3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM CAPITAL, CONTAGION, AND FINANCIAL CRISES: WHAT STOPS A RUN FROM SPREADING? Nicholas K. Tabor* and Jeffery Y. Zhang** After the 2008-09 financial crisis, policymakers around the world focused on enacting improvements that would make the emergence of a financial crisis less likely (ex ante) and recovery from one more rapid (ex post). This Article identifies a gap in both the academic literature and the current financial regula- tory framework in exploring how to limit the damage—to other firms, and to the financial system—when a crisis is ongoing. Policymakers cannot predict the origins of every future crisis, just as firefighters cannot predict the origins of every future fire. Once one begins, how can they keep the damage from spreading? * Nicholas K. Tabor, JD/MBA, is a member of the Policy Planning Section in the Division of Supervision and Regulation of the Board of Gov- ernors of the Federal Reserve System. ** Jeffery Y. Zhang, JD/PhD, is a member of the Banking Regulation and Policy Group in the Legal Division of the Board of Governors of the Federal Re-serve System. They previously served in the White House, respectively, on the staffs of the National Economic Council and Council of Economic Advisors. The authors wish to thank their many colleagues who provided thoughtful comments, and participants in the University of Pennsylvania Law School Law and Economics Seminar, as well as Howell Jackson, Hal Scott, Andrew Metrick, David Zaring, Emilie Feldman, Michael Sinkinson, Jeremy Kress, Emil Pitkin, Natasha Sarin, David Skeel, Gary Gorton, Morgan Ricks, Shameel Ahmad, Susannah Barton Tobin, Steve Schaus, Jennifer Reich, Jo- seph Bretschneider, Alexander McCammon, Carl West-phal, Rebecca Green, and Catherine Shaw. The views expressed in this Article are the authors’ alone and do not necessarily reflect the views of the Federal Reserve Board or the United States government. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 576 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 The academic theory on financial crisis “firefighting” di- vides into two main camps. The “capital view” claims that runs on financial institutions are fundamentally rational, and that investors care mainly about solvency. Under this view, the best way to fight runs is to raise capital requirements ahead of time, to multiples of current levels. The “contagion view” claims in- stead that the lack of liquid assets both defines and causes bank runs; an institution’s access to cash (and instruments like it) determines whether and when investors will withdraw funding. Under this view, the best way to fight runs is for gov- ernments to lend banks money—freely, at high rates, and against good collateral—and to promise to do so well before a crisis starts. In this Article—the first to directly address this question empirically—we show that neither view fits the most cata- strophic financial shock of the last ninety years: the 2008 Leh- man Brothers bankruptcy. In some cases, banks with more cap- ital and liquidity were actually more exposed, not less, to the market panic following Lehman’s collapse. By contrast, we show that simple market correlation was a powerful predictor of exposure to the Lehman run. We also show that market val- uations of large banks are more highly correlated today than they were in September 2008, creating a potential unaddressed conduit for an unexpected shock to metastasize into a conta- gious run. I. Introduction ...................................................................... 577 II. Primer on Post-Crisis Bank Regulations ....................... 585 A. Capital, Liquidity, and Runs: How to Make (or Break) a Bank ...................................................... 585 B. The Post-Lehman Reforms .................................... 589 C. Prior Literature ..................................................... 597 III. Analysis ......................................................................... 605 A. Overview of Methodology and Research Design ... 605 B. Summary of Results .............................................. 614 1. Simple Panel Fixed-Effects Regression Results ............................................................ 614 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 577 2. Compound Panel Fixed-Effects Regression Results ............................................................ 617 3. Overall Results ................................................ 618 IV. Policy Implications ........................................................ 619 A. Bank Runs Aren’t (Always) About Cash ............... 619 B. Leverage—But Not Regulatory Capital—Can Predict How a Run Spreads ................................ 622 C. Contagion Theory: Capital, Complexity, and Information Scarcity ............................................ 625 D. Storms, Fires, and Correlation Channels ............. 626 V. Implications for Regulatory Design ................................ 629 A. Supervisory Stress Tests ....................................... 629 B. “Monoculture Risk” in the Financial Sector ......... 633 C. Revisiting Post-Crisis International Capital Standards ............................................................ 634 D. Preparing Disclosures in Advance ........................ 635 E. Herd Behavior and Market Structure .................. 636 VI. Conclusion ...................................................................... 637 VII. Appendix....................................................................... 639 A: Definitions of Key Regulatory Terms and Ratios . 639 B: Descriptive Statistics ............................................ 641 C: Basel III Common Equity Tier 1 (“CET1”) Capital Proxies and Robustness Checks .......................... 642 D: In-Sample Institutions .......................................... 645 E: Simple Panel Fixed-Effects Regression Results ... 647 F: Multiple Panel Fixed-Effects Regression Results . 649 G: Simple Regression of Selected Explanatory Variables on Cumulative Changes in Share Price ............................................................................. 650 H: Criteria for Additional Tier 1 Capital (Basel III) . 653 I. INTRODUCTION A decade removed from the peak of the 2008-09 financial crisis and after hundreds of published articles, academics and policymakers are still working to understand systemic risk in the financial sector. This long progression is as expected; “it 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 578 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 was not until 30 years after the Great Depression that Milton Friedman and Anna Schwartz published A Monetary History of the United States in 1963, with its now canonical critique of Federal Reserve Board policy in the 1930s . . . . [O]n the time scale needed to fully absorb the significance of major financial disasters, we are in the early days and should not expect to reach immediate consensus on either diagnoses or progno- ses.”1 What all practitioners and scholars agree is that a crisis begins with an unexpected shock. The unexpected shock that arrived in the early hours of September 15, 2008, was one of the most consequential in American financial history. Lehman Brothers Holdings, Inc. (“Lehman”) had filed for bankruptcy the night before, launch- ing an insolvency process that would affect hundreds of bil- lions of dollars in financial assets.2 The previous Friday—the most recent time U.S. markets were open for trading—com- mon wisdom held that another investment bank would pur- chase Lehman, perhaps with public assistance, as had been the case when JPMorgan Chase & Co. purchased Bear Stearns six months earlier.3 Instead, the day’s trading began with news that Lehman had gone under, and that the ac- counts of Lehman’s British and Japanese brokerage opera- tions had been frozen.4 1 Howell F. Jackson, Introduction: Thinking Hard About Systemic Risk, in SYSTEMIC RISK IN THE FINANCIAL SECTOR: TEN YEARS AFTER THE GREAT CRASH 1, 2 (Douglas W. Arnert et al. eds., 2019). 2 The voluntary insolvency petition for Lehman’s U.S. holding com- pany was filed at 1:45AM on Monday, September 15th. See Matt Egan, Leh- man Brothers: When the Financial Crisis Spun Out of Control, CNN BUS. (Sept. 14, 2018), https://www.cnn.com/2018/09/30/investing/lehman-broth- ers-2008-crisis/index.html [https://perma.cc/LG48-5K8X]. 3 See Alexandra Twin, Stocks Struggle on Bank Woes, CNN MONEY (Sept. 12, 2008), https://money.cnn.com/2008/09/12/markets/mar- kets_newyork/index.htm [https://perma.cc/Z7N2-R2WT]. 4 See ANDREW ROSS SORKIN, TOO BIG TO FAIL: THE INSIDE STORY OF HOW WALL STREET AND WASHINGTON FOUGHT TO SAVE THE FINANCIAL SYS- TEM—AND THEMSELVES 536 (2010). See also Jennifer Hughes, Winding Up Lehman Brothers, FIN. TIMES (Nov. 7, 2008), 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 579 Investors were, quite understandably, unsure how to re- act.5 However, their behavior was not indiscriminate. In the previous months, through the failure of Bear Stearns and the nationalization of Fannie Mae and Freddie Mac, the equity and debt markets had a fairly uniform view of large financial institutions; their credit default swap (“CDS”) spreads, the price of insuring against a default on their debt, remained tightly clustered, and their share prices remained highly cor- related. Lehman’s bankruptcy shattered that uniformity. By the end of the day, share prices and CDS spreads on large fi- nancial institutions had splintered, with some on a vastly more adverse trajectory than others.6 Fears about Lehman had spread, but they had not spread evenly. This Article seeks to identify gaps in the design of post- crisis regulatory reforms, by clarifying why some financial in- stitutions experienced greater stress than others in the imme- diate aftermath of the Lehman bankruptcy. In doing so, we fill a gap in the academic literature and policy discussion on how to limit the damage to the financial system after a panic starts, but before it spreads. A wide range of work has exam- ined the effect of post-crisis reforms on preventing a financial crisis ex ante and hastening recovery from one ex post. How- ever, vanishingly little scholarship examines the effectiveness of capital and liquidity reforms on a third goal: preventing an ongoing run at one bank from spreading to others.7 In an https://www.ft.com/content/e4223c20-aad1-11dd-897c-000077b07658 [https://perma.cc/2XYF-ATH4]. 5 See, e.g., Alexandra Twin, Stocks Get Pummeled: Wall Street Sees Worst Day in 7 Years, with Dow Down 504 Points, as Financials Implode, CNN MONEY (Sept. 21, 2008), https://money.cnn.com/2008/09/15/mar- kets/markets_newyork2/ [https://perma.cc/CZK8-UN2G] (“You have to throw out the history books because there's really nothing to compare this to . . . . We’ve never witnessed this before . . . [t]here’s no road map for this.” (internal quotation marks omitted)). 6 See infra Figures 1–2. 7 A preliminary note on vocabulary: a “run,” for purposes of this Arti- cle, refers only to a single institution event and is not defined specifically by the behavior of creditors. For a further discussion of the definition we adopt, see infra note 101 and accompanying text. We use the terms “panic” and 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 580 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 atmosphere of market uncertainty, no policy task could be more important. Figure 1: Share Prices Circa Lehman Bankruptcy8 “crisis,” for which there are no precise consensus definitions, to indicate a contagious run that extends to other financial institutions. Our use of these terms draws on empirical work that finds “panics are systematic.” See Gary Gorton, Banking Panics and Business Cycles, 40 OXFORD ECON. PAPERS 751, 773 (1988). For a broad-based discussion of runs, as well as runs that took place during the financial crisis, see BEN S. BERNANKE, THE FEDERAL RE- SERVE AND THE FINANCIAL CRISIS 64–97 (2013). See also BEN S. BERNANKE ET AL., FIREFIGHTING: THE FINANCIAL CRISIS AND ITS LESSONS 15 (2019) (“A fi- nancial crisis is a bank run writ large, a crisis of confidence throughout the system.”). 8 Equity Price Data, BLOOMBERG LP, https://www.bloomberg.com/pro- fessional/ [https://perma.cc/5C2B-XGQP]. Each line represents one of the fourteen largest U.S. banks in September 2008. The first vertical bar indi- cates September 13–14 (markets closed). The second vertical bar indicates September 15 (date of Lehman filing). 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 9/1 9/3 9/5 9/7 9/9 9/11 9/13 9/15 9/17 9/19 9/21 9/23 9/25 9/27 9/29 N o rm al iz ed S h ar e P ri ce s 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 581 Figure 2: CDS Spreads Circa Lehman Bankruptcy9 The academic literature on this subject, until now, has di- vided into two main schools of thought. The “capital view”— expressed by Daniel Tarullo of Harvard Law School and Anat Admati of the Stanford Graduate School of Business, and by many central bankers around the globe—claims that runs are fundamentally rational, and that investors care mainly about the solvency of a financial institution.10 When stress emerges, the institution’s level of capital—how much of its funding comes from equity and similar instruments—determines whether investors will extend or withdraw funding from that institution. Under this view, the best way to fight runs is to raise capital requirements ahead of time, to multiples of their current levels.11 The “contagion view”—expressed by Hal Scott of Harvard Law School and Gary Gorton of the Yale School of 9 CDS Spread Data, BLOOMBERG LP, https://www.bloomberg.com/pro- fessional/ [https://perma.cc/5C2B-XGQP]. 10 See Anat R. Admati et al., infra note 80, at 43. 11 See Daniel K. Tarullo, Governor, Fed. Reserve Sys., Speech at Princeton University: Departing Thoughts (Apr. 4, 2017), https://www.fed- eralreserve.gov/newsevents/speech/tarullo20170404a.htm [https://perma.cc/U9N3-4KFL]. 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 9/1 9/3 9/5 9/7 9/9 9/11 9/13 9/15 9/17 9/19 9/21 9/23 9/25 9/27 9/29 N o rm al iz ed C D S S p re ad s 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 582 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Management, and by many in the private sector—claims that runs can start for virtually any reason, not just insolvency worries. Instead, the lack of liquid assets both defines and causes a bank run; an institution’s access to liquidity deter- mines whether and when investors will withdraw funding.12 Under this view, the best way to fight runs is for central banks to lend money freely at high rates against good collateral, and to promise to do so well before a crisis starts.13 Our results are surprising and complicate both prevailing views of contagious bank runs: • Contrary to the capital view, institutions with higher levels of regulatory capital experienced more funding stress. One factor seems to explain this puzzling relationship: The higher an institution’s 2008 regulatory capital, the more it relied on the use of debt—a relationship that persists today and, in some cases, is worse than that observed in 2008. • Contrary to the contagion view, balance-sheet li- quidity levels did not explain any variation in run exposure. • By contrast, two simple measures strongly pre- dicted funding stress: a simple leverage ratio, and the correlation between an institutions’ share price and Lehman’s. This suggests that simple measures matter more to investors in an emergency, and that tighter correlations can correspond to a faster- spreading run. Notably, and perhaps worryingly, the share prices of large U.S. financial institutions are more highly correlated with each other today than they were with Lehman a decade ago. 12 See Cardiff Garcia, “Misunderstanding Financial Crises”, a Q&A with Gary Gorton, FIN. TIMES (Oct. 25, 2012), https://ftal- phaville.ft.com/2012/10/25/1223861/misunderstanding-financial-crises-a- qa-with-gary-gorton/ [https://perma.cc/E8GH-C959]. 13 See Hal S. Scott, This Should Be Trump’s Top Priority on Financial Reform, CNBC (Jan. 26, 2017), https://www.cnbc.com/2017/01/26/this- should-be-trumps-top-priority-on-financial-reform-harvard-law-professor- commentary.html [https://perma.cc/5738-E873]. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 583 Our results have important implications for the design of financial regulation, both domestically and internationally, and for the post-crisis capital and liquidity standards that are now subscribed to by over 100 countries.14 First, our results suggest that a central analogy for finan- cial crises is wrong or, at best, incomplete. Discussions of fi- nancial “shocks” often treat contagious runs like a sudden un- predicted storm, which hits an entire neighborhood and spares only the strongest houses from destruction. By con- trast, our findings suggest that a contagious run is more like a fire, which starts inside a single home. Certain factors (e.g., fire-proofing, sprinklers, smoke alarms) can keep the blaze from starting—but once it does, they are irrelevant as to whether it consumes the neighborhood. Nearby homes may burn, or an updraft, flaming debris, or burning embers could carry the flames to houses clear across town. Fighting the fire requires an entirely different set of tools—from firehoses and firebreaks, to evacuation plans and zoning laws. Second, our findings suggest that simpler measures matter more in a financial crisis; that only certain capital measures are associated with the transmission of runs from one institu- tion to another; and that the relationship between liquidity and run exposure is more complex than it might first appear. These implications are consistent with an old strand of the law-and-economics literature—specifically, the efficiency of information about firms, and the transaction costs involved in obtaining such information. These themes should figure prominently in our accounts of financial crises, when distin- guishing fact from rumor is most difficult. Third, capital and liquidity on an institution’s balance sheet play critical roles before and after an idiosyncratic shock; the existing literature is clear on both those points, and nothing in our results qualifies or contradicts it. However, our results suggest that investors treat capital and liquidity very 14 See Roberta Romano, For Diversity in the International Regulation of Financial Institutions: Critiquing and Recalibrating the Basel Architec- ture, 31 YALE J. REG. 1, 3 (2014). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 584 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 differently during a crisis than in normal times, and that dif- ferent regulatory tools are necessary to stop a bank run from spreading. Identifying, monitoring, and addressing the “corre- lation channels” that carry stress between institutions ap- pears to be one such tool. Further research is necessary to identify others, and to gauge the effect of regulatory disclo- sures, trading automation, and risk-weighting reforms on our findings. Finally, our results suggest that the decade since 2008 has seen a convergence among the largest financial institutions. Before the Lehman bankruptcy, the equity returns of several firms in our sample were highly correlated, and the higher the regulatory capital ratios of those firms, the more leveraged they were. Today, many of those returns are even more highly correlated, and the negative relationship between Basel III’s highest-quality risk-based capital ratio (Common Equity Tier 1) and simple leverage (common equity/total assets) is even stronger. Other supervisory measures, such as resolution planning, may have altered the relationships that this Article examines, by making bank activities safer, business models more uniform, runs less likely, and recovery more orderly.15 However, our findings suggest those measures may involve a subtle trade-off—greater safety before a run occurs, but greater vulnerability once one begins. Our Article proceeds as follows. Part II provides a primer of the fundamental concepts involved in our research, de- scribes the regulation of capital, liquidity, and short-term funding instruments before and after the fall of Lehman, and reviews the post-crisis literature on the causes of runs. Part III outlines our methods and research design and summarizes our results; Part IV describes the policy implications of those 15 The penultimate Part of this paper contains a short discussion of the impact of resolution planning; however, we are not aware of any empirical research looking at the specific impact of resolution planning on run behav- ior, which limits our ability to offer a substantive evaluation of its impact. See infra Part IV. