Argentina - Bank Failure Resolution : Recent Developments and Issues for Discussion

CONFIDENTIAL Report No. 19082-AR Argentina: Bank Failure Resolution Recent Developments and Issues for Discussion GREEN COVER March 25, 1999 Augusto de la Torre Finance, Private Sector and Infrastructure Latin America and the Caribbean Region The World Bank FILE coP TABLE OF CONTENTS Page No 1. Background: From 1991 to the Present.........................................................1 2. The Emerging New Model of Bank Resolution-the case of Banco A lm afuerte .................................................................................................7 3. The Emerging Model of Bank Failure Resolution-Issues for D iscussion....................................................**. ----------...-...---. ---..**.*..---.----9 3.1 The "all-deposits" constraint............................................................10 3.2 Criteria to construct the liability side of the "good" bank.....................11 3.3 Criteria to construct the asset side of the "good" bank........................13 3.4 The "least cost" criterion and Its implcations.....................................14 3.5 The asset-liability equivalence constraint ..........................................16 3.6 Institutional arrangements..............................................................16 3.7 Legal Protection ............................................................................. 17 3.8 Rules versus discretion ................................................................... 18 3.9 Other issues .................................................................................. 19 ANNEXES: Annex 1: Example of Bank Resolution under the Emerging Model (h la Almafuerte) Annex 2: Example of Bank,Resolution with (I) an Incremental Rule for Inclusion of Deposits in the "Good" Bank, and (Ii) a Deposit Guarantee Conditional on Asset Insufficiency Annex 3: Example of Bank Resolution with (I) an Incremental Rule for Inclusion of Deposits In the "good" Bank, and (ii) a Least Cost Criterion for the contribution of the Deposit Guarantee Agency (DGA) to the "Good" bank March 1999 ARGENTINA: BANK FAILURE RESOLUTION Recent Developments and Issues for Discussion By Augusto de la Torre' This note has been prepared to serve as an Input to the Interagency Committee that will be created to review Argentine bank failure resolution processes, including the relations between the Superintendency of Banks, Central Bank, and the Deposit Guarantee Corporation (SEDESA). The creation of such Committee is a condition for second tranche release under the Structural Adjustment Loan approved by the World Bank Board in November 1998. A plan of action, satisfactory to the Bank, to Implement such Committee's recommendations is a condition for the release of the third tranche. The note is structured in three sections. The first reviews the evolution of bank failure resolution in Argentina since 1991. The second describes In some detail a recent case, that of Banco Almafuerte, as an illustration of the bank resolution model emerging in present day Argentina. The third and most extensive section reviews key general and Argentine-specific issues in bank failure resolution, issues that the Interagency Committee may wish to discuss in depth. 1. Background: From 1991 to the Present Bank failure resolution policies and procedures have evolved significantly in Argentina since the introduction of Convertibility in 1991. This evolution has reflected a remarkable ability quickly to adapt -within the constraints imposed by Convertibility Law- the legal, normative, organizational, and procedural frameworks to deal with bank failures in the midst of the stress caused by two adverse exogenous shocks: the major Tequila shock of 1995 and the relatively more benign financial turbulence unleashed * Lead Specialist, Finance, Latin American and the Caribbean Region, The World Bank. The author wishes to acknowledge the useful comments provided by Stefan Alber, Aquiles Almansi, Federico Caparros-Bosch, Gerard Caprio, Asli Demirguc-Kunt, Jonathan Fiechter, Paul Levy, and Marjory Waxman. The normal disclaimers apply. 2 since the second half of 1997 by the South East Asian and Russian crises. In adapting the bank failure resolution, the authorities have sought to strike an appropriate balance between two objectives: averting contagion and limiting moral hazard. In 1991, at the beginning of the Convertibility Plan, Argentina had a bare-bones system of bank failure resolution: troubled banks unable to restore compliance with key prudential norms, particularly those associated with solvency and/or liquidity, would automatically see their license removed and go on to judicial liquidation.' Under such system -not unlike those existing in many Latin American countries- there was neither an explicit deposit insurance scheme nor resolution procedures other than judicial liquidation, implying that, in the event of a bank failure, all deposits would immediately turn Illiquid and value-Impaired. Depositors would have to wait in line for a long time (months and even years) to have access to liquidation proceeds. Given the protracted, legally assailable, and cumbersome nature of the judicial liquidation process, a sharp destruction of asset value would take place compared to their going concern value. Under these circumstances, depositors derived little consolation, if at all, from their relatively high position in the priority of claims ladder -just below labor claims. By featuring a clear risk of loss to depositors, this bare-bones system of bank resolution sought to minimize moral hazard -an objective that was gladly embraced in reaction to the huge Central Bank losses and runaway inflation that had resulted from previous episodes of banking crises and bailouts. The system aimed at encouraging depositors to monitor the condition of banks and impose discipline on them, not only by asking for higher interest rates when depositing in banks perceived to be riskier but also by precipitating, through a run on deposits, the closure of banks perceived to be unviable. This type of system was particularly stringent in the case of Argentina because of Convertibility, which severely limits the authorities' room to elude or significantly delay the liquidation of a troubled bank by keeping it afloat with Central Bank liquidity.2 1 The Central Bank has the power to establish terms and conditions for rehabilitation plans (planes de regularizaci6n y saneamiento) under which troubled banks are given the opportunity to correct their problems so as to restore financial viability. Noncompliance with such plans is a cause for removal of the license and, hence, the beginning of judicial liquidation. 