1 A theory of debt

The role of debt and debt contracts has been one of the focal points in corporate finance. Thus, going through the main theoretical explanations of debts and the role of financial contracting is crucial for understanding the role of new bail-inable instruments.

Corporate finance scholars firstly approached the debt from the perspective of the firm, investigating its optimal capital structure (ie which combination of equity and debt financing maximises the value of the firm). Modigliani and MillerFootnote ¹¹ demonstrated that, under a certain set of assumptions,Footnote ¹² the choice between equity and debt do not affect the overall value of the firm (the irrelevance proposition).Footnote ¹³ This basic yet revolutionary idea has been refined, questioning its assumptions. For our interests here, the refinements related to agency and incomplete contract theory are discussed. In fact, agency theory, among other features, explains the fundamentals of the role of debt-holders in the corporation, while incomplete contract theory takes a step forward, explaining why people decide to lend money, and the crucial role that contracts have on that decision.

Jensen and Meckling, relaxing the assumption of symmetric information, showed that the risk (and costs) of bankruptcy are partly shifted from shareholders to bondholders (risk-shift). Therefore, being shareholders in charge, risk-taking tends to be excessive since they bear only a small fraction of the expected bankruptcy costs.Footnote ¹⁴ “Gambling for resurrection” incentives represent the extreme consequence of this phenomenon: once financial distress approaches, shareholders are rationally prone to undertake negative net present value projects, yielding high returns with low probabilities, since they would fully gain the benefits without bearing any costs.Footnote ¹⁵

Building on agency theory, the “costly contracting hypothesis”Footnote ¹⁶ of debt argues that controlling the bondholder-shareholder conflicts can increase the value of the firm. In other words, creditors can play a positive role in the governance of the firm. In fact, although financial contracting is costly, parties are willing to bear these costs (eg draft a covenant) as long as the benefits deriving from financial contracting outweigh the bargaining and enforcement costs, at the margin.

The literature analysed so far explains the role of debt for the value of the firm, but fails to expain why people are willing to lend and become creditors. This link is crucial, as corporate governance has been defined as the ways in which suppliers of finance (including creditors) to corporations assure themselves of getting a return on their investment.Footnote ¹⁷

In the understanding of how creditors assure themselves of getting a return on their investment, the introduction of contract incompleteness signalled a groundbreaking change.Footnote ¹⁸ Aghion and Bolton modelled debt contracts as a mechanism to allocate control, contingent on future situations that are difficult or impossible to describe ex ante.Footnote ¹⁹ Accordingly, the capital structure of a firm is relevant for managerial incentives as long as contracts are incomplete, ie their enforcement is constrained by the lack of compelling provisions for each and every possible course of action.

Contingent allocation of control evokes, first and foremost, statutory bankruptcy, where the ability of the shareholders to retain control is contingent on meeting the debt obligation. Apart from that, one can think of debt covenants restricting the courses of action available to the firm and thus limiting the control power of shareholders.Footnote ²⁰ Again, secured debt can be seen as a way of allocating control over specific assets to a creditor, contingent on default of a contractual obligation.

2 Bank debt and its peculiarities

In the specific context of banks, the understanding of contingent allocation of control needs to be broadened, since debt is excessively difficult to restructureFootnote ²¹ so that short-term debt can also be seen as a way of allocating control.Footnote ²² In fact, given the impossibility of designing long-term contracts ex ante, parties agree on short-term exposures with a view to rolling them over as long as the firm’s situation does not deteriorate.Footnote ²³

Considering this peculiarity, one could also wonder whether, and to what extent, such a theory of debt applies to banks. In fact, banks are acknowledged to be, somehow, “special” for a wide array of reasons, among which at least two are crucial here.Footnote ²⁴

First, banks perform a maturity transformation of their assets, as they borrow money short term and lend long term, exposing banks to (retail and wholesale) run.Footnote ²⁵ This provides a sound argument for deposit guarantee which, in turn, begs for intrusive regulation substituting for the lack of monitoring.Footnote ²⁶ As a by-product, banks implicitly rely on public bailouts, strengthening the moral hazard problem.