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 585 results; and Part V offers a path forward on regulatory design. Part VI concludes. II. PRIMER ON POST-CRISIS BANK REGULATIONS A. Capital, Liquidity, and Runs: How to Make (or Break) a Bank Firms are investment vehicles that accept money from in- vestors and use it to engage in activities that pay a return.16 In exchange, investors typically gain the right to a specific measure of value from a firm. For example, an investor can purchase a right to the residual value of a firm’s assets or al- locations of a firm’s profits.17 Alternatively, an investor can purchase a right to the value of his or her initial investment, plus some kind of interest.18 The first of these obligations usu- ally is called equity; however in the context of a bank, it and a 16 See R.H. Coase, The Nature of the Firm, 4 ECONOMICA 386 (1937). 17 See Fin. Accounting Standards Bd., Equity (Topic 505): Overall (2018) [hereinafter, Fin Accounting Standards Bd., Equity], https://asc.fasb.org/section&trid=2208765 [https://perma.cc/J8TN-R3YA]; Fin. Accounting Standards Bd., Equity (Topic 505): Stock Dividends and Stock Splits (2012), https://asc.fasb.org/section&trid=2208795 [https://perma.cc/UG8D-87DG]. 18 See FIN. ACCOUNTING STANDARDS BD., ACCOUNTING STANDARDS UP- DATE NO. 2016-19: TECHNICAL CORRECTIONS AND IMPROVEMENTS 16 (2016), https://asc.fasb.org/imageRoot/54/108316354.pdf [https://perma.cc/92FB- XM77] (“[A] receivable or payable (collectively referred to as debt) repre- sents a contractual right to receive money or a contractual obligation to pay money on demand or on fixed or determinable dates that is already included as an asset or a liability in the creditor’s or debtor’s balance sheet at the time of the restructuring.”). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 586 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 variety of similar financial instruments are often called capi- tal.19 The latter is usually called debt.20 Equity generally comes with no guaranteed return; if a firm invests poorly, its equity might lose all or nearly all of its worth, and investors holding such equity (called “sharehold- ers”) typically cannot recover their investment from a firm in court. Debt, by contrast, generally comes with a contractually obligated return; even if a firm invests poorly, it retains a duty to repay the investors who hold its debt (called “creditors”). In the event that a firm files for bankruptcy, the claims of se- cured creditors have priority over those of common sharehold- ers.21 Creditors’ debt is often secured by firms’ remaining as- sets, like equipment or real estate, and creditors are entitled to a share of the proceeds from the sale of those assets.22 The equity and debt of a firm often trade in public markets, and when new information becomes available about a firm, the price of those financial instruments can change.23 For ex- ample, when the expected value of a firm falls, the market 19 Many sources either explicitly or implicitly conflate bank capital and equity. See, e.g., William Alden, What Is Bank Capital, Anyway?, N.Y. TIMES, (July 10, 2013), https://dealbook.nytimes.com/2013/07/10/what-is- bank-capital-anyway [https://perma.cc/XC5S-L72X]. However, while the term “capital” almost always includes common equity, the two terms are not precisely coterminous. See, e.g., FED. RESERVE BANK OF S.F., WHAT IS BANK CAPITAL AND WHAT ARE THE LEVELS OR TIERS OF CAPITAL? (2003), https://www.frbsf.org/education/publications/doctor-econ/2001/septem- ber/bank-capital [https://perma.cc/YW5E-XFC2]. 20 For a comparison between the features of debt and equity, see Fin. Accounting Standards Bd., Distinguishing Liabilities From Equity (Topic 480): Overview and Background (2017), https://asc.fasb.org/sec- tion&trid=2175789 [https://perma.cc/3B62-NPG8]. For a list of exceptions relating to hybrid interests, see Fin. Accounting Standards Bd., Distin- guishing Liabilities From Equity (Topic 480): Scope and Scope Exceptions (2017), https://asc.fasb.org/section&trid=2175795#SL109262033-110874 [https://perma.cc/XS3F-ML6R]. 21 See 11 U.S.C. § 507(b) (2018). 22 See 11 U.S.C. §§ 725–26. 23 See Christopher Paul Saari, The Efficient Capital Market Hypothe- sis, Economic Theory and the Regulation of the Securities Industry, 29 STAN. L. REV. 1031, 1035–41 (1977). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 587 price of its equity may also fall. When the probability that a firm will repay its debt falls, the market price of its debt may also fall. To protect against this latter risk, a firm’s creditors (or, for that matter, anyone else) might enter into a CDS con- tract with a separate financial institution, which will pay the holder of the CDS if the firm defaults on its debt.24 The price of a CDS contract is known as the “CDS spread”—the higher the probability that a credit event will occur, such as a default, the greater the spread typically becomes.25 In ordinary times, with respect to the funding available to it, a financial institution26 is much like any other business. The main output of a bank is credit; it uses outside investment (e.g., deposits, bonds, common stock) to fund the creation of financial assets (e.g., loans).27 Ideally, the bank makes more money off those assets than their investors demand for fund- ing them. If that holds true, the bank is able to repay its cred- itors and earn a profit, and its shareholders’ equity grows in value. 24 See Credit Default Swaps, PAC. INV. MGMT. COMPANY, https://www.pimco.com/en-us/resources/education/understanding-credit- default-swaps [https://perma.cc/UH3C-RRDZ] (last visited June 1, 2020). 25 See Mark J. Flannery et. al., Credit Default Swap Spreads As Viable Substitutes for Credit Ratings, 158 U. PA. L. REV. 2085, 2088 (2010). The parties to a CDS contract may agree to expand the definition of a credit event might to include other events, such as firm downgrades. See, e.g., Jer- emy C. Kress, Credit Default Swaps, Clearinghouses, and Systemic Risk: Why Centralized Counterparties Must Have Access to Central Bank Liquid- ity, 48 HARV. J. LEGIS. 49, 52 (2011); Daniel Hemel, Empty Creditors and Debt Exchanges, 27 YALE J. REG. 159, 162 (2010). 26 In the remainder of this Section, we use the term “bank” as short- hand to refer to any financial institution. As discussed in the following Sec- tion, and reflected in our sample and results, banks (i.e., deposit-taking in- stitutions) are not the only financial institutions that can experience a run. See infra Section II.B. 27 Depositors, notably, are bank creditors who provide principal (in the form of deposits) that the bank must repay, typically with a share of inter- est. See GARY B. GORTON, MISUNDERSTANDING FINANCIAL CRISES: WHY WE DON’T SEE THEM COMING 5–6 (2012); Jeanne Gobat, Banks: At the Heart of the Matter, FIN. & DEV., June 1, 2018, at 56, 56. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 588 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 However, banks also have unique characteristics that ex- pose them to unique risks. Banks typically use short-term sources of funding, like deposits, to invest in long-term pro- jects, like thirty-year mortgages or ten-year business loans.28 In a “fractional reserve” banking system, the amount of money a bank invests in these projects can exceed the money it re- ceives in funding.29 As long as a bank’s investors do not with- draw their short-term funding at once, the bank can operate normally. By contrast, if too many investors in a bank demand too much cash at once, the bank can face a “run.”30 To pay some creditors, it can sell its liquid assets for cash on short notice, at a price close to their economic worth.31 As investor demands mount, however, a bank may have to sell other, less liquid assets at less—perhaps far less—than their actual worth.32 Doing so can satisfy some short-term creditor de- mands, but it results in losses that can further erode creditor confidence, leading to even more demands for cash.33 28 For a less condensed discussion of this activity (on maturity trans- formation, liquidity transformation, and other core functions of financial in- termediation), see Laura E. Kodres, What Is Shadow Banking?, FIN. & DEV., June 1, 2013, at 42. 29 For detail on the mechanics of this mechanism, see Michael McLeay et al., Money Creation in the Modern Economy, 2014 BANK ENG. Q. BULL. 14, 17–18. See also MORGAN RICKS, THE MONEY PROBLEM: RETHINKING FINAN- CIAL REGULATION 67–72 (2016). 30 See supra note 6 and accompanying text. 31 See, e.g., Stanley Fischer, Vice Chairman, Fed. Reserve Sys., Is There a Liquidity Problem Post-Crisis? 1 (Nov. 15, 2016), https://www.bis.org/review/r161118d.htm [https://perma.cc/3MVL-TBCL] (defining market liquidity as “the ability to rapidly execute sizable securi- ties transactions at a low cost and with a limited price impact”); SUSAN MCLAUGHLIN ET AL., BANK FOR INT’L SETTLEMENTS, FOREIGN EXCHANGE LI- QUIDITY IN THE AMERICAS iii (2017), https://www.bis.org/publ/bppdf/bispap90.htm [https://perma.cc/YX6U- YFE8] (defining a market as liquid if “an investor wishing to execute a transaction of a desired size can do so at or near the prevailing market price, relatively quickly, and with no material price impact”). 32 See GORTON, supra note 27, at 45–46. 33 See id. at 46. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 589 Repeated enough times, a firm that was solvent when a run began can become insolvent before it ends.34 Liquidity is tied closely to both the probability and severity of a run, since a less liquid bank must sell more of its assets at a loss to satisfy creditors. However, illiquidity and suscep- tibility to a run are not the same thing. The difference lies in the source and duration (or, “fragility”) of a bank’s funding. For example, assume a bank gets all of its funding in over- night credit (which must be renewed on a daily basis) and keeps half that funding in reserve as cash. That bank is highly liquid, but also highly runnable, since its creditors could with- draw funding on less than a day’s notice. By contrast, assume a bank gets all of its funding in ninety-day loans, uses 95% of that funding to issue sixty-day consumer loans, and holds 5% in reserve as cash. That bank is highly illiquid, but not highly runnable, since its creditors have no contractual right to with- draw funding before the bank’s assets mature. B. The Post-Lehman Reforms Because the sources of bank funding are diverse, the chan- nels that can give rise to a bank run are also diverse.35 In the 2008 financial crisis, the run on large, diversified financial in- stitutions occurred principally in the sale-and-repurchase (or, “repo”) market.36 To borrow in this market, an institution would typically offer investors a securitized bond, often backed by the stream of payments from a group of mortgage 34 For a more detailed description of the dynamic in this paragraph that incorporates an account of deposit insurance and includes working def- initions of “market” and “funding liquidity,” see id. 35 See Matthew Pritsker, The Channels for Financial Contagion, in IN- TERNATIONAL FINANCIAL CONTAGION 67, 70–77 (Stijn Claessens & Kristin J. Forbes eds., 2001) (describing potential channels of financial contagion); see also Gary Gorton & Andrew Metrick, Securitized Banking and the Run on Repo, 104 J. FIN. ECON. 425, 430–33 (2012). 36 See id. at 425–26. Note that, while the “run on repo” involved non- banking financial institutions, large multi-line financial institutions and in- vestment banks were major participants. See Darrell Duffie, Prone to Fail: The Pre-Crisis Financial System, 33 J. ECON. PERSP. 81, 88–90 (2019). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 590 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 loans, as collateral.37 If a borrowing institution failed to repay its (usually short-term) loan by repurchasing the bond at a premium, the investor could sell the bond.38 However, as the mortgage market began to deteriorate, doubts about the qual- ity of those securitized bonds and the mortgage loans backing them increased, as did the cost of repo borrowing, until such borrowing ceased almost entirely.39 The story from here is familiar. Starting in late 2007, gov- ernments intervened.40 Intervention fostered expectations of future intervention.41 The September 15, 2008, bankruptcy of Lehman violated those expectations.42 Markets reacted 37 See Duffie, supra note 36. 38 See id. at 91. 39 See id. at 90–92. 40 In the United States, pre-Lehman interventions included the Term Auction Facility (launched Dec. 12, 2007), the Single-Tranche Open Market Operations Facility (launched Mar. 7, 2008), the Term Securities Lending Facility (launched Mar. 11, 2008), and the Primary Dealer Credit Facility (launched Mar. 16, 2008). See HAL S. SCOTT, CONNECTEDNESS AND CONTA- GION: PROTECTING THE FINANCIAL SYSTEM FROM PANICS 75 (2016). The U.S. government also provided multiple lines of support to the government-spon- sored enterprises focused on the housing market (principally Fannie Mae and Freddie Mac), ultimately resulting in their conservatorship under the Federal Housing Finance Agency. See BERNANKE ET AL., supra note 7, at 230; History of Fannie Mae & Freddie Mac Conservatorships, FED. HOUSING FIN. AGENCY, https://www.fhfa.gov/Conservatorship/Pages/History-of-Fan- nie-Mae--Freddie-Conservatorships.aspx [https://perma.cc/75QJ-MTDE] (last visited May 29, 2020). Other substantial interventions took place in the United Kingdom and Europe. See Communication From the Commission to the European Council: A European Economic Recovery Plan, COM (2008) 800 final (Nov. 26, 2008). 41 Combatting these expectations was, at the time, among the stated intentions of the bankruptcy. See Secretary Hank M. Paulson Jr., Statement by Sec. Paulson on Economy, C-SPAN (Sept. 15, 2008), https://www.c- span.org/video/?281125-2/statement-sec-paulson-economy [https://perma.cc/EK55-V2JW] (“Moral hazard is something I don’t take lightly.”). 42 We discuss the possible role of private information in our results in- fra Part III, but both financial market performance and the timeline of “Leh- man weekend” support this conclusion. On Friday, September 12, 2008, eq- uity markets were stable, and the Dow Jones Industrial Average ultimately 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 591 poorly.43 The result was further government intervention. Most crisis-era public programs offered financial institutions greater access to more liquid assets, such as cash or sovereign bonds, by pledging less liquid assets as collateral.44 After closed up 1.8% on the week. See Steven Russolillo, This Day in Crisis His- tory: Sept. 12, 2008, WALL ST. J. (Sept. 12, 2013), https://blogs.wsj.com/mon- eybeat/2013/09/12/this-day-in-crisis-history-sept-12-2008/ [https://perma.cc/9X3B-ZR5W]. The CEOs of the largest U.S. banks spent the weekend attempting to negotiate an assistance package for the sale of Lehman to Barclays, which appeared to be near consummation until last- minute discussions with the U.K. Financial Services Authority revealed that, under London Stock Exchange rules, Barclays would need an affirma- tive shareholder vote before Monday morning to move forward with the deal. See Andrew Ross Sorkin, Lehman Files for Bankruptcy; Merrill Is Sold, N.Y. TIMES (Sept. 14, 2008), https://www.ny- times.com/2008/09/15/business/15lehman.html [https://perma.cc/JQ2P- YQEK]; Too Big to Fail: Expectations and Impact of Extraordinary Govern- ment Intervention and the Role of Systemic Risk in the Financial Crisis Be- fore the Fin. Crisis Inquiry Comm’n, 111th Cong. 7–8 (2010) (statement of Thomas C. Baxter, Executive Vice President and General Counsel of the Federal Reserve Bank of New York). In response to this news, on the follow- ing Tuesday, September 16, the Dow closed down 4.4%, and the S&P 500 index had fallen 4.7%. See Tom Lauricella et al., Dow, Markets in Europe Post Big Falls, WALL ST. J. (Sept. 16, 2008), https://www.wsj.com/arti- cles/SB122152873162140589 [https://perma.cc/R23G-LDCE]. 43 See Lauricella, supra note 42. Markets hit their crisis-era lows in March 2009, with the Dow Jones Industrial Index and S&P 500 falling to their lowest points since the late 1990s. See Alexandra Twin, For Dow, An- other 12-Year Low, CNN MONEY (Mar. 9, 2009), https://money.cnn.com/2009/03/09/markets/markets_newyork/ [https://perma.cc/UAB4-CNQM]. 44 In addition to the programs described supra note 40, the U.S. gov- ernment created a secured revolving credit facility available to insurer AIG (launched September 16, 2008), the Asset-Backed Commercial Paper Money Market Mutual Fund Liquidity Facility (launched September 22, 2008), the Temporary Guarantee Program (launched September 29, 2008), the Com- mercial Paper Funding Facility (launched October 27, 2008), and the Term Asset-Backed Securities Loan Facility (launched November 25, 2008). See SCOTT, supra note 40, at 76. Federal Home Loan Bank advances also became an important source of liquidity. See JONATHON ADAMS-KANE & JAKOB WIL- HELMUS, MILKEN INST., THE REAL STORY BEHIND THE SURGE IN FHLB AD- VANCES: MACROPRUDENTIAL POLICY CHANGED HOW BANKS BORROW 4 fig.3 (2017). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 592 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Lehman, new programs also provided institutions with capi- tal.45 Still others directly supported markets in specific finan- cial products.46 Ultimately—and consequently—the financial system avoided collapse.47 Early debates about the causes of the crisis focused on the adequacy of bank capital.48 Regulators have long required banks to fund a certain proportion of their assets with money derived from equity or equity-like instruments, which could “absorb” losses if their assets lost value.49 These requirements 45 U.S. capital-based programs included the Troubled Asset Relief Pro- gram (passed as part of the Emergency Economic Stabilization Act of 2008, Pub. L. No. 110-343, tit. 1, 122 Stat. 3765, 3767–800 (codified as amended at 12 U.S.C. §§ 5211–41 (2018)) and the broader Capital Purchase Program. SCOTT, supra note 40, at 76. 46 The Federal Reserve also instated unlimited swap lines with four foreign central banks, and the FDIC raised its deposit insurance limit. See SCOTT, supra note 40, at 76–77. 47 The role of the government in staving off collapse is still subject to debate. For an accessible overview of this debate, and an argument for the paramount importance of public assistance programs during periods of fi- nancial crisis, see ALAN S. BLINDER, AFTER THE MUSIC STOPPED: THE FINAN- CIAL CRISIS, THE RESPONSE, AND THE WORK AHEAD (2013). 48 Improving bank capital was a commitment listed in the Leaders’ Statement following the 2009 Pittsburgh G20 meeting, the focus of the first post-crisis supervisory stress tests, and the subject of countless statements from crisis-era policymakers. See Leaders’ Statement, The Pittsburgh Sum- mit 8 (Sept. 24–25, 2009), https://www.treasury.gov/resource-center/inter- national/g7-g20/Documents/pittsburgh_summit_leaders_state- ment_250909.pdf [https://perma.cc/7CPE-CVGK]; Ben S. Bernanke, Chairman, Fed. Reserve Sys., The Supervisory Capital Assessment Pro- gram (May 11, 2009), https://www.federalre- serve.gov/newsevents/speech/bernanke20090511a.htm [https://perma.cc/3K6S-Q3UU]; John Fell, Directorate Gen., European Cent. Bank, Address at the Conference on Bank Structure and Competition: Stress Testing in a Crisis—The European Experience (May 10, 2012). 