Too big to fail and too big to save: dilemmas for banking reform.

Barth, James R. ; Wihlborg, Clas

'Too big to fail' traditionally refers to a bank that is perceived to generate unacceptable risk to the banking system and indirectly to the economy as a whole if it were to default and be unable to fulfill its obligations. Such a bank generally has substantial liabilities to other banks through the payment system and other financial links, which can be sources of 'contagion' if a bank fails. The main objectives in this paper are to identify the different dimensions of too big to fail and evaluate various proposed reforms for dealing with this problem. In addition, we document the various dimensions of size and complexity, which may contribute to or reduce a bank's systemic risk. Furthermore, we provide an assessment of economic and political factors shaping the future of too big to fail.

Keywords: financial crises; too big to fail; systemically important banks; Dodd-Frank Act; Financial Stability Board; regulation and supervision

JEL Classifications: G18; G21; G28

I. Introduction

'Too big to fail' traditionally refers to a bank that is perceived to generate unacceptable risk to the banking system and indirectly to the economy as a whole if it were to default and unable to fulfil its obligations. Unlike non-financial firms banks generally have substantial liabilities to other banks through the payment system and other financial links, which can be sources of contagion if a bank fails.

In the United States, the practice of treating troubled big banks differently from troubled small ones dates back to the 1984 bailout of Continental Illinois Corporation. That taxpayer-funded rescue was based on fears that a bank collapse of Continental's magnitude would destabilise the entire financial system (see, for example, Kaufman, 2002; Shull, 2010; and Barth, Prabha, and Swagel, 2012). Those same fears prompted far bigger bank bailouts, both in the US and Europe, during the recent global financial crisis. In the wake of that experience, regulators and banking experts almost unanimously agree that regulatory reform is essential to ensuring that no bank is ever again too big to fail.

After the Great Recession too big to fail has become a concern for financial firms more broadly, since there is substantial interconnectedness among financial institutions involved in financial market trading activities that generate liquidity in the markets for a variety of financial instruments. The financial crisis in 2007-9, which contributed to the Great Recession, originated outside the traditional banking system when the liquidity in and value of financial instruments held by and issued by banks and financial institutions dropped dramatically.

Outside the US the policy responses to banking crises have long been characterised by general bailouts of banks' creditors (and sometimes shareholders as well) in the form of, for example, blanket guarantees, unlimited liquidity support or state nationalisation, as documented in Caprio et al. (2005) and Honohan and Klingebiel (2003). One explanation is that few countries outside the US have had established special legal procedures for bank insolvencies until very recently. Failing banks in most countries had to be resolved under general corporate insolvency law. Resolution under such laws is time-consuming and not suitable for failing banks with liabilities that support liquidity in the economy. (1)

After the Continental Illinois failure the US resolution procedures for banks were strengthened with the implementation of the Federal Deposit Insurance Corporation Improvement Act in 1991. Under these procedures small and midsized banks in the US have been allowed to fail with consequences for banks' uninsured creditors. The bailout of large banks but not small and midsized banks during the financial crisis established too big to fail and the differential treatment of such banks.

Financial reforms outside the US since the crisis include the implementation of special procedures for bank resolution, in particular in the EU, with the intention to address the too big to fail problem as well. In the US the Dodd-Frank Wall Street Reform and Consumer Protection Act (Dodd-Frank Act) includes procedures for resolving large banks. We argue that the credibility of these procedures is critical for the future of too big to fail.

The differential treatment of large banks has been formalised in the designation of banks as globally systemically important banks (G-SIBs) and a larger number of systemically important financial institutions (SIFIs). These designations have implications for their regulation and supervision, as well as for the degree to which these financial institutions may enjoy implicit subsidies.

We use the concept of too big to fail to incorporate 'too complex to fail' as well. We discuss factors affecting complexity and the impact of these factors on systemic risk. Although complexity and size are far from perfectly correlated, increased complexity essentially implies that a financial institution can become too big to fail at a smaller size.

Our main objectives are to identify the different dimensions of too big to fail and evaluate various proposed reforms for dealing with this problem. In addition, we document the various dimensions of size and complexity, which may contribute to or reduce a bank's systemic risk.

Some observers argue that the size of an individual bank failing is less important for a government's decision whether to bail out a bank than the share of the banking system under stress. For example, Brown and Ding (2011) emphasise 'too many to fail.' There are reasons to believe in a separate effect of too big to fail. One is that the too big to fail bank can be an independent source of crisis. Furthermore, many small banks are unlikely to fail at exactly the same time in response to a large shock outside the banking system. A resolution authority may be able to manage the failures of many banks one by one over some time without having to fear the same degree of contagion as in the case of one large bank's failure.

One major cost generated by banks being too big to fail is that competition between banks of different sizes and degrees of complexity becomes distorted. If the non-insured creditors of some banks perceive that these banks are less likely to fail than other banks, the costs of funding for the relatively safe banks become relatively low. The International Monetary Fund (April 2014) estimates that the implicit subsidies given just to G-SIBs in 2011-12 represent $15-$70 billion in the US, $25--$110 billion in Japan, $20-$110 billion in the United Kingdom, and up to $90-$300 billion in the Euro Area.

The existence of an implicit subsidy implies also that banks have strong incentives to achieve sufficient size and/or complexity to enjoy this competitive advantage. The importance of these incentives relative to incentives generated by economies of scale and scope as explanations for the trend towards increased size and complexity is an open question that we will come back to.

A second cost is associated with 'regulatory capture.' A few very large banks are likely to have a very strong influence on regulators, supervisors and legislatures. Regulatory capture may be the result of explicit lobbying efforts of politicians and financial support in elections. Less deliberately, regulators and supervisors have a tendency to develop shared objectives and values with those being regulated and supervised. (2)

A third cost associated with too big to fail is that it creates a link between bank risk and sovereign risk. The fiscal costs of bailing out a large bank can be enormous and, thereby, contribute to a country's fiscal crisis. For example, the fiscal cost of injecting capital in the Anglo-Irish bank in Ireland in 2009 amounted to 35 per cent of Ireland's GDP. Bank rescues in Spain in 2011 similarly contributed to the debt crisis in that country.

The capital injections in relation to GDP in the US banking system during the financial crisis were far from the magnitudes of the Irish and the Spanish rescue operations but the political resistance to government aid to banks grew strong from both ends of the political spectrum. A political consensus that banks must never be too big to fail developed. This consensus is reflected in the Dodd-Frank Act. It explicitly states that no bank must be too big to fail but there is widespread scepticism whether the specific reforms are sufficiently strong to allow a large bank to be resolved in such a way that uninsured creditors must share in losses. (3) There are similar questions with respect to the credibility of so-called bail-in rules in the EU Bank Recovery and Resolution Directive. (4)

The fiscal and political costs associated with bank bailouts during the financial crisis 2007-9 and the euro debt crisis thereafter suggest that the large banks in many countries are not only too big to fail but also 'too big to save.' This view is reflected in the many proposals to eliminate the too big to fail problem.