Second, large and interconnected banks generate systemic risk, meaning that the failure of a financial institution results in much bigger costs than the ones borne by the institution, as the effects of one failure propagate to the rest of the financial system (eg contagion risk, trust issues, etc).Footnote ²⁷

The post-crisis wave of reforms has confronted both of these issues, introducing the concept of loss of absorbency capacity of banks’ liabilities, ie banks should bear in their balance-sheets sufficient instruments to be internally recapitalised if in trouble, minimising the impact on financial stability.Footnote ²⁸ Those instruments can be either capital, as defined by the Basel III accords,Footnote ²⁹ or non-capital instruments that are apt to absorb losses.Footnote ³⁰ What makes an instrument apt to absorb losses are its qualitative characteristics, ie its contractual arrangements.

In the EU, the regulatory framework for loss-absorbing instruments is composed by the Capital Requirement Regulation (CRR),Footnote ³¹ laying down qualitative and quantitative requirements for capital instruments; and the BRRD, dealing with the qualitative requirements for non-capital instruments that are considered “eligible” to be counted as bail-inable liabilities.Footnote ³² Once the bank goes into financial distress and becomes “failing or likely to fail”,Footnote ³³ the resolution authority can decide to write-down or convert bail-in eligible claims through the so-called bail-in mechanism, recapitalising the bank without the need for a State bailout.Footnote ³⁴

Therefore, from an ex ante perspective, loss-absorbing instruments are designed to work as a self-insurance buffer against the risk of financial distress, insofar as the losses arising from financial distress are allocated to investors in bail-inable securities and not to the general public. In a world of complete contracts,Footnote ³⁵ this should also optimise the risk-taking incentives of bankers, since each future course of action would be governed in a manner that is consistent with the fulfilment of the contractual obligations attached to bail-inable securities, ie not becoming “failing or likely to fail” so to avoid losses for bail-inable debt-holders.Footnote ³⁶

Nonetheless, in a world of incomplete contracts such an outcome is dubious, bearing in mind that qualitative requirements on loss-absorbing instruments and the administrative discretion of the resolution authority further constrain the possibility of efficiently allocating control. This paper, therefore, discusses the qualitative requirements for bail-in eligible liabilities from the perspective of the investor, ie as a set of constraints on contractual possibilities.

1 A brief overview of qualitative requirementsFootnote ³⁹

The BRRDFootnote ⁴⁰ requires institutions to meet at all times a Minimum Requirement for Own FundsFootnote ⁴¹ and Eligible LiabilitiesFootnote ⁴² (MREL). Quantitatively, MREL is a tailor-made requirement, while the amount of own funds is statutorily required in the CRR.

Before individually addressing each class of items, three general features should be highlighted. First, for absorbing losses, eligible instruments must provide a genuine contribution of wealth; thus, they cannot be owned by, secured by, funded by or guaranteed by the issuing institution. In fact, if that happened, the losses would be indirectly borne by the issuer itself. Second, eligible instruments are subordinated to all the others and, within eligible instruments, each class of instruments has different seniority. Third, Art 48 BRRD regulates the sequence of subordination, allocating the losses to CET1 holders first, then AT1, T2 and eventually other eligible liabilities, as described in Figure 1.

Figure 1 Capital and bail-in eligible liabilities waterfall

For explanatory purposes, non-Tier 2 instruments are analysed first, and subsequently Tier 2 and Additional Tier 1. Indeed, to be eligible, junior liabilities must comply with senior liabilities’ requirements, and with further requirements specific to the senior class. The analysis covers rules on subordination, maturity of instruments, repayments and distributions. All the requirements are then summarised in Figure 2.

Figure 2 Outline of the main qualitative requirements for bail-in eligible liabilities

MREL eligible instruments must have a remaining maturity of at least one year.Footnote ⁴³ For instance, a bond whose maturity is 31 January 2025 will be fully counted as MREL eligible until 31 January 2024. Moreover, Art 44(2) BRRD exempts from the application of bail-in some classes of liabilities that are incompatible with the aim of the tool, such as covered deposits.Footnote ⁴⁴ MREL eligible instruments are not required to comply with specific rules on repayments of principal amount and distribution of interests, while these aspects represent a crucial part of qualitative requirement for more junior capital instruments.

