Dodd–Frank not only cut back on the Fed’s ability to serve as lender of last resort, it also cut back on the FDIC’s ability to expand deposit insurance or create other guarantees during a crisis. A strong guarantee system is an important complement to a strong lender of last resort, and indeed even more important when the power of the lender of last resort has been seriously weakened. Nonetheless, increasingly short-term liabilities have moved out of the banking system and uninsured short-term liabilities of nonbanks have increased. So we need not only to restore the powers of the government to insure bank funding, we need to give serious thought to expanding such funding to the shadow banking sector, particularly money market funds. The expansion of guarantees outside the banking sector raises significant challenges.
Dodd–Frank §1105 limits the FDIC’s ability to provide the guarantees that helped stabilize the financial system during the financial crisis, such as (1) the Transaction Account Guarantee Program (“TAGP”) to provide temporary unlimited guarantees or increases in insurance limits to domestic non–interest-bearing transaction deposits and (2) the Debt Guarantee Program (“DGP”) to temporarily guarantee the issuance of senior debt. The FDIC implemented the TAGP and DGP by invoking a general authority to mitigate instability in the financial system once the Treasury Secretary made a systemic risk determination under section 13(c)(4)(G) of the Federal Deposit Insurance Act, an authority now removed.
Dodd–Frank §1106 further diminishes the FDIC’s powers to protect bank creditors by eliminating the open bank assistance that had allowed the FDIC to provide loans, purchase assets, assume liabilities, and provide cash contributions to prevent an insured bank from failing. Now, a bank subsidiary must enter an FDIC receivership for the FDIC to have the authority to guarantee the short-term liabilities of the bank subsidiary. The FDIC may only do so if the Secretary of the Treasury (in consultation with the President and upon the recommendation of two-thirds of the Board of Directors of the FDIC and the Board of Governors of the Federal Reserve,) determines that doing so is necessary to avoid or mitigate “serious adverse effects on financial stability.”
Finally, new Section 131 of the Emergency Stabilization Act, as a result of the TARP legislation, prevents the Treasury from enacting the same temporary guarantee program that it used successfully to stabilize the money market mutual fund industry during the financial crisis. Congress has barred the Treasury from using the Economic Stabilization Fund to establish any future guarantee program for the industry.
Again, a starting point for reform would be to eliminate the changes made to the guarantee system by Dodd–Frank. But more needs to be done to further strengthen the system. The objective of this chapter is to explore the various problems in designing a better and broader insurance regime but take no position on the best approach. The problems are complicated and deserve much further study.
Insurance for customer deposits administered by the FDIC has formed an integral element of depository banking regulation in the United States since 1934.1 Deposit insurance is credited with stabilizing the depository banking system after it collapsed in the early 1930s.2 Indeed its application has not been confined to the United States: explicit deposit insurance is a recurring worldwide feature of modern banking regulation utilized in more than 88 countries (excluding countries that employ an “implicit” guarantee of bank deposits that is not formalized through the provision of a discrete insurance fund).3 The economic efficiencies of deposit insurance have been demonstrated by Diamond and Dybvig4 and Carnell, Macey, and Miller (2009),5 among others.