The dilemma facing policymakers is that if ongoing reforms do not credibly eliminate the too big to fail problem the too big to save problem may become the cause of an inability to deal effectively with the resolution of large banks in another crisis. The problem may become particularly acute for banks with large operations in several countries with conflicting interests with respect to bearing the burdens of crisis management.

Relatively small countries with large banks with international operations are particularly vulnerable to the too big to save problem. The Icelandic banking crisis in 2009 is so far the prime example of too big to save. Iceland simply did not have the financial resources to rescue three banks with large international operations.

The paper proceeds as follows. In Section 2 we present a brief history of too big to fail. In Section 3 we describe how banking systems in a number of countries and globally have become increasingly concentrated and dominated by relatively few giant banks. In Section 4 we focus on complexity as it contributes to systemic risk and discuss the link between too big to fail and complexity. We present measures that describe different dimensions of complexity. The criteria developed by the Financial Stability Board (FSB) to identify G-SIBs are discussed. Section 5 reviews various policy proposals with the objective to reduce size and complexity of banks. Costs and benefits of proposals are discussed. These depend to a large extent on what factors have been driving the development towards size and complexity. We emphasise that an efficient organisation of the financial system requires that the ongoing regulatory reforms strengthen market discipline on banks. Section 6 concludes with an assessment of economic and political factors shaping the future of too big to fail.

2. The big keep getting bigger: a brief history of too-big-to-fail

In view of all the concern today over some banks being too big for regulatory comfort, it is instructive to look back at the circumstances that gave rise to the decision that bank size per se could be a key factor in treating big banks that encounter financial difficulties differently from smaller banks.

The concept of too big to fail seems to have been established in the US as a result of the $4.5 billion bailout of Continental Illinois National Bank & Trust Co. of Chicago in 1984. The Wall Street Journal reported that the Comptroller of the Currency, C. Todd Conover, told Congress that "the federal government won't currently allow any of the nation's 11 largest banks to fail." The article noted that the response to Conover's statement was that the "government had created a new category of bank: the 'TBTF' bank, for Too Big To Fail". The article pointed out that this was the first time a government official acknowledged the existence of such a policy (Carrington, 1984).

The origin of too big to fail in the US may be explained by the observation that the US was the only advanced country where banks were allowed to fail at the time. The much more concentrated banking systems in Europe were generally viewed as akin to public utilities before the setting of interest rates and the allocation of credit were liberalised during the 1980s. The banks dominated the financial systems and a special insolvency law for banks did not exist or was not applied. Heavily regulated banks were assumed not to fail. (5)

Even when banking crises erupted in Europe after liberalisation in the early 1990s it was taken for granted that banks would not be allowed to fail. For example, in the early 1990s the Scandinavian countries used a combination of nationalisation, blanket guarantees of bank liabilities and capital injections to manage banking crises.

The attitude in Europe towards bank bailouts did not change until after the Great Recession in 2007-9. This crisis also triggered an awareness among taxpayers, policymakers and regulators that too big to fail was an issue in the public interest that needed to be addressed.

Going back to US developments, the original eleven banks subsequently shrank to four, while the eleven BHCs were reduced to five. Only Citibank and Wells Fargo remain with their original holding company. Continental Illinois was eventually seized by the Federal Deposit Insurance Corp. and sold to Bank of America. This meant that despite the earlier bailout of a too-big-to-fail bank, regulators eventually allowed the same bank to be acquired by a bigger bank, which became even bigger. The four remaining BHCs accounted for nearly 50 per cent of the total assets of all BHCs at year-end 2013. In contrast, at year-end 1983 the holding companies of the eleven original too-big-to-fail banks accounted for 33 per cent of the total assets of all BHCs.

3. Just how big are the world's biggest banks?

There are several different ways to measure 'big'. The relevance of each particular measure depends on the objective of the analysis. Size per se does not necessarily inform us about the importance of a bank within the financial system or about the vulnerability of the real economy from a bank's failure.

We start by focusing on the 100 biggest publicly traded banks in the world by their total assets, as of the second quarter, 2015. The banks are headquartered in 25 different countries and the biggest is Industrial & Commercial Bank of China, with $3,667 billion in assets. The smallest is Charles Schwab Corporation of the United States with $164 billion in assets. The biggest bank is therefore 22 times the size of the 100th biggest bank in the world. There are 22 'trillion dollar banks' in the world, and four are US banks.

Focusing on total assets, however, does not produce an apples-to-apples comparison. There is no uniform worldwide accounting standard for measuring assets. Most of the home countries for banks rely on International Financial Reporting Standards (IFRS), but the US and others rely on their own Generally Accepted Accounting Principles (GAAP). Countries that use IFRS, and some that use GAAP, report derivatives on a gross rather than a net basis. When an adjustment is made to measure total assets on a comparable basis, the result is a significant change for several of the world's biggest banks. In particular, JP Morgan Chase reports total assets of $2.5 trillion under US GAAP, but $3.3 trillion under IFRS, and the bank jumps from sixth place to second place among the world's largest banks. Likewise, Bank of America leaps from ninth place to fifth. More importantly, US GAAP treatment may be understating the assets of all US banks by a total of $3.4 trillion.

There are a total of 1,226 publicly traded banks in the 25 countries with total assets of $94 trillion. The 100 biggest banks account for 8 per cent of all the banks, but 83 per cent of the total assets. The US has the most publicly traded banks of any nation, with 736 banks. Japan and Russia rank second and third with 97 and 60 banks, respectively. China is the country whose banks collectively have the most total assets--$19.2 trillion. The banks in the United States and Japan are ranked second and third in terms of total assets at $18.2 trillion and $10.0 trillion, respectively. The total assets of the banks in these three countries total $47.4 trillion, or 50 per cent of assets of all the banks.

Another way to view the world's 100 biggest banks is not simply in terms of total assets but their total assets relative to the total assets of all banks. This may be viewed as a measure of concentration in banking. In terms of individual countries, the ratio of the total assets of the biggest banks to total bank assets ranges from a low of 28 per cent in Taiwan to a high of 99 per cent in Sweden and the United Kingdom. In the case of the United States, the comparable ratio is 67 per cent.

Still another way to view the world's 100 biggest banks is by measuring total assets as a share of GDP in order to measure the burden on the economy of the failure of large banks. By that measure, Swiss big banks are by far the world's biggest, with assets equal to 340 per cent of Switzerland's GDP. Russia's banks would be the smallest, with assets equal to only 52 per cent of GDP. The fifteen biggest US banks are at the lower end of the range, with assets equal to 105 per cent of US GDP.

Now consider the distribution of the world's 100 biggest banks by number and total assets across the 25 countries. The United States and China, with fourteen, have more banks than any other nation on the list. Japan ranks second with eight banks and Canada and Spain tie for third with six banks. The remaining 20 countries account for 52 banks. Measured by total banking assets, however, China's big banks lead the world with $19 trillion. The United States' banks come in second, with combined assets of $12 trillion, and the United Kingdom ranks third with $8 trillion. Banks in the remaining countries have $39 trillion in assets, or 50 per cent of the worldwide total.