Coming to capital instruments, Tier 2 liabilities must be issued with an initial maturity of at least five years.Footnote ⁴⁵ The instrument can be redeemed (called, repurchased or repaid) in advance after five years from the issuance and with prior supervisory approval.Footnote ⁴⁶ No contractual provisions can provide incentives for early repayment;Footnote ⁴⁷ and no contractual clause can give the holder the right to accelerate the repayment of interests and the principal amount of the instruments.Footnote ⁴⁸

Whereas AT1 and T2 items share the provisions governing redeemability, and constraints on contractual clauses, the main difference between the two lies in the perpetuity of AT1 items. Moreover, distributions shall be paid out of distributable items (profits, retained distributable earnings) and the issuer has full discretion in cancelling distributions

Lastly, loss-bearing provisions out of a resolution procedure complement the principle of seniority in resolution.Footnote ⁴⁹ In fact, according to the CRR framework, AT1 items shall be written down or converted into CET1 when the level of CET1 falls below 5.125% of RWA, allowing the institution to have new equity when it would be difficult or too expensive to raise. The trigger event for write down or conversion can be higher than the statutory one, if provided in the contract.Footnote ⁵⁰

2 Compliance of contractual devices

This section aims to bring together the framework developed Section II and the analysis of the qualitative requirements discussed in section 1 above. It argues that many of the usual contractual devices are largely unavailable in banking because such devices do not comply with existing regulation.

First of all, it is easy to rule out the possibility of complying with capital and MREL requirements through short-term or secured debt for financial stability purposes. In fact, short-term exposures are prone to run in case of idiosyncratic or systemic turmoil. On the other hand, the stability of financial institutions, as well as the feasibility of resolution procedures, beg for long-term non-runnable items on which other short-term investors, depositors and the market as a whole can rely. In a similar vein, eligible liabilities need to actually absorb losses. Therefore, secured items are radically incompatible with the aim of capital regulation, and the principles of resolution, as the loss-absorbency capacity would only be fictional, and the losses would still be allocated to the institution.

The situation is more entangled when it comes to covenants. Covenants’ salience can be grasped by referring to Patrick Bolton: “In short, the area where the issue of contractual incompleteness perhaps matters most, when it comes to debt contracts, is the design and enforcement of debt covenants”.Footnote ⁵¹

First, for the sake of financial stability and public trust, using a covenant breach as an event of default to force debt renegotiation represents an unavailable option.Footnote ⁵² In fact, contractually granting the opportunity to force a financial institution into bankruptcy in case of a breach would provide the creditor with disproportionate power, putting her in the position to generate significant negative externalities for the whole economy. Covenants can also have other functions, namely to speed up repayment in case of a breach, so to end an exposure that is considered to be too risky and to impose behavioural remedies upon covenant breaches.Footnote ⁵³

As previously described, speeding-up repayment is unavailable for capital instruments, while there are no constraints for non-capital eligible items. One could say that the regulator has extensive control over capital items both through regulatory tools, such as the ban on the acceleration of repayment, and supervisory tools, such as the necessary permission to call items earlier than the original maturity.

Instead, behavioural covenants (eg restriction on payment of dividends) are, in principle, compliant with the requirements of both capital and non-capital items. Nonetheless, it should be noted that covenants are costly to draft and enforce. Therefore, it might well be that, even if available, their costs outweigh the benefits. In fact, banks assets and operations are, by definition, opaque, so monitoring costs are higher than in non-financial firms. In addition, there is a set of rules mimicking the possible behavioural covenants that could be imposed:Footnote ⁵⁴ think, for instance, about the restrictions on dividends provided by Art 141 of the Capital Requirement Directive (CRD IV).Footnote ⁵⁵ Thus, in comparison with non-financial firms, imposing covenants would be more costly, because of the higher cost of compliant drafting, monitoring and enforcement. On the other hand, benefits, if any, would amount to the marginal increase of mismanagement deterrence that a tailor-made covenant can yield on top of the supervisory rules.Footnote ⁵⁶ To sum up, consistent with the costly contracting hypothesis, qualitative requirements leave sufficient room for covenants even if complementary rules and the nature of financial institutions shrink the scope for efficient covenants.