The federal deposit insurance system arose as a consequence of the Great Depression, with federal officials recognizing the efficacy of using deposit insurance to assure depositors and preemptively forestall bank runs.6 The leading supporter for federal deposit insurance was Representative Henry Steagall.7 Throughout the Depression, Representative Steagall indicated that deposit insurance would dissuade depositors from running on a potentially distressed bank, thereby resulting in durable “stability” for the United States’ banking system.8 Opponents, including Senator Carter Glass, countered that the failures of prior state deposit insurance regimes showed the probable ineffectiveness of this approach.9 Strong public support for the proposal ultimately persuaded those opponents to accept the establishment of federal deposit insurance.10 In June 1933, the Banking Act of 1933 was accordingly enacted; Section 8 of the Banking Act stipulated the creation of the Federal Deposit Insurance Corporation (“FDIC”).11
For deposit-taking banks, the role of liability insurer is filled by the FDIC, but only in the context of depository borrowing under a limit (currently $250,000); it does not cover nonbanks.12 Although depository insurance is rightly regarded as a critical stabilizing attribute of financial regulation, innovation in financial technology over the past three decades and increasing nonbank intermediation in the modern financial system have now rendered the coverage it provides highly incomplete. This was proven most dramatically during the financial crisis. At the beginning of the crisis, short-term creditors of financial institutions assumed the existence of an implicit government guarantee of all short-term liabilities and appeared to be largely justified in doing so. The government’s assisted rescue of Bear Stearns in March 2008 in partnership with JPMorgan Chase and its subsequent effective nationalization of the government-sponsored enterprises (“GSEs”) Freddie Mac and Fannie Mae in July of the same year are likely to have reinforced belief among market participants (including short-term creditors) in the existence of an unlimited implied public guarantee of large US financial institutions. But then by allowing Lehman Brothers to fail in September 2008, the government was seen as canceling or at least weakening the guarantee. According to this interpretation, the anti-bailout signal transmitted by the failure of Lehman, not the failure itself, triggered the spread of contagion effects in markets for short-term institutional borrowing by withdrawing protection that market participants had assumed they would receive.13
Based on this lesson from the financial crisis, an important part of a solution to contagion may be a more complete public guarantee of short-term nondeposit financial liabilities, whether held by banks or by nonbank financial institutions.14 Such a system of more universal insurance for short-term financial liabilities would assure short-term creditors automatic protection through assessments on issuers,15 removing the element of uncertainty tied to discretionary emergency lending or politically contingent (and unpopular) bailouts. The costs of supplying a public guarantee could be internalized through the use of insurance premiums or through some other form of assessment, either before or after they are triggered.
Of course, the concern with short-term liability insurance arises from the same moral hazard problem created in all insurance regimes (or actual bailouts as discussed later in this book): insured creditors, like any policyholders protected from loss, have little incentive to monitor risk-taking by issuers.16 Short-term creditors that are insured against losses will also require a lower interest rate on the credit provided to financial institutions. The resulting lower cost of short-term debt will give financial institutions further incentive to increase short-term funding by taking advantage of the cheaper funding. More short-term funding corresponds with a higher likelihood of failure, and therefore a higher expected payout by the insurance provider. This economic cost of moral hazard can in theory be internalized by optimizing the premiums extracted from policyholders, but it is questionable whether the pricing of insurance on short-term liabilities can be perfected.
Given the current levels of insured liabilities in the financial system, a very challenging question for an insurance regime is what caps, if any, to apply to guarantees of covered liabilities. To estimate the size of short-term liabilities, we include money market mutual fund shares, repurchase agreements, commercial paper, and securities lending. This is consistent with the scope of “shadow liabilities,” as defined by Pozsar et al. (2012).17
Cumulatively, as shown in table 12.1, nondeposit uninsured short-term liabilities totaled $8 trillion at the end of 2014. While this represents a reduction from their peak level in 2008 of about $9.9 trillion, these nondeposit liabilities (all uninsured) are comparable to total deposits of about $10.4 trillion, and greater than insured deposits of $6.1 trillion.
Deposit and nondeposit US financial system liabilities, 1950 to 2014
Note: Bd. of Governors of the Fed. Reserve Sys., Flow of Funds Accounts of the United States, (June 6, 2013), http://www.federalreserve.gov/releases/z1/.
The tabulation is based on the convention in Pozsar et al., supra note 17, at 5 n.4 (defining “shadow bank liabilities” as sum of MMMF shares outstanding [line 13, L.121119], open market paper [line 1, L.208], federal funds and repo liabilities [line 1, L.207], net securities loaned [line 20, L.130129], GSE liabilities [line 2125, L.124123], agency- and GSE-backed pool securities [line 6, L.125123], and ABS issuer liabilities [line 11, L.126]).124]); Quarterly Banking Profile: First Quarter 2013, 7 FDIC Quarterly 2 (2013) (2012 data found in table I-B).