To provide a more historical perspective on the importance of big banks, one can focus on just the 50 biggest banks in the world. Doing so, one finds that their total assets relative to world GDP increased to 83 per cent in the second quarter of 2015 from only 15 per cent in 1970. For purposes of comparison, the ratio of total assets of the 50 biggest US banks relative to US GDP increased to 83 per cent in the second quarter of 2015 from only 25 per cent in 1970. These figures illustrate what has been called the financialisation of many advanced economies in the form of expansion of market-oriented financial activities that are not directly linked to intermediation. (6)

Turning to concentration in the banking industry, the largest 25 banks account for about half of the total assets of all banks worldwide. The assets of the same banks are almost two-thirds of world GDP. A relatively few of the world's 1,862 publicly traded banks are truly big in terms of either their share of the global bank assets or global GDP.

There are still other ways in which to rank the size of banks. Two of these ways are to measure individual banks by their assets as a share of either total banking assets in their home countries or as a share of their home-country GDP. There is a weakness with these figures in that they do not allow us to distinguish between domestic and foreign assets of the banks. To measure the importance of a bank within a country's banking system we should include only domestic assets but this figure is not available for all banks. We return to banks' cross-border activities in the next section on complexity.

We now consider the size of individual banks relative to total banking assets in their respective countries. ING Groep NV is at top with 88 per cent of Netherland's total banking assets. Among US banks, JP Morgan Chase was in first place with 13.5 per cent of the total. In the global context, however, JP Morgan Chase ranks only 57th among the world's biggest 100 banks. This means that the biggest US banks are relatively small when compared to the world's other 100 biggest banks on the basis of the share of an individual bank's total assets relative to all the banking assets in the bank's home country.

The other way to compare individual banks is by their total assets relative to the home country's GDP. (7) In this case, Danske Bank ranks number 1 with assets equal to 178 per cent of Denmark's GDP. JP Morgan Chase, with total assets equal to 13.6 per cent of US GDP, ranks 67th worldwide. In short, the biggest US banks are relatively small players in their own country when compared to many of their counterparts elsewhere in the world. Thus, US banks are less likely to be considered too big to save than many smaller banks in smaller countries.

Regardless of how they are organised, most of the world's big banks have a mix of businesses with very different kinds of assets as well as different involvement in cross-border activity. The mix of business areas within a bank is one factor affecting its complexity.

4. Dimensions of complexity

4.1. Complexity and systemic risk in the literature

In this subsection we discuss the link between complexity in our terminology and systemic risk, and review recent literature on this topic. We then discuss different dimensions of complexity with the FSB's criteria for G-SIBs as a starting point.

Not surprisingly, financial markets assess the value of big banks in very different ways. In the United States, for example, the ratio of market value to book value for the fifteen biggest banks ranges from a high of 3.66 for American Express, with a close second of 3.47 for Charles Schwab Corporation, to a low of 0.74 for Bank of America, as of the end of the second quarter, 2015. The markets clearly recognise that the fifteen big US banks represent a range of different business models, which suggests that they should not be viewed as the same when it comes to tackling the problem of too big to fail, especially any proposals to break them up. (8)

An important aspect of the complexity of a bank is lack of consistency between legal, functional and financial organisation. Carmassi and Tierring (2015) emphasise that most banks are organised functionally with little resemblance to the legal organisation. As a result, the valuation and identification of a failing (legal) entity's assets and liabilities becomes time-consuming and complex. (9) The operations of a failing functional entity may be conducted in a large number of different domestic and foreign legal entities, which at the same time serve other functional entities. Thus, one functional entity may not be easily resolved in insolvency proceedings without dragging other functional entities into the proceedings. The complexity of resolution is amplified if the bank operates cross-border in different legal jurisdictions.

Lack of financial independence of functional and legal entities can be added as another source of complexity in resolution. It is common that a BHC issues a large part of the debt that is being allocated to different functions and subsidiaries. It is also common that equity investment in subsidiaries is debt financed by the parent. This financial complexity is a source of financial contagion between separate legal entities.

Most large banks in the US and Europe are legally organised in hundreds, sometimes thousands, of legally separate subsidiaries with little resemblance to the functional organisation. In order to solve problems of excessive size and complexity we have to understand why banks have grown to the size and complexity they have. Carmassi and Herring (2015) refer to a number of possible reasons, including economies of scale and scope, regulation and tax rules.

Carmassi and Herring (2013) and Laeven et al. (2014) use the number of subsidiaries of a BHC as an important indicator of complexity. This measure clearly cannot capture all dimensions of complexity. Its value as a proxy can be related to the lack of correspondence between legal and functional organisation. If the legal subdivision into subsidiaries had corresponded to the functional subdivision of the firm, a large number of subsidiaries could have been an indicator of a simplified and more transparent organisation since each legal and functional unit could be resolvable as a stand-alone unit.

4.2. Dimensions of complexity across countries and banks

The FSB uses 'complexity' as one of several factors capturing systemic risk effects and for the designation of some banks as G-SIBs. The same criteria for identifying G-SIBs are used by the BCBS to determine the stricter regulatory capital requirements for such banks. The G-SIB designation is based on five separate characteristics: size, interconnectedness, substitutability and financial institution infrastructure, cross-jurisdictional activity, and complexity. Complexity in the terminology of the FSB depends only on OTC derivatives activity, trading and available-for-sale (AFS) securities, and Level 3 assets (BCBS, 2014).

Complexity in the FSB's terminology is strongly related to interconnectedness since derivatives as well as trading activities, which are captured in complexity, explain a large share of the number of links to other financial institutions. The FSB measures interconnectedness using the wholesale funding ratio, and intra-financial system assets and liabilities. Cross-jurisdictional activity is also treated as separate from complexity. Such activity contributes strongly to complexity as measured by the number of subsidiaries in the literature noted above.

Complexity also includes Level 3 assets, which are assets that are not traded in markets and can be assessed in terms of 'fair value' only by means of a bank-specific valuation model. Traditional bank loans would be included in this category. They cannot be easily valued and sold in bankruptcy. Thus, systemic risk arises both in traditional bank loans and involvement in securitisation and market-based activities more generally. The weights of different factors contributing to systemic risk are therefore important for the relative systemic risk ranking of banks with different business models.

Substitutability captures that contagion to the financial and real sectors which depends on the structure of the financial system and the existence of other institutions and services that may substitute for operations of a failing bank. Substitutability would decline if a large bank provides, for example, credit services through traditional bank lending as well as an underwriter of securities. The separation of commercial and investment banking would increase substitutability.