49 For a primer on global capital regulations and the rationales behind them, see MOODY’S ANALYTICS, REGULATION GUIDE: AN INTRODUCTION (2011), https://www.moodysanalytics.com/-/media/whitepaper/2011/11-01- 03-regulation-guide-introduction.pdf [https://perma.cc/E67K-FGY6]; ANAT ADMATI & MARTIN HELLWIG, THE BANKERS’ NEW CLOTHES: WHAT’S WRONG WITH BANKING AND WHAT TO DO ABOUT IT 94–95 (2013). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 593 are typically articulated as “capital ratios,” in the general form of: Before the crisis, the prevailing international standards for capital regulation were the Basel II accords.50 For the numer- ator of the capital ratios, Basel II distinguished between three different categories of capital—including a Tier 1 category consisting of common equity and similar instruments.51 For the denominator, Basel II required institutions to use risk- weighted assets (“RWA”) instead of the total book value of all their financial assets.52 Risk-weighting applies a coefficient to the value of each asset; the higher the probability the asset will lose value, the greater the coefficient, and the greater the increase in the denominator of the capital ratio. Basel II al- lowed two methods for calculating risk-weighted assets: an “advanced approach,” which allowed larger institutions to use their internal risk models to conduct the calculation, and a “standardized approach,” which did not.53 Under both ap- proaches, when a bank’s assets are riskier, Basel II requires the bank to fund those assets with more equity and equity- like instruments.54 50 Since the 1970s, financial regulators have collaborated to produce standards for the oversight of internationally active banking institutions. These standards are called the “Basel accords,” after the Swiss city where the first such agreement was negotiated (and where continued work on these standards is based). See History of the Basel Committee, BANK INT’L SETTLEMENTS, https://www.bis.org/bcbs/history.htm [https://perma.cc/SQ8Z-GCJL] (last visited May 29, 2020). 51 BASEL COMM. ON BANKING SUPERVISION, INTERNATIONAL CONVER- GENCE OF CAPITAL MEASUREMENT AND CAPITAL STANDARDS 244 (2006) [here- inafter BASEL COMM., BASEL II]. The other two specific inclusions in Basel II Tier 1 capital are disclosed reserves and non-cumulative perpetual pre- ferred stock. Id. at 244–45. 52 Id. at 244. 53 Id. at 12–149. 54 Id. 𝑐𝑎𝑝𝑖𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 594 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 The numerator and denominator of the required capital ra- tios changed under the post-crisis Basel III accords.55 For the numerator, Basel III created a new Common Equity Tier 1 (“CET1”) requirement, meant to hew more closely to equity capital, and introduced a new version of Tier 1 capital.56 For the denominator, Basel III introduced a new version of the standardized approach and limited the discretion associated with the internal ratings-based approach.57 Basel III also cre- ated a new leverage requirement, which used no risk- weighting and included both on-balance sheet assets and off- balance sheet exposures.58 Finally, the accords included sev- eral additional capital “buffers.”59 Liquidity regulation also changed dramatically as a result of the crisis. Basel III created a new Liquidity Coverage Ratio (“LCR”), the first liquidity regulation in the Basel accords, meant to ensure that institutions had enough “high-quality liquid assets” to meet their demands for cash over a thirty-day period.60 To tackle the run risk associated with an over-reli- ance on short-term funding, Basel III also created a measure 55 BASEL COMM. ON BANKING SUPERVISION, BASEL III: A GLOBAL REGU- LATORY FRAMEWORK FOR MORE RESILIENT BANKS AND BANKING SYSTEMS 2–3 (2011) [hereinafter BASEL COMM., BASEL III]. 56 Id. at 13–17; see also Daniel K. Tarullo, Member, Bd. of Governors of the Fed. Reserve Sys., Remarks at the 2016 Financial Stability Confer- ence: Financial Regulation Since the Crisis 8 n.9 (Dec. 2, 2016), https://www.bis.org/review/r161205f.pdf [https://perma.cc/959Q-DN2G] (“In addition to increasing minimum capital ratios, post-crisis reforms also placed more emphasis on the quality of regulatory capital by introducing the common equity tier 1 capital ratio, which reflects the focus by bank in- vestors and counterparties during the crisis on common equity.”). 57 BASEL COMM., BASEL III, supra note 55, at 3–4, 51–54. 58 Id. at 4. These two measurements were seen as complementary. If risk-weighting was either too complex or too easily manipulated, the weights would fail to capture actual credit exposures—but without risk- weighting, a financial institution could seek higher returns by undertaking riskier loans for the same equity funding requirements. See infra notes 137– 39 and accompanying text. 59 BASEL COMM., BASEL III, supra note 55, at 54–60. 60 BASEL COMM. ON BANKING SUPERVISION, BASEL III: THE LIQUIDITY COVERAGE RATIO AND LIQUIDITY RISK MONITORING TOOLS 1 (2013). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 595 to ensure that banks had enough long-term funding to cover their long-term assets.61 In November 2007, disclosure re- quirements from the Financial Accounting Standards Board (“FASB”) also took effect, requiring publicly traded companies in the United States to disclose their total “Level 1, 2, and 3” assets.62 Assets are divided into “levels” roughly according to how liquid they are.63 The Basel accords are non-binding international agree- ments that apply to internationally active banks; however, na- tional regulators implement and enforce Basel standards through their own domestic regulations.64 In the United States, for instance, regulators promulgated new capital and liquidity requirements under the Dodd-Frank Wall Street Re- form and Consumer Protection Act.65 For the Federal Reserve Board (the “Board”), these regulations included a host of “en- hanced supervision and prudential standards,”66 with capital and liquidity requirements roughly increasing with the size and complexity of financial institutions.67 For the Office of the Comptroller of the Currency, they included a set of “height- ened expectations” for large institutions.68 For all U.S. 61 See BASEL COMM. ON BANKING SUPERVISION, BASEL III: THE NET STA- BLE FUNDING RATIO 1–3 (2014) [hereinafter BASEL COMM., NET STABLE FUND- ING RATIO]. 62 FIN. ACCOUNTING STANDARDS BD., STATEMENT OF FINANCIAL AC- COUNTING STANDARDS NO. 157: FAIR VALUE MEASUREMENTS 12–17 (2010). 63 See id. 64 See BASEL COMM. ON BANKING SUPERVISION, CHARTER (2013). 65 Dodd-Frank Wall Street Reform and Consumer Protection Act §§ 115(b)(1), 165(b)(1)(A), 12 U.S.C. §§ 5325(b)(1), 5365(b)(1)(A) (2018). 66 Id. 67 Note in particular that the federal banking agencies imposed an “en- hanced Supplementary Leverage Ratio” for large financial institutions. Reg- ulatory Capital Rules: Regulatory Capital, Enhanced Supplementary Lev- erage Ratio Standards for Certain Bank Holding Companies and Their Subsidiary Insured Depository Institutions, 79 Fed. Reg. 24,528 (May 1, 2014) (to be codified at 12 C.F.R. pts. 6, 208, 217, and 324). 68 See OCC Guidelines Establishing Heightened Standards for Certain Large Insured National Banks, Insured Federal Savings Associations, and Insured Federal Branches; Integration of Regulations, 79 Fed. Reg. 54,518 (Sept. 11, 2014) (codified at 12 C.F.R. pts. 30, 168, and 170). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 596 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 prudential regulators69 the new regulations also mandated the development of firm recovery and resolution plans, de- signed to avoid or facilitate the liquidation of a large, complex financial institution without interrupting critical financial market operations or requiring public financial support.70 Many Basel member jurisdictions, including the United States, also crafted new “stress-testing” regimes, formalizing a tool first deployed in the throes of the crisis.71 Typically, a regulatory stress test involves a set of macro-level “stress sce- narios”—e.g., a fall of X% in GDP, a rise of Y% in unemploy- ment, or some combination of factors—that regulators and banks use to model potential future changes to banks’ balance sheets.72 In the U.S., regulators launched a stress-testing pro- gram focused on capital at large financial institutions and re- quired those institutions to separately run their own periodic stress tests simulating liquidity shocks.73 69 These regulators include the Office of the Comptroller of the Cur- rency, the Federal Deposit Insurance Corporation, and the Board of Gover- nors of the Federal Reserve System. 70 See Resolution Plans Required, 76 Fed. Reg. 67,323 (Nov. 1, 2011) (codified at 12 C.F.R. pts. 243, 381). 71 See Dodd Frank Act § 165(i), 12 U.S.C. § 5365(i) (2018) (United States); BANK OF ENG., THE BANK OF ENGLAND’S APPROACH TO STRESS TEST- ING THE UK BANKING SYSTEM 5–8 (2015) (United Kingdom); cf. Takako Taniguchi & Finbarr Flynn, Japan Will Leave Banks to Carry Out Their Own Stress Tests, BLOOMBERG (May 9, 2016), https://www.bloom- berg.com/news/articles/2016-05-09/japan-to-scrutinize-banks-stress-tests- stopping-short-of-fed?sref=m42vRBnI [https://perma.cc/56J6-N5R8] (Ja- pan). 72 See, e.g., Press Release, Bd. of Governors of the Fed. Reserve Sys., Federal Reserve Board Releases Scenarios for 2018 Comprehensive Capital Analysis and Review (CCAR) and Dodd-Frank Act Stress Test Exercises and Issues Instructions to Firms Participating in CCAR (Feb. 1, 2018), https://www.federalreserve.gov/newsevents/pressre- leases/bcreg20180201a.htm [https://perma.cc/R229-V7CY]. 73 For an overview of this system as enacted shortly after the crisis, see Daniel K. Tarullo, Member, Bd. of Governors of the Fed. Reserve Sys., Speech at the Federal Reserve Bank of Chicago Annual Risk Conference: Developing Tools for Dynamic Capital Supervision (Apr. 10, 2012), 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 597 C. Prior Literature These new regulatory and supervisory measures had a dominant stated motivation at the time they were created: “[N]o more taxpayer-funded bailouts, period.”74 On these grounds alone, regulators had good reason to focus on capital. Both before and after the crisis an extensive academic litera- ture has explored the benefits of higher capital ratios—from minimizing the moral hazard associated with deposit https://www.federalreserve.gov/newsevents/speech/tarullo20120410a.htm [https://perma.cc/7FVF-TV86]. 74 Helene Cooper, Obama Signs Overhaul of Financial System, N.Y. TIMES (July 21, 2010) (internal quotation marks omitted), https://www.ny- times.com/2010/07/22/business/22regulate.html [https://perma.cc/6VKL- PA7J]. The closely related mantle of “too big to fail” took hold in public dis- course soon after the financial crisis. See, e.g., Jonathan Macey, Brave New Fed, WALL ST. J. (Mar. 31, 2008), https://www.wsj.com/arti- cles/SB120692412871875675 [https://perma.cc/JV5W-C66B] (describing the Bear Stearns sale as contravening the intent of existing public policies to minimize the use of the “too-big-to-fail” doctrine); Neil Irwin, Paulson To Urge New Fed Powers, WASH. POST (June 19, 2008), http://www.washing- tonpost.com/wp-dyn/content/article/2008/06/18/AR2008061803225.html [https://perma.cc/3W2R-XPK8] (“We must limit the perception that some in- stitutions are either too big to fail or too interconnected to fail . . . . If we are to do that credibly, we must address the reality that some are.” (internal quotation marks omitted)); President Barack Obama, Remarks on Financial Rescue and Reform at Federal Hall (Sept. 14, 2009), https://obamawhitehouse.archives.gov/the-press-office/remarks-president- financial-rescue-and-reform-federal-hall [https://perma.cc/TQ2T-KBUQ] (“Those on Wall Street cannot resume taking risks without regard for con- sequences, and expect that next time, American taxpayers will be there to break their fall.”). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 598 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 insurance,75 to reducing the probability of insolvency,76 to im- proving lending volumes in the wake of a shock.77 75 See, e.g., João A.C. Santos, Bank Capital Regulation in Contempo- rary Banking Theory: A Review of the Literature, 10 FIN. MKTS., INSTS. & INSTRUMENTS 41, 49–52 (2001). 76 See, e.g., Laura Chiaramonte & Barbara Casu, Capital and Liquid- ity Ratios and Financial Distress: Evidence from the European Banking In- dustry, 49 BRIT. ACCT. REV. 138, 139, 156–57 (2017) (describing the associa- tion between higher capital ratios and a lower probability of “failure and distress” at large EU banks). 77 The overwhelming majority of the empirical literature on bank cap- ital has focused on the relationship between capital and lending. That liter- ature is outside the scope of this Article, but much of it suggests that capital has a smoothing effect, dampening lending during high points in the busi- ness and credit cycles and preserving lending during low points. See, e.g., Ben S. Bernanke, Non-Monetary Effects of the Financial Crisis in the Prop- agation of the Great Depression, 73 AM. ECON. REV. 257, 263–65, 267–68, 272–74 (1983); Joseph Noss & Priscilla Toffano, Estimating the Impact of Changes in Aggregate Bank Capital Requirements on Lending and Growth During an Upswing, 62 J. BANKING & FIN. 15, 15–17, 25–26 (2016) (finding that a slight increase in capital requirements results in a slight decrease in lending by UK banks); Mark Carlson et al., Capital Ratios and Bank Lend- ing: A Matched Bank Approach, 22 J. FIN. INTERMEDIATION 663, 674–79, 682–86 (2013) (finding that banks whose capital ratios were relatively high had strong long growth during the 2008-2010 financial crisis); Marko Košak et al., Quality of Bank Capital and Bank Lending Behavior During the Global Financial Crisis, 37 INT’L REV. FIN. ANALYSIS 168, 174–76 (2015) (finding greater continuity in lending for banks with higher Tier 1 capital and retail deposit levels, but not for banks with higher Tier 2 capital or in- terbank deposit levels); Leonardo Gambacorta & Paolo Emilio Mistrulli, Does Bank Capital Affect Lending Behavior?, 13 J. FIN. INTERMEDIATION 436, 436–41, 449–52 (2004) (finding that the credit supply of well-capital- ized Italian banks is not dependent on the business cycle); Sudipto Karmakar & Junghwan Mok, Bank Capital and Lending: An Analysis of Commercial Banks in the United States, 128 ECON. LETTERS 21, 21–23 (2015) (finding “a moderate relationship between capital ratios and business lending” through multiple business cycles); cf. Matthew Osborne et al., In Good Times and in Bad: Bank Capital Ratios and Lending Rates, 51 INT’L REV. FIN. ANALYSIS 102, 102–103, 107–12 (2017) (finding that better-capi- talized banks are more likely to engage in secured household lending during upturns than in downturns); Jonathan Bridges et al., The Impact of Capital Requirements on Bank Lending 3–4, 16–21, 23 (Bank of Eng., Working Pa- per No. 486, 2014) (finding an initial decrease in lending after an increase 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 599 A separate, largely theoretical literature also argues that higher capital levels can help prevent runs by limiting doubts about an institution’s ability to pay its short-term debts.78 These claims about an institution’s ex ante vulnerability to a run also entail a theory about investor behavior during a run—that the “deeper reason” for investors’ actions is insol- vency. Before the financial crisis, under this theory, banks were highly indebted. When banks suffered losses, investors, including other financial institutions, lost confidence and cut off funding, fearing that banks might become unable to repay their debts. The Lehman Brothers bankruptcy itself height- ened investors’ concerns by showing that even a large finan- cial institution might not be bailed out, and therefore that de- fault of such an institution was a real possibility.79 in capital requirements, followed by recovery within three years); Jose M. Berrospide & Rochelle M. Edge, The Effects of Bank Capital on Lending: What Do We Know, and What Does it Mean? 2–3, 13–17 (Fed. Reserve Bd. Fin. & Econ. Discussion Series, Working Paper No. 2010-44, 2010) (finding a relatively weak relationship between capital levels and loan growth); Shekhar Aiyar et al., The International Transmission of Bank Capital Re- quirements: Evidence from the UK, 113 J. FIN. ECON. 368, 368, 375–80 (2014) (finding decreased interbank lending to “non-core” countries following the introduction of higher capital requirements in the UK). However, as Berro- spide and Edge note, the time frame and geographic scope of these studies seem to play a role in their results. See Berrospide & Edge, supra note 77, at 3–5. 78 See, e.g., Anat R. Admati & Martin F. Hellwig, Bank Leverage, Wel- fare and Regulation, in SYSTEMIC RISK IN THE FINANCIAL SECTOR 217 (Doug- las W. Arner et al. eds., 2019); cf. Gorton, supra note 7, at 774–79 (deter- mining, from an analysis of pre- and post-Federal Reserve Act banking crises that “[d]epositors panic when the liabilities signal is strong enough,” and rejecting a “sun spot” hypothesis of depositor behavior). 79 See, e.g., ADMATI & HELLWIG, supra note 49, at 12; see also Daniel K. Tarullo, Member, Bd. of Governors of the Fed. Reserve Sys., Speech at the Clearing House Business Meeting and Conference: The Evolution of Capital Regulation 2 (Nov. 9, 2011), https://www.bis.org/review/r111110c.pdf [https://perma.cc/TJG3-MJBT] (“In the fall of 2008, there was widespread doubt in markets that the common equity of some of our largest institutions was sufficient to withstand the losses that those firms appeared to be facing. This doubt made investors and counterparties increasingly reluctant to deal 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 600 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 The most prominent (and voluble) counterargument to this view claims that, while higher capital might prevent runs, it also comes with substantial private and social costs.80 How- ever, another counterargument based in contagion theory also exists. It claims that the capital-focused account of runs is based on a faulty premise.81 Contagion theory agrees that a run can begin with doubts about the quality of a bank’s assets. It argues, however, that a run can begin for virtually any reason, from concerns about asymmetric information, to a change in the elasticity of in- vestment, to simple herd behavior or randomness.82 Under this theory, what both causes and distinguishes a run is a lack of liquidity—that is, a mismatch between the cash that bank with those firms, contributing to the severe liquidity strains that character- ized financial markets at the time.”). 80 See, e.g., Douglas Elliott et al., Assessing the Cost of Financial Reg- ulation 13–56, 67 (Int’l Monetary Fund, Working Paper No. 12/233, 2012), https://www.imf.org/en/Publications/WP/Issues/2016/12/31/Assessing-the- Cost-of-Financial-Regulation-40021 [https://perma.cc/9ASR-GBKA]; cf. Anat R. Admati et al., Fallacies and Irrelevant Facts in the Discussion of Capital Regulation, in CENTRAL BANKING AT A CROSSROADS: EUROPE AND BE- YOND 33, 39–40 (Charles Goodhard et al. eds., 2014). We largely elide a third counterargument on the relationship between leverage and agency rents, where empirical research remains scant. See, e.g., Gary Gorton & Andrew Winton, Liquidity Provision, Bank Capital, and the Macroeconomy, 49 J. MONEY, CREDIT & BANKING 5, 6 (2017); Thierno Amadou Barry et al., Own- ership Structure and Risk in Publicly Held and Privately Owned Banks, 35 J. BANKING & FIN. 1327, 1335–39 (2011) (observing a direct relationship in sample of European banks). 81 See SCOTT, supra note 40, at xv (defining contagion as “an indiscrim- inate run by short-term creditors of financial institutions that can render otherwise solvent institutions insolvent due to the fire sale of assets that are necessary to fund withdrawals and the resulting decline in asset prices”). 