies were required to hold only one third of it in cash.14 The national banks also effectively excluded trust companies from the New York Clearing House Association (NYCHA), a private organization that facilitated clearing and that could provide emergency lending to its members in times of crisis (Gorton, 1985).  Trust companies were permitted to gain access to the NYCHA by clearing through a member bank, but only if they maintained a minimum level of cash reserves, which most found unacceptably high.15 When the panic arose, there was no established mechanism to facilitate cooperation among New York’s trust companies, or to provide loans to a trust company that faced a liquidity problem.16

Onset of the Panic

The events of the Panic of 1907 that had the most severe consequences for financial markets were the widespread runs on trust companies that began in October.  Importantly, these runs were precipitated by events that had no direct connection to any trust company.  Instead, they were triggered by a failed attempt to corner the shares of United Copper Company, a mining concern, which resulted in significant losses for the speculators involved.  Historical accounts suggest that the runs on trust companies were driven by depositors’ fears that these institutions may have suffered losses in the speculation, which were later proven unfounded.17  This characteristic of the crisis is especially important for establishing an effect of the financial panic on the outcomes of non-financial firms, as it suggests that the runs were not related to revelations about the quality of the trust companies’ corporate clients.  In this section, we provide a brief account of the onset of the crisis, and present an econometric analysis of the determinants of the different trust companies’ deposit
losses during the panic.

Mining entrepreneur Augustus Heinze, along with speculators E. R. Thomas and Charles W. Morse, were at the center of the failed speculation.  These individuals had gained control of a series of small banks and used some portion of their resources to finance their ventures.18  These banks suffered losses when the attempt to corner the shares of United Copper, which was undertaken to engineer a “bear squeeze,” failed spectacularly.19 On October 16, a run began on the Mercantile National Bank, which was under the control of Heinze, Morse and Thomas, who appealed to the NYCHA for aid.  The NYCHA provided a loan to Mercantile, and publicly pledged to support the other member banks connected to those men as well.  As a condition for this aid, the NYCHA required the resignation of the entire board of directors of Mercantile, and demanded that Morse, Thomas and Heinze resign from all other clearing banks where they held directorships.20  The very public support from the NYCHA and the change in management ended the run on Mercantile, although it was liquidated the following January.  It is possible that the expulsions of these individuals from New York’s banking industry contributed to the perception that they had embezzled funds or committed fraud.

No trust company was directly involved in the failed United Copper corner.  However, the well-known financier Charles T. Barney, president of Knickerbocker Trust and director of Trust Company of America, two of the largest trust companies in the city, was known to have been involved in earlier business dealings with Morse, and held a board seat with the National Bank of North America, controlled by Morse. Moreover, Morse, Thomas, and Augustus’ brother Arthur Heinz held directorships with other trust companies.  The business connections among these individuals, and the board seats they held, were widely reported in the press.21 The losses and runs suffered by the Mercantile National Bank and other banks controlled by Morse, Thomas and the Heinzes likely raised concerns among depositors about whether these men had also endangered the solvency of the trust companies with which they or their associates were affiliated.

The connections between the men at the center of the United Copper speculation and various financial institutions are illustrated in Figure 1. Morse, Thomas, Barney and the Heinzes held seats on the boards of five trust companies; we identify these institutions as having a direct connection to these men.  However, those five trust companies were, in turn, closely associated with three other trust companies, because they had at least two directors in common with those three firms.22  These three trust companies are therefore identified as having an indirect connection to Morse, Thomas, Barney and the Heinzes.  The degree to which the different trust companies were associated with those men may have influenced the intensity of the runs they faced during the panic, and we formally test this hypothesis below.

The runs on trust companies began silently around October 16, when Knickerbocker Trust started to face heavy withdrawals. Knickerbocker was one of the few trust companies that chose to maintain sufficient reserves to gain access to the NYCHA through a member of the clearinghouse, National Bank of Commerce.  When Knickerbocker depositors began to withdraw their funds by depositing checks on their accounts in other banks, the National Bank of Commerce was responsible for those checks.  Facing a debit balance at the NYCHA of $7 million and the prospect of even larger debits, on October 21 the National Bank of Commerce announced that it would no longer act as Knickerbocker’s clearing agent.23  On that same day, Knickerbocker Trust announced that it had dismissed Charles T. Barney from the office of its Presidency, because of his “personal position in the directorate of certain institutions recently under criticism,” and “in particular because of his connection with Mr. Morse.”24