There is no similar international standard for the designation of a bank as a Systemically Important Financial Institution (SIFI), as discussed in Chouinard and Ens (2013). Each country chooses what is a SIFI and the specific regulation and supervision they are subject to. In the US the Dodd-Frank Act specifies an asset size of $50 billion as the borderline for being subject to special regulation and resolution procedures. There is an inconsistency between the criteria for designations of SIFIs and G-SIBs (see Barth and Sau, 2015).

As already noted, cross-border comparisons of complexity and systemic risk are difficult at either the individual or aggregate bank level. Largely because of differences in regulation between countries, the assets of big banks can include different mixtures of bank loans, securities, insurance policies and other products (see Barth, Caprio and Levine, 2006). This diversity is wide even among US BHCs. When one examines the composition of assets and liabilities of the world's 100 largest banks, it is found that the share of traditional bank assets (loans) and liabilities (deposits) varies even more across countries than across banks. The composition of revenues varies as well. Non-interest revenues and gains from trading and derivatives activities play a much greater role for US and UK banks than for most other large banks with the exception of the two large Swiss banks and Deutsche Bank.

Items that can be used to describe complexity according to our more general terminology incorporating organisation takes the form of subsidiaries, cross-border activity, interconnectedness and involvement in market-oriented, non-traditional banking activities. In particular, we find using complexity factors that are available for the 100 largest banks that size and the number of subsidiaries are significantly and positively correlated. Further results confirm the findings of Laeven et al. (2014) that size is not correlated with complexity in all its dimensions. The only significant positive correlations are with the number of subsidiaries, and derivatives and trading positions. Other dimensions of complexity are correlated to some extent with each other indicating that banks with relatively little involvement in traditional banking are strongly involved in securities markets, and in derivatives and trading.

We ask next if the factors associated with size and complexity also affect banks' risk of failure. This is done by performing regressions with each bank's accounting based Z-score and ROA as dependent variables, and with data for the 100 banks in 2014. We find that several variables indicating high complexity are associated with relatively high bank risk as well as systemic risk. Laeven, et al. (2014) found that size, number of subsidiaries and involvement in market-based activities increased systemic risk. With the exception of size, our results indicate that these variables are also associated with greater bank risk.

5. Addressing too big to fail and complexity

The purpose of the regulatory reforms being proposed or already being carried out is to prevent future banking crises whenever possible and to lessen the severity of those that do occur. (10) Too big to fail is viewed as a source of implicit protection of banks' creditors and, thereby, excessive risk-taking that contributes to the likelihood of a crisis. This situation is exacerbated, moreover, due to the moral hazard that arises when bank depositors are protected from losses by a government deposit insurance system. Complexity of bank organisations is also viewed as a source of excessive risk-taking as a result of opaqueness to supervisors and resolution authorities, as well as to other actors in the financial system. The problems associated with the resolution of large and complex banks may increase the severity of a crisis and the fiscal costs of crisis management.

The reforms attempt to tackle too big to fail in four ways: 1) restricting the size of banks; 2) restricting the scope of bank activities through separation of different activities into separate legal and functional entities (ring-fencing); 3) requiring higher capital levels for systemically important institutions; 4) providing an orderly framework for shutting down and resolving troubled banks while minimising the risk of contagion. The different approaches are not entirely independent. Restrictions on scale and scope can be difficult to separate. Some reform proposals may have effects in more than one of these dimensions. For example, so-called 'living wills' may affect both the scope and scale of activities, and the ability of authorities to resolve a failing bank.

5.1. Restricting the size of banks

The first type of reform involves restricting the size of banks. The Dodd-Frank Act limits the size of banks by prohibiting bank mergers or acquisitions if the resulting bank would hold more than 10 per cent of total nationwide bank deposits or more than 10 per cent of the aggregate consolidated liabilities of all financial companies. These limits could impede future mergers and acquisitions in the banking industry. (11)

As regards the potential impact of the merger restriction based on deposits in the US, only the largest US bank, JP Morgan Chase, already exceeds the limit on deposits and, therefore, would be prohibited from any further external growth. Other banks still have room for expansion, but it would be limited. Meanwhile, US Federal Reserve Governor Tarullo (2009) has suggested limiting non-deposit liabilities of US banks to a specified percentage of US GDP. If those liabilities were limited to 2 per cent of GDP, as proposed by some lawmakers in the Safe, Accountable, Fair and Efficient (SAFE) Banking Act of 2012, (12) five banks would be immediately prohibited from any further mergers and acquisitions. Another ten banks would have some leeway for further external growth.

Johnson and Kwak (2010, pp. 214-15) also state that "the simplest solution [to the TBTF problem] is a hard cap on size: no financial institution would be allowed to control or have an ownership interest in assets worth more than a fixed percentage of US GDP". They add that "as a first proposal, this limit should be no more than 4 per cent of GDP, or roughly $570 billion in assets today". For investment banks, they state that "as an initial guideline, an investment bank (such as Goldman Sachs) should be effectively limited in size to 2 per cent of GDP, or roughly $285 billion today." Based upon GDP in the second quarter of 2015, 4 per cent amounts to $716 billion. JP Morgan Chase and Bank of America are the only two banks that have assets that exceed this amount. Moreover, Goldman Sachs and Morgan Stanley both have assets that exceed 2 per cent of GDP, or $358 billion.

The problem, of course, is that there is no bright line that enables one to distinguish easily between big banks that pose a systemic risk and those that do not pose such a risk. To the extent that the demarcation line that is chosen is then adjusted to account for different degrees of 'bigness', the end result might once again lead back to a too big to fail problem. Indeed, the G-SIBs are of different asset sizes and some of these banks are smaller than other banks not identified as so. Moreover, some of the US banks are identified as G-SIBs even though their asset size is substantially less than 4 or even 2 per cent of US GDP.

A potential cost of limits on size is that they could reduce economies of scale in banking, not to mention economies of scope since increasing size often is associated with diversification from, for example, traditional banking to more market-based activities. Thus, size limits could increase the cost of banking services unless economies of scale and scope are not generated by reduced funding costs from being too big to fail. In this regard, Wheelock and Wilson (2012, p. 171) found that "... as recently as 2006, most U$ banks faced increasing returns to scale, suggesting that scale economies are a plausible (but not necessarily the only) reason for the growth in average bank size". In addition, Hughes and Mester (2011, p. 23) found "... evidence of large scale economies at smaller banks and even larger economies at large banks....". They added that these measured economies of scale did not result from implicit subsidies from being considered too big to fail. To the extent that US banks are limited in size, they may also be at a competitive disadvantage to the big banks in other countries that do not impose such limits. The implication is clear: one should not rush to limit bank size unless one can be confident that the benefits outweigh the costs. (13) International comparability of limits to bank size should also be considered.

To the extent that increasing scale is associated with diversification into additional financial services we must consider the evidence with respect to economies of scope.