Finally, some attention has to be devoted to the hypothesis of statutory bankruptcy and the special ways in which it materialises for banks. Once the bank reaches the point of non-viability, there are two main options: liquidation (gone concern) and resolution (going concern).Footnote ⁵⁷ For what is of interest here, gone concern options can be neglected, as the procedure of bank liquidation has no governance implication, ie no contingent shift in control. On the contrary, in a going concern scenario, investors can have a potentially relevant governance impact when the resolution authority decides to convert eligible liabilities into shares.Footnote ⁵⁸

In fact, some authors have argued that converting debtors into shareholders may influence the incentives of managers towards risk-taking.Footnote ⁵⁹ Moreover, it is likely that the conversion of other instruments follows the heavy dilution or even the complete write-off of outstanding shares.Footnote ⁶⁰ Hence, a radical shift in control would happen. Some preliminary observations are worth discussing.

First, as opposed to clearly defined economic rights, governance characteristics of converted shares are not clearly regulated. Indeed, eligible items ought to be converted into common shares, as preferences in distributions and repayment are not allowed under the subordination requirement of CET1.Footnote ⁶¹ On the contrary, in neither the CRR nor the BRRD is the issue of governance rights of the converted items addressed.Footnote ⁶²

Second, the conversion would happen at the point of non-viability, so that the incentives of new shareholders would not sharply differ from the old ones’ in terms of risk shifting and gambling for resurrection.Footnote ⁶³ No rational reasons, indeed, lead us to believe that allegedly risk-averse bondholders are sticking to their risk appetite in the event of conversion. In fact, risk appetite cannot be assumed to be immutable,Footnote ⁶⁴ but is contingently changing according to the investor’s payoffs.Footnote ⁶⁵

Ultimately, the decision of public authorities over the most desirable course of action to undertake, ie the probability of conversion, remains – in the large part – discretionary.

Nonetheless, contracts might come to hand once again. In fact, it is possible to write contractual clauses for the conversion of eligible liabilities into CET1 items once a pre-established trigger event materialises (contingent convertibles). Therefore, theoretically, investors can bargain out of regulatory discretion.

1 Rationale and design features

Contingent convertibles (or Cocos) represent an academic invention that dates back to the early 2000s,Footnote ⁶⁶ and became popular in the aftermath of the financial crisis, as they have been deemed be innovative items that could potentially disrupt the discourse on financial stability.Footnote ⁶⁷ Cocos consist of financial instruments that are written down or converted into equity when a contractually predetermined event (trigger) occurs, with the immediate consequence of recapitalising and deleveraging a bank that is going towards financial distress.Footnote ⁶⁸

From a legal perspective, no EU regulation, neither CRR nor BRRD,Footnote ⁶⁹ provides for a special regulatory regime for contingent convertibles even though they are part of bail-inable securities and, usually, of regulatory capital. A sharply different approach has been employed by the Swiss regulator, where Cocos are the only instruments, other than CET1, that count as regulatory capital. Swiss regulation requires large banks to issue Cocos amounting to at least 9% of RWAs, of which at least two thirds have a trigger point higher than 7% of capital over RWAs.

In the EU, Cocos count for regulatory capital as long as their characteristics comply with qualitative requirements for specific classes, as discussed in Section III. Hence, they count as AT1 if they are perpetual and distributions are discretionary, or as T2 if they are long-term, or even non-capital bail-in eligible if they have maturity shorter than five years or they embed covenants that are incompatible with capital instruments. Yet, it is clear that issuing non-capital Cocos would make little sense since the bank would need to issue costly instruments twice. Thus, it is reasonable to argue that Cocos are either AT1 or T2, depending on their characteristics,Footnote ⁷⁰ with the additional constraint that the trigger event of AT1 cannot be lower than 5.125% of CET1 over RWAs.Footnote ⁷¹

Looking closer at the literature on convertible items, it appears clear that “contingent convertible” represents an umbrella concept that comprehends a wide variety of items, whose goal and main characteristics are similar while sharply differing in their design. While a comprehensive survey of various proposed designs cannot be provided here, some crucial aspects ought to be introduced.