While uninsured deposits hit a low of $1.6 trillion in 2010, this was caused in large part by the extension of unlimited insurance to all bank transaction accounts from 2008 to 2012. With the end of such unlimited insurance in December 2012, there has been a substantial growth in uninsured bank deposits, standing at $4.2 trillion in 2014, more than two-thirds the size of insured deposits. In 2014 only 59.1 percent of domestic bank deposits were insured, compared with roughly 80 percent from 2010 to 2012. So the level of uninsured bank deposits is significant, and represents a substantial exposure to a run. As illustrated in the final row of table 12.1, the total of uninsured short-term funding was around $12.3 trillion in 2014, almost triple the $4.2 trillion in insured deposits. This implies that only around 33 percent of short-term liabilities are insured, around half of the fraction that was insured in 1970. So, not only do we have substantial contagion risk at banks, such risk also exists outside banks, as we witnessed during the crisis.
It should be noted that this methodology produces some double counting. This is because the our estimate includes money market funds and commercial paper, repos, and uninsured depositors, even though money market funds are large investors in commercial paper, repos, and uninsured deposits (e.g., CDs). This means that the fraction of these short-term debt instruments held by money market funds would be double counted. However, government MMFs also invest directly in Treasuries and GSE securities, so all MMMF shares ($2.688 trillion) are not double counted. According to Crane Data, government MMFs hold approximately $1.024 trillion in such government securities directly.18 Thus, to most accurately eliminate double-counting through the inclusion of money market funds, $1.664 trillion in MMMF shares ($2.688 trillion − $1.024 trillion) should be subtracted from the $12.273 trillion total of uninsured short-term liabilities. Total uninsured short-term liabilities would therefore stand at $10.61 trillion at the end of 2014. By this measure, 63.4 percent of all short-term liabilities are uninsured.
Does this mean we need to insure all short-term liabilities to prevent runs on the financial system? According to Crane Data, all money market funds hold approximately $630 billion in repos (or 17.5 percent of outstanding repo), $369 billion in commercial paper (40 percent of outstanding commercial paper), and $511 billion in CDs,19 so that insuring only MMF investments might be sufficient to deter runs in those markets. However, doing so would still leave at least $9.1 trillion in uninsured short-term liabilities. This includes repos, commercial paper, and uninsured deposits not held by money market funds.
The Financial Stability Board and the International Organization of Securities Commissions have suggested that investment funds such as ordinary mutual funds, may also pose run risks.20 At the end of 2013, the US mutual fund industry had around $15 trillion in assets.21 Including mutual funds would therefore increase the level of short-term “liabilities” to approximately $25 trillion. Large-scale redemption by mutual fund shareholders could force a fund to liquidate assets, depressing the market price of those assets. However, this “fire-sale” scenario is unlikely to impose losses on significant financial institutions that do not hold the long-term bonds and equities that are typically held by mutual funds.
One core principle of a nondeposit liability insurance system is that covered institutions should internalize their costs by making payments to the insurance provider (the government and/or private sector) that reflect the cost of providing the guarantee. There are several different timing mechanisms by which these insurance fees may be collected; specifically, institutions could pay for coverage before, during, or after the guarantee is used.
Under a system funded ex ante, covered institutions pay a periodic risk-based fee, or insurance premium, in exchange for receiving nondeposit short-term liability insurance. The main reason for this approach is that it provides a private fund from which insurance payments could be made, when necessary, thus avoiding public expenditures.22 Recurring and risk-based fees also help mitigate the moral hazard that can arise from guaranteed liabilities. By pricing insurance to reflect the risk of covered institutions’ activities, regulators can attempt to control moral hazard.
While some of the challenges of pricing insurance for nondepository liabilities are new, many of these issues have been addressed for decades in the analogous case of pricing FDIC deposit insurance. Examining how FDIC insurance is priced is therefore an instructive place to begin the analysis.