5.2 Restrictions on the scope of activities

The second type of reform involves requiring banks to legally and functionally separate certain particularly risky activities or simply barring banks from those activities altogether. The Liikanen report produced by the European Commission proposes the separation of proprietary trading of securities and derivatives, and certain other activities linked to those markets, from deposit-taking banks within a banking organisation. The Vickers report proposes 'ring-fencing' of retail banking from wholesale/ investment banking in the UK. The ring-fenced banks would take retail deposits, provide payments and services, and supply credit to households and businesses. In the US, the Volcker Rule prohibits an insured depository institution or its affiliates from engaging in 'proprietary trading'. It also prohibits insured institutions from sponsoring or acquiring ownership interests in hedge funds or private equity funds.

There are several theoretical arguments underlying the mentioned separation proposals. One is that simpler banks pose less risk to the financial system and the broader economy, because some activities are inherently more risky and simpler organisations are more transparent for both market participants and bank supervisors. A second argument is that banks involved in, for example, investment banking and trading benefit from deposit insurance systems, which are intended primarily to protect creditors in the retail banking system. A third argument is that operational and financial separation of non-traditional banking activities would simplify resolution and allow entities involved in these activities to fail without the need to bailout creditors. Thereby, market discipline on risk-taking in these activities would be strengthened. In this sense restrictions on activities can be viewed as complementary to resolution procedures.

If these arguments in favour of restrictions on activities are accepted, it is still necessary to balance the benefits against potentially lost economies of scope. According to Laeven et al. (2014) there is evidence that there are diseconomies of scope when banks move from traditional commercial banking into market-based activities. These diseconomies would be generated by agency costs (Boot and Ratnovski, 2012) and potential conflicts of interest (Schmid and Walter, 2009). On the other hand, a large literature following Kroszner and Rajan (1994) points to information advantages of banks involved in both traditional banking and market-based activities.

The verdict on economies of scope and which particular activities benefit from them is still out. This is an argument for allowing the market to determine the most efficient organisational structure. However, the market will accomplish this only if there are no implicit subsidies in banking and banks are allowed to fail with consequences for shareholders and creditors.

The argument that ring-fencing of commercial banking from market-based financial services would enhance market discipline by removing explicit and implicit government support from market-based activities can also be questioned. The argument is based on the premise that governments are compelled to bail out commercial banks as sources of contagion but not financial firms involved in market-based activities. The financial crisis demonstrated that financial system contagion occurs through market-based activities as much as from traditional commercial banking. The pressure on governments to bail out, for example, investment banks may be as large as the pressure to bail out commercial banks. The Lehman Brothers experience may have strengthened this pressure. If so, the remaining benefits from ring-fencing could be increased transparency and lower resolution costs, while the costs could be reduced economies of scope.

Goodhart (2014) notes that proposals to dismantle universal banks into separate retail and wholesale parts are based on a "misreading of causes of the financial crisis". Similarly, the Volcker rule has been motivated by an opinion that proprietary trading was a significant factor in the recent financial crisis. However, the losses that led to problems at Lehman Brothers, Bear Stearns, IndyMac, Washington Mutual and other failed institutions were mainly connected to mortgage-backed securities and real estate, rather than to losses from the kind of trading that would be targeted by the Volcker Rule. (14) Nor is there clear evidence that separating commercial banking from investment banking as suggested by the Vickers report would increase safety. Despite strong separation between the two businesses in the 1980s under the Glass-Steagall Act, several big banks nevertheless almost failed because of bad loans in Latin America. Likewise, legions of savings-and-loans failed due to real estate loans. This suggests it is unlikely that simply reinstating Glass-Steagall would prevent problems at big banks in the future. In a sense, it is not even easy to pinpoint the problem that the Volcker Rule would solve. This is not to say that there could not be benefits from it. On the face of it, simpler institutions may become more transparent, less prone to excessive risk-taking and less subject to conflicts of interest. On the other hand, regulated reorganisation may lead to a conflict relative to incentives to exploit economies of scope. If so banks may try to avoid consequences of regulation in non-transparent transactions and activities.

Some evidence regarding trading losses might be helpful in this regard (see Barth and McCarthy, 2012). Since 1990, there have been fifteen instances when traders at different firms lost at least $ 1 billion (in 2011 dollars). The losses totalled nearly $60 billion and ranged from a low of $1.1 billion on ill-fated foreign exchange derivatives at a Japanese subsidiary of Shell Oil to a high of $9 billion on credit default swaps at Morgan Stanley. Four of the firms were banks, two were investment banks, two were hedge funds, one was a local government, and six were manufacturing or petrochemical firms. In other words, almost half the losses were not at financial services firms but at institutions that typically use financial products for hedging purposes.

Look at the same fifteen losses above in relation to the equity those institutions had at the time. The losses at the banks were less threatening to financial stability than those at the non-bank firms. Relative to equity, the largest losses were at non-banks. Thus, the Volcker Rule may be targeting the wrong firms. The more regulators limit banking activities, moreover, the more likely they are to create incentives for those same activities to take place in the so-called shadow banking system. In the process, however, risks may also shift from the banking industry to the shadow banking system.

The post-crisis regulatory regime embodied in the Dodd-Frank Act does not seek to break up big banks or to reinstitute Glass-Steagall barriers between commercial and investment banking. This perhaps reflects the observation that the failures of banks in the crisis are not well correlated with the end of the Glass-Steagall restrictions. Bear Stearns and Lehman Brothers both suffered failures but both were essentially pure investment banks. By contrast, JP Morgan Chase combined investment and commercial banking but weathered the crisis well. An alternative to Glass-Steagall-like restrictions would be for regulators to focus on activities that appear to pose particular risks, and to act more pre-emptively to head off systemic problems. This approach is embodied in the creation of the Financial Stability Oversight Council, an umbrella group of federal regulators that is meant to watch over the entire financial system. One, however, is right to question whether this new approach will indeed be successful. (15)

The relatively strong case for separation in the form of operational and financial ring-fencing exists for cross-border banking since there is no agreement on how to coordinate conflicting interests and resolution procedures. The case for economies of scale and scope across borders appears weak. Operational ring-fencing would make it possible to allow an entity in one country to fail without serious operational repercussions for other parts of a financial group. New Zealand has implemented an operational ring-fencing rule stating that subsidiaries of foreign banks must be operationally separable within 24 hours. Financial ring-fencing would protect the capital of a bank in one country from failures in other countries. Several countries including the US have implemented such ring-fencing for foreign branches as well as subsidiaries.

As a general observation, complexity would be reduced by greater correspondence between functional and legal organisation. To achieve such reorganisation based on scale and scope economies among functions it is necessary to obtain a better understanding of why banks choose to organise with great complexity. As noted, Carmassi and Herring (2015) refer to banks' ability to take advantage of regulatory frameworks and tax rules in combination with economies of scope and scale to explain the current structure, but explicit knowledge of this issue is lacking.