2 To be sure, under Argentine Convertibility there is a margin, equivalent to at least one-third of the money base, for Central Bank liquidity assistance to the banking system. The Central Bank can issue money to advance liquidity to banks inasmuch as disposable international reserves exceed money base. Furthermore, disposable reserves are defined to include dollar denominated 3 The 1995 Tequila quickly made it plain that the bare-bones system was ill suited to manage bank failures -particularly in the case of large banks- while adequately limiting contagion risk. At the outset of the Tequila, the system led to the liquidation of some banks, including some middle size ones [mention some names]. But it was soon observed that a resolution framework purely consisting of judicial liquidation was Intensifying deposit runs, dangerously magnifying the erosion of confidence and propagating distress. Under such circumstances, the authorities introduced in the first half of 1995 three important Institutional Innovations directly relevant to bank failure resolution. Two of these innovations -the reforms to the Financial Institutions Law and the creation of a Deposit Guarantee Scheme- were meant to be permanent, i.e., to endure beyond the Tequila juncture. The third, consisting of the creation of a Bank Capitalization Trust Fund, was seen as a transitional crisis management tool, although its two-year life was subsequently extended until February 2000. As regards the first innovation, the core reform of the Financial -Institutions Law (No. 21.526) introduced in 1995 is embodied in Article 35 bis. This article empowers the Central Bank to carry out a version of a "good bank/bad bank" resolution procedure before the liquidation of a falling bank. This procedure consists of: (i) separating from the liability side of the failing bank the privileged liabilities, namely, liabilities to labor, deposits, and debts to the Central Bank; (ii) carving out assets from the asset side; and (ii) transferring the assets and liabilities thus separated to another existing financial institution, ensuring "equivalence" between the transferred assets and deposits. The procedure can be applied, at the sole discretion of the Central Bank, to a bank deemed to be unviable and prior to the removal of its license,3 with the result that only the residual or "bad" bank (containing the assets and liabilities that were not separated and transferred) would go on to judicial liquidation. - This type of good bank/bad bank surgery is underpinned by a priority of claims queue for unsecured claims, according to which employees stand ahead of depositors, (and internationally traded) bonds Issued by the Republic of Argentina, provided that such bonds do not exceed one third of total disposable reserves. 3 A bank may be deemed unviable, and thus subject to the revocation of its license when, in the sole* opinion of the Central Bank, Its solvency and/or liquidity position is so weak that it cannot be restored to health via a rehabilitation plan (see Article 44 of Law No. 21.526). 4 depositors before the Central Bank, and the Central Bank ahead of other claims on the failing bank. There is no room for bailing out bank shareholders under Article 35 bis; shareholders are always left in the residual bank. Additionally, the Central Bank is given in the Law what appears to be ample authority to eliminate or substantially dilute the property rights of shareholders, including by writing down their capital against identified losses and required increases in provisions, or by revoking their right to be in the banking business, even prior to the application of Article 35 bis. Article 35 bis would seem to imply that claims with priority lower than that of the Central Bank must always stay in the residual ("bad") bank and await the results of the liquidation. As a consequence, the good bank/bad bank separation provided for under this Article must normally be performed with the bank closed; for if it were open, non- depositor creditors would engineer a run to avoid the losses associated with being relegated to the residual bank. The second institutional innovation consisted In the creation of a Deposit Guarantee Scheme (Sistema de Seguro de Garantia de los Dep6sitos Bancarios) via Law No. 24.485. This scheme is composed of a Fondo de Garantia de Dep6sitos (FGD) and a corporation, the Seguro de Dep6sitos Sociedad An6nima (SEDESA), which administers the FDG. SEDESA offers a guarantee, per depositor, of up to US$30,000, provided that the interest rate paid on such deposits does not exceed a certain level.5 The FGD is fed by mandatory contributions paid by the banking industry. The Central Bank has flexibility in setting the rate of contribution within a range -from a minimum of 0.015 percent to a maximum of 0.06 percent of average daily deposits- and may require the advanced payment of up to two years of contributions. The Investment and uses of the FGD are governed by an Executive Committee made up by representatives from the contributing financial institutions; the Central Bank has one representative in such Committee, who lacks voting power but has veto power. I More precisely, the procedure of Article 35 bis to create and transfer the "good" bank is applied while the failing bank is under a so-called periodo de suspensi6n transitoria, which suspends its operation. After this period, the license of the failing bank is removed and judicial liquidation of the residual ("bad") bank begins. s Not covered by the guarantee are also related-party deposits and deposits made by other financial institutions. 5 In the event of a bank failure, SEDESA could use FGD resources to pay guaranteed deposits in cash, in which case it has the special right to get payments from the liquidation before all non-guaranteed depositors. However, SEDESA is not confined to the cash payout option when honoring the deposit guarantee. The Executive Committee may also authorize the use of FGD resources to inject capital into, or make a loan, including a non-reimbursable one, to: (i) a financial institution under a rehabilitation plan; (ii) a financial Institution that purchases the assets and assumes the deposits of a bank that is under the Article 35 bis procedure; or (iii) a financial institution that buys a bank that is under a rehabilitation plan. Any of these three operations may be authorized if, in doing so, the direct cost to the FGD is lower than what it would be if the cash payout route were followed. Clearly, the second of these three alternatives to a cash payout of guaranteed deposits has the potential of enhancing significantly the flexibility of the good bank/bad bank resolution procedure contemplated in Article 35 bis -a flexibility that has been often seized by the authorities. The third innovation introduced in the first half of 1995 was the creation, by Decree 445/95, of a Bank Capitalization Trust Fund (Fondo Fiduciario de Capitalizaci6n Bancaria). This Trust Fund has been fed mainly by budgetary transfers financed via a special bond issue (the Bono Argentina 1998) as well as the proceeds from loans granted by multilateral organizations, including, importantly, by the World Bank under the 1995 Argentina Bank Reform Loan. The Trust Fund was designed to strengthen the hand of the authorities in dealing with the banking system turmoil created by the Tequila. According to the decree, the Trust Fund is empowered to inject capital into financial institutions directly, Including by making loans convertible into equity, or indirectly, by buying the illiquid assets of distressed financial entities. In practice, the Trust Fund has utilized only two instruments in assisting financial institutions: a long term collateralized loan; and a noncollateralized subordinated loan, convertible into a negotiable obligation. The authorities have used the Trust Fund to foster consolidation in the banking system, via assisted bank mergers, acquisitions, and capitalization.6 Where these transactions turned out to be nonviable, the Trust Fund has absorbed 6 With the assistance of the Bank Capitalization Trust Fund, 19 transactions involving private sector financial institutions have been performed. The Trust Fund has provided some US$750 million in loans, of which only US$89 million have been definitively lost in transactions that turned out to be nonviable. At present, the Trust Fund is no longer active and virtually has no funds. 