82 See id. at 9–13. Douglas W. Diamond and Philip H. Dybvig’s seminal argument posits that demand deposit contracts have multiple equilibria, one of which is a bank run, and that “almost anything” can cause a “shift in expectations” and a move to the run equilibrium. Douglas W. Diamond & Philip H. Dybvig, Bank Runs, Deposit Insurance, and Liquidity, 91 J. POL. ECON. 401, 402–04, 416–18 (1983). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 601 creditors demand, and the cash that a bank owns or can get.83 If too few of a bank’s assets are liquid, even a few creditor re- demptions can exhaust them, leading to the sale of illiquid as- sets at a loss and further demands for cash. This positive-feed- back loop continues until the bank is insolvent—but with enough liquidity, it never occurs at all. In this view, “crises are about cash and not capital,”84 and regulators who seek to pre- vent contagious runs should focus on making liquidity freely available, rather than increasing the ex ante proportion of funding a bank gets from equity.85 Much of the contagion literature is theoretical, but some historical evidence is consistent with the contagion hypothe- sis. First, banking crises were prevalent in the United States even in the 19th century, when bank capital ratios hovered above 50%.86 Second, although balance sheet measures of reg- ulatory capital have increased substantially since the 2008 crisis, some market-based measures of volatility and risk re- main the same or higher than they were a decade ago.87 Third, 83 See GORTON, supra note 27, at 153. 84 Id. See also SCOTT, supra note 40, at 10 (describing contagion as “a liquidity-driven phenomenon” that is “not conditioned on insolvency”). These sources are ultimately rooted in Walter Bagehot’s dictum to “lend freely,” at “a very high rate of interest,” on “good banking securities.” WAL- TER BAGEHOT, LOMBARD STREET: A DESCRIPTION OF THE MONEY MARKET 31, 97 (Richard D. Irwin Inc., 1962) (1873). Importantly, however, Scott grants an important role for capital in protecting specifically against “[a] correlated negative shock [that] causes the failure of many large financial institutions at the same time,” since in such an event banks would lack “adequate col- lateral” to be eligible for credit from a lender of last resort, such as a central bank. SCOTT, supra note 40, at 181. 85 See ANDREW METRICK ET AL., GRP. OF THIRTY, MANAGING THE NEXT FINANCIAL CRISIS: AN ASSESSMENT OF EMERGENCY ARRANGEMENTS IN THE MA- JOR ECONOMIES 14–16 (2018), http://group30.org/images/uploads/publica- tions/Managing_the_Next_Financial_Crisis.pdf [https://perma.cc/J97R- PYLY] (arguing for an expansion of the Federal Reserve’s emergency lend- ing powers). 86 GORTON, supra note 27, at 161. 87 Natasha Sarin & Lawrence H. Summers, Understanding Bank Risk through Market Measures, BROOKINGS PAPERS ECON. ACTIVITY 57, 75–76, 88–101 (2016). Sarin and Summers “suspect that without increases in 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 602 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 other non-liquidity explanations for Lehman seem to fall short; for example, other major financial institutions had rel- atively low asset and liability exposures to Lehman, suggest- ing that the run was not purely a function of direct counter- party risk.88 The claims of the capital and contagion literature fall into a gap in the academic literature. The theoretical literature fo- cuses not on reducing the probability of a run, nor on mitigat- ing its damage, but on what can stop a run already underway from spreading. Relatedly, the capital and contagion litera- ture also discusses investor decision making, especially those decisions to withdraw funding on extremely short notice. By contrast, most academic work on capital has focused on its role capital requirements, levels of volatility would have increased even more than [they] observe,” but note that their results could be due to gaps be- tween the definitions of regulatory and true economic capital. Id. at 59–60. Consistent with the former explanation, several papers have argued that socially optimal levels of Tier 1 regulatory capital are substantially higher than the current required level. See, e.g., Simon Firestone et al., An Empir- ical Assessment of the Costs and Benefits of Bank Capital in the US 1–2, 5 fig.1 (Fed. Reserve Bd. Fin. & Econ. Discussion Series, Working Paper No. 2017-034, 2017), https://papers.ssrn.com/sol3/papers.cfm?ab- stract_id=2946814 [https://perma.cc/AJ8Y-M67H] (advocating for the mini- mum required Tier 1 capital range to be increased to 13–26%); Martin Brooke et al., Measuring the Macroeconomic Costs and Benefits of Higher UK Bank Capital Requirements 7, 12–24 (Bank of Eng., Working Paper No. 35, 2015), https://www.bankofengland.co.uk/-/media/boe/files/financial-sta- bility-paper/2015/measuring-the-macroeconomic-costs-and-benefits-of [https://perma.cc/JMY6-4RWT] (10–14%); David Miles et al., Optimal Bank Capital, 123 ECON. J. 1, 26–31, 28 tbl.9 (2013) (16–20%); Jihad Dagher et al., Benefits and Costs of Bank Capital 11–20 (Int’l Monetary Fund, Staff Discussion Note No. SDN/16/04, 2016), https://www.imf.org/exter- nal/pubs/ft/sdn/2016/sdn1604.pdf [https://perma.cc/5R73-2U7N] (15–23%); Wayne Passmore & Alexander H. von Hafften, Are Basel's Capital Sur- charges for Global Systemically Important Banks Too Small?, 15 INT’L J. CENT. BANKING 107, 136–44 (2019) (6.5–14.75%); FED. RESERVE BANK OF MINNEAPOLIS, THE MINNEAPOLIS PLAN TO END TOO BIG TO FAIL 43–44, 49–51 (2017), https://www.minneapolisfed.org/publications/special-studies/end- ingtbtf/final-proposal [https://perma.cc/VZ66-NJNX] (23.5%). 88 SCOTT, supra note 40, at 29–58. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 603 before and after a crisis (i.e., in preventing a crisis or has- tening recovery from one).89 The Lehman bankruptcy is foundational to both the capital and contagion narratives, not only because of its magnitude and historical significance, but also because the decision to file was unexpected, and thus was plausibly exogenous to subse- quent investor behavior. We are aware of only one study that examines how capital and liquidity levels affected the re- sponses of investors in financial institutions to the Lehman failure.90 This study found that large banks with lower lever- age and higher degrees of reliance on deposit funding had higher post-Lehman stock returns—and that pre-Lehman regulatory capital ratios and liquidity measures did little to explain those returns.91 However, several attributes of this study limit its specific relevance to runs. First, its market and balance-sheet data are included on an annual basis, which is too infrequent to cap- ture run behavior in funding markets.92 Second, its outcome variable only captures the behavior of shareholders in equity markets.93 Equity market capitalization is an easily 89 See supra notes 75–77 and accompanying text. 90 Asli Demirguc-Kunt et al., Bank Capital: Lessons from the Financial Crisis (Int’l Monetary Fund, Working Paper No. WP/10/286, 2010), https://www.imf.org/external/pubs/ft/wp/2010/wp10286.pdf [https://perma.cc/F7K6-P8K2]. Using CDS spread data, Nicolas Dumontaux and Adrian Pop also found that the negative effect of the Lehman bank- ruptcy on both share value and CDS spreads was “correlated with [the] fi- nancial conditions of the surviving institutions.” Nicolas Dumontaux & Adrian Pop, Contagion Effects in the Aftermath of Lehman’s Collapse: Meas- uring the Collateral Damage 2, 16–31 (Laboratoire d’Economie et de Man- agement Nantes, Working Paper No. 2012/27, 2012), https://hal.archives- ouvertes.fr/hal-00695721/document [https://perma.cc/3TC9-ZE4K]. How- ever, their measures of financial condition were focused on institutions’ loan books, specifically “the ratio of loan loss reserves to total loans” and “the ratio of non-performing assets as a fraction of total assets.” Id. at 21. These balance-sheet measures, as well as their regression specifications have the same limitations as those of Demirguc-Kunt. 91 See Demirguc-Kunt et al., supra note 90, at 9–11. 92 Id. at 7–8. 93 Id. at 4. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 604 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 observable component of enterprise value, and since it is a highly liquid measure, it can reflect concerns from a wide range of investors.94 However, debt markets are, by definition, where a run occurs, and as discussed, institutions are funded by both equity and debt. The interests of creditors and share- holders are not necessarily aligned, and their reaction to a shock also might be different.95 Third, the study uses Basel II measures of capital, and thus sheds little light on whether Ba- sel III and other reforms might have changed the relevance of regulatory capital ratios to investors during a run.96 94 See Sarin & Summers, supra note 87, at 58 (discussing the relation- ship of bank equity and debt market measures to bank exposures and asset quality). 95 See, e.g., Antonio S. Mello & John E. Parsons, Measuring the Agency Cost of Debt, 47 J. FIN. 1887 (1992). 96 Demirguc-Kunt’s analysis also contains a methodological flaw: Its outcome variable measures the raw change in a sample bank’s stock price, controlling for the stock’s beta (defined as the covariance between the stock’s return and the return of the host country’s stock market). Demirguc-Kunt et al., supra note 90, at 7. This approach has several shortcomings. First, failing to normalize raw price changes to an institution’s ticker price or mar- ket capitalization can make meaningful comparison impossible. For exam- ple, Bank A with 100 shares trading at $10 has the same market capitali- zation as Bank B with 1000 shares trading at $1. If the price of both banks’ shares fall $0.10, Bank A loses 1% in value, while Bank B loses 10%. Second, this approach fails to isolate capitalization from earnings data. That is, Bank A may have had a higher return on equity (“ROE”) than Bank B before the Lehman bankruptcy, and its ROE may be higher afterwards. However, the same high leverage that improved Bank A’s pre-Lehman ROE could have set it on a worse trajectory afterwards—precisely as basic corporate finance suggests it would. See, e.g., Troy Adkins, Optimal Use of Financial Leverage in a Corporate Capital Structure, INVESTOPEDIA (Apr. 4, 2019), https://www.investopedia.com/articles/investing/111813/optimal-use-finan- cial-leverage-corporate-capital-structure.asp [https://perma.cc/39K9- M3SY]. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 605 III. ANALYSIS A. Overview of Methodology and Research Design This Article intends to fill the gap in the academic litera- ture and policy discussions by using Lehman’s unexpected bankruptcy to examine the impact of capital and liquidity lev- els on large financial institutions’ susceptibility to a run.97 Re- call that the Friday prior to Lehman’s bankruptcy—the most recent time U.S. markets were open for trading prior to Leh- man’s demise—investors could reasonably expect that an- other bank would purchase Lehman, perhaps with public as- sistance, as had been the case when JPMorgan Chase & Co. purchased Bear Stearns six months earlier.98 That did not oc- cur. Instead, Monday’s trading began with news that Lehman had gone under, and that the accounts of Lehman’s British and Japanese brokerage operations had been frozen.99 We set up the econometric specification as follows. First, our outcome variables—share prices and CDS spreads—draw on infor- mation from both debt and equity markets.100 Importantly, 97 We cannot reject fully the absence of private ex ante information about Lehman’s planned filing, and thus completely eliminate concerns about endogeneity. However, as discussed above, Lehman counterparties and government officials worked to negotiate a transaction that would have avoided the filing until late on Sunday, September 14; markets were closed over the entire “Lehman weekend”; and the Lehman filing itself came shortly before the Monday, September 15, market opening. See supra notes 2–3, 42 and accompanying text. The likelihood that the market already re- flected Lehman’s failure before that Monday is low, as the subsequent splin- tering of equity prices and CDS spreads of large institutions suggests. See supra notes 4, 43 and accompanying text. 98 See supra note 3 and accompanying text. 99 See supra note 4 and accompanying text. 100 Several studies have used similar techniques to examine other as- pects of investor behavior during the 2008 crisis. See, e.g., Jian Yang & Ying- gang Zhou, Credit Risk Spillovers Among Financial Institutions Around the Global Credit Crisis: Firm-Level Evidence, 59 MGMT. SCI. 2343 (2013); Barry Eichengreen et al., How the Subprime Crisis Went Global: Evidence from Bank Credit Default Swap Spreads, 31 J. INT’L MONEY & FIN. 1299 (2012). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 606 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 these variables are proxies for run behavior, constructed from publicly available data, rather than direct measures. As dis- cussed above, the definition of a bank run is typically limited to the withdrawal of debt financing—in a prototypical case, the withdrawal of deposits.101 Direct measures of such with- drawals are unavailable publicly, either today or in 2008, at the level of granularity required to isolate the effect of the Lehman failure. Nevertheless, we believe our market-based proxies are credible: runs on financial institutions occur when there is no trust that those institutions will repay their liabil- ities. That breakdown in trust is captured by movements in the market. CDS spreads, for example, capture the market price of insuring against a credit event, including default.102 Equity prices complement CDS measures, since equities trade in thicker, more complete markets; and, because a bank can be funded mostly from debt (viz., deposits), a bank could We also explored the possibility of measuring “run behavior” through an institution’s use of public liquidity programs on September 15, 2008. How- ever, of those programs available to U.S. financial institutions, only one was (a) operating and accessible on that date and (b) has data publicly available on its use by individual institutions: the Primary Dealer Credit Facility. This program offered liquidity to primary dealers (six, specifically, on that day), who in turn provided liquidity to non-primary dealers through the repo market. See Primary Dealer Credit Facility (PDCF), BD. OF GOVERNORS FED. RES. SYS. (Mar. 18, 2020), https://www.federalreserve.gov/regreform/re- form-pdcf.htm [https://perma.cc/73X2-RB58]; Transaction Data, Primary Dealer Credit Facility (“PDCF”), BD. OF GOVERNORS FED. RES. SYS., https://www.federalreserve.gov/regreform/reform-pdcf.htm [https://perma.cc/73X2-RB58]. As such, the direct exposures of those pri- mary dealers provides little meaningful information on their own proximate liquidity needs. Detailed transaction information on discount window lend- ing, meanwhile, is unavailable publicly prior to 2010. See Discount Window Lending, BD. OF GOVERNORS FED. RES. SYS. (Apr. 1, 2020), https://www.fed- eralreserve.gov/regreform/discount-window.htm [https://perma.cc/MD9R- P6K8]. 101 See, e.g., GORTON, supra note 27, at 9 (“A financial crisis in its pure form is an exit from bank debt, a bank run.”). 102 See Peipei Wang & Ramaprasad Bhar, Information Content in CDS Spreads for Equity Returns, 30 J. INT’L FIN. MKTS., INSTS., & MONEY 55, 57 (2014); CHRISTOPHER L. CULP ET AL., CREDIT DEFAULT SWAPS 158 (2018). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 607 theoretically experience a fire-sale in equity markets without experiencing similar strain in debt markets.103 Second, we use a panel fixed-effects regression approach for our baseline model, with standard errors clustered by in- stitution, to capture changes in the trajectory of share prices and CDS spreads in response to the Lehman failure.104 This dynamic model utilizes daily market data to estimate the im- pact of Lehman’s bankruptcy. The regression holds constant firm-specific outcomes and time-specific outcomes—that is, any trends associated with a particular firm or day before the Lehman run occurred. Keeping these trends “fixed” adds con- fidence that our model is appropriately attributing the 103 We can imagine one scenario that might practically fit this descrip- tion: a preemptive guarantee, perhaps by a deposit insurer, to leave cus- tomer deposits intact. Such a planned resolution would reflect significant financial strain, and likely would even trigger payment on a CDS contract— but it would not fit the traditional definition of a run, since creditors would not be withdrawing any funding, and direct cost of debt measures (e.g., cost of deposits) might not reflect the strain at all. 104 Our specification comes from Daron Acemoglu et al., Women, War, and Wages: The Effect of Female Labor Supply on the Wage Structure at Midcentury, 112 J. POL. ECON. 497 (2004), and employs clustered standard errors. Clustering standard errors is a now commonplace technique used to address potential serial correlations among results. See A. Colin Cameron & Douglas L. Miller, A Practitioner’s Guide to Cluster-Robust Inference, 50 J. HUM. RES. 317 (2015). Clustering shrinks the effective size of our sample, reducing it from several hundred observations (the share price and CDS spread of each institution, on each day) to several dozen (several weeks of daily share prices and CDS spreads, for each institution). Given the normal- ization of share prices in our sample, and the uniform application of the Lehman shock across our entire population, there is an argument against using clustering. See Alberto Abadie et al., When Should You Adjust Stand- ard Errors for Clustering? (Nat’l Bureau Econ. Research, Working Paper No. 24003, 2017), https://www.nber.org/papers/w24003 [https://perma.cc/H787-L6SM]. Notably, we also conducted our analysis of U.S. institutions without clustering and, separately, with a different clus- tering method (bootstrapped clustered standard errors). Our results are di- rectionally identical and robust under all three specifications, and are avail- able on request. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 608 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 observed variation during our time window to the Lehman bankruptcy.105 𝑦𝑖,𝑡 = 𝛼𝑖 + 𝛼𝑡 + 𝛼𝑖 ∙ 𝑡 + 𝛾 ∙ 𝑑𝐿𝐸𝐻 + 𝜙 ∙ 𝑑𝐿𝐸𝐻 ∙ 𝑚𝑖 + 𝜀𝑖,𝑡 On the left-hand side, 𝑦𝑖,𝑡 is the share price, five-year CDS spread, or one-year CDS spread of financial institution i on day t, normalized by its value on the first trading day of Sep- tember 2008. Our window for this data runs from September 1 to 19, capturing the rest of the trading week after the Leh- man filing and the two weeks beforehand. On the right-hand side of the regression specification, 𝛼𝑖 captures firm-specific characteristics; 𝛼𝑡 captures aggregate time effects; 𝛼𝑖 ∙ 𝑡 cap- tures firm-specific time trends; and 𝑑𝐿𝐸𝐻 is a dummy variable for the period following September 15, 2008, the day when Lehman collapsed. The dummy variable equals zero on the days prior to September 15, 2008, and switches to one on that day and afterward. The coefficient of interest is 𝜙, which cor- responds to the interaction term between the Lehman failure dummy and the regulatory ratio of interest at financial insti- tution i prior to Lehman’s failure, captured by 𝑚𝑖. 106 105 Studies looking at similar phenomena sometimes employ an abnor- mal return event study, which looks at the deviation of a firm’s equity re- turns from an underlying normal return (a “normal” return is usually de- fined as the return on a broad-based index, like the S&P 500). For a detailed description of event study methodologies, see Norman Strong, Modelling Abnormal Returns: A Review Article, 19 J. BUS. FIN. & ACCT. 533 (1992). An abnormal return event study is, in a sense, a narrower specification of the panel regression we use. However, our specification has two advantages. First, by focusing on interaction effects, it sheds light on changes in the first derivative of equity returns, showing how trends in those returns changed following the Lehman filing. Second, it avoids the challenge of fixing a ref- erence return for the underlying market, which was affected substantially itself by the performance of the financial institution returns of interest. See infra notes 4, 43 and accompanying text. 106 We also tested a related regression specification that employs daily changes, as opposed to cumulative changes, of share prices and CDS spreads. In that specification, the dummy variable equals one1 on Septem- ber 15, 2008, and zero otherwise. The results are qualitatively identical and are available on request. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 609 Third, we test a broad set of explanatory variables that could have impacted investor behavior prior to Lehman’s col- lapse, using balance sheet measures from the most recent quarterly filings prior to the Lehman bankruptcy, or more re- cent market information, as available.107 It is worth noting that our core results, described below, are visible even without this panel fixed-effects regression ap- proach or variations thereof. Appendix G contains a series of bivariate correlation charts, with the cumulative change in share prices on the vertical axis and the explanatory variable on the horizontal axis.108 The fitted lines in these charts show that, counterintuitively, banks with greater balance sheet li- quidity and regulatory capital experienced more funding strain, not less, and that share price correlation with Lehman was intuitively associated with greater funding strain. For capital, we first test the Tier 1 Capital ratio as reported (“AR T1”) by the firms listed in Appendix D (AR T1/RWA).109 These numbers have an important caveat: in fall 2008, large U.S. institutions were still reporting figures under the older (and less granular) Basel I accords, rather than Basel II.110 Because public accounting and regulatory disclosures do not correspond to the Basel II risk-weighting categories, we can- not reconstruct the Basel II capital denominator at the time, and are thus limited to the 2008 risk-weighted denominator. However, we can and do construct the Basel II Tier 1 107 See infra Table 1 for a summary of measures used and infra Appen- dix B for descriptive statistics. 108 See infra Appendix G. 109 See infra Appendix D. 110 U.S. regulators finalized the Basel II capital rules for advanced ap- proach banks in November 2007 and required a minimum of four quarters of “parallel run,” in which institutions would calculate both Basel I and Ba- sel II capital ratios, but would only report Basel I figures. See Risk-Based Capital Standards: Advanced Capital Adequacy Framework — Basel II 72 Fed. Reg. 69,288, 69,301–02 (Dec. 7, 2007) (codified at 12 C.F.R. pts 3, 208, 225, 325, 559–60, 563, and 567). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 610 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 numerator, reflecting the then-imminent regulatory require- ment for U.S. banking organizations.111 Next, we run a series of tests to examine how investors weighed the regulatory capital measures available to them at the time of the Lehman filing. To isolate the Basel II numera- tor, we test a ratio of Basel II Tier 1 capital (“B2 T1”) over total assets (“TA”) (B2 T1/TA). To isolate the Basel I denominator, we test a ratio of common equity (“CE”) over 2008 risk- weighted assets (CE/RWA). To assess whether the new Basel III measures better pre- dict investor behavior, we also calculate and test a proxy for the Basel III numerator measures: Basel III Tier I capital (“B3 T1”) and Basel III CET1 (“CET1”).112 We test each proxy over 2008 risk-weighted assets (B3 T1/RWA, CET1/RWA) and total assets (B3 T1/TA and CET1/TA).113 Finally, we leave the Basel requirements behind entirely and test a simple leverage measure (CE/TA). For liquidity, as discussed above, pre-Lehman investors had access to roughly a year’s worth of new FASB fair value accounting measures for most institutions.114 To examine how relevant these measures were to investor behavior, we con- struct and test a Level 1 asset ratio (L1/TA). To see if investor behavior reflected more conventional indicators of liquidity, we also test holdings of cash and cash equivalents (Cash + Equivalents/TA). For funding fragility, we test each institution’s overall re- liance on short-term wholesale funding (“STWF”) a measure of credit that investors can withdraw on short notice, which is 111 Notably, the Basel II Tier 1 definition was almost identical to the Basel I Tier 1 definition, adding only non-cumulative perpetual preferred stock to Basel I’s paid-up share capital/common stock and disclosed re- serves. Compare BASEL COMM., BASEL II, supra note 51, at 244–45, with BA- SEL COMM. ON BANKING SUPERVISION, INTERNATIONAL CONVERGENCE OF CAP- ITAL MEASUREMENT AND CAPITAL STANDARDS 14 (1988). 112 See infra Appendix C. 113 It is not possible to reconstruct a Basel III risk-weighted denomina- tor proxy using data publicly available in 2008. 114 See supra note 62 and accompanying text. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 611 now addressed as part of Basel III (STWF/TA).115 In theory, however, large redemptions of STWF do not endanger a bank with enough liquid assets to cover those redemptions. To test whether this interaction of funding and liquidity predicted in- vestor behavior, we also test ratios of liquid assets to STWF (L1/STWF and Cash + Equivalents/STWF), expecting less run exposure at an institution with a ratio near or above one.116 We also investigate a market-based measure available to all investors and the general public at the time of the crisis: correlation in equity returns. We construct a proxy (hereinaf- ter, the “Lehman Correlation”) by calculating the correlation between each bank’s day-over-day change in share price and the equivalent changes for Lehman over a long window pre- ceding the Lehman bankruptcy filing.117 A higher correlation suggests that market shocks affect the share prices of two in- stitutions in similar sign and magnitude. If the correlation has high explanatory power, it suggests that investors may have relied more heavily on high-level market proxies—rather than balance sheet measures—to assess an institution’s exposure to the Lehman shock. Finally, to account for the possibility that investors acted on private information about Lehman itself, we assess the di- rect exposure of each institution in our sample to Lehman. Us- ing information from the Lehman U.S. bankruptcy trustee, we aggregate the total amount awarded on claims either held by or transferred from an institution in our sample (or one of its subsidiaries or affiliates). At best, this measure is a loose proxy of actual counterparty exposures, or the actual amounts 115 See BASEL COMM., NET STABLE FUNDING RATIO, supra note 61, at 1. 116 These measures follow the broad form of the working capital or “current” ratio (current assets/current liabilities); however, those defini- tions cover maturities of one year or less. See Fin. Accounting Standards Bd., Debt (Topic 470): Overall (2016). 117 The specific formula used to calculate the daily change in the share price is log(pt /(pt − 1)), applied to end-of-day share prices from January 2007 through July 2008. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 612 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Lehman owed to other parties at the time of its failure.118 However, given the statutory penalties for misstating claims in the liquidation of a securities dealer,119 this data may be the best available measure of relative counterparty exposures to Lehman at the time of insolvency. 118 The reasons for this looseness include that the proxy (a) does not distinguish between claims payable to the institution or its transferee and trustee or custody claims payable to one of the institution’s customers; (b) potentially double-counts claims transferred from one in-sample institution to another; and (c) does not reflect claims on non-U.S. Lehman entities. The greatest shortcoming, however, is that our proxy is based on the total amount awarded (which could understate true exposures) rather than the total amount claimed (which could overstate true exposures). Lehman’s Chapter 11 voluntary petition suggests that initial claims may have even less of a relationship to investor behavior than the ultimate award figures we used. Lehman’s largest unsecured creditor in that petition was Citibank, N.A., with an astonishing $138 billion in exposure—more than eleven times the second-largest unsecured creditor. Voluntary Petition at Schedule 1, In re Lehman Bros. Holdings Inc., 404 B.R. 752 (Bankr. S.D.N.Y. 2009) (No. 08-13555), 2008 WL 4200597. Its sixth-largest unsecured creditor was also a branch of Citibank, N.A., with another $275 million in exposures. Id. Based on call report filings, these claims were 11.9% of Citibank, N.A.’s un- weighted assets, and 147% of its reported Basel II Tier 1 capital. See Citi- bank N.A., Call Report, at 10, 57 (June 30, 2009). This initial petition only included one other bank from our U.S. sample and four other banks from the separate non-U.S. Global Systemically Important Bank (“G-SIB”) sam- ple, and it excluded other in-sample banks that ultimately filed award claims. See Voluntary Petition, supra note 118, at Schedule 1. Based only on the information in this petition, we would expect Citibank to experience a faster pace of withdrawals than any other institution by several orders of magnitude. Instead, the fastest pace was at institutions like Goldman Sachs and Morgan Stanley, who were not listed among Lehman’s largest unse- cured creditors. 119 Lehman bankruptcy proceedings occurred under the Securities In- vestor Protection Act, which bars false statements of account and acts of fraudulent conversion. See In re Lehman Bros. Inc., No. 08-01420JMPSIPA, 2008 WL 5423214 (Bankr. S.D.N.Y. Nov. 26, 2008); 15 U.S.C. §§ 78jjj(c)(1)(C)(ii), (c)(2) (2018). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 613 Table 1: Explanatory (RHS) Variables Under Consid- eration RWA Denominator “Capital Ratios” TA Denominator “Leverage Ratios” STWF Denominator and Various Proxies CE / RWA CE / TA Level 1 / STWF AR Tier 1 / RWA B2 Tier 1 / TA Cash / STWF B3 Tier 1 / RWA B3 Tier 1 / TA (“sim- ple leverage ratio”) Lehman Correlation B3 CET1 / RWA B3 CET 1 / TA Lehman Claims Level 1 / TA (“liquid- ity proxy”) Cash / TA (“liquidity proxy”) STWF / TA For our initial sample, we include the twenty-seven largest U.S. banking institutions by total consolidated assets as of September 2008, as well as those investment banks that later converted to bank holding companies (which were then receiv- ing public assistance, and are now subject to the same post- crisis prudential reforms as the other twenty-seven institu- tions in our sample120).121 120 See Sorkin, infra note 177. 121 Where “total assets” are indicated above, for firms filing the Y-9C we specifically use total consolidated assets. Note, however, that we exclude Charles Schwab and E*TRADE Financial, which are both savings & loan holding companies, in the baseline panel regressions. See The Charles Schwab Corporation, FED. FIN. INSTS. EXAMINATION COUNCIL (2019), https://www.ffiec.gov/nicpubweb/nicweb/Institution- Profile.aspx?parID_Rssd=1026632&parDT_END=99991231 [https://perma.cc/HMD6-C2J4]; E*TRADE Financial Corporation, FED. FIN. INSTS. EXAMINATION COUNCIL (2019), https://www.ffiec.gov/nicpubweb/nic- web/Institution- Profile.aspx?parID_Rssd=3412583&parDT_END=99991231 [https://perma.cc/TN6D-76NN]. We omitted from our sample twenty-two non-U.S. G-SIBs, as identified by the Financial Stability Board. See FIN. STABILITY BD., 2017 LIST OF GLOBALLY SYSTEMICALLY IMPORTANT BANKS (G- 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 614 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 B. Summary of Results The full results of our analysis can be found in Appendices E and F.122 Here, we briefly review our results for each explan- atory variable. In short, only Basel II Tier 1 leverage and the Lehman correlation predicted run behavior in the intuitive di- rection across all markets. By contrast, balance sheet liquidity and regulatory capital either had no statistically significant relationship to run behavior, or an unexpected relationship— that is, the more an institution had, the more exposed it was to the run. 1. Simple Panel Fixed-Effects Regression Results For capital, we would expect the estimated coefficients to be significant, and to imply that higher capital levels are con- sistently associated with higher share prices and lower CDS spreads (i.e., positive and negative, respectively). For no measure of regulatory capital was this true. The estimated coefficient on reported Tier 1 capital (AR T1/RWA) was found to be significant only for one-year CDS spreads. Even in this limited result, the coefficient points in the wrong direction—suggesting that institutions that were better capitalized for regulatory purposes were more exposed to run behavior. For every other capital measure with risk- weighted assets in the denominator, the results are even starker: no risk-based measure was associated with run expo- sure in equity markets, and in CDS markets, they pointed to- wards greater exposure to the Lehman run. Abandoning risk-weighting and adopting the post-crisis definition of capital produces very different results, but only SIBS) (2017), https://www.fsb.org/wp-content/uploads/P211117-1.pdf [https://perma.cc/47FA-RXTW]. Data in this non-U.S. sample is highly het- erogeneous, representing a wide set of jurisdictions, regulatory mecha- nisms, and policy decisions in the implementation of the various Basel ac- cords. As such, it cannot adequately support any broad conclusions about non-U.S. jurisdictions or institutions. 122 See infra Appendices E–F. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 615 in CDS markets. Simple leverage (CE/TA) and Basel II Tier 1 leverage (B2 T1/TA) are both intuitive and statistically signif- icant for one-year and five-year spreads, implying that insti- tutions with less reliance on debt were less exposed to the Leh- man run. When used with an unweighted capital denominator (TA), the new Basel III numerators (B3 T1, B3 CET1) exhib- ited the same relationship. Of all four measures, however, only Basel II Tier 1 leverage had predictive power in equity markets. Together, these results suggest the challenges associated with Basel I (several of which had been noted before the cri- sis123) may have resided in its risk-weighting system, rather than its definition of capital.124 Bolstering this suggestion, simple leverage (CE/TA) and the reported Tier 1 capital ratio (AR T1/RWA) were negatively correlated before the Lehman filing. That is, the higher an institution’s Basel I ratio, the more leveraged it was likely to be.125 Since Basel III was aimed at making regulatory capital more equity-like, we would expect the Basel III definitions of capital to be associated with lower run exposure. However, the Basel III numerator proxies (B3 T1/TA and CET1/TA) instead performed worse than reported Basel I measures in 123 See Mark E. Van Der Weide & Jeffrey Y. Zhang, Bank Capital Re- quirements after the Financial Crisis, in THE OXFORD HANDBOOK OF BANK- ING 707, 708 (Allen N. Berger et al. eds., 3d ed. 2019). 124 For more on the potential challenges associated with risk- weighting, see Andrew G. Haldane, Exec. Dir., Bank of Eng,, Speech at the Federal Reserve Bank of Atlanta: Constraining Discretion in Bank Regula- tion 4 (Apr. 9, 2013), https://www.bankofengland.co.uk/-/me- dia/boe/files/paper/2013/constraining-discretion-in-bank-regula- tion.pdf?la=en&hash=46E1F9BF24E99D85DBC28746A190ABBC5DA785 E2, [https://perma.cc/NF7D-RQ2Q] (“At least at an aggregate level, bank risk weights appear to have borne, at best, a tenuous relationship with risk. At worst, they were a contrarian indicator.”); John Vickers, Keynote Ad- dress at the International Conference of Banking Supervisors: Safer, But Not Safe Enough 3 (Nov. 29, 2018), https://www.bis.org/bcbs/events/icbs20/vickers.pdf [https://perma.cc/R9HK- EB3B]. 125 See infra Table 2. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 616 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 explaining investor behavior around Lehman. A separate test shows why: for institutions in our U.S. sample, the Basel III Tier 1 ratio (B3 T1/RWA) is even more negatively correlated with simple leverage (CE/TA) today than the Basel I Tier 1 ratio was before Lehman’s failure.126 The more regulatory cap- ital an institution has under Basel III, the greater its reliance on debt. Table 2: Correlation Between Basel Capital Ratios and Simple Leverage, 2008 and 2018 2008Q2 (CE/TA, AR T1/RWA) 2018Q2 (CE/TA, B3 T1/RWA) U.S. G-SIB -0.047 (n=24) -0.270 (n=27) Non-U.S. G-SIB 0.387 (n=20) -0.161 (n=19) For balance sheet liquidity, the results are similarly coun- terintuitive. Neither liquidity proxy (L1/TA, Cash + Equiva- lents/TA) is consistently positive and significant across out- come variables. Instead, both measures are negative and statistically significant in equity markets—suggesting equity investors withdrew funding more quickly from institutions with more balance sheet liquidity. The lone exception (Cash + Equivalents/TA, in one-year CDS markets) suggests a greater sensitivity to the most liquid assets among investors. For funding fragility, a greater reliance on short-term wholesale funding was associated with a lower share price, but not with wider one-year or five-year CDS spreads. As far as the interaction between liquidity and funding fragility, re- sults are mixed. Only one proxy explains run exposure as ex- pected, and in only one market (Cash + Equivalents/STWF, in one-year CDS spreads). By contrast, our other proxy (Level 1/STWF) was associated with lower share prices in equity markets, and had no explanatory power in other markets. 126 See id. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 617 By contrast, simple return correlation is uniformly statis- tically significant at 𝛼 = 0.01 (***) across all markets, as is the Lehman claims proxy variable. 2. Compound Panel Fixed-Effects Regression Results Three measures are robust and directionally intuitive in every permutation of our U.S. sample: Basel II Tier 1 leverage (B2 T 1/TA), the Lehman correlation, and Lehman claims. Next, we test the overlap among these variables using a mod- ified regression specification: 𝑦𝑖,𝑡 = 𝛼𝑡 + 𝛼𝑖 ∙ 𝑡 + 𝛽1 ∙ 𝑚1,𝑖 + 𝛽2 ∙ 𝑚2,𝑖 + 𝛾 ∙ 𝑑𝐿𝐸𝐻 + 𝜙1 ∙ 𝑑𝐿𝐸𝐻 ∙ 𝑚1,𝑖 + 𝜙2 ∙ 𝑑𝐿𝐸𝐻 ∙ 𝑚2,𝑖 + 𝜀𝑖,𝑡 Several variables are identical to our prior specification, including 𝑦𝑖,𝑡 (normalized share price/CDS spread of institu- tion i on day t); 𝛼𝑡 (aggregate time effects); 𝛼𝑖 ∙ 𝑡 (institution- specific time trends); and 𝑑𝐿𝐸𝐻 (a dummy for September 15, 2008). However, we have removed 𝛼𝑖 (institution-specific char- acteristics) and replaced it with 𝑚1,𝑖 and 𝑚2,𝑖, two specific reg- ulatory ratios of interest for institution 𝑖. 𝜙1 and 𝜙2 are the coefficients of interest, capturing the interaction between the Lehman failure dummy and 𝑚1,𝑖 and 𝑚2,𝑖, respectively. If 𝜙1 remains statistically significant after controlling for 𝜙2, it sug- gests that 𝑚1,𝑖 has explanatory power above and beyond that of 𝑚2,𝑖 (and vice versa). Our results are a powerful endorsement of the explanatory power of return correlation. When controlling for the Lehman correlation, both Lehman claims and Basel II Tier 1 leverage (B2 T1/TA) lost significance, with only one exception: Lehman claims retained weak significance in equity markets. This sug- gests that the Lehman correlation captures the variation in run exposures suggested by the two other variables, except in equity markets, where counterparty information still retained some independent explanatory power. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 618 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 3. Overall Results Of all fifteen right-hand side variables, only three were sta- tistically significant and directionally intuitive in both CDS and equity markets: one leverage measure (B2 T1/TA), one correlation measure (the Lehman correlation), and one coun- terparty exposure measure (Lehman claims). Of these, the correlation measure was the most powerfully predictive of run exposure. • For capital, every measure with 2008 risk-weighted assets in the denominator showed either statisti- cally insignificant or counterintuitive results: The higher the ratio, the more quickly investors reduced their equity exposure, and the higher the implied probability of default. Focusing only on the numer- ator, the most stringent Basel III definition of capi- tal (CET1) had less explanatory power than its an- tecedent Basel II measure. • Similarly, our balance sheet liquidity proxies were not intuitively predictive of investor behavior. Our proxies behaved just as the regulatory capital vari- ables did: The higher the levels, the greater an in- stitution’s exposure to the Lehman run. • Our measure of funding fragility only had explanatory power in equity markets, not CDS markets. • Our correlation measure performed as well as—or better than—our proxy for private information about Lehman’s direct counterparty exposures. Finally, as an intuitive check, we plot a subset of explana- tory variables against cumulative changes in share price over our post-crisis window. These simple results are consistent with our analysis: correlation, claims, and leverage explain a substantial amount of the variation observed in run expo- sures, in the intuitive direction; CET1 risk-based capital also explains a substantial amount of the variation in run expo- sures, but in the counterintuitive direction; and liquidity 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 619 results are still highly clustered. In short, the institutions most exposed to the Lehman run were not those with the low- est levels of capital or balance sheet liquidity. IV. POLICY IMPLICATIONS Our results target a specific question—namely, what actu- ally mattered during the Lehman market panic? How did the “fire” spread once it started? To be sure, our results are limited by the nature and context of the Lehman bankruptcy, and they do not speak to the ex ante role of capital or liquidity in preventing a crisis, nor to their ex post role in hastening re- covery from one. The existing literature is clear on both those points, and nothing in our results qualifies or contradicts them. Instead, our results address the window after a fire has started and is about to spread. This window is small, but it has outsized consequences for the path a financial crisis will take, as well as the public and private costs of that crisis. As such, our results bear directly on the capital versus contagion debate; shed light on the causes of systemic bank runs; and suggest critical policy steps that could be taken to address run vulnerability at large financial institutions. A. Bank Runs Aren’t (Always) About Cash Balance sheet liquidity did little to stop a run on institu- tions in the wake of Lehman’s filing. On the contrary, in sev- eral markets and by several measures, institutions with a greater share of liquid assets on their balance sheets experi- enced faster outflows. This result is unexpected, and despite two strong counterarguments, it is difficult to dismiss en- tirely. The first counterargument would limit our results to insti- tutions with low levels of balance sheet liquidity. Financial in- stitutions had relatively few liquid assets entering 2008,127 127 In 2008, across the entire banking sector highly liquid assets were less than 15% of total assets, compared to approximately 25% today. See BD. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 620 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 and by the time Lehman filed for bankruptcy, markets had already experienced months of volatility.128 Investors might have (quite reasonably) assumed that post-Lehman redemp- tions would dwarf whatever liquid assets banks had on hand, and accordingly discounted the most recent balance sheet li- quidity figures they had available to them entirely. However, the market had another critical source of liquid- ity at the time Lehman fell: the United States government, which followed Bagehot’s dictum and was operating four rele- vant liquidity facilities in addition to the discount window in September 2008.129 Five months earlier, when Bear Stearns failed, Lehman itself had borrowed an average of $2.2 billion a day (and $15.2 billion total) from just one of those facili- ties.130 On September 14, the Board publicly announced a “sig- nificant broadening” of the collateral eligible to be posted at two of their liquidity facilities, and increased the frequency and quantity of funds institutions could borrow from those fa- cilities.131 The day it filed, Lehman took out another $28 bil- lion in overnight loans from the Federal Reserve, and other primary dealers took out $13.25 billion.132 Thus, although banks’ balance sheet assets may have been highly illiquid, the large banks themselves were seemingly awash in liquidity. While the Federal Reserve did not disclose full and precise fig- ures contemporaneously, investors would have known this OF GOVERNORS FED. RESERVE SYS., FEDERAL RESERVE SUPERVISION AND REG- ULATION REPORT 6 fig.7 (2018). 128 See VIX Index Historical Data, CHI. BD. OPTIONS EXCH., http://www.cboe.com/products/vix-index-volatility/vix-options-and-fu- tures/vix-index/vix-historical-data [https://perma.cc/BY4Z-F5MK]. 129 See SCOTT, supra note 40, at 75. 130 See Transaction Data, supra note 100. 131 Press Release, Fed. Reserve Bd. of Governors, Federal Reserve Board Announces Several Initiatives to Provide Additional Support to Fi- nancial Markets, Including Enhancements to its Existing Liquidity Facili- ties (Sept. 14, 2008), https://www.federalreserve.gov/newsevents/pressre- leases/monetary20080914a.htm [https://perma.cc/NJ8E-EN62]. 132 See Transaction Data, supra note 100. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 621 government provided liquidity was available for large finan- cial institutions’ use.133 This high level of public support creates a second counter- argument: Investors were indifferent to balance sheet liquid- ity issues because institutions could access potentially unlim- ited amounts of government cash. In other words, instead of holding only at low levels of liquidity, our results would only hold at high levels of liquidity. However, this counterargu- ment would imply two inconsistent facts: that investors (a) ran on large financial institutions because the government was no longer guaranteed to support them, and (b) disre- garded the balance sheets of those institutions because the government was guaranteed to support them. Even if those facts were both true, then under contagion theory there would have been no reason for investors to run at all. That is, if the liquidity needs of those large financial institutions were sure to be met, then even Lehman Brothers would still have been standing.134 At a minimum, then, the relationship between liquidity and run behavior is more complex than contagion theory might suggest. During the Lehman run, any number of other factors may have affected that relationship—from doubts 133 The Federal Reserve disclosed details on the entities that received assistance from crisis-era special facilities in December 2010. See id. In 2012, it began regularly publishing details regarding discount window lend- ing activity, roughly two years after the activity takes place. See Credit and Liquidity Programs and the Balance Sheet, BD. OF GOVERNORS FED. RES. SYS. (Aug. 13, 2019), https://www.federalreserve.gov/monetarypol- icy/bst_lendingdepository.htm [https://perma.cc/GJ44-56DA]; Transaction Data, Discount Window Lending, BD. OF GOVERNORS FED. RES. SYS., https://www.federalreserve.gov/regreform/discount-window.htm [https://perma.cc/YPW7-RNUM]. 134 There is, of course, a counterfactual argument as well that the runs on Lehman and other financial institutions would have been much worse without access to government-offered liquidity. No natural experiment ex- ists to test this argument, however we do find it plausible given the combi- nation of our results and the sheer volume of liquidity support that was made available at the time of the crisis. Thus, we cannot reject the null hy- pothesis that liquidity levels play no role in run behavior. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 622 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 about the fidelity of bank balance sheets to uncertainty sur- rounding the nature of public support for financial markets. To the extent those other factors matter, however, it is diffi- cult to argue that the Lehman run was “all about cash.” B. Leverage—But Not Regulatory Capital—Can Predict How a Run Spreads By contrast, simple balance sheet measures of capital per- formed remarkably well in explaining an institution’s suscep- tibility to a run. Basel II Tier 1 leverage was robust across tests, even more consistently than simple leverage (CE/TA), suggesting that investors paid attention to both common eq- uity and loan-loss reserves, despite well-documented issues with pre-crisis reserve rules.135 However, regulatory capital measures led to two disconcerting results—the first regarding risk-weighting and the capital denominator, and the second regarding reforms agreed under Basel III. First, we found no specification, in any sample or in any market, where a higher risk-weighted capital ratio corre- sponded to less run exposure. In fact, the opposite was true; investors generally withdrew funding faster from institutions with more capital as a proportion of risk-weighted assets. This result made little sense, until we found that the Tier 1 ratio as reported in 2008 (with a Basel I risk-weighted denomina- tor) was negatively associated with common-equity leverage (with an unweighted denominator).136 In other words, the bet- ter-capitalized an institution was under outstanding regula- tory measures in 2008, the more highly leveraged it was. This result suggests a surprising regulatory design issue, even after accounting for the well-known shortcomings in pre- 135 See RAJ GNANARAJAH, CONG. RESEARCH SERV., R45339, BANKING: CURRENT EXPECTED CREDIT LOSS (CECL) 1–2 (2018); Eugene A. Ludwig & Paul A. Volcker, Banks Need Long-Term Rainy Day Funds, WALL ST. J. (Nov. 16, 2012), https://www.wsj.com/arti- cles/SB10001424127887324556304578120721147710286 [https://perma.cc/HS28-9S2T]. 136 See infra Table 2. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 623 crisis risk-weighting rules. Leverage limits—which have been a part of U.S. banking regulation since 1981, before the first Basel accords—are intended to complement risk-based capital requirements.137 Leverage ratios impose the same “regulatory capital charge” for every asset,138 giving institutions an incen- tive to hold riskier assets (which pay a higher return). Risk- weighting limits a bank’s ability to act on that incentive, but even if risk weights are incorrect, a leverage requirement can still serve as a “backstop,” and keep a bank from assuming too high an overall level of debt. In short, an institution with a higher risk-weighted capital ratio should hold more equity— not less.139 Basel I and II did not include a leverage minimum, but Ba- sel III did—and made extensive changes to risk-weighting— to serve as precisely this kind of backstop.140 As such, we 137 See Michael Brei & Leonardo Gambacorta, The Leverage Ratio Over the Cycle 2 (Bank for Int’l Settlements, Working Paper No. 471, 2014), https://www.bis.org/publ/work471.pdf [https://perma.cc/Z6QG-MQEY]. 138 See Van Der Weide & Zhang, supra note 123, at 726. 139 See BERNANKE ET AL., supra note 7, at 25 (describing capital as “the flip side of leverage; the more an institution relies on borrowing, the lower its capital levels, and the greater its exposure to shocks”). 140 See BANK INT’L SETTLEMENTS, BASEL III LEVERAGE RATIO FRAME- WORK 1 (2017) (“The leverage ratio is also intended to reinforce the risk- based capital requirements with a simple, non-risk-based ‘backstop.’”); Jaime Caruana, Gen. Manager, Bank for Int’l Settlements, Promontory An- nual Lecture: Financial Regulation, Complexity and Innovation 1 (June 4, 2014), https://www.bis.org/speeches/sp140604.htm [https://perma.cc/E8YK-GLKB] (describing “the leverage ratio as a backstop to the risk-weighted measure”); Daniel K. Tarullo, Member, Fed. Reserve Bd. of Governors, Opening Statement to Meeting of Board of Governors (Apr. 8, 2014), https://www.federalre- serve.gov/newsevents/press/bcreg/bcreg20140408a-tarullo-statement.htm [https://perma.cc/SEV6-9EFQ] (saying the “leverage ratio serves as a criti- cal backstop to the risk-based capital requirements,” immediately prior to the Board of Governors vote on the enhanced supplementary leverage ratio); Mervyn King, Governor, Bank of Eng., Banking—From Bagehot to Basel, and Back Again 6–7 (Oct. 20, 2010), https://www.bis.org/re- view/r101028a.pdf [https://perma.cc/T2XB-VLXM] (noting that “the regula- tory framework needs to contain elements that are robust with respect to changes in the appropriate risk weights, and that is why the Bank of 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 624 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 would expect Basel III to change these fire-sale relationships, so that an institution holding higher levels of regulatory cap- ital would correspond with (a) lower leverage and (b) less run vulnerability. We cannot fully test Basel III capital require- ments against run vulnerability, since it is not possible to cre- ate a rough proxy for the Basel III denominator using public data. However, Basel III did not change the relationship be- tween regulatory capital and simple leverage. In fact, that re- lationship has gotten even stronger: at U.S. banks, simple lev- erage (CE/TA) and the reported Tier 1 capital ratio (T1/RWA) are even more negatively correlated today than they were in 2008—and now, unlike then, non-U.S. G-SIBs display the same negative correlation.141 Our findings also suggest these issues may extend to the Basel III numerator. Again, Basel III redefined Tier 1 Capital as having two components: Common Equity Tier 1 and Addi- tional Tier 1. Common Equity Tier 1 was intended to closely reflect common equity. However, even without a risk- weighted denominator, our two Basel III numerator proxies failed to predict consistently post-Lehman run behavior across markets. Higher Basel III Tier 1 and CET1 levels, in other words, were not always associated with a lower risk of a run in our 2008 sample, and in equity markets, their relation- ship to run exposure was weaker than that observed of the equivalent Basel II measures. England advocated a simple leverage ratio as a key backstop to capital re- quirements”). 141 Critically, no U.S. advanced approach institutions are in parallel run today. Recent empirical work on loan-levels has found evidence of po- tential manipulation in risk-weighting. See Matthew C. Plosser & João A.C. Santos, Banks’ Incentives and the Quality of Internal Risk Models, 31 REV. FIN. STUD. 2080 (2018) (finding a downward bias in risk-rating at lower- capital banks after reviewing a sample of post-crisis Shared National Cred- its); Giovanni Ferri & Valerio Pesic, Bank Regulatory Arbitrage Via Risk Weighted Assets Dispersion, 33 J. FIN. STABILITY 331 (2017) (finding a larger degree of risk-weight manipulation at less capitalized banks in a sample of 239 institutions); see also Mike Mariathasan & Ourada Merrouche, The Ma- nipulation of Basel Risk-Weights, 23 J. FIN. INTERMEDIATION 300 (2014) (ex- amining risk-weight manipulation vis-a-vis Basel II requirements). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 625 C. Contagion Theory: Capital, Complexity, and Information Scarcity The policy changes made from Basel II to Basel III may shed light on the reasons for this discrepancy. The Basel III CET1 capital numerator includes the Basel II definition of Tier 1 capital, but adds retained earnings and accumulated other comprehensive income (i.e., unrealized gains) and sub- tracts three categories of intangible assets.142 In turn, Basel III Tier 1 capital includes the definition of CET1, but adds “additional Tier 1 capital,” Additional Tier 1 Capital under Basel III is defined by a list of fourteen criteria, several of which include sub-criteria.143 In one interpretation of our results, investors placed less faith in the specific measures Basel III added to regulatory capital (or, placed more faith in the measures Basel III stripped away). However, a simpler explanation could account for almost all of our capital-related findings, one that policy- makers have raised elsewhere: more complex measures might just matter less in the middle of a run.144 Contagion theorists claim that creditors become more in- formation-sensitive during a crisis, sparked by the possibility that others have private information about the value of their previously safe, cash-like assets.145 Our results suggest that instead—or, at least, as a corollary—time constraints and other transaction costs might place some limits on that 142 See infra Appendix A. 143 See infra Appendix H. 144 See, e.g., Andrew G. Haldane, Exec. Dir., Bank of Eng., Remarks at the American Economic Association: Capital Discipline (Jan. 9, 2011), https://www.bis.org/review/r110325a.pdf [https://perma.cc/UP73-EHZ2]; William Coen, Ring-Fencing’s Global Impact, BANKING PERSP., Summer 2018, at 36, 39 (“Simply put: One cannot and should not relentlessly pursue risk sensitivity as a goal in itself; it must be balanced with simplicity and comparability.”). 145 See Gary Gorton & Guillermo Ordoñez, Collateral Crises, 104 AM. ECON. REV. 343, 344 (2014); Tri Va Dang et al., Ignorance, Debt, and Finan- cial Crises 3–4 (Apr. 1, 2015) (unpublished manuscript) (on file with the Columbia Business Law Review). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 626 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 information-sensitivity. That is, if a measure is simple and in- formation-rich, investors may be more sensitive to changes in it. If a measure is too complex or opaque, or if it requires ad- ditional vetting to make it trustworthy, they may disregard it, or even take it as cause for concern. In this view, investors don’t just care about greater information during a run; they care about the efficiency of that information, as well as the cost of obtaining it, verifying it, and making it meaningful. D. Storms, Fires, and Correlation Channels Another result supports the idea that informational effi- ciency matters: the simplest explanatory variable in our sam- ple was also the most robust. Return correlation predicted run behavior as well as any balance sheet measure and better than our proxy for actual Lehman counterparty exposures. Correlation retained explanatory power even when holding leverage constant, and its significance persisted across sam- ples and markets. In other words, it matters that the system- wide run began at Lehman Brothers—and if it had begun at another institution, it would have happened differently. Based on the strength of this result, the analogy between contagious runs and “shocks” might be inapt, or at least in- complete. A shock implies an exogenous event that buffets firms equally, destroying some and preserving others accord- ing to their individual “shock absorbers” (like capital).146 A common shock can certainly hit one or more institutions sim- ultaneously, like the failure of a single counterparty, the 146 The definition of an exogenous shock in the economic literature is not fixed. See Panos Varangis et al., Exogenous Shocks in Low Income Coun- tries: Economic Policy Issues and the Role of the International Community 2–4 (Nov. 20, 2004) (unpublished manuscript) (on file with the Columbia Business Law Review). However, the analogy of capital to a “shock absorber” is still common. See, e.g., Paul J. Davies, This New Banking Shock Absorber Might Fail to Impress, WALL ST. J. (Dec. 29, 2015), https://www.wsj.com/ar- ticles/this-new-banking-shock-absorber-might-fail-to-impress-1451391539 [https://perma.cc/7588-6ET2]. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 627 downgrade of a common creditor, or a cybersecurity attack.147 But a shock need not be common to cause a contagious run; instead, all that’s needed is a conduit that carries investor con- cerns (and withdrawals) from one financial institution to an- other. A separate analogy clarifies this distinction. The “shock” analogy treats contagious runs like a storm hitting a neigh- borhood with little warning. Chance plays a role in the homes that are spared, along with factors that a homeowner can’t necessarily foresee or control, like poor weather-proofing or a weak foundation. Generally, though, the storm affects the en- tire neighborhood, and only the strongest houses survive. By contrast, our results suggest that a contagious run is more like a fire, which starts inside a single neighborhood home. Factors such as fire-proofing, sprinklers, and smoke alarms are all relevant to whether the blaze starts, but after it does, they are irrelevant to whether it consumes the neigh- borhood. The fire can move directly to adjacent houses, but proximity is hardly the only way it can spread; an updraft, flaming debris, or burning embers can carry the flames clear across town. Residents in neighboring properties would prob- ably flee their homes, but so would other residents who knew they were vulnerable—for example, if they knew they lived downwind from the blaze.148 Run behavior in 2008 displayed a similar dynamic. Leh- man Brothers (figuratively) caught fire the day it filed for 147 A dissent to the final report of the Financial Crisis Inquiry Com- mission highlighted an alleged common shock as a cause of the 2008 crisis. FIN. CRISIS INQUIRY COMM’N, THE FINANCIAL CRISIS INQUIRY REPORT 419 (2011) (dissenting statement of Vice Chairman Bill Thomas and Commis- sioners Keith Hennessy and Douglas Holtz-Eakin). The Commission major- ity declined to adopt this view. See id. at xviii–xx. 148 Several key figures from the financial crisis have recently embraced the same analogy. See BERNANKE ET AL., supra note 7, at 21, 112 (“Fire pre- vention had failed. Now the fate of the [financial] system would depend on fire-fighting. . . . [T]he U.S. economy and financial system today may be less prone to modest brush fires but more vulnerable to a major inferno if, de- spite updated and improved fire codes, a conflagration were to begin.”). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 628 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 bankruptcy, and investors looked hurriedly for who would burn next. Counterparty exposures were one way for with- drawals, collateral calls, haircuts, and other run-related be- havior to spread, but those exposures were mostly private in- formation and opaque to market participants. Instead, investors may have drawn conclusions about their vulnerabil- ity from the simplest available measure—that is, whether a particular financial institution had followed Lehman’s trajec- tory in the past. These “correlation channels” were reliable conduits for post-Lehman run risk. From an investor’s perspective, their role is as intuitive as Keynes’s “beauty contest”: if the market sees two banks as closely linked, then naturally, when the value of one falls, the value of the other will fall, too.149 Re- gardless of why the link between them exists; regardless of whether they actually owe money to each other or a common third party; regardless of whether their obligations are guar- anteed in bankruptcy; regardless of whether the government may lend them money, or even invest in them—regardless of all of this—if two banks are tied together, and one is faltering, what rational investor would remain exposed to the other? 149 Keynes developed his famous hypothetical to explain herd behavior in equity markets. In short: A newspaper contest requires contestants to pick out the six most attractive photos out of many. The winner is the en- trant whose list most closely resembles everyone else’s most popular selec- tions. See JOHN MAYNARD KEYNES, THE GENERAL THEORY OF EMPLOYMENT, INTEREST AND MONEY 147–64 (1936). What is a contestant’s optimal strat- egy? Per Keynes: It is not a case of choosing those [faces] which, to the best of one’s judgment, are really the prettiest, nor even those which average opinion genuinely thinks the prettiest. We have reached the third degree where we devote our intelli- gences to anticipating what average opinion expects the av- erage opinion to be. And there are some, I believe, who prac- tise the fourth, fifth and higher degrees. Id. at 156. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 629 V. IMPLICATIONS FOR REGULATORY DESIGN A. Supervisory Stress Tests The identification of correlation channels offers several im- plications for improving the design of our post-crisis regula- tory framework—the first of which pertains to government “stress tests” of financial institutions. At the height of the 2008-09 financial crisis, the Federal Reserve created the Su- pervisory Capital Assessment Program to estimate the poten- tial losses at large banks in the event economic and financial conditions worsened further.150 Since then, the Federal Re- serve moved to the current stress testing assessment—the Comprehensive Capital Analysis and Review—to evaluate whether the largest financial institutions with operations in the United States have sufficient capital to absorb future po- tential losses and continue to lend under stressed condi- tions.151 In this process, the Federal Reserve simulates mac- roeconomic scenarios like a recession in which GDP falls and the unemployment rate rises significantly. In the 2019 stress test cycle, for example, the Federal Reserve tested banks against a hypothetical global recession in which the unem- ployment rate in the United States rose to 10%.152 The stressed banks were required to show that they could continue to meet minimum capital requirements in the face of those hypothetical macroeconomic shocks.153 Our results strongly suggest that it is not enough to simply stress test financial institutions using broad-based aggregate 150 See Ben S. Bernanke, Chairman, Fed. Reserve Sys., Speech at the 2009 Financial Markets Conference in Jekyll Island, Georgia: The Supervi- sory Capital Assessment Program (May 11, 2009), https://www.federalre- serve.gov/newsevents/speech/bernanke20090511a.htm [https://perma.cc/8UZU-YMMD]. 151 See supra notes 71–73 and accompanying text. 152 See BD. OF GOVERNORS FED. RESERVE SYS., 2019 SUPERVISORY SCE- NARIOS FOR ANNUAL STRESS TESTS REQUIRED UNDER THE DODD-FRANK ACT STRESS TESTING RULES AND THE CAPITAL PLAN RULE 4–5 (2019). 153 See id. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 630 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 economic shocks, nor even by using direct or indirect counter- party exposures. A bank’s ability to survive a stress event de- pends, in part, on the specific nature and location of that event within the financial sector—especially since financial crises often start with an idiosyncratic stress at a specific institu- tion, rather than a broad shock to the system itself.154 A lack of widespread direct counterparty exposure to a failing bank forecloses one avenue towards financial strain, but it does not guarantee safety from a run. Correlations can change over time—they can either form or disappear, strengthen or weaken—with substantial impli- cations for contagious runs. In the weeks and months before the Lehman bankruptcy, the share prices of many large finan- cial institutions were already highly correlated.155 These cor- relations read like a topographical map for the near-term cri- sis that followed, despite the extent to which prices and CDS spreads splintered on the day of the Lehman filing.156 Alarm- ingly, the share prices of many of these institutions are even more highly correlated today, exceeding even their highest 2008 correlation with Lehman.157 154 See ADMATI & HELLWIG, supra note 49, at 75 (“[T]he question of whether banks should be allowed to fail rarely arises as a matter of princi- ple. Rather a particular bank is in trouble and the authorities must decide whether to let it go into bankruptcy or a similar process or to allow it to continue operating, possibly after an injection of public money.”). 155 See infra Table 3. 156 See supra note 6 and accompanying text. Notably, existing research suggests that market correlations increase during a financial crisis, making the divergent behavior of specific institutions’ equity and CDS values even more notable. See, e.g., Silvio Contessi et al., How Did the Financial Crisis Alter the Correlations of U.S. Yield Spreads? (Fed. Reserve Bank of St. Louis, Working Paper No. 2013-005D, 2014), https://files.stlou- isfed.org/files/htdocs/wp/2013/2013-005.pdf [https://perma.cc/HC8P- CWN5]. 157 Separately, we also examined two measures of overall financial sec- tor equity correlation. The first, share price synchronicity, is a fixture in development economics literature as a proxy for capital market thickness. See ROBERTO R. ROCHA ET AL., FINANCIAL ACCESS AND STABILITY: A ROAD MAP FOR THE MIDDLE EAST AND NORTH AFRICA 313–16 (2011), 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 631 Monitoring these channels at a high frequency—more than the annual frequency of supervisory stress tests—would offer a more precise insight into which institutions would be vul- nerable should a shock occur and spread. This would be akin to firefighters monitoring firebreaks in a neighborhood they serve; epidemiologists and public health experts monitoring the use of water sources in an area vulnerable to cholera; or cybersecurity experts monitoring data hubs in advance of a cyber attack. In all four cases, the information provided can help direct scarce public resources to where they can be most effective, when the potential impact is highest and the time for planning is lowest. http://documents.worldbank.org/cu- rated/en/343771468052798123/pdf/649370PUB0Fina00Box361550B00Pub- lic0.pdf [https://perma.cc/MUY4-SEBM]. One version of this measure re- gresses the return of stock 𝑖 at time 𝑡 on broad market returns at time t, using a market-wide index, and takes the resulting 𝑟2 value as a measure of stock co-movement. See Randall Morck et al., The Information Content of Stock Markets: Why Do Emerging Markets Have Synchronous Stock Price Movements?, 58 J. FIN. ECON. 215 (2000). We adapt this measure using a mix of KBW Nasdaq and Dow financial sector and banking indices, and find high synchronicity in the period immediately before the Lehman bank- ruptcy and low synchronicity immediately after, with synchronicity between those levels for the equivalent dates in 2018. However, a simpler measure— the cross-sectional standard deviation of daily changes in the share price of our in-sample institutions—reveals current co-movement at levels similar to that observed immediately before the financial crisis. Results of these analyses are available upon request; however, as discussed above, we be- lieve these aggregate measures fail to capture important information about share price correlation between specific institutions. See infra Table 3. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 632 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Table 3: Pairwise Correlation Coefficients – 2008 and 2018158 January 2007 - July 2008 JPM BAC C WFC GS MS DB LEH JPM 1.00 BAC 0.84 1.00 C 0.77 0.82 1.00 WFC 0.81 0.83 0.76 1.00 GS 0.74 0.69 0.76 0.69 1.00 MS 0.75 0.74 0.79 0.73 0.83 1.00 DB 0.68 0.64 0.70 0.63 0.72 0.68 1.00 LEH 0.68 0.70 0.73 0.67 0.80 0.79 0.63 1.00 January 2017 - July 2018 JPM BAC C WFC GS MS DB JPM 1.00 BAC 0.92 1.00 C 0.87 0.84 1.00 WFC 0.76 0.73 0.71 1.00 GS 0.81 0.79 0.77 0.63 1.00 MS 0.86 0.86 0.80 0.68 0.82 1.00 DB 0.58 0.55 0.56 0.47 0.54 0.55 1.00 158 Equity Price Data, supra note 8. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 633 B. “Monoculture Risk” in the Financial Sector Our results also speak to the fact that correlation does not exist in a vacuum. Correlation can reflect risks, exposures, or activities that are common among financial institutions. Since the crisis, the largest stand-alone investment banks have be- come bank holding companies,159 and other banking institu- tions have made substantial cuts to their investment banking units.160 Increased correlation may indicate that the activities of large, consolidated banks are converging, as other research indicates, toward a more retail-focused, stable, and profitable business model.161 Importantly, this convergence involves an intuitive trade-off between greater stability in normal times, and greater risk in the event of a contagious run. A close anal- ogy is monoculture: planting the same bountiful crop year af- ter year, while risking that a single pest or pathogen may spoil your entire yield. To the extent this “monoculture risk” exists, it represents a profound challenge at the very heart of the post-crisis macroprudential framework—in part because it in- creases the chances that regulators will face the simultaneous failure of several large financial institutions during a future crisis.162 159 See Sorkin, infra note 177. 160 See Mark DeCambre, Barclays is About to Make it Official: Invest- ment Banking is Dead Almost Everywhere, QUARTZ (May 8, 2014), https://qz.com/207478/barclays-is-about-to-make-make-it-official-invest- ment-banking-is-dead-almost-everywhere/ [https://perma.cc/SV5R-S2KR]. 161 See Rungporn Roengpitya et al., Bank Business Models: Popularity and Performance 15–19 (Bank of Int’l Settlements, Working Paper No. 682, 2017), https://www.bis.org/publ/work682.pdf [https://perma.cc/DF5M- 7CD5] (identifying, in a panel of 178 banks, a post-crisis trend in institu- tions transitioning to a retail banking model, and observing that such retail- focused institutions experience lower cost-to-income ratios and higher and more stable return on equity). 162 See BERNANKE ET AL., supra note 7, at 121 (arguing that new reso- lution powers are “likely to be more effective in managing the failure of a Lehman-type firm in an otherwise stable environment than when other firms are also in danger and the entire system is on the edge of panic”). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 634 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 C. Revisiting Post-Crisis International Capital Standards Our findings also reiterate the point that international capital standards are not evergreen, including those crafted since the last financial crisis. Regulators’ existing experience also reflects this point. In 2011, then-Federal Reserve Gover- nor Daniel Tarullo criticized the old Basel II definition of risk- based capital in a speech: [A]t least some of the instruments that qualified as “Tier 1 capital” for regulatory purposes [under Basel II] were not reliable buffers against losses, at least not on a going concern basis. It is instructive that during the height of the crisis, counterparties and other mar- ket actors looked almost exclusively to the amount of tangible common equity held by financial institutions in evaluating the creditworthiness and overall stabil- ity of those institutions. They essentially ignored the Tier 1 and total risk-based capital ratios in regulatory requirements. In the fall of 2008, there was wide- spread doubt in markets that the common equity of some of our largest institutions was sufficient to with- stand the losses that those firms appeared to be fac- ing. This doubt made investors and counterparties in- creasingly reluctant to deal with those firms, contributing to the severe liquidity strains that char- acterized financial markets at the time.163 Nearly a decade after those remarks, our results suggest that the Basel III definition may now face similar issues, which merits further investigation. By definition, the inverse relationship between our Basel III-based leverage measures (B3 T1/TA and CET1/TA) and simple leverage (CE/TA) re- flects a substitution away from common equity and toward other Basel III-eligible funding. What was the relationship be- tween those other forms of funding and run behavior during the Lehman run? Did some predict run behavior especially 163 See Tarullo, supra note 79, at 3. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 635 well or poorly? Risk weights also changed dramatically from Basel II to Basel III. Would these risk weights have altered the inverse relationship between the Basel II Tier 1 capital ratio (B2 T1/RWA) and post-Lehman run behavior? More broadly, is it possible to improve the existing risk-based capi- tal framework, enabling it to stanch an ongoing contagious run, in addition to making the occurrence of one less likely and the consequences of one less dire? D. Preparing Disclosures in Advance Our results also suggest that improving disclosures, par- ticularly those surrounding credit exposures and credit risk, could enormously impact institutions’ pre-crisis behavior. Re- call that, to account for the possibility that investors acted on private information about Lehman itself, we assessed the di- rect exposure of each institution in our sample to Lehman in the United States. Using information from the Lehman U.S. bankruptcy trustee, we aggregated the total amount awarded on claims either held by or transferred from an institution in our sample (or one of its subsidiaries or affiliates). If our hypothesis about informational efficiency is correct, these kind of counterparty exposures could have been more relevant than return correlation in the post-Lehman panic, yet have been too expensive and cumbersome to obtain. If that information had been public in 2008—for example, Citibank N.A.’s unsecured Lehman claim of more than $138 billion, or more than 147% of its reported Tier 1 capital164—would measures like correlation and leverage have been so closely associated with investor behavior? In turn, would limiting counterparty exposures have also limited the spread of the Lehman run? Or would second- and third-degree exposures, and the uncertainty around them in a chaotic market, have caused independent damage? How would these same counter- factuals apply to resolution planning, and to the more detailed 164 See Voluntary Petition, supra note 118, at Schedule 1. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 636 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 view regulators now have of the legal structure of large finan- cial institutions? E. Herd Behavior and Market Structure Finally, our results hint at the potential danger of a critical trend that has emerged since 2008: the widespread adoption of algorithmic trading. By some accounts, algorithmic trading and other forms of automated investment now drive around 85% of daily U.S. trading volumes, much of it based on mo- mentum strategies that could encourage herd behavior.165 Au- tomation could allow investors to process extremely complex information on short notice, or it could encode their existing beliefs—biased or not—and amplify market swings, which may also involve investor runs through the correlation chan- nels. We conclude by noting that network and agent-based mod- els of financial stress might have the potential to help answer these questions.166 However, our results also show the explan- atory value of very simple and transparent models in compli- cated times. Regulators, investors, and institutions all under- standably struggle to decide what information is trustworthy during a crisis. Certainty may be an ideal condition, but it is often scarce when markets are highly volatile. In its absence, 165 See, e.g., Gregory Zuckerman et al., Behind the Market Swoon: The Herdlike Behavior of Computerized Trading, WALL ST. J. (Dec. 25, 2018), https://www.wsj.com/articles/behind-the-market-swoon-the-herdlike-be- havior-of-computerized-trading-11545785641 [https://perma.cc/UD8P- 25U7]. 