When the Banking Act of 1933 established the FDIC, insured institutions were covered for $2,500 for each depositor (a limit that was subsequently raised) and paid premiums as a fixed percentage of insurable deposits.23 While the basic assessment rates were adjusted over the initial years of the FDIC, under the permanent system, rates settled out at 1/12th of 1 percent (8.33 basis points) of total deposits below the insurance ceiling, some portion of which was credited back to covered banks if the insurance fund’s anticipated losses were covered beyond a specified percentage.24 Despite the lack of sensitivity in this insurance pricing system to risk, it is credited with largely preventing banking panics for over 50 years after its implementation.25
Bank failures related to the late 1980s savings and loan crisis caused the FDIC Bank Insurance Fund to become insolvent by $7 billion. This shortfall, and the recognition that the FDIC had likely underpriced its insurance coverage for large institutions, prompted Congress to make key changes to the structure and insurance policies of the FDIC in what became the Federal Deposit Insurance Corporation Improvement Act of 1991 (“FDICIA”).26 FDICIA required the FDIC to set deposit insurance premiums in relation to the risk the bank poses to the insurance fund.27 The assessment base was still set at banks’ insurable deposit base but now assessment rates depended on (1) banks’ capital ratings (classified in three groups), and (2) banks’ “Supervisory Group,” a determination based on a bank’s score under the CAMELS rating system (a composite score of different bank health metrics).28 Premium payments ranged from a low of 2 bps for the most well-capitalized banks to 43 bps for undercapitalized banks with low CAMELS ratings.29 While this system sensibly reflected an attempt to have the cost of insurance reflect banks’ risk to the BIF, it had several problems. Carnell et al. note that the risk-based pricing arguably still did not account for the degree of failure risk for the riskiest banks,30 perhaps reflected by the fact that the large banks paid the lowest possible rates under this scheme.31 Moreover, when the BIF reached the target designated reserve ratio (1.25 percent of insured deposits) in 1995, most banks stopped paying premiums altogether; in 1999, for example, 93 percent of FDIC-insured institutions paid no premiums at all.32
The Dodd–Frank Act mandated a third wave of changes to the FDIC insurance pricing system,33 largely prompted by the incurred and expected losses in the DIF (the Deposit Insurance Fund, successor to the BIF) related to the financial crisis. In February 2011 the FDIC promulgated its new rules implementing these Dodd–Frank requirements.34 These rules affected both of the primary inputs for deposit insurance pricing. First, the assessment base for all banks was changed “from [a system] based on domestic deposits to one based on assets.”35 This new assessment base highlights a fundamental change in the FDIC’s definition of risk: while historic approaches had been focused on an institution’s covered deposits (which directly represent the maximum potential loss to the DIF), the new approach recognizes that an increase in balance sheet risk—even if it is not funded by additional insured deposits—could increase risk to the DIF from an increase in the probability of losses.
The new rule has modified assessment rates as well, although these rates still depend on similar key inputs, namely capital adequacy and supervisory ratings. The assessment rate system is bifurcated: institutions defined as “small insured depository institutions” (those having less than $10 billion in assets) pay one set of premiums based on risk classifications. These risk classifications are a function of their risk level (which depends on tangible equity and a supervisory evaluation) and the level of reserves in the DIF, as set out in table 12.2.
DIF assessment rates
Note: Assessments, Large Bank Pricing, 76 Fed. Reg. 10,717–20 (Feb. 25, 2011); Barbara R. Mendelson and Marc-Alain Galeazzi, Morrison Foerster, Client Alert: FDIC approves final rule of assessments, dividends, assessment base and large bank pricing, at 2 (Feb. 10, 2011) [hereinafter Morrison Foerster LLP, Summary of New FDIC Rules].