6 losses, thereby facilitating the construction of the "good" bank under Article 35 bis procedures. Taken jointly, the three institutional innovations mentioned above contributed significantly to preventing what could have otherwise been a banking system meltdown unleashed by the Tequila. The increase in moral hazard that may have resulted from the bailout of large depositors that these innovations afforded in some cases, can be rightly considered to have been a small price paid in order to limit contagion risk under the turbulent financial environment unleashed by the Tequila. Moreover, a systemic crises was avoided with very little use of public sector resources, particularly when compared to international experiences. In part due to these institutional innovations, since 1995 the authorities have broadly succeeded in promoting an orderly process of consolidation and strengthening of the banking system,'7 with striking results.8 In the period July 1995 to December 1998 the Argentine authorities resolved 16 financial Intermediaries using the powers under Article 35 bis. The liabilities of the resolved institutions added to more than US$4.3 billion, with the median size bank in the group (Banco Argencoop) featuring liabilities of around US$210 million, and the largest case having been that of Banco Mayo, with liabilities of some US$1.3 billion. In 10 of these 16 cases, FGD resources were used to facilitate the execution of Article 35 bis procedure -essentially by completing the asset side of the "good" bank- for a sum total of around US$600 million. The Capitalization Trust Fund provided, to the same end, a total of around US$325 million, which were distributed over 8 cases. The resolution processes under Article 35 bis, while representing a dramatic improvement over the judicial liquidation alternative, often proved to be protracted. ' To be sure, many other factors have played a major role in the system's consolidation and strengthening. These factors include: a marked improvement in prudential oversight and its enforcement; the enhancement of market monitoring via, for instance, the requirement that banks issue subordinated debt on a regular basis; the privatization of public sector banks; the unrestricted opening of the banking sector to foreign capital; tougher-than-Basle capital requirements; and stringent liquidity requirements. For details see Argentina: Financial Sector Review, Report No. 17864-AR, September 28, 1998, The World Bank. 8 The number of financial institutions in the Argentine system fell from 205 at end-1994 to around 130 at end-1998. Over the same period, the number of foreign-owned banks rose from 30 to 42, with their share in the system's deposits increasing from 16 percent to about 55 percent. In respect of publicly owned banks, BANADE, the former development bank, was closed, and 15 provincial banks were privatized. 7 Over time, however, the Argentine authorities have accumulated experience, continuously refined tools and procedures, and increasingly streamlined execution. The recent case of Banco Almafuerte may be seen as an illustration that bank failure resolution in Argentina has come of age. However, the same case puts into sharp focus a number of issues that need careful assessment. 2. The Emerging New Model of Bank Resolution-the case of Banco Almafuerte Banco Almafuerte Coop. Ltdo. was a small bank by Argentine standards, with a book value of assets of about US$200 million. Its failure was effectively resolved over a weekend, a major achievement even after considering that It was the case of a small bank, for It stands in sharp contrast with almost all prior cases that took weeks or months. This is all the more impressive considering that -as was mentioned earlier- the resolution-procedure of Article 35 bis can be performed realistically only with the failing bank's doors closed. A long period of closed doors Is bound to cause not only severe irritation among the bank's creditors but also a deterioration in the value of its assets. Under unsettled financial markets, furthermore, delays in executing the Article 35 bis procedure could also lead to contagious deposit runs. In the case of Almafuerte, these problems were avoided, as its doors were for only a couple days and during a time (a weekend) when they would have been closed anyhow. In resolving Almafuerte, the Central Bank authorities combined Innovative financial engineering with the full range of powers under Article 35 bis, as well as with the powers granted to the Executive Committee of the FGD. Annex 1 presents a simplified numerical example of a bank resolution based on the approach used in the case of Almafuerte. The example -whose numbers bear no resemblance to the actual case- should help the reader visualize more clearly the description of Almafuerte's resolution provided below. A main source of delays In executing the good bank/bad bank surgery contemplated in Article 35 bis has been the time required for due dilligence In order reasonably to estimate the value of assets. In the case of Almafuerte, this hurdle was overcome by a form of asset securitization, that is, by placing the assets (other than cash) of Almafuerte into a Trust (Fideicomiso) that, in turn, issued bonds backed by those assets. 8 The Trust issued three classes of bonds: A, B, and C. Bonds type A were structured to be senior to B, and B Bonds senior to C Bonds. That is, in the event of a shortage of resources in the Trust, B Bonds would be serviced only if there are resources left over after servicing A Bonds, with C Bonds standing a similar relation of subordination relative to B Bonds. To reduce the probability of default, particularly with respect to A Bonds, the book value of assets placed in the Trust was substantially higher than the total face value of bonds it issued (a ratio of assets to bonds of about 1.50 to 1.00). This version of asset securitization permitted a quick construction of the "good" bank. In effect, 8 good banks were assembled during Almafuerte's resolution, because there were 8 existing financial Institutions that participated in acquiring the assets and assuming the deposits of such "good" banks. The "good" banks were put together in line with the geograplc distribution of the Almafuerte's branch network. Each "good" bank contained, In the liability side, the debts to labor and the deposits corresponding to the branches In a given geographic area and, in the asset side, three assets: cash, A Bonds, and an IOU from SEDESA for an amount up to the corresponding guaranteed deposits.9 There were, consequently, as many Trusts containing Almafuerte's assets as "good" banks. As required by Article 35 bis, the asset and liability sides of each "good" bank were matched, Implying a zero accounting net worth, with the amount of the IOU from SEDESA essentially set as a residual to ensure that assets match liabilities. Each "good" bank was transferred "as is" to the acquirer, who did not make any explicit payment in respect of the franchise value of the "good" bank. This in part reflected the concerted nature of the operation, with the Central Bank playing a major coordinating role. Each acquirer bank took on, for a fee, the responsibility of managing, recovering, or selling the assets in the corresponding Trust. Central Bank claims on Almafuerte, resulting from the liquidity assistance it had provided to that bank before its failure, were not part of the liability side of the "good" banks, nor did they stay in the residual ("bad") bank, in the sense that the Central Bank did not go as a creditor into the liquidation phase. Instead, the Central Bank ceded to the Trust the assets It had received as collateral to the mentioned liquidity assistance and, in exchange, received B Bonds, thereby standing as the second claimant in line vis- 9 SEDESA contributed an IOU (backed by future FGD receivables), rather than cash, because the FGD's liquid funds had been fully drained in previous cases of bank failures. 