These events came as a shock to Knickerbocker’s depositors.  The end of the clearing relationship meant that other banks would no longer cash the trust company’s checks and, more importantly, that the NYCHA would not aid Knickerbocker if the firm encountered liquidity problems.25 The dismissal of Barney, even though it was accompanied by assurances that the firm was in sound condition, may have created the impression that Barney had done something improper or used the funds of Knickerbocker to help finance the speculative schemes of Morse.  A severe run on the Knickerbocker ensued, and the firm could not withstand the heavy withdrawals without receiving external assistance.  It never did, and on October 22, Knickerbocker was forced to close its doors.

Panic quickly spread as “wild rumors circulated” regarding the financial condition of other trust companies.26  These rumors often focused on possible connections between trust companies and the men at the center of the failed corner scheme; the chairman of the Trust Company of America went so far as to issue a public statement that his “company had no business relations, directly or indirectly, with Charles W. Morse, as the rumors had intimated.”27  Within a few days, all of the trust companies where Thomas, the Henizes, or Morse held directorships announced their resignations.28 By October 23, runs had spread to the Trust Company of America and Lincoln Trust, and several other trust companies also faced heavy deposit withdrawals.29  To address the fears of depositors, some trust companies stated that they had no connection to the men associated with the scandal in their advertising.30 All trust companies began to call in loans and liquidate assets to build up their cash reserves.

The total losses of deposits of the 38 trust companies in New York City between August 22 and December 19 of 1907 are depicted in Figure 2.31  All of the trusts either directly or indirectly associated with Morse, the Heinzes, Thomas, or Barney lost substantial amounts of their deposits, although several others with no apparent direct or indirect connection to these men did as well.  It is worth noting that the size of the different trusts prior to the crisis, measured as their total assets as of June 1907, was generally uncorrelated with the percentage decline in deposits.

Table 1 analyzes the determinants of the percentage change in each New York trust company’s deposits between August and December of 1907.  In column (1), we regress the change in deposits on indicator variables for whether the trusts were directly or indirectly connected to Morse, the Heinzes, Thomas or Barney, as defined in Figure 1.  Each variable has large negative effects and they jointly account for about 40 percent of the variation in the dependent variable. Relative to a trust company with no connection to the men at the center of the scandal, one with an indirect connection experienced a loss in deposits that was about 22 percent larger.  The loss for a trust company with a direct connection was even greater, as their deposits declined almost by 34 percent more than for trusts not associated to the scandal.

These results suggest that associations with the scandal triggered the runs on the trusts companies, but it is also possible that they may have been driven by differences in the financial solidity of the trust companies at the time of the crisis.  In column (2), we add several balance sheet ratios calculated from the trust companies’ financial statements of June 1907—measures of net worth, cash reserves relative to deposits, the percentage of their assets invested in securities, and their overall size—as well as the log of the age of the trust.32  The estimated correlations with these variables generally have the expected signs, with the firms’ net worth and cash holdings having particularly large and positive magnitudes, suggesting that more solvent trusts faced fewer withdrawals of deposits.  Controlling for these characteristics in the regression, however, does not diminish the size of the estimated effect of the two indicator variables for association with the tainted
bankers.  The available balance sheet information does not capture differences in the depositor clienteles of the trust companies.  Hansen (2011) notes that New York’s trust companies were divided between those located in the vicinity of Wall Street, which generally received large deposits from corporations and institutions, and those located in uptown Manhattan, which more aggressively solicited deposits from individuals. He argues that the uptown firms experienced greater deposit losses in the panic because small individual depositors were more likely to participate in runs.  In order to address this possibility, in column (3) we include an indicator variable for whether the trust company had an uptown headquarters.33 The results indicate that the uptown firms did indeed lose a greater proportion of their deposits, largely confirming Hansen’s argument.  But importantly, controlling for an uptown location does not substantially diminish the size of the estimated effect of the indicator variables for the strength of the connection with the men associated with the United Copper corner.  Interestingly, the uptown variable does diminish the estimated effect of some of the balance sheet measures, indicating that the effects estimated in column (2) may have resulted partly from correlation between the financial solidity of trust companies and their type of depositor
clientele. In sum, our results indicate that the deposit losses can be regarded in large measure as a response to an association with men involved in a scandal, and thus minimize concerns of reverse causality for the empirical analysis that follows.