First, setting the correct trigger event is crucial for efficient design, this a conundrum that financial economists are busy discussing. Here it is sufficient to emphasise that there are two main policy options: Cocos with high or low triggers. High trigger indicates the fact that the triggering event is relatively far from the point of non-viability,Footnote ⁷² while low trigger means the opposite. Second, Cocos can contractually provide both for conversion or write down of their principal amount once the event is triggered.Footnote ⁷³ Third, a further variation is to design re-convertible items, ie to contractually establish a second event that, if triggered, converts back the shares into Cocos. The re-conversion can be automatic or in the form of an option in favour of the converted shareholders. As Section IV.3 illustrates, the design with a reconversion option has beneficial effects from a corporate governance perspective.

Given the rationale and the design features of Cocos, the following section explores the impact of issuing Cocos on risk-taking incentives.

2 Cocos and risk-taking

There is no consensus in the economic literature about the impact of Cocos on risk-taking, since there is no generalised view on how to approach the issue, and the outcome is highly dependent on the assumptions of the various models.Footnote ⁷⁴ The overall effect on risk-taking appetite can be disentangled into three main dimensions.

First, the immediate consequence of converting Cocos into ordinary shares is to deleverage the balance sheet of the bank which, in turn, reduces the incentives for “risk shifting” and “gambling for resurrection” behaviours. It should be noted that this can be a persistent effect only if the bank is, latu sensu, profitable. In fact, if the value of shares keeps decreasing, the beneficial effect of deleveraging will not last.

Second, there is an effect on incentives toward monitoring. In fact, Cocos holders bear losses with relatively high probability, so that their incentives to monitor banks’ activities in general, and the trigger event-related issues in particular, increase.Footnote ⁷⁵

Third, and most relevant for our purposes, there might be “pure” governance effects, ie contractual design of convertible items positively copes with the agency problems among corporate constituencies, providing for ways to assure a return on investmentFootnote ⁷⁶ better aligned with risk-decreasing attitude. In the following paragraphs we discuss the potential for beneficial governance effects and the necessary contractual features for them to materialise.

3 Governance feature of Cocos design

The crucial issue in assessing the governance benefits deriving from contingent convertibles rests upon the type of shares Cocos are converted into.Footnote ⁷⁷ As a first approximation, one can think of two main possibilities: first, converting Cocos into plain ordinary shares that do not differ from the pre-existing ones in the voting and economic rights; second, converting Cocos into different kind of shares with different voting and economic rights.

In fact, as discussed above in Section III.2, if the converted shareholders have the same position as the pre-existing ones there is no rational ground for arguing that their risk incentives are going be more risk-averse just because they were bondholders beforehand.

However, as the previous discussion may have clarified, designing financial instruments for banks is all but straightforward and unrestricted, so it is issuing CET1 instruments. Thus, for qualifying as CET1, shares must not be in any way preferred to other CET1 items, either in distributions of dividends or in liquidation.Footnote ⁷⁸ This hampers, at least in the EU framework, the argument put forward by Coffee about positive governance effects if Cocos are converted into preferred stocks with considerable voting rights.Footnote ⁷⁹ Within equally subordinated shares, CRR makes it possible to differentiate the quantity of voting and economic rights. In fact, banks, in principle, can issue multiple, restricted or non-voting shares and economic rights can be adjusted accordingly; ie higher economic rights can be attached to shares with fewer or no voting rights.Footnote ⁸⁰

Given this framework, governance effects arise both before and after conversion. In good times, shareholders are averse to conversion if that means a considerable dilution of their voting or economic rights.Footnote ⁸¹ Moreover and most importantly, once the bank runs into financial distress and the conversion happens, Cocos holders are playing an active governance role whose content is highly dependent on the contractual design of converted shares. In fact, if risk appetite is context dependent – so that one cannot rationally expect that Cocos holders will stick to their risk-averse attitude once converted – this is also true for contractual entitlements that vary the context and so the appetite toward risk.