5.3. Requiring higher capital levels for systemically important banks

The third type of reform involves requiring banks to hold additional equity capital. This is meant to ensure that firms have a bigger buffer against losses and a greater ability to survive a crisis. More equity capital would also provide more protection for taxpayers against future bailouts. The guidelines for new and more stringent capital requirements under Basel III are phased in over the period 2013 to 2019. The leverage ratio will be 3 per cent in 2019, while the risk-based capital requirement will be as high as 13 per cent for some banks, and even as high as 16.5 per cent for G-SIBs. The FSB, moreover, has confirmed its final proposal for Total Loss Absorbing Capacity (TLAC) calling for G-SIBs to hold, on top of the required minimum CET1 of 4.5 per cent, an additional 11.5 per cent of 'loss absorbency' in the form of Tier 1 and Tier 2 capital relative to risk-weighted assets, rising to 13.5 per cent by 2022. Importantly, TLAC is focused on meeting this requirement in part through long-term, unsecured debt that can be converted into equity when a bank fails or reaches a critical market-value trigger (contingent convertible debt or CoCos) (16). This is meant as an additional measure to put an end to too big to fail by forcing bondholders to inject capital into a big bank that fails rather than taxpayers.

In a first, the Basel III agreement among international bank regulators, as just noted, calls for a minimum leverage ratio. This ratio is not risk-based, like the other capital guidelines. According to Haldane (2012, p. 19), "... the leverage ratio [should play] the frontstop role [in Basel III] given its simplicity and superior predictive performance". He adds that "the more complex the bank, the stronger is this case". Furthermore, we concur with Hoenig (2012), who states that "an effective capital rule should result in a bank having capital that approximates what the market would require without the safety net in place. The measure that best achieves these goals is what I have been calling the tangible equity to tangible assets ratio". During the US financial crisis, this seemed to be the only ratio that anyone paid attention to insofar as banks in general were amply capitalised by nearly all the other capital ratios.

Additional capital requirements for big or systemically important banks provide a disincentive for size (and perhaps also for complexity or interconnectedness). These requirements might also be seen as a 'tax' that offsets the possible funding advantages of big and complex banks - a disincentive for size and complexity as defined by the criteria for G-SIBs and SIFIs. How blunt this 'tax instrument' is as a disincentive for size and complexity depends on how capital requirements depend on size and complexity. SIFIs are so far identified primarily by size, while G-SIBs are identified by complexity characteristics as well. We have discussed the difficulty of analysing the relation between complexity and systemic risk. Another difficulty is to specify a relationship for the marginal tax (capital requirement) of increasing size and complexity. If a tax is imposed only at specific size and complexity triggers, the marginal tax of increasing size and complexity beyond the trigger is zero. In this case the capital requirement tax is ineffective as a disincentive to increase size and complexity. The extra capital requirement tax becomes primarily a payment for the latent negative externality, though the implicit revenue from the tax accrues to private suppliers of capital rather than to the government.

As already noted, it should be kept in mind that big banks provide benefits as well as costs to society, a point discussed by the Clearing House Association (2011). Moreover, the capital charge, as usual with a tax, results in a deadweight loss in the form of reduced lending and economic activity. The quantitative importance of this impact remains a subject of considerable debate. Admati et al. (2010, 2013) see little negative impact of higher capital requirements and recommend an equity ratio as high as 30 per cent. Miles et al. (2013) argue along with Admati et al. that an optimal capital ratio is much higher than current levels. But Kashyap, Stein, and Hanson (2010) see a meaningful impact on bank funding costs during the transition period as banks raise additional equity capital, and then a modest ongoing impact. Research by regulators points to modest impacts, while banks and their associations point to greater impacts. In the wake of the recent crisis, it is certain that big banks will hold more capital, both at the insistence of regulators and of their own volition. Given the considerable changes in the banking industry and its more stringent regulation, the ongoing impacts of higher capital standards will be understood only over time.

One danger of setting the capital requirement very high as recommended by Admati et al. (2010), is that it creates a large discrepancy between required and desired equity financing. Such a discrepancy is likely to create strong incentives to manipulate, evade and avoid the requirement. The stronger these incentives are the greater are the costs to supervise and examine banks' compliance with the regulation. These costs are likely to be particularly high within the Basel Capital Accord framework for risk-weighting based on internal models that enable substantial manipulation of risk-estimates.

5.4. What to do about big bank failures?

The fourth type of reform involves changes to the framework for dealing with the collapse of big or systemically important banks. There are two motivations behind such policies: first, to better ensure the stability of the system; second, to alert market participants that banks are more likely to be allowed to fail and that creditors will be forced to take losses. That awareness may help remove advantages that big banks have previously enjoyed by being perceived as too big to fail and, thereby, strengthen market discipline on banks' risk-taking. (17)

The Dodd-Frank Act requires banks to devise their own 'living wills', or plans for possible recovery of a bank's core business by disinvestment in other financial activities and an orderly shutdown if the bank actually fails. This could prove to be a symbolic step, because no one knows how or if the plans will work in the event of an actual crisis. Even so, however, the preparation of a living will may provide an additional signal that regulators will let banks collapse rather than bail them out in the future. Living wills are subject to regulators' approval and they allow regulators to intervene to separate activities. Thus, the living wills have the potential to lead to a process of reduced complexity and greater correspondence between legal and functional organisations.

The new orderly liquidation authority in the Dodd-Frank Act, as well as the Bank Recovery and Resolution Directive within the EU's banking union, could fundamentally change the way in which failures at big banks are resolved. (18) Bondholders and other creditors are now more likely to incur losses if a bank fails, even though the deployment of government resources is allowed to support a bank and slow its demise through the Orderly Liquidation Fund. Absent additional congressional action (which is now hard to imagine, given the unpopularity of the Troubled Asset Relief Program), in the case of a future failure of a big bank that involves the resolution of the holding company beyond simply the insured depository institutions, bondholders will incur losses. The EU's Bank Recovery and Resolution Directive also allows for the deployment of a fund in resolution while specifying a minimum of 'bail-in' of unsecured creditors.

It is difficult to predict how the new resolution authority under the Dodd-Frank Act and the Single Resolution Mechanism in the EU will be used. (19) In the US the FDIC might use its new authority to arrange a debt-for-equity swap that recapitalises the failing bank, turning the former bondholders into the new owners. The possibility of having such a swap imposed on bondholders should affect the terms under which potential creditors are willing to provide funding to banks that might be put through a resolution. One risk is that the new resolution authority could give providers of funding an incentive to flee at the first hint of trouble. The threat of such bank runs is an important disciplining device, but it could also lead to more hair-trigger responses and inadvertently prove destabilising.

Either way, however, the resolution authority will be incomplete and perhaps unworkable until there is more international coordination of bankruptcy regimes. In the case of Lehman's failure, for example, the UK bankruptcy regime disrupted the operations of many US-based firms when it froze their overseas assets. Of the thirteen US banks of the 100 biggest banks for which information is available, seven have foreign assets. Among the latter banks, Citigroup has the largest share of foreign assets at 46 per cent, while Capital One has the smallest share at 4.5 per cent. Importantly, international coordination of regulatory regimes for both normal times and during resolution or bankruptcy procedures will be crucial for the continued evolution of the global financial system. (20)

One aspect of international coordination is the issue of single-entry point vs multiple entry point resolution. Strongly integrated banks without clear separation of banks across countries and activities require single-entry point resolution. To achieve agreement on a single entry point for international banks it is necessary that regulators in countries where the banks are operating mutually recognise resolution regimes. Conflicting interests of countries in a crisis situation implies that such mutual recognition is currently not in sight.