Rescues organized by J.P. Morgan

In response to the growing crisis, on October 19 J.P. Morgan began to organize teams of bankers he trusted, and charged them with determining the solvency of the financial institutions that came under pressure.
34  The most powerful and best-connected man in American financial markets, Morgan’s own interests and influence were quite far reaching.35  He had previously organized interventions that benefited markets generally, for example by providing emergency lending to the U.S. Treasury and intervening in foreign exchange markets in 1895 to keep the Dollar on the gold standard.  During the Panic of 1907, Morgan coordinated a series of rescues of trust companies, securities dealers, and the City of New York that were instrumental in resolving the financial crisis.

The first institution to appeal to Morgan for aid was Knickerbocker Trust.  On Monday October 21, Morgan committed to provide aid the following day only if it was determined that the institution was solvent.37 On October 22, with panicked depositors forming long lines outside of its branches, Knickerbocker paid out all of its $8 million in cash.  Morgan’s men, who examined Knickerbocker’s books throughout the morning, said they were unable to determine whether the trust was in fact solvent.  Therefore no aid was provided, and at 12:30 PM Knickerbocker had no choice but to close its doors. The receivers appointed to take over Knickerbocker later determined that the institution was in fact solvent, and its depositors received the amounts owed them in full, although over a period of several months.38 On the afternoon of October 23, Morgan organized emergency loans to the Trust Company of America, after a series of dramatic scenes in which its securities were rushed to Morgan’s offices and evaluated as collateral for loans from the large commercial banks closely associated with him.  These loans, as well as others that Morgan organized the following week, enabled this institution to stay open.  On the night of Sunday November 3, Morgan hosted a meeting of nearly all the city’s trust company presidents in his library, famously locking them inside until they collectively pledged $25 million for the aid of the Trust Company of America and other failing trust companies.

The run on the Trust Company of America was one of the most severe up to that point in American history: the firm paid out more than $34 million in deposits in just a few weeks.  But it never closed, and thanks to the various rescues organized by J.P. Morgan and his associates, the only New York City trust company to fail was Knickerbocker.39  Morgan’s ability to organize these rescues was a consequence of his firm’s resources and credibility, which enabled him to stand behind the emergency loans provided by institutions like National City Bank to the Trust Company of America based solely on his men’s assessment of their collateral.  But it also resulted from his power and influence within financial markets.  In times of panic, it may be contrary to a financial institution’s narrow self-interest to extend a loan to a failing competitor, even if it is in that institution’s interest for the panic to be halted.  Morgan’s power enabled him to “dragoon” other
financial institutions into taking actions that were privately costly, but beneficial for the markets as a whole (Sprague, 1910).40

Morgan cannot be regarded as an entirely disinterested actor in these events.  Among the many rescues he organized was a rescue of the investment bank Moore & Schley, which had used a large block of stock in the Tennessee Coal & Iron Railway as collateral for loans which it suddenly needed to repay.  Morgan helped arrange for U.S. Steel, a firm he had helped create and a competitor of Tennessee Coal & Iron, to purchase that block of its stock. Morgan’s associates even received assurances from President Roosevelt that the transaction would not be held in violation of antitrust laws.  This transaction averted a crisis on the NYSE, but it also benefited U.S. Steel and, therefore, J.P. Morgan.

Morgan’s decision to allow Knickerbocker Trust to fail, while working assiduously to save the Trust Company of America, may also have been motivated by self-interest.41  Morgan himself was a director of the National Bank of Commerce, the institution that stopped clearing for Knickerbocker, so he could have intervened on behalf of Knickerbocker.42  However, our board interlock data suggest that Knickerbocker had few ties to clients of J.P. Morgan, whereas the directors of Trust Company of America served on the boards of several railroads and industrial firms closely associated with Morgan.43  Thus, it is likely that J.P. Morgan & Company underwrote many of the securities held by the Trust Company of America. Morgan’s partners may have been concerned about the consequences of liquidating the trust company’s holdings of those securities, or any other negative consequences that may have resulted from the association between their firms and a failed institution, thereby making them more favorably inclined towards the valuation of those
securities as collateral for loans. Indeed, Morgan’s associates publicly announced they would provide support for the Trust Company of America well before they were able to determine whether it was solvent, whereas aid to Knickerbocker was made contingent on establishing that trust’s solvency.44 On the other hand, Morgan may simply have miscalculated the consequences of permitting Knickerbocker to fail, and acted to save the Trust Company of America the next day in response to deteriorating conditions in banking markets.