9 A-vis the Trust. This treatment of Central Bank liquidity credits can be construed to be consistent with the priority of claims rules specified in Articles 49 and 53 of Financial Institutions Law (No. 21.526) for judicial liquidation. Similarly, the claims of SEDESA on Almafuerte's assets -resulting from the addition of a SEDESA IOU to the "good" bank- did not form part of the liability side of the "good" bank nor did it integrate the residual bank. Instead, SEDESA received C Bonds, thereby standing third in the queue vis-h-vis the Trust. This treatment of SEDESA can be construed to be consistent with the privileges it has under the Law only to the extent that SEDESA has better chances of recovering its resources by standing third in line vis- h-vis the Trust than by standing ahead of non-guaranteed deposits vis-h-vis the judicial liquidation. That this may be possible can be rather surprising for a non-Latin American observer. However, the truth is that in Argentina, as in most of Latin America, the liquidation process -whether judicial or extra-judicial- tends to be staggeringly inefficient, generally implying an excessive destruction of asset value. The resolution of Almafuerte minimized the size of the residual ("bad") bank that would go on to judicial liquidation. In the liability side, the residual bank contained only non-privileged liabilities, i.e., Almafuerte's liabilities left after excluding those to labor, depositors, the Central Bank, and SEDESA. In the asset side, the residual bank included only the assets that the Trusts would put back -assets deemed to be non-recoverable or without value. 3. The Emerging Model of Bank Failure Resolution-Issues for Discussion The resolution structure applied in the case of Banco Almafuerte was, undoubtedly, skillfully executed. The securitization of assets achieved by the Trust arrangement was innovative and effective, in that it facilitated the quick transfer of the "good" bank. In turn, the speedy nature of the operation substantially reduced the depositor uncertainty and distress often associated with bank closures, thereby minimizing contagion risk. It may be also argued that this type of rapid, nondestabilizing resolution of troubled banks has had and will continue to have a positive feedback on banking discipline. Quick, effective bank resolution removes concerns that supervisory authorities may have about the systemic implications of closing a bank and, thus, is likely to foster tougher 10 enforcement and earlier intervention and resolution of unviable banks. Moreover, the authorities' current effort to standardize contracts and procedures involving the Trust scheme, should further enhance the resolution process. However, as the authorities undertake a systematic review of bank failure resolution, a number of general as well as Argentine-specific issues arise, pointing towards areas where improvements may be considered. 3.1. The "all-deposits" constraint Important issues revolve around the fact that Article 35 bis contains a crucial constraint that increases moral hazard (while reducing contagion risk) and tends to deplete rapidly the liquid resources of the FGD. Article 35 bis requires the incorporation of all the deposits, regardless of their size, in the liability side of the "good" bank -a constraint that henceforth will be referred to as the "all-deposits constraint." Sufficient assets must be found to match the total amount of deposits, otherwise Article 35 bis cannot be applied and the resolution defaults to judicial liquidation. This constraint thus implies that the FGD's liquid resources would tend to be drained quickly in the resolution of a middle-size or large bank -as the recent experience with the case of Banco Mayo has shown. In effect, according to the current legislation, if the value of non-cash assets under judicial liquidation approaches zero, in each resolution SEDESA could contribute to the asset side of the "good" bank up to an amount equal to the failing bank's guaranteed deposits, with the maximum contribution tending to be required in most resolution cases in order to match all the deposits of the failing bank. With cash (or IOUs) from SEDESA used to complete the required amount of assets in the "good" bank, what was intended to be a limited deposit guarantee becomes de facto a full deposit guarantee, with the FGD standing to loose ahead of large depositors. (Recall in this connection that SEDESA gets C bonds, compared to depositors whose resources are backed by A Bonds.) Annex 1 depicts this type of situation, illustrating also that a maximum contribution from SEDESA to the "good" bank may be required even in cases where the failing bank's insolvency is not particularly deep -equivalent to 10 percent of total liabilities in the Annex's example. There are, of course, ways to relax the all-deposits constraint and to control better the contribution of the Deposit Guarantee Agency to the asset side of the good bank. Given 11 the trade-off between moral hazard and contagion risk, it would be presumptuous and plain wrong to think that a simple recipe for a "right" balance exists. However, in assessing possible changes and refinements to their bank failure resolution system, the Argentine authorities may wish to review the range of feasible policy choices and issues along the lines of the paragraphs below. 3.2. Criteria to construct the liability side of the "good" bank Key issues concern the procedures and criteria used to construct the asset and liability sides of the "good" bank during the resolution phase. As regards the liability side of the "good" bank, the all-deposits constraint reflects a well-known legal tradition whereby different classes of creditors have different priorities of claim via-h-vis available assets, but there are no priority of claim differences within a given class of creditors. Accordingly, depositors as a class of creditors have a priority of claim over, say, subordinated debt holders but individual depositors are treated alike, sharing pro-rata in the pool of available assets. An explicit, limited deposit insurance scheme alters this scheme somewhat, as it gives guaranteed deposits a privilege over the non-guaranteed ones in the event of a bank failure. The logic of these priority of claims rules, together with a limited deposit guarantee scheme, leads to the conclusion that a "good" bank may be constructed in the resolution phase if it contains, in the liability side, either all deposits or only guaranteed deposits, with no choices in between. To escape these corner solutions, a rule would be needed to establish a priority of claim sequence among depositors, effectively revising the pro-rata criterion. A natural candidate that would give maximum flexibility would be an incremental rule, according to which the limit per depositor to be included in the "good" bank would be decided at the discretion of the resolution authorities. The authorities would have the freedom to raise such limit gradually, starting from the guaranteed amount, up to the point where the amount per depositor is such that liability side of the "good" bank is no greater than its asset side. The examples provided in Annexes 2 and 3 use this incremental rule in constructing the "good" bank; however, as deposit amounts cannot be shown along a continuum, they are shown in tranches (e.g., more than $30,000 but less or equal to $50,000, etc.). 