The organisational complexity discussed here has led to a discussion of 'single-entry point' vs 'multiple entry point' resolution procedures. A single entry point for a large global bank implies recognition of the near-impossibility of resolving individual functional and legal entities. It is clearly appropriate if a bank is organised as a universal bank supplying a variety of financial services within different divisions of one legal unit. The different divisions can be thought of as branches without financial independence and functionally integrated to the extent that there are economies of scope to be exploited.

At the other extreme of bank organisation we can imagine a large number of subsidiaries defined by separate functions. Firewalls would separate the operations and the subsidiaries would be responsible for their own debt financing. The different units would essentially be stand-alone units with common ownership. Multiple entry point resolution is appropriate for this type of organisation. There would be relatively little complexity since each legal unit could default on its own without severe repercussions for other subsidiaries within the banking group. On the other hand, in this organisation economies of scope between functional areas cannot be exploited. The systemic risk consequences of a failing entity within this kind of bank organisation are relatively easier to manage.

A key problem facing the Orderly Liquidation procedures in the US as well as the bail-in rules in the EU's Single Resolution Mechanism is to achieve credibility that creditors of large banks will not be bailed out in a crisis. Loopholes in the rules for activation of the procedures will always exist and allow regulators to bailout creditors for fear of contagion. It is possible that a large bank must fail without a bailout in order to achieve credibility of the resolution procedures. (21)

6. Summary and conclusion

That some banks are too big to fail is not new. Neither is the challenge for policymakers to implement reforms that eliminate the need to bailout big banks. The regulatory regimes for big banks in many countries are undergoing changes from those that prevailed before the global financial crisis. Banks will now be required to hold more capital, have more robust access to liquidity, undergo increased regulatory scrutiny, and face restrictions on certain activities. Many of these changes are still evolving as some reforms are being implemented and proposals for additional reforms are still being evaluated.

A fundamental problem is that we do not understand well why the world's financial systems are dominated by relatively few large and organisationally complex banks. It is uncertain whether complexity to a large extent is driven by economies of scope, by the implicit subsidy associated with high bailout probability or by tax incentives and the design of regulation, including capital requirements. The most effective approach to complexity cannot be assessed without greater knowledge of these factors.

Removal of the implicit subsidy associated with size and complexity would provide information and incentives for banks to organise themselves over time based on economies (and diseconomies) of scale and scope. Current regulatory approaches to complexity include requirements for separation or ring-fencing of particular activities, implementation of resolution procedures that would enable banks to fail with a minimum of bailouts of creditors and a minimum of contagion, and additional capital requirements for systemically important banks.

There is little doubt that the various measures to separate activities within banking groups, including living wills, will induce large banks to adjust their organisation. The extent to which banking structures will be simplified sufficiently to enable resolution of failing banks without bailouts depends very much on regulators' and supervisors' behaviour. The credibility that the new procedures will be applied in a crisis of one or more large banks has not been established, and most likely will not be until there is another crisis.

Big banks do possess considerable power that may be used to influence the regulatory authorities to pursue policies that increase the risk of a systemic crisis. The regulatory authorities, moreover, may also pursue such policies based upon a bias in favour of banks. In the absence of evidence that the benefits exceed the cost of breaking up big banks, policymakers may simply have to monitor the incremental reforms they have already begun to implement and make adjustments as the results become available. Barth, Caprio and Levine (2012) point out that given the poor past performance of the regulatory authorities, it may also be prudent to establish procedures to hold them more accountable for achieving stability in the future.

NOTES

(1) See Wihlborg (2012).

(2) See Barth et al. (2012) for an analysis of regulatory capture in the US.

(3) See Barth, Prabha and Wihlborg (2015).

(4) See European Shadow Financial Regulatory Committee, Statement No 39, 14 November, 2014.

(5) Philip Davis (1995) provides an overview of the evolution of financial systems and their fragility since 1970 with an emphasis on the period 1988-93.

(6) Beck et al. (2014) find that GDP growth is strongly linked to financial intermediation while financialisation is not.

(7) One might wish to distinguish between the ratio of domestic assets to domestic GDP and the ratio of foreign assets to the GDPs in which the assets are located.

(8) For additional information on different business models of European banks, see Ayadi et al. (2012).

(9) Carmassi and Herring (2013) discuss how the Lehman Brother's bankruptcy was complicated and contagious among different Lehman subsidiaries as a result of the way assets and liabilities were booked in subsidiaries without correspondence to the functional organisation.

(10) This section draws upon Barth, Prabha and Swagel (2012).

(11) The Dodd-Frank Act provides exceptions to these limits in the case of mergers and acquisitions of troubled institutions.

(12) This is a bill introduced by Senator Sherrod Brown (see http:// www.brown.senate.gov/newsroom/press/release/ brown-introduces-bill-to-end-too-big-to-fail-policies-prevent-mega-banks-from-putting-our-economy-at-risk).

(13) See Saunders and Walter (2010) for a more general discussion on this issue.

(14) See Barth et al. (2009).

(15) For further discussion of this and related issues, see Barth, Caprio and Levine (2012).

(16) See Calomiris and Herring (2013).

(17) See Mullineux (2012) for a discussion of the role of governance with respect to too big to fail banks.

(18) For a discussion of resolution procedures in other countries, see Wihlborg (2012).

(19) Brown and Din? (2011) find that a government is less likely to take over or less likely to close a failing bank if other banks in that country are weak. They further argue that this too-many-to-fail effect was present in the US Savings and Loan Crisis of the 1980s and the Japanese Banking Crisis of the 1990s.

(20) See, for example, Prabha and Wihlborg (2012) for a discussion of this issue as it relates to global bank organisational structure.

(21) See, for example, Barth et al. (2015) and European Shadow Financial Regulatory Committee (2014).

REFERENCES

Admati, A. and Hellwig, M. (2013), The Bankers' New Clothes, Princeton University Press.

Admati, A.R., DeMarzo, P.M., Hellwig, M.F. and Pfleiderer, P. (2010), 'Fallacies, irrelevant facts, and myths in the discussion of capital regulation: why bank equity is not expensive', Stanford GSB Research Paper No. 2063, September.

Ayadi, R., Arbak, E. and De Groen, W.P. (2012), Regulation of European Banks and Business Models: Towards a New Paradigm? Center for European Policy Studies.

Barth, J.R., Caprio, G. and Levine, R. (2006), Rethinking Bank Regulation: Till Angels Govern, Cambridge University Press.

--(2012), Guardians of Finance: Making Regulators Work for Us, MIT Press.