Three possible designs can be considered, to discuss the governance role of converted shareholders.

(1) No voting rights: in this scenario, converted shareholders have no say in the decision-making process of the bank, so their direct governance impact is negligible, whereas the “deleveraging effect” stays the same. It should be noted that higher economic rights, compliant with CET1 requirements, could be attached to non-voting shares so that the ex ante aversion of pre-existing shareholders to conversion can be slightly amplified.

(2) Same voting and economic rights: in this situation, the argument about context-dependent risk appetite applies; hence, converted shareholders have the same risk incentives as the pre-existing ones. Yet, if controlling shareholder extracts detrimental private benefits from control,Footnote ⁸² the conversion can still have beneficial governance consequences. In fact, if the number of converted shares is big enough to alter the existing control power, converted shareholders can push for more of a shareholder’s value-oriented management, for instance substituting the incumbent managers. Moreover, the conversion scenario in this situation would amplify the ex ante averseness towards conversion of controlling shareholders that may opt for limiting the private benefit extracted from the bank. Yet, absent any regulatory requirement to issue Cocos, a simpler solution is not to issue them in the first place.

Even out of the latter scenario, the incentive parallelism can be modified by providing converted shareholders with an option to reconvert to Cocos once a second trigger event is breached. So, in case of a high-trigger Cocos that converts into shares, once the turmoil is overcome, the converted shareholders have the opportunity to opt for reconversion to a relatively safer asset, undercutting the shift toward larger risk appetite. This would elicit the risk-preference of converted shareholders, so that the genuinely risk-averse ones will push for a more cautious approach to get out of distress and reconvert to bonds. Thus, if the trigger is high enough not to provide “gambling for resurrection” incentives and the number of converted shares is big enough to impact on the control of the bank significantly, the reconversion option can be a powerful tool.

(3) Limited voting rights: it is relatively common for companies to issue classes of shares whose voting rights are limited to specific issues (usually strategic decision, such as mergers) or contingent on some event (eg dividends are not distributed for two years in a row). In our context, limited voting rights seem more likely to occur. In such a scenario, most of the arguments proposed for (unrestricted) voting shares are applicable, including the beneficial effect of an option for reconversion.

What sharply differs from the latter design is that restricted voting shares are not apt to perform a shift in control, since they do not usually provide for voting rights on the election of the board. Therefore, the magnitude of the beneficial impact is more limited than for the previous design, especially with regard to limiting the private benefit of control. Finally, limited voting shares can provide enhanced economic rights, which ex ante increase the aversion towards conversion of pre-existing shareholders.

While it is impossible to provide a comprehensive assessment of each and every possible design feature, the three mentioned above are, arguably, the paradigms of general application. Indeed, in addition to the main designs discussed, one can think of a wide variety of more “exotic” variations, for instance converting Cocos into a class of shares that have the specific right to appoint one or more directors or, again, the case for multiple voting rights shares. However, these cases are highly jurisdiction-dependent, and it is not possible to analyse all of them. In fact, it is not only capital requirements that impose qualitative restrictions on shares, but also national corporate laws. So, for instance, it is relatively easy to issue multiple voting shares in Anglo-Saxon jurisdictions, but that is not the case in civil law oriented ones, where the principle of “one-share-one-vote” still consistently applies. The same argument goes for the specific rights to elect directors: it could be relatively simple in Italy, where the slate voting procedure for board election represents a well-established legal tool, but it may turn out to be virtually impossible in Germany, where the co-determination principle makes the composition of the board a particularly delicate equilibrium.

To sum up, the contractual design of Cocos matters for corporate governance. Design features that differentiate the incentives of pre-existing and converted shareholders are potentially able to decrease the overall risk appetite of the bank as long as the trigger event of conversion is set relatively high and the number of converted shares is significant enough to impact on corporate decision-making. This markedly differentiates Cocos from other bail-inable securities, as their design, given the current regulatory framework, is unfit to discipline the risk-taking behaviours of managers.