Consequences of the Panic

On October 26, in the face of heavy withdrawals from out-of-town banks, the New York Clearing House issued “clearing house loan certificates” in order to provide liquidity to its members, and New York’s banks soon after suspended the convertibility of their deposits into currency.  The banks in the rest of the country soon followed, with some receiving legal sanction of their state governments.  By early November, the trust companies in New York apparently began making payments via certified checks payable at the NYCHA, rather than in cash.45  Full convertibility of deposits by the nation’s banks was not restored until January 1908.  The suspension likely made important transactions more difficult (see James, McAndrews and Weiman, 2011).  On the other hand, the suspension likely halted the spread of the banking panic and averted a total collapse of the banking system, as in 1930-33 (Friedman and Schwartz, 1968).

The contraction of lending that occurred during the panic in New York was heavily concentrated within trust companies.  Prior to the panic, the aggregate volume of New York trust company loans was similar to that of New York’s national banks. However, during the panic total loans at trust companies contracted by $247.6 million, or 37 percent, between August and December (Moen & Tallman, 1992).  During the same period, the loans of national banks in New York fell by only 2 percent.  Contemporary observers noted the consequences:  “It is obvious that every trust company is protecting itself to the full extent of its powers, and the small borrowers, however solvent, necessarily suffer at such a time.”46
 
The panic occurred at a time when credit markets were already under stress, and produced a significant overall contraction in liquidity.  Yet the crisis was most severe for New York’s trust companies, and for a handful of those in particular.  To determine the effect of the crisis on nonfinancial firms, we use an empirical strategy that exploits the variation in the deposit losses among trust companies, which was largely driven by factors unrelated to the performance of the trusts’ client firms.

3. Data

Board data and ties to financial firms

We identify the connections between a trust company and a non-financial firm by the presence of a director of the trust on the board of the non-financial firm.  To observe these relationships, we collected the names of all directors and managers of all NYSE-listed industrials and railroads as reported in Moody’s Manuals over several years around the panic. To identify directors of trusts, we obtained lists of directors of commercial banks and trust companies from the Rand McNally Bankers’ Directory.  Cross-referencing the names of bankers with those of corporate directors enables us to identify the presence of trust company directors on boards of non-financial firms. We match the names of corporate directors to those of bankers based on last name, first name, second initial, and suffix. A valid concern is that matching on names may lead to erroneous matches.  This procedure may overestimate the degree of interlocking across institutions if, for example, two different people with the same name held a directorship in an industrial company and a trust company.  However, we have implemented this same procedure for subsequent years when sources such as the Pujo
Committee Report, which identified the interlocks of directors of a substantial number of banks and non-financial companies around 1912, is available. Our matching procedure produces a nearly identical outcome to the Pujo report.47
 
Table 2 displays summary statistics for the data on trust company connections to firm boards in 1907.  From our 77 NYSE-traded railroads, 84% had at least one trust company representative on its board.  From our sample of 109 industrial companies, 70% had a trust company representative among its directors. The prevalence of trust company directors among the directors of non-financials may to some extent reflect the desire of trusts to form alliances with important firms—that is, for the trust to invite an industrialist or a railroad manager to serve on its own board.  But a substantial number of these cases were more likely the trust company directors serving on the non-financial’s board. Moreover, the extraordinarily high rate at which these interlocks occurred indicates that trust companies were indeed very prominent prior to the panic.

 An important feature of our data is that it allows us to identify the connections between specific trust companies and non-financials at the firm level through board linkages.  Figure 3 illustrates the connections between the five trust companies identified as most prominently connected to the scandal of Heinze, Morse, and Barney’s failed cornering scheme (those directly connected to the scandal in Figure 1), and NYSE-traded firms.  Directors of these trust companies held 39 board seats with NYSE-traded firms, including many prominent railroads and industrials.

Stock Price Data

The standard datasets of stock prices, such as CRSP, do not cover the earliest decades of the twentieth century.   For all firms in our sample, we use the New York Times to collect the closing prices of common shares traded on the NYSE at the end of each week from the end of August 1907 to December 1907.

Accounting Data

No readily available dataset of accounting information exists for early twentieth century