12 This incremental rule would mitigate moral hazard by giving operational meaning to the notion that smaller depositors should have a priority of claim over larger ones. The risk of loss In the event of a bank failure for large depositors would give them an incentive to monitor the condition of banks. With appropriate legal drafting, the application of the incremental rule could be circumscribed to the resolution phase, leaving the application of the traditional pro-rata criterion for deposits left in the residual ("bad") bank that moves on to liquidation. The Argentine authorities have recently proposed to Congress certain legal reforms, including a rule that would relax the all-deposits constraint of Article 38 bis, by permitting the non-inclusion of amounts above $100,000 per depositor in the "good" bank. More precisely, under the reformed legislation, the liability side of the "good" bank would always have to incorporate at least deposit amounts up to US$100,000. But it there is an excess of assets in the "good" bank, deposit amounts above that figure could be also included on a pro-rata, rather than an incremental, basis. The authorities may wish to consider the merits of revising the proposed legal amendment in order to accommodate the full flexibility and operational simplicity of an incremental rule. Another issue in the construction of the liability side of the "good" bank concerns the definition of eligible deposits. Such a definition would imply that there are certain types of deposits that would in no case be included in the "good" bank; they would have to remain always in the residual bank. The simplest way to tackle this issue would be by following the same logic used in defining insured deposits (see page 4 above, including footnote 5), with the result that related-party deposits, deposits by other financial institutions, and deposits whose interest rate exceed certain level would all be ineligible -i.e., they would be condemned to stay in the residual bank In the event of a bank failure. In the absence of a clear delimitation of eligible deposits, interbank deposits related to the subordinated debt scheme (i.e., the "B" of the BASIC program) could end- up in the "good" bank, balled-out at the expense of FGD resources. An unambiguous definition of eligible deposits that would be relevant to the application of Article 35 bis would undoubtedly enhance bank monitoring and limit moral hazard. It may, however, pose certain operational complications that could hamper somewhat the speed of bank failure resolution. 13 3.3. Criteria to construct the asset side of the "good" bank The key issue in assembling the asset side of the "good" bank concerns the determination of whether the Deposit Guarantee Agency should make a contribution and the size of It. Following are the main options and issues in this regard. An option that would maximize protection to the resources of the Deposit Guarantee Agency (DGA) and, hence, minimize moral hazard, would be to define the deposit guarantee as conditional on asset Insufficiency, implying that the DGA would make a contribution to the assets of the "good" bank only if there are not enough assets in the failing bank to cover guaranteed deposits. The DGA conditional contribution, if triggered, would be for an amount no greater than what is needed to honor the deposit guarantee. To further protect Its resources, the DGA could be given a first-in-line claim In the liquidation of the residual bank.10 And to enhance the effectiveness of bank failure resolution process, the incremental rule mentioned above could be used for the determination of the liability side of the "good" bank. A deposit guarantee conditional on asset Insufficiency implies that non-guaranteed deposits would not benefit at the expense of the DGA and that bank failures would not drain systematically the DGA's liquid resources. The trade-off is, of course, that such guarantee is relatively less effective in the management of systemic stress, as it provides relatively less flexibility to control contagion risk. Annex 2 provides a simplified example of a bank failure resolution under a regime that combines a deposit guarantee conditional on asset insufficiency, on the one hand, and an incremental rule, on the other. The example illustrates that, under that regime, the DGA would tend not to be exposed to a risk of loss even in the case of a resolution of a deeply insolvent bank (the bank in the example has an insolvency "hole" equivalent to 50 percent of total liabilities), and even if asset value declines significantly under the liquidation phase (in the example, non-cash assets loose 60 percent of their value as a result of moving on to liquidation), but provided that guaranteed deposits amounts are not too high. 10 This alternative could be relaxed without undermining its spirit by allowing the DGA to advance liquid resources to facilitate the resolution process, provided that within a certain period (and in any case prior to the liquidation of the residual bank) the DGA recovers in full such resources. 14 An the other extreme from the previous option would be a full DGA's contribution, meaning that the DGA would always inject resources in the asset side of the "good" bank in an amount equal to the guaranteed deposits of the failing bank, independently of the value of available asset in this bank. In exchange, the DGA would stand as a creditor in the liquidation of the residual ("bad") bank. Under this regime, an incremental rule would be useful only in case the "good" bank could not accommodate all the deposits of the failing bank, even after the contribution from the DGA. The example in Annex 1 has been constructed to illustrate this regime. A good bank/bank bank procedure based on a full contribution by the DGA to the "good" bank would provide flexibility to protect all deposits in most cases, thereby reducing the trauma of a bank closure and minimizing the risk of contagious runs on other banks. This flexibility would be valuable particularly in the management of systemic stress, e.g., In the context of a simultaneous or sequential failure of several banks. The trade-off is, of course, the increase in moral hazard because large depositors would get bailed out at the expense of DGA's loss." A corollary disadvantage is that bank failure resolution would systematically and substantially drain the liquid resources of the DGA. Hence, the more bank failure resolution approaches this polar alternative, the greater would be the need for substantial backstop financing for the DGA.12 3.4. The "least cost" criterion and Its Implications Between the two extremes mentioned above, there are a continuum of alternatives where the contribution of the DGA to the "good" bank is controlled through some form of a least cost criterion. For the purposes of this note, the standard definition of this criterion implies that a good bank/bad bank type resolution should be implemented only if it imposes a cost to the DGA that Is no greater that the cost that the DGA would incur if it were to pay out guaranteed deposits in cash, in exchange for a fairly privileged claim on the liquidation. Four variables affect the potential size of the DGA st Moral hazard could be offset In some degree, however, through tight supervision and, particularly, early intervention and resolution. Early resolution takes away from the shareholders and administrators of an insolvent bank the "captive" depositor funds, thus depriving them of the wherewithal with which they could take on excessive risks or loot. 12 Under Argentine-type convertibility, backstop financing to a DGA could not entail money creation by the Central Bank; it would have to come from domestic or, preferably, external debt. 15 contribution to the "good" bank under a least cost criterion: (i) the size of the insolvency "hole" of the failing bank -caeteris paribus, the greater the insolvency the greater the potential contribution by the DGA; (ii) the priority of claim accorded the DGA in the liquidation -the higher such priority, the smaller the potential contribution by the DGA; (iii) the share of guaranteed deposits in the liabilities of the