In contrast, “large” and/or “highly complex”36 financial institutions pay another set of premiums, which, under the new rule, are not based on risk categories but instead rely on the use of a new risk-based “scorecard” that determines an institution’s “performance score” and “loss severity score.”37 The combined scores produce an assessment rate. An institution’s performance score is a “weighted average of … three components: the weighted average CAMELS rating score; the ability to withstand asset-related stress score; and the ability to withstand funding-related stress score.”38 Its loss severity score “measures the relative magnitude of potential loss to the FDIC in the event of the insured depository institution’s failure.”39 These combine to determine an institution’s new risk-adjusted score.
This new FDIC approach to pricing has an important implication for insuring nondeposit liabilities. The FDIC now correctly recognizes that expected loss exposure to the DIF depends on the total net assets of the bank, and not just on insured deposits. This suggests that any scheme insuring nondeposit liabilities should similarly take into account the overall riskiness of the asset side of a financial institution’s balance sheet, and not simply the size or particular types of instrument on the liability side of the ledger.40 This type of pricing exercise could be implemented by performing historical credit analysis on portfolios, as the Basel capital adequacy system does, although one must recognize as well the limitations of that system.41
The FDIC’s approach to pricing insurance starts from the premise (familiar to credit analysts) that the expected cost to the government of providing insurance is equal to the expected losses the insurance fund will incur by providing insurance over time.42 This expected loss to the insurance fund is contingent on two factors: the probability of loss (failure of the insured institution) and the magnitude of loss given the failure of the institution.43 However, both of these factors are difficult to quantify accurately. First, the probability of an individual financial institution failing is a complex function of what is going on in the broader financial system; the loss given default will also be affected by such institutional interdependencies. While one might think that looking at the historical default performance of financial institutions would be instructive in this regard, models designed to predict the frequency of such “tail-risk” events often suffer from having insufficient historical data for back-testing (particularly in the case of new financial instruments)44 and their over-reliance on assumptions about the normality of outcome distributions.45 Second, both the probability of failure and the expected loss given failure are affected by the very existence of insurance itself; estimating these variables in a system with insurance versus one without insurance is a particularly difficult task.46
While this credit analysis framework may provide a conceptually useful way of thinking about the insurer’s loss exposure, it is not particularly helpful in quantifying precise prices for the insurance itself. In this regard it may be more helpful to draw on options-pricing theory. From the insurers’ perspective, providing a guarantee is analogous to writing a put option on the asset value of the bank, struck at the value of the firm’s nonequity liabilities. Indeed Robert Merton has made this argument for a long time, and suggested a quantitative model for pricing deposit insurance based on the same readily observable inputs widely used in contemporary options pricing theory.47 The essential insight from Merton’s analysis is that the option price (i.e., the actuarially fair insurance price) is a function of the volatility of the underlying asset owned by the financial institution; the degree of “moneyness” of the option accordingly corresponds to the equity position of the owners of the firm—when the option, which is owned by the short-term creditors of the institution, is extremely “out of the money,” the equity has positive value, and vice versa.48
While this option-pricing approach to insurance pricing is elegant and has intuitive appeal, there are some theoretical and practical difficulties with it, which may explain why the FDIC has never adopted this approach to risk-based pricing. On the theoretical front, the option-pricing formula requires an assumption about the shape of the distribution of returns for bank asset valuations.49 On the practical front, since most banks in the United States, and smaller institutions in particular, do not have publicly traded equity, obtaining accurate asset-pricing information could be difficult, although this problem is mitigated if the insurance coverage is limited to larger institutions.50
A different approach than the one currently used would be for the government to issue a guarantee of the short-term nondeposit liabilities of financial institutions above the designated asset threshold, and to fund this guarantee on an ex post basis, by levying a charge on covered institutions (or their stakeholders) after the crisis has passed. From the perspective of short-term creditors who would benefit from the guarantee, these two approaches would provide the same protection; the difference is that there would be no permanent “insurance fund” under the ex post system, so the payout to short-term creditors of a failed covered institution would first be covered by the government, which would then later recoup the cost of the guarantee through assessments on covered institutions. On the one hand, an advantage of the ex post funded system is that the guarantee does not cost anything unless a covered institution actually fails and the guarantee is used. On the other hand, institutions would not pay for the benefit of avoiding contagious runs through the availability of the guarantee. The mechanics of how an ex post charge would work presents challenges in terms of who would pay the charge and what the size of the charge would be.