166 See, e.g., Jeremy Oldfather et al., Bank Complexity: Is Size Every- thing?, FEDS NOTES (July 15, 2016), https://www.federalreserve.gov/econ- resdata/notes/feds-notes/2016/bank-complexity-is-size-everything- 20160715.html [https://perma.cc/5W28-M7EB]; Fabio Caccioli et al., Net- work Models of Financial Systemic Risk: A Review, 1 J. COMPUTATIONAL SOC. SCI. 81 (2017) (summarizing computational science research into network models of financial crises); Richard Bookstaber & Mark Paddrik, An Agent- based Model for Crisis Liquidity Dynamics (Office of Fin. Research, Working Paper No. 15-18, 2015), https://www.financialresearch.gov/working-pa- pers/files/OFRwp-2015-18_Agent-based-Model-for-Crisis-Liquidity-Dy- namics.pdf [https://perma.cc/834G-UC89]. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 637 it seems, investors may see simplicity as the best available substitute—and by some of the simplest measures, the risk of a contagious run is higher today than it was a decade ago. VI. CONCLUSION This Article focused squarely on the unexpected 2008 Leh- man bankruptcy as a case study to examine the evolution of a market panic. Our analysis shows that institutions with higher levels of regulatory capital experienced more funding stress during the weeks immediately after Lehman’s failure. Balance sheet liquidity levels did not seem to predict exposure to stress. In contrast, two simple measures strongly predicted the discriminate actions of participants in equity and debt markets: a simple leverage ratio, and the correlation between institutions’ share price and Lehman’s. This suggests that simple measures matter more to investors in an emergency, and that tighter correlations can correspond to a faster- spreading run. Our results have important implications for the design of financial regulation, both domestically and in- ternationally, and for the post-crisis capital and liquidity standards that are now subscribed to by over 100 countries.167 We would be remiss to not mention the recent and historic volatility in equity and debt markets caused by COVID-19. U.S. market volatility in March matched that observed around the Lehman bankruptcy.168 Equity prices tumbled, and CDS spreads of financial institutions widened sharply.169 At the time of writing, this crisis is ongoing. However, early signs strongly support the central findings discussed in this Article. Market participants have not behaved indiscrimi- nately, even when the Dow Jones Industrial Average fell 10% 167 See Romano, supra note 14. 168 See VIX Index Historical Data, CHI. BD. OPTIONS EXCH., http://www.cboe.com/products/vix-index-volatility/vix-options-and-fu- tures/vix-index/vix-historical-data [https://perma.cc/BY4Z-F5MK]. 169 See IÑAKI ALDASORO ET AL., BANK FOR INT’L SETTLEMENTS, EFFECTS OF COVID-19 ON THE BANKING SECTOR: THE MARKET’S ASSESSMENT 2 fig.1 (2020). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 638 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 on one day and jumped 9% the next day.170 The situation de- mands a fuller analysis when the crisis subsides. However, we remain confident that business models and correlations mat- ter, and they matter disproportionately during a panic.171 170 See Stan Choe & Alex Veiga, Dow Closes Down 10% in Worst Day on Wall Street Since 1987, CHI. TRIBUNE (Mar. 12, 2020), https://www.chicagotrib- une.com/coronavirus/ct-nw-coronavirus-stock-market-reaction-20200312- c736p4egzjak3nnqblzj46db6a-story.html [https://perma.cc/68UT-KWDB]; Alexa Veiga & Damian J. Troise, Stocks Surge on Wall Street, Almost Erasing Thursday’s Historic Dow Plunge, CHI. TRIBUNE (Mar. 13, 2020), https://www.chicagotribune.com/coronavirus/ct-nw-coronavirus-stock-market- reaction-20200313-quxsrtlae5bzhl6o2v672ijcei-story.html [https://perma.cc/DT6L-BUT8]. 171 See Vipal Monga, Oil Crash Is Bad News for Regional Banks That Went Big on Energy, WALL ST. J. (Mar. 16, 2020), https://www.wsj.com/arti- cles/oil-crash-is-bad-news-for-regional-banks-that-went-big-on-energy- 11584352803 [https://perma.cc/KQ8Y-83XK] (describing the vulnerability of banks whose business models focus on the energy sector); David Benoit & Leslie Scism, The Fed Cut Is the Deepest for Banks and Insurers, WALL ST. J. (Mar. 3, 2020), https://www.wsj.com/articles/the-fed-cut-is-the-deepest- for-banks-and-insurers-11583277221 [https://perma.cc/HJQ7-PM4P] (sug- gesting that banks more dependent on net interest income will suffer more than their peers in the very low interest rate environment). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 639 VII. APPENDIX A: Definitions of Key Regulatory Terms and Ratios Table 4: Definitions172 Tier 1 Capital (Basel II numerator) Common Equity + Disclosed Reserves + Non-Cumulative Perpetual Preferred Stock Tier 1 Capital (Basel III numerator) Common Equity Tier 1 Capital + Additional Tier 1 Capital173 Common Equity Tier 1 Capital (Basel III numera- tor)174 Common Equity+ Stock Surplus (Share Premium) from Common Equity + Disclosed Reserves + Retained Earnings + Accumulated Other Comprehensive In- come net of [Goodwill + Net Deferred Tax Assets + Other Intangible Assets] 172 Basel II ratios are taken from BASEL COMM., BASEL II, supra note 51. Basel III ratios are taken from BASEL COMM., BASEL III, supra note 55. 173 See infra Appendix H for the full criteria of what may be included as “Additional Tier 1 Capital”. 174 The elements in this row are abbreviated for clearer comparison to the Basel II Tier 1 measure. See BASEL COMM., BASEL II, supra note 51, at 13–15 (providing a full definition of CET1, as well as fourteen criteria for inclusion of instruments as “common shares”). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 640 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Tier 1 Capital Ratio (Basel II) Tier 1 Capital (Basel II) Risk Weighted Assets (Basel II) Common Equity Tier 1 Ratio (Basel III) CET1 Risk Weighted Assets (Basel III) Tier 1 Leverage Measure (Basel III) Tier 1 Capital (Basel III) Total Leverage Exposure 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 641 B: Descriptive Statistics Table 5: Descriptive Statistics RHS Variable Mean Median Std. N AR T1/RWA 0.093 0.089 0.031 24 B2 T1/TA 0.068 0.065 0.019 22 B3 CET1/RWA 0.061 0.054 0.027 22 B3 CET1/TA 0.050 0.045 0.032 27 B3 T1/RWA 0.068 0.061 0.024 22 B3 T1/TA 0.054 0.050 0.031 27 CE/RWA 0.115 0.105 0.052 21 CE/TA 0.091 0.084 0.035 27 L1/TA 0.031 0.007 0.048 25 Cash/TA 0.044 0.025 0.052 27 STWF/TA 0.112 0.079 0.087 27 L1/STWF 0.222 0.140 0.241 24 Cash/STWF 0.580 0.288 0.808 26 Lehman Correlation 0.642 0.656 0.080 26 Lehman Claims 15.676 16.329 4.275 18 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 642 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 C: Basel III Common Equity Tier 1 (“CET1”) Capital Proxies and Robustness Checks Y-9C Common Equity Tier 1 (“CET1”) Proxy for U.S. Banking In- stitutions175 + BHCK3230 Common stock (par value) + BHCK3240 Surplus (exclude all surplus related to pre- ferred stock) + BHCK3247 Retained earnings + BHCKB530 Accumulated other comprehensive income + BHCKA130 Other equity capital components176 - BHCK3163 Goodwill - BHCK2148 Net deferred tax assets - BHCK0426 Other intangible assets Table 6: U.S. CET1 Proxy Comparison Using Y-9C Bank Ratio Bank Ratio Bank Ratio Bank Ratio ALLY 94% CMA 91% JPM 99% RF 87% AXP 94% COF 89% KEY 90% SCHW 98% BAC 97% DFS 96% MS 101% STI 90% BBT 93% ETFC 97% MTB 88% STT 109% BK 111% FITB 96% NTRS 102% SYF 102% C 95% GS 100% NYCB 98% USB 99% CFG 89% HBAN 94% PNC 112% WFC 99% For all but two U.S. institutions, our proxy is based on the Federal Reserve Y-9C line items described above. As a 175 For a description of these proxies, see BD. OF GOVERNORS FED. RE- SERVE SYS., INSTRUCTIONS FOR PREPARATION OF CONSOLIDATED FINANCIAL STATEMENTS FOR HOLDING COMPANIES: REPORTING FORM FR Y-9C (2020). 176 The Y-9C calls for this item to be reported as a negative value. See id. at 31. As such, although the definition calls for it to be subtracted, it is added here. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 643 robustness check, we construct our CET1 proxy using 2018:Q2 for each firm listed above, and divide the resulting measure by that firm’s publicly reported 2018:Q2 CET1 value (e.g., if the CET1 proxy is $35 billion and the publicly reported CET1 value is $38 billion, the ratio reported above is 35/38 = 92%). In the table above, the mean of the twenty-eight ratios is 97%, and the median is 96%. Bloomberg Common Equity Tier 1 (“CET1”) Proxy for G-SIBs + BS_SH_CAP_AND_APIC + BS_RETAIN_EARN + ARD_ACC_OTH_COMPREHENSIVE_INC - BS_GOODWILL - OTHER_INTANGIBLE_ASSETS_DETAILED - BS_DEF_TAX_LIAB Table 7: G-SIB CET1 Proxy Comparison Using Bloom- berg Bank Ratio Bank Ratio Bank Ratio Bank Ratio 1288 HK Eq- uity 108% 939 HK Equity 103% DB US Equity 127% NDA SS Equity 110% 3988 HK Eq- uity 105% ACA FP Eq- uity 105% GLE FP Equity 127% RBS LN Equity 112% 8306 JP Equity 95% BARC LN Equity 117% HSBA LN Eq- uity 116% RY CN Equity 100% 8316 JP Equity 94% BNP FP Eq- uity 109% HVM GR Eq- uity 111% SAN SM Equity 96% 8411 JP Equity 99% CSGN SW Equity 72% INGA NA Eq- uity 108% UBSG SW Eq- uity 135% For institutions in our non-U.S. G-SIB sample, our CET1 proxy is based on the Bloomberg ticker items described above. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 644 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 We also use this CET1 proxy for Goldman Sachs and Morgan Stanley, which converted to bank holding companies shortly after the Lehman bankruptcy and (consequently) did not re- port Y-9C values before it.177 As a robustness check, we again construct our CET1 proxy using 2018:Q2 for each firm listed above, and divide the re- sulting measure by that firm’s publicly reported 2018:Q2 CET1 value. The exceptions are 1288 HK Equity, the reported value for which comes from 2017:Q4, and HVM GR Equity, the reported value for which comes from 2016:Q3. In the table above, the mean of the twenty ratios is 107% and the median is 108%. 177 See Andrew Ross Sorkin, As Goldman and Morgan Shift, a Wall St. Era Ends, N.Y. TIMES (Sept. 21, 2008), https://dealbook.ny- times.com/2008/09/21/goldman-morgan-to-become-bank-holding-compa- nies/ [https://perma.cc/8WVJ-N2N2]. 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 645 D: In-Sample Institutions Table 8: Sample Institutions178 Institution Share Price/Mkt. Cap. Data Sept. 2008 CDS Spreads Data Sept. 2008 JPMorgan Chase 1 1 Bank of America 1 1 Citigroup 1 1 Wells Fargo 1 1 Goldman Sachs 1 1 Morgan Stanley 1 1 Bank of New York Mellon 1 0 State Street 1 0 Northern Trust 1 0 U.S. Bancorp 1 0 PNC Financial 1 0 Capital One 1 1 Charles Schwab 1 0 BB&T Corp. 1 0 SunTrust Inc. 1 0 American Express 1 1 Ally Financial 0 1 Citizens Financial 0 0 Fifth Third 1 0 KeyCorp 1 0 M&T Bank 1 0 Huntington 1 0 Discover Financial Services 1 0 178 Information obtained from institutions’ filed FR Y-9Cs. See Holding Company Data, FED. RES. BANK OF CHI., https://www.chicagofed.org/bank- ing/financial-institution-reports/bhc-data [https://perma.cc/4KRW-Z52J]. 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 646 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Synchrony Finan- cial 0 0 Comerica Inc. 1 0 E*TRADE Finan- cial 1 0 SVB Financial Group 1 0 NY Community Bancorp 1 0 Total 26 9 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 647 E: Simple Panel Fixed-Effects Regression Results Capital Measure Share Price Five-Year CDS Spread One-Year CDS Spread CE/RWA �̂� = −0.034 𝑠. 𝑒. = 0.400 𝑅2 = 0.765 𝑛 = 21 �̂� = 3.496 𝑠. 𝑒. = 3.103 𝑅2 = 0.679 𝑛 = 6 �̂� = 4.577 𝑠. 𝑒. = 3.456 𝑅2 = 0.803 𝑛 = 6 AR T1/RWA �̂� = −0.555 𝑠. 𝑒. = 0.644 𝑅2 = 0.766 𝑛 = 22 �̂� = 2.394 𝑠. 𝑒. = 2.738 𝑅2 = 0.665 𝑛 = 7 �̂� = 5.750 ∗ 𝑠. 𝑒. = 2.289 𝑅2 = 0.820 𝑛 = 6 B3 T1/RWA �̂� = −1.150 𝑠. 𝑒. = 0.833 𝑅2 = 0.773 𝑛 = 22 �̂� = 4.883 ∗ 𝑠. 𝑒. = 2.118 𝑅2 = 0.694 𝑛 = 7 �̂� = 7.071 ∗∗ 𝑠. 𝑒. = 2.183 𝑅2 = 0.846 𝑛 = 6 B3 CET1/RWA �̂� = −0.701 𝑠. 𝑒. = 0.862 𝑅2 = 0.769 𝑛 = 22 �̂� = 3.391 𝑠. 𝑒. = 2.202 𝑅2 = 0.681 𝑛 = 7 �̂� = 6.357 ∗∗ 𝑠. 𝑒. = 2.159 𝑅2 = 0.841 𝑛 = 6 CE/TA �̂� = 0.914 𝑠. 𝑒. = 0.606 𝑅2 = 0.771 𝑛 = 24 �̂� = −3.596 ∗ 𝑠. 𝑒. = 1.821 𝑅2 = 0.672 𝑛 = 9 �̂� = −8.956 𝑠. 𝑒. = 5.561 𝑅2 = 0.754 𝑛 = 8 B2 T1/TA �̂� = 2.081 ∗ 𝑠. 𝑒. = 1.013 𝑅2 = 0.780 𝑛 = 21 �̂� = −15.366 ∗∗ 𝑠. 𝑒. = 5.822 𝑅2 = 0.715 𝑛 = 6 �̂� = −18.452 ∗ 𝑠. 𝑒. = 7.629 𝑅2 = 0.832 𝑛 = 6 B3 T1/TA �̂� = 0.109 𝑠. 𝑒. = 0.581 𝑅2 = 0.765 𝑛 = 24 �̂� = −3.584 𝑠. 𝑒. = 2.058 𝑅2 = 0.652 𝑛 = 9 �̂� = −5.682 ∗∗ 𝑠. 𝑒. = 2.168 𝑅2 = 0.733 𝑛 = 8 B3 CET1/TA �̂� = 0.114 𝑠. 𝑒. = 0.536 𝑅2 = 0.765 𝑛 = 24 �̂� = −3.261 ∗ 𝑠. 𝑒. = 1.638 𝑅2 = 0.650 𝑛 = 9 �̂� = −4.386 ∗ 𝑠. 𝑒. = 2.254 𝑅2 = 0.731 𝑛 = 8 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 648 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 Non-Capital Measure Share Price Five-Year CDS Spread One-Year CDS Spread Liquidity L1/TA �̂� = −0.961 ∗∗ 𝑠. 𝑒. = 0.349 𝑅2 = 0.792 𝑛 = 22 �̂� = 2.824 ∗ 𝑠. 𝑒. = 1.435 𝑅2 = 0.671 𝑛 = 8 �̂� = 4.785 ∗ 𝑠. 𝑒. = 2.426 𝑅2 = 0.772 𝑛 = 7 Cash + Equiva- lents/TA �̂� = −0.279 ∗∗ 𝑠. 𝑒. = 0.129 𝑅2 = 0.766 𝑛 = 24 �̂� = −2.194 𝑠. 𝑒. = 1.559 𝑅2 = 0.652 𝑛 = 9 �̂� = −4.788 ∗ 𝑠. 𝑒. = 2.380 𝑅2 = 0.749 𝑛 = 8 Funding Fragility STWF/TA �̂� = −0.533 ∗∗∗ 𝑠. 𝑒. = 0.170 𝑅2 = 0.778 𝑛 = 24 �̂� = 0.728 𝑠. 𝑒. = 0.739 𝑅2 = 0.649 𝑛 = 9 �̂� = −0.797 𝑠. 𝑒. = 1.902 𝑅2 = 0.729 𝑛 = 8 L1/STWF �̂� = −0.151 ∗ 𝑠. 𝑒. = 0.079 𝑅2 = 0.790 𝑛 = 21 �̂� = 0.457 𝑠. 𝑒. = 0.301 𝑅2 = 0.664 𝑛 = 8 �̂� = 0.810 𝑠. 𝑒. = 0.452 𝑅2 = 0.763 𝑛 = 7 Cash + Equiva- lents/STWF �̂� = −0.003 𝑠. 𝑒. = 0.017 𝑅2 = 0.767 𝑛 = 23 �̂� = −0.293 𝑠. 𝑒. = 0.186 𝑅2 = 0.653 𝑛 = 9 �̂� = −0.437 ∗∗ 𝑠. 𝑒. = 0.179 𝑅2 = 0.739 𝑛 = 8 Correla- tion Lehman Correlation �̂� = −0.747 ∗∗∗ 𝑠. 𝑒. = 0.195 𝑅2 = 0.788 𝑛 = 24 �̂� = 3.540 ∗∗∗ 𝑠. 𝑒. = 0.958 𝑅2 = 0.718 𝑛 = 8 �̂� = 4.580 ∗∗∗ 𝑠. 𝑒. = 0.896 𝑅2 = 0.833 𝑛 = 7 Lehman Claims �̂� = −0.014 ∗∗∗ 𝑠. 𝑒. = 0.004 𝑅2 = 0.807 𝑛 = 17 �̂� = 0.061 ∗∗ 𝑠. 𝑒. = 0.019 𝑅2 = 0.698 𝑛 = 8 �̂� = 0.094 ∗∗∗ 𝑠. 𝑒. = 0.024 𝑅2 = 0.791 𝑛 = 8 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 649 F: Multiple Panel Fixed-Effects Regression Results 𝑚1,𝑖 𝑚2,𝑖 U.S. Banking Institution Sample Share Price Five-Year CDS Spread One-Year CDS Spread Leh- man Cor- rela- tion Leh- man Claim s �̂� = −0.702 ∗∗ 𝑠. 𝑒. = 0.342 𝑅2 = 0.848 𝑛 = 17 �̂� = −0.008 ∗ 𝑠. 𝑒. = 0.004 𝑅2 = 0.848 𝑛 = 17 �̂� = 4.373 ∗∗∗ 𝑠. 𝑒. = 1.663 𝑅2 = 0.820 𝑛 = 7 �̂� = −0.025 𝑠. 𝑒. = 0.027 𝑅2 = 0.820 𝑛 = 7 �̂� = 5.517 ∗∗∗ 𝑠. 𝑒. = 1.819 𝑅2 = 0.879 𝑛 = 7 �̂� = −0.022 𝑠. 𝑒. = 0.027 𝑅2 = 0.879 𝑛 = 7 B2 T1/T A Leh- man Corre- lation �̂� = 0.375 𝑠. 𝑒. = 0.564 𝑅2 = 0.824 𝑛 = 21 �̂� = −0.705 ∗∗∗ 𝑠. 𝑒. = 0.203 𝑅2 = 0.824 𝑛 = 21 �̂� = 6.982 𝑠. 𝑒. = 9.373 𝑅2 = 0.826 𝑛 = 6 �̂� = 5.472 ∗∗ 𝑠. 𝑒. = 2.479 𝑅2 = 0.826 𝑛 = 6 �̂� = 10.226 𝑠. 𝑒. = 12.482 𝑅2 = 0.887 𝑛 = 6 �̂� = 7.022 ∗∗ 𝑠. 𝑒. = 3.532 𝑅2 = 0.887 𝑛 = 6 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 650 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 G: Simple Regression of Selected Explanatory Variables on Cumulative Changes in Share Price -40% -30% -20% -10% 0% 10% 20% 30% 50% 60% 70% 80% 90% Lehman Correlation -40% -30% -20% -10% 0% 10% 20% 30% 0% 5% 10% 15% Risk-Based Capital (B3 CET1/RWA) 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 651 -40% -30% -20% -10% 0% 10% 20% 30% 0 5 10 15 20 25 Lehman Claims (Natural Logarithm) -40% -30% -20% -10% 0% 10% 20% 30% 0% 5% 10% 15% 20% Liquidity (Level 1/TA) 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 652 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 -40% -30% -20% -10% 0% 10% 20% 30% 0% 5% 10% 15% 20% Simple Leverage (CE/TA) -40% -30% -20% -10% 0% 10% 20% 30% 0% 10% 20% 30% 40% Short-Term Wholesale Funding /Total Assets 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 653 H: Criteria for Additional Tier 1 Capital (Basel III)179 1) Issued and paid-in; 2) Subordinated to depositors, general creditors, and subordi- nated debt of the bank; 3) Is neither secured nor covered by a guarantee of the issuer or related entity or other arrangement that legally or econom- ically enhances the seniority of the claim vis-à-vis bank credi- tors; 4) Is perpetual, i.e., there is no maturity date and there are no step-ups or other incentives to redeem; 5) May be callable at the initiative of the issuer only after a minimum of five years: A) To exercise a call option a bank must receive prior su- pervisory approval; and B) A bank must not do anything which creates an expecta- tion that the call will be exercised; and C) Banks must not exercise a call unless: 1) They replace the called instrument with capital of the same or better quality and the replacement of this capi- tal is done at conditions which are sustainable for the income capacity of the bank180; or 2) The bank demonstrates that its capital position is well above the minimum capital requirements after the call option is exercised.181 179 The entirety of Appendix H is borrowed from BASEL COMM., BASEL III, supra note 55, at 15–17. 180 Id. at 16 n.15 (“Replacement issues can be concurrent with but not after the instrument is called.”). 181 Id. at 16 n.16 (“Minimum refers to the regulator’s prescribed mini- mum requirement, which may be higher than the Basel III Pillar 1 mini- mum requirement.”). 3_2020.2_TABOR_ZHANG (DO NOT DELTE) 10/21/2020 1:56 PM 654 COLUMBIA BUSINESS LAW REVIEW [Vol. 2020 6) Any repayment of principal (e.g., through repurchase or re- demption) must be with prior supervisory approval and banks should not assume or create market expectations that super- visory approval will be given 7) Dividend/coupon discretion: A) The bank must have full discretion at all times to cancel distributions/payments182; B) Cancellation of discretionary payments must not be an event of default C) Banks must have full access to cancelled payments to meet obligations as they fall due; and D) Cancellation of distributions/payments must not impose restrictions on the bank except in relation to distributions to common stockholders. 8) Dividends/coupons must be paid out of distributable items; 9) The instrument cannot have a credit sensitive dividend fea- ture, that is a dividend/coupon that is reset periodically based in whole or in part on the banking organisation’s credit stand- ing; 10) The instrument cannot contribute to liabilities exceeding assets if such a balance sheet test forms part of national insol- vency law; 11) Instruments classified as liabilities for accounting pur- poses must have principal loss absorption through either (i) 182 Id. at 16 n.17 (“A consequence of full discretion at all times to cancel distributions/payments is that ‘dividend pushers’ are prohibited. An instru- ment with a dividend pusher obliges the issuing bank to make a divi- dend/coupon payment on the instrument if it has made a payment on an- other (typically more junior) capital instrument or share. This obligation is inconsistent with the requirement for full discretion at all times. Further- more, the term ‘cancel distributions/payments’ means extinguish these pay- ments. It does not permit features that require the bank to make distribu- tions/payments in kind.”). 3_2020.2_TABOR_ZHANG (DO NOT DELETE) 10/21/2020 1:56 PM No. 2:575] CAPITAL, CONTAGION, AND FINANCIAL CRISES 655 conversion to common shares at an objective pre-specified trigger point or (ii) a write-down mechanism which allocates losses to the instrument at a pre-specified trigger point. The write-down will have the following effects: A) Reduce the claim of the instrument in liquidation; B) Reduce the amount re-paid when a call is exercised; and C) Partially or fully reduce coupon/dividend payments on the instrument; 12) Neither the bank nor a related party over which the bank exercises control or significant influence can have purchased the instrument, nor can the bank directly or indirectly have funded the purchase of the instrument; 13) The instrument cannot have any features that hinder re- capitalization [sic], such as provisions that require the issuer to compensate investors if a new instrument is issued at a lower price during a specified time frame; and 14) If the instrument is not issued out of an operating entity or the holding company in the consolidated group (e.g., a spe- cial purpose vehicle), proceeds must be immediately available without limitation to an operating entity183 or the holding company in the consolidated group in a form which meets or exceeds all of the other criteria for inclusion in Additional Tier 1 capital. 183 Id. at 17 n.18 (“An operating entity is an entity set up to conduct business with clients with the intention of earning a profit in its own right.”).