failing bank -the higher such share, the higher the potential contribution by the DGA; and (iv) the degree of loss in the value of assets as they move into judicial liquidation -the greater the loss, the larger the potential contribution by the DGA. Rather than examining in detail the effects of the various combinations of these four variables on the contribution by the DGA to the "good" bank, it appears preferable to lay out two sets of general observations that are helpful In clarifying the policy options. First, only benefits can be obtained from Improving and enforcing prompt corrective action and early Intervention rules, in order to minimize the probability that insolvent banks continue to operate, which would strengthen the soundness of the banking system. Similarly, major benefits could be obtained from streamlining the liquidation process, as this would imply a lesser loss of value for assets under liquidation and, thus, a lower risk of loss for the DGA under a least cost criterion. Regardless of whether the DGA stands first in line or not vis-h-vis the liquidation, and notwithstanding the importance of minimizing the size of the residual bank for effective resolution, the least cost criterion would yield a high cost to the DGA If asset value collapses in the liquidation phase. The foregoing considerations suggest that, in reviewing bank failure resolution, the Argentine authorities would do well in casting a wide net, so as to identify improvements to the antecessor phase (prompt correction and early intervention) and successor phase (liquidation) of bank resolution proper. The World Bank could assist the Argentine authorities by providing references of international best practices in connection with these areas. Second, given the way in which least cost is calculated, the DGA's priority of claim in the liquidation makes a significant difference in Its exposure to risk of loss. The good bank/bad bank resolution examples in Annex 3 illustrate the two alternatives for the DGA's priority ranking vis-h-vis the liquidation, namely, for the DGA to be (I) in the same situation as non-guaranteed depositors, sharing with them on a pro-rata basis," 13 This alternative is featured in the United States Federal Deposit Insurance system. 16 or (ii) first in line, i.e., ahead on non-guaranteed depositors. Sharing pro-rata with non- guaranteed depositors in the liquidation necessarily entails that the DGA offers some protection to these depositors. The way this plays out In a good bank/bad bank resolution procedure is that the least cost calculation leads to a significant contribution by the DGA to the asset side of the "good" bank, as is Illustrated by the example in Alternative A of Annex 3. In contrast, a first-in-line priority of claim in the liquidation severely reduces the DGA's exposure to risk of loss, even if the failing bank is deeply insolvent and the loss of asset value under the liquidation phase is substantial. This is Illustrated by the example in alternative B of Annex 3, where the least cost calculation leads to no losses for the DGA in a resolution of a failing bank that has an insolvency "hole" equivalent to 40 percent of liabilities and whose non-cash assets loose 60 percent of their value by virtue of entering the liquidation stage. Again, by moving from the first to the second alternative In Annex 3, flexibility to control contagion risk is sacrificed for the sake of reduced moral hazard. 3.5. The asset-liability equivalence constraint Article 35 bis explicitly requires that "equivalence" be maintained between the assets and liabilities that are separated from the failing bank and then transferred. While it is not clear why such equivalence is called for in the Law, one can certainly think of cases where the constraint would unnecessarily hinder maximum asset value preservation and recovery. Assume the failure of a bank where the value of "good" assets significantly exceeds deposits. The equivalence constraint would force us to leave "good" assets in the residual bank, with social costs resulting from the loss of value of such assets under liquidation. Clearly, non-depositor creditors would be better off and depositors would not be worse off if: (i) a "good" bank containing an excess of assets over deposits were constructed and transferred; and (ii) the proceeds from such transfer (net of transactions costs) would be placed -as they should- in favor of the liquidation. 3.6. Institutional arrangements The effectiveness of bank failure resolution is not independent of the organizational and institutional arrangements that house it. This raises the general issue of whether the responsibility for bank resolution should rest with an independent, specialized institution, 17 separate from the banking supervision agency. A number of incentives-related reasons appear to militate in favor of a separate entity which, in practice, has tended to be a publicly administered Deposit Guarantee Agency -such is the case, for instance, of the United States FDIC and Spanish FOGADE. It may be argued that a separate deposit guarantee/resolution agency would have (i) a natural incentive to protect the integrity of the accumulated guarantee fund; (ii) no psychological resistance for early intervention and resolution of a failing bank, as opposed to bank supervisors who may tend to perceive their role as one of preventing bank closures, possibly even interpreting those closures as an admission of supervisory failure; and (iii) specialized skills in bank failure resolution, skills that are quite different than those required for bank supervision. Furthermore, it would seem that involving the Central Bank too deeply in bank failure resolution processes could damage the credibility of monetary and bank supervision policies. These arguments have to be weighed against important reasons, particularly practical ones, that can be advanced in favor or keeping bank failure resolution powers with the supervisory agency. There is, in the case of Argentina, the credibility that that the Central Bank brings to the application of Article 35 bis, on account of its recognized independence and proven professionalism. More generally, there are likely to be coordination gains and economies of scale (particularly where skilled human resources are in shortage and substantial learning by doing has already been accumulated) in having the supervisory and resolution functions under the same roof. Be it as It may, the issue of organizational and institutional arrangements is of great importance and should be included in the comprehensive review of bank failure resolution that the Argentine authorities plan to undertake. 3.7. Legal protection Resolving and closing a failed bank are extreme expressions of "enforcement" which, if delayed, gives rise to high social costs. Hence, the prompt and effective execution of resolution procedures requires solid legal ground as well as clear legal protection for the bank failure resolution process Itself (lawsuits should not suspend It) and for the authorities charged with the responsibility. Authorities need to be protected against lawsuits initiated against them for actions carried out in good faith in the performance of 18 their official duties; otherwise they would fall to act forcibly and in a timely manner. Such protection, which is consistent with the first of the Basle Core Principles for Effective Bank Supervision, is unfortunately weak in most Latin American countries, including Argentina. A separate note on this particular topic is under preparation and will be shared with the Argentine authorities. 