Barth, J.R., Li, T., Lu, W., Phumiwasana, T. and Yago, G. (2009), The Rise and Fall of the US Mortgage and Credit Markets: A Comprehensive Analysis of the Meltdown, John Wiley & Sons.

Barth, J.R., Li, T., Prabha, A.P., Chiang, Y. and Li, L. (2012), 'Industrial loan companies: where banking and commerce meet', Financial Markets, Institutions & Instruments, 21 (1), pp. 1-69.

Barth, J.R. and McCarthy, D. (2012), 'Trading losses: a little perspective on a large problem', Milken Institute, November.

Barth, J.R., Prabha, A.P. and Swagel, P. (2012), 'Just how big is the too big to fail problem?', Journal of Banking Regulation, 13, pp. 265-99.

Barth, J.R., Prabha, A.P. and Wihlborg, C. (2015), 'The Dodd-Frank Act: key features, implementation progress, and financial system impact', in Barth, J.R. and Kaufman, G.G. (eds), The First Great Financial Crisis of the 21st Century: A Retrospective, December.

Barth, J.R. and Sau, M. (2015), 'Designating bank SIFIs: an arbitrary threshold for risk', Milken Institute Report, November.

Beck, T., Degryse, H. and Kneer, C. (2014), 'Is more finance better? Disentangling intermediation and size effects of financial systems 'Journal of Financial Stability, 10, February, pp. 50-64.

Boot, A.W.A. and Ratnovski (2012), 'Banking and trading', IMF Working Paper, 12/238.

Brown, C.O. and Ding, I.S. (2011), 'Too many to fail? Evidence of regulatory forbearance when the banking sector is weak', Review of Financial Studies 24.4, pp. 1378-405.

Calomiris, C.W. and Herring, R.J. (2013), 'How to design a contingent convertible debt requirement that helps solve our too-big-to-fail problem', Journal of Applied Corporate Finance, 25, Spring, pp. 39-62.

Caprio, G., Klingebiel, D., Laeven, L. and Noguera, G. (2005), 'Banking crisis database', in Honohan, P. and Laeven, L. (eds), Systemic Financial Crises: Containment and Resolution, New York, NY: Cambridge University Press.

Carmassi, J. and Herring, R.J. (2013), 'Living wills and cross-border resolution of systemically important banks 'Journal of Financial Economic Policy, 5, 4.

--(2015, forthcoming), Corporate Structures, Transparency and Resolvability of Global Systemically Important Banks, Systemic Risk Council, Jan.

Carrington, T. (1984), 'US won't let 11 biggest banks in nation fail', Wall Street Journal, 20 September.

Chouinard, E. and Ens, E. (2013), 'Assessing the systemic importance of financial institutions', Bank of Canada, Financial System Review, December.

Clearing House Association (2011), 'Understanding the economics of large banks', November, http://www.theclearinghouse.org/ index.html?f=073071.

Davis, E.P. (1995), Debt, Financial Fragility and Systemic Risk, Oxford University Press.

European Shadow Financial Regulatory Committee (2014), 'Complexity and credibility in the Single Resolution Mechanism', Statement No 39, London, 14 November, www.esfrc.eu.

Goodhart, C. (2014), 'The parlous state of macroeconomics and the optimal financial structure', International Review of Financial Analysis, 36, pp. 78-83.

Haldane, A.G. (2012), 'The dog and the frisbee', Bank of England, August.

Hoenig, T.M. (2012), 'Back to basics: a better alternative to Basel capital rules', speech to the American Banker Regulatory Symposium, Washington, D.C., 14 September, http://www.fdic. gov/news/news/speeches/chairman/spsep_1412_2.html.

Honohan, P. and Klingebiel, D. (2003), 'The fiscal cost implications of an accommodating approach to banking crises', Journal of Banking & Finance, 27, (8), August, pp. 1539-60.

Hughes, J.P. and Mester, L.J. (2011), 'Who said large banks don't experience scale economies? Evidence from a risk-return-driven cost function', Working Papers No. 11 -27, Federal Reserve Bank of Philadelphia.

International Monetary Fund (2014), Global Financial Stability Report, April.

Johnson, S. and Kwak, J. (2010), 13 Bankers: The Wall Street Takeover and the Next Financial Meltdown, New York: Pantheon.

Kashyap, A.K., Stein, J.C. and Hanson, S. (2010), 'An analysis of the impact of 'substantially heightened' capital requirements on large financial institutions'. Harvard University mimeo, May.

Kaufman, G. (2002), 'Too big to fail in banking: what remains?', The Quarterly Review of Economics and Finance, 42, pp. 423-36.

Kroszner, R.S. and Rajan, R.G. (1994), 'Is the Glass-Steagall Act justified? A study of the US experience with universal banking before 1933', American Economic Review, 84, (4), pp. 810-32.

Laeven, L., Ratnovski, L. and Tong, H. (2014), 'Bank size and systemic risk', IMF Staff Discussion Note, SDN/14/04.

Miles, D., Yang, J. and Marcheggiano, G. (2013), 'Optimal bank capital', The Economic Journal, 3, pp. 1-37.

Mullineux, A. (2012), 'The governance of 'too big to fail' banks' in Barth, J.R., Lin, C. and Wihlborg, C. (eds), International Research Handbook on Banking and Governance, Edward Elgar Publishing.

Prabha, A. and Wihlborg, C. (2012), 'Cross-border banking in subsidiaries and branches; organization, supervision and resolution', in Caprio, G. (ed.), Encyclopedia of Financial Globalization, Elsevier.

Saunders, A. and Walter, I. (2010), 'Financial architecture, systemic risk, and universal banking', Financial Markets and Portfolio Management, 24, I, edited by Markus Schmid, Springer, March.

Schmid, M.M. and Walter, I. (2009), 'Do financial conglomerates create or destroy economic value?', Journal of Financial Intermediation, 18, (2), pp. 193-216.

Shull, B. (2010), 'Too big to fail in financial crisis: motives, countermeasures, and prospects', the Levy Economics Institute of Brad College Working Paper, No. 601.

Tarullo, D.K. (2009), 'Confronting too big to fail', speech at the Exchequer Club, Washington, D.C., 21 October.

Wheelock, D.C. and Wilson, P.W. (2012), 'Do large banks have lower costs? New estimates of returns to scale for US banks', Journal of Money, Credit and Banking, 44(1), pp. 171-99.

Wihlborg, C. (2012), Developing Distress Resolution Procedures for Financial Institutions', SUERF (Societe Universaire Europeenne de Recherches Financiers) Study, 2012/4, August.

James R. Barth * and Clas Wihlborg **

* Lowder Eminent Scholar in Finance at Auburn University, and Senior Fellow at Milken Institute. E-mail: barthjr@auburn.edu. ** Fletcher Jones Chair of International Business at Chapman University. E-mail: wihlborg@chapman.edu. This is a short version of a much longer paper with tables and figures, which is available on request. The authors are grateful for helpful comments from two anonymous referees and the excellent assistance of Yanfei Sun.