3.8. Rules versus discretion A deeper, general issue that permeates most of the discussion in this section is that of rules versus discretion. Bank failure resolution is an complex subject where sound judgement is essential, which argues for a degree of discretion. Too much discretion, however, contributes to lack of transparency and erodes accountability. The "right" balance is of course elusive, not the least because even in a rules-intensive system aimed a minimizing moral, the too-big-to-fail phenomenon would tend to undermine the viability -and hence the credibility- of such tight rules. These complications notwithstanding, the Argentine authorities may wish to explore this issue in their review of bank failure resolution processes, by weighing the pros and cons of the two prototypical approaches briefly described below. The first approach emphasizes the option value of discretion. As a consequence, it has a bias in favor of a bank resolution framework that would leave considerable flexibility to control contagion risk. This may be accomplished by, say, allowing for the inclusion of all deposits and even certain non-deposit liabilities in the liability side of the "good" bank and defining the size of the contribution of the DGA to the asset side of the "good" bank independently of the value of available assets of the failing bank. To limit the attendant moral hazard, however, this approach practices "constructive ambiguity," by maintaining alive the threat of not carrying out the Article 35 bis-type procedure and, thus, allocating losses to non-guaranteed deposits through traditional liquidation, in the event of a bank failure. The second approach, by contrast, emphasizes the virtues of rules. As a consequence, it favors a bank failure resolution framework that minimizes moral hazard and the risk of 19 loss to the DGA. This can be done by, say, defining the deposit guarantee as conditional on asset insufficiency and according the DGA a first-in-line priority of claim in the liquidation. However, conscious of the too-big-to-fail phenomena and the need to control contagion risk under extreme circumstances, this approach provides for a "rule" that would permit the abandonment of "normal rules." This extraordinary rule could take the form of a contingent clause In the Law that specifies the conditions under which, when systemic risk is perceived by the relevant authorities, the deposit* guarantee limit can be, by exception, raised to cover up to all deposits and, if needed, even up to all bank liabilities. 14 3.9. Other issues Finally, some more specific issues are worth mentioning. The first concerns the current lack of liquid resources In the DGF managed by SEDESA. These funds were depleted with the resolution of Banco Mayo; moreover, as a result of the most recent resolutions, several months of the Fund's future receivables have already been mortgaged. Such lack of liquid resources erodes the credibility of the Deposit Guarantee and hinders the effectiveness of bank failure resolution processes. In this connection, the Argentine authorities may wish to analyze the convenience and/or feasibility of backstop financing for SEDESA. Given the constraints Imposed by Convertibility, an appropriate backstop financing could take the form of a contingent line of external credit to SEDESA, provided by multilateral sources or private capital markets. However, as was noted before, the size of possible backstop financing cannot be determined independently of the criteria and procedures that guide the construction the "good" bank. Second, the Argentine authorities may wish to assess the utility and appropriateness of including additional mechanisms to the bank failure resolution toolkit. In particular, they could consider the figure of a "bridge bank" as featured, for instance, in the United States FDIC Improvement Act of 1991. A bridge bank is a temporary license for the operation of a "good" bank, which may come in handy where, say, the resolution 14 This second approach was adopted in United States by the FDIC Improvement Act of 1991. It introduced a contingent clause according to which, when handling a failing bank, the FDIC can extend coverage to bank liabilities beyond the insured amount if there is a joint determination by the Board of Governors of the Federal Reserve, the FDIC Board, and the Treasury Secretary (after consulting with the President of the United States) that the failure of such bank would entail systemic risk. 20 must be completed quickly to avoid confidence erosion but time is still needed to achieve a successful sale of the "good" bank. Additionally, the Argentine authorities could consider suitable normative amendments to widen the arsenal of contractual arrangements that can facilitate the transfer of a "good" bank's assets -for instance, "loss-sharing" agreements that may improve the buyer's incentives to maximize asset recovery. The review of these alternatives could prove even more useful considering that the Trust scheme used in the case of Banco Almafuerte appears to lack adequate Incentives to promote maximum asset recovery.15 Third, the particular financial engineering utilized in the resolution of Banco Almafuerte raises Issues in connection to the potential payment failure by the Trust and the treatment of A Bonds for regulatory purposes. To be sure, the overcollateralization of the Bonds issued by the Trust clearly reduces the risk of default by the latter, particularly In respect of the (senior) A Bonds. Nonetheless, It seems reasonable to argue that there is a nonzero probability of nonpayment. Recognizing this, the Central Bank recently has created a contingent fund that would make payments on A Bonds in case the corresponding Trust were unable to do so. This fund is the result of a "controlled" regulatory forbearance scheme, according to which the banks that acquire "good" banks (in the context of the application of Article 35 bis) are allowed to constitute part of the regulatory minimum liquidity with public bonds that yield a greater interest rate than that earned on deposits in the Central Bank; the earnings attributable to such difference in rates feed the mentioned fund. Given this scheme, the Central Bank allows for A Bonds to be recorded in the balance sheets at face value, without the need for provisions. 'Ingenious though it is, the scheme does reduce systemic liquidity, insignificantly at this stage, but it could do It in a noticeable manner in the event of a failure of a large bank or of several small to mid-size ones. The Interagency Committee may wish to discuss alternative mechanisms to the contingent fund scheme. World Bank User Q\Argentina\ArgFailResl.rtf 03/20/99 3:40 PM 15 In effect, early observations on the performance of the Trust scheme indicate a tendency for the Trust process to be "captured" by lawyers, with the consequent sacrifice of market-based financial criteria. Banks that acquired the "good" banks and that administer the Trusts do not seem to have any incentive to recover assets beyond what is needed to make payments on A Bonds. ANNEX 1 Example of Bank Resolution under the Emerging Model (i la Almafuerte) Simplifying assumptions: Transactions costs are zero; non-cash assets are net of provisions; the going concern value of assets is preserved during the resolution; assets become valueless under judicial liquidation. The falling bank's balance sheet: Falling bank Assets Liabilities Cash 10 2 Uabilities to labor Non-cash assets 80 20 Guaranteed deposits 48 Non-guaranteed deposits 10 Uabilities to Central Bank 20 Other liabilities -10 Capital The falling bank's non-cash assets are securitized via a Trust, as follows: Trust Assets Llabilities Non-cash assets 80 40 A Bonds (senior to B Bonds) 10 B Bonds (senior to C Bonds) 20 C Bonds 10 Excess collateral for bonds A "good" bank Is formed for all deposits: "Good" bank Assets Liabilities Cash 10 2 Uabilities to labor A Bonds 40 20 Guaranteed deposits IOU from Deposit Guarantee Agency 20 48 Non-guaranteed deposits 0 Capital The failing's bank debt to the Central Bank Is settled with B Bonds: The Central Bank releases to the Trust the failing bank's assets that had been the collateral for a $10 past liquidity loan, and receives $10 in B Bonds in payment for such loan. The Deposit Guarantee Agency (DGA) receives C Bonds for Its contribution to the "good" bank: A $20 IOU from the DGA, equal to the amount of guaranteed deposits, is added to the assets of the "good' bank. In exchange, the DGA receives C Bonds from the Trust. The DGA In not worse off compared to a cash payout of guaranteed deposits in exchange for a claim against the liquidation because, by assumption, no value is recoverable through judicial liquidation. The residual ("bad") bank, containing the rest the falling bank's liabilities, moves on to judicial liquidation... Residual bank Assets Liabilities Non-cash assets 0 20 Other liabilities The assets of the residual bank are those not retained in the Trust because they are deemed valueless. ANNEX 2 Example of Bank Resolution with (I) an Incremental Rule for Inclusion of Deposits in the "Good" Bank, and (li) a Deposit Guarantee Conditional on Asset Insufficiency Simplifying assumptions: Transactions costs are zero; non-cash assets are net of provisions; the going concern value of assets is preserved during the resolution; non-cash good assets loose 60 percent of their value under judicial liquidation. The falling bank's balance sheet: Falling bank Assets Liabilities Cash 5 5 Liabilities to labor Non-cash 'good" assets 45 30 Deposits amounts : $30,000 (guaranteed) Non-cash "bad" assets 0 15 $30,000 < Deposit amounts $50,000 20 $50,000 < Deposit amounts . $75,000 20 Deposit amounts > $75,000 10 Other liabilities -50 Capital Available good assets imply that not all deposits can be included In the "good" bank; an incremental (not prorata) rule is used to select deposits for the "good" bank Since good assets in the failing bank exceed guaranteed deposits, the Deposit Guarantee Agency makes no contribution to the "good" bank "Good" bank Assets Liabilities Cash 5 5 Liabilities to labor Non-cash "good" assets 45 30 Deposits amounts 5 $30,000 (guaranteed) $ from DGA 0 15 $30,000 < Deposit amounts s $50,000 0 Capital The incremental rule dictates that only up to $50,000 per depositor can be included In the liability side of the "good" bank. Larger deposit amounts take a hit in the liquidation (see residual bank below). The Deposit Guarantee is conditional on assets being insufficient to cover guaranteed deposits, hence, in this case, It makes no contribution to the "good" bank. Due to the Incremental rule and the preservation of asset value within the resolution, smaller non-guaranteed deposits are clearly better off than under the liquidation, while larger depositors are not necessarily worse off. The residual ("bad") bank containing larger deposit amounts and other liabilities moves on to liquidation... Residual bank Assets Liabilities Non-cash "bad" assets 0 20 $50,000 < Deposit amounts s $75,000 $ from sale of "good" bank ? 20 Deposit amounts > $75,000 0 Liability to DGA 10 Other liabilities ? Capital ANNEX 3 Example of Bank Resolution with (I) an Incremental Rule for Inclusion of Deposits in the "Good" Bank, and (ii) a Least Cost Criterion for the contribution of the Deposit Guarantee Agency (DGA) to the "Good" bank Simplifying assumptions: Transactions costs are zero; non-cash assets are net of provisions; the going concern value of assets is preserved during the resolution; non-cash good assets loose 60 percent of their value under judicial liquidation. The falling bank's balance sheet: Falling bank Assets Liabilities Cash 5 5 Liabilities to labor Non-cash "good" assets 60 20 Deposits amounts : $30,000 (guaranteed) Non-cash "bad" assets 0 40 $30,000 < Deposit amounts $75,000 15 $75,000 < Deposit amounts : $125,000 15 Deposit amounts > $125,000 5 Other liabilities -40 Capital Alternative A: Available good assets Imply that not all deposits can be included In the "good" bank; an Incremental (not prorata) rule Is used to give priority to smaller depositors for Inclusion in the "good" bank For the least cost criterion, it Is assumed that, In the liquidation, the DGA has the same priority of claim as that of non-guaranteed deposits, sharing with them prorate "Good" bank Assets Liabilities Cash 5 5 Liabilities to labor Non-cash "good" assets 60 20 Deposits amounts s $30,000 (guaranteed) $ from DGA 15 40 $30,000 < Deposit amounts 5 $75,000 15 $75,000 < Deposit amounts : $125,000 0 Capital Under this alternative, given available assets, the incremental rule dictates that only up to $125,000 per depositor can be included in the liability side of the "good" bank. Larger deposit amounts take a hit in the liquidation (see residual bank below). The least cost criterion implies that the DGA's contribution to the "good" bank could not exceed a maximum (MaxC), usually defined as the cost to the DGA of paying out guaranteed deposits in cash in exchange for a claim against the liquidation. Given the assumptions regarding the loss of asset value and the DGA's priority of claim in the liquidation, MaxC may be calculated as follows: I MaxC = GD - (GD/L)*(CA + LVNCA - LL) MaxC = 20 - (20/100)*[5 + (1 - 0.6)*60 - 5] = 15.2 GD e guaranteed deposits; L w total liabilities; CA m cash assets; LVNCA a liquidation value of non-cash assets; LL a liabilities to labor. The residual ("bad") bank containing large deposit amounts and other liabilities, Including the claim of the DGA, moves on to lIquidation... Residual bank Assets Liabilities Non-cash "bad" assets 0 15 Deposit amounts > $125,000 $ from sale of "good" bank ? 15 Liability to DGA 5 Other liabilities ? Capital ANNEX 3 (cont.) Alternative B: Available assets for "good" bank Imply a need to exclude certain (high) deposit amounts; such exclusion is performed according to an "Incremental rule" For the least cost criterion, It is assumed that the DGA has priority of claim over depositors In the liquidation, Le., that It stands first in line (after labor claims) "Good" bank Assets Liabilities Cash 5 5 Liabilities to labor Non-cash "good" assets 60 20 Deposits amounts 5 $30,000 (guaranteed) $ from DGA 0 40 $30,000 < Deposit amounts a $75,000 0 Capital Under this alternative, given available assets, the incremental rule dictates that only up to $75,000 per depositor can be included in the liability side of the "good" bank. Larger deposit amounts take a hit in the liquidation (see residual bank below). The least cost criterion implies that the DGA's contribution to the "good" bank could not exceed a maximum (MaxC), usually defined as the cost to the DGA of paying out guaranteed deposits in cash in exchange for a claim against the liquidation. Given the assumptions regarding the loss of asset value and the DGA's priority of claim in the liquidation, MaxC may be calculated as follows: MaxC = GD - X X w (CA + LVNCA - LL) If (CA + LVNCA - LL) 5 GD X w GD If (CA + LVNCA - LL) > GD (CA + LVNCA - LL) = [5 + (1 - 0.6)*60 - 5] = 24 > GD = 20 a MaxC = 20 - 20 = 0 GD i guaranteed deposits; L a total liabilities; CA * cash assets; LVNCA g liquidation value of non-cash assets; LL liabilities to labor. The residual ("bad") bank containing large deposit amounts and other liabilities, including the claim of the DGA (If any), moves on to judicial liquidation... Residual bank I Assets Liabilities Non-cash "bad" assets 0 15 $75,000 < Deposit amounts s $125,000 $ from sale of "good" bank ? 15 Deposit amounts > $125,000 0 Liability to DGA 5 Other liabilities ? Capital CATALOGUERS/ILE CONFIDENTIAL Report No.: 19082 AR Type: ER