Determining who would pay the charge requires two levels of analysis: which institutions would pay the charge and which stakeholders of those institutions would be liable.52 One might start from the principle that those who benefit from the insurance system should bear the cost of it; on this view, short-term creditors of the failed institution would appear to be the obvious beneficiaries and thus the responsible parties. However, there are two reasons to favor applying the charge more broadly, with respect to both institutions and stakeholders. First, the potential benefits of a guarantee are much broader than any particular troubled institution or its own stakeholders; other stakeholders (besides short-term creditors) in the covered institution benefit because they do not have the value of their investments impaired by contagious run behavior during crises. Short-term creditors and longer term stakeholders in other covered financial institutions also benefit because of the reduced likelihood of contagious runs. Second, if the charge were focused only on short-term creditors of particular troubled institutions, this could actually cause these creditors to run in anticipation of a levy on that institution, thereby exacerbating rather than mitigating the problem of contagious panic. Both of these arguments suggest that the charge should be broadly based with respect to both institutions and stakeholders. Additionally, imposing small charges on a wider range of stakeholders would reduce the risk of distorting the markets for short- or long-term capital (compared with imposing larger, more concentrated charges on particular groups of stakeholders).
Determining the size of the charge under an ex post system would also pose challenges. Perhaps the most obvious way to size the ex post charge would be to simply assess a charge equivalent to the realized cost of guarantee (which would be known after resolving a failed covered institution). Under this approach, charges would only be assessed in the (hopefully) relatively rare event that covered institutions actually fail. While this is arguably attractive, because payments to cover the cost of the short-term liability guarantee would only be required in the rare case that they have been paid out, it is not clear that the magnitude of the necessary ex post assessments can optimally reduce ex ante moral hazard for covered institutions and their stakeholders.
Still another option for an insurance scheme would be to implement a guarantee program for short-term nondeposit creditors of financial institutions only when it becomes apparent that there is a crisis involving heightened systemic risk. Under such a system there would be no standing guarantee for nondeposit creditors (as in either of the previous two approaches) or standing insurance fund (as in the ex ante funded approach). Rather, the government (perhaps led by FSOC) would monitor systemic risk and issue a guarantee to a relevant set of short-term nondeposit creditors53 when it is perceived to be necessary. The guarantee itself would be mandatory for all institutions determined to be within the scope of coverage; this would ensure maximum protection against contagious panic. By contrast, Treasury’s guarantee of money market funds during the financial crisis was voluntary; money market funds (and not their shareholders) could opt to join by paying a fee (discussed in the next paragraph). While the practical impact of Treasury’s guarantee was ultimately quite similar to a mandatory guarantee—nearly all major money market funds, representing over $3 trillion in industry assets under management, opted to join—it seems reasonable to expect that a mandatory guarantee would be even more effective at minimizing the possibility of systemic contagion, particularly as applied to a potentially broader set of institutions than just money market funds and depository institutions. Such a system could be funded through assessments on covered institutions at the time the guarantee is furnished; to the extent such assessments were insufficient to cover the cost of the guarantee, the government would have to employ an ex post approach as discussed above, but the size of such an assessment would be less given the charges to those covered in the crisis.
A major problem with this approach, as with the ex post approach, is the cost of the insurance would not take into account the benefit of the option to have such insurance in a crisis, when the option never has to be used because of its very presence.
A potential challenge for an expanded insurance regime is achieving international participation. This is important in order to prevent short-term creditors from transferring funding out of financial institutions in risky and nonguaranteed jurisdictions into safer insured institutions in jurisdictions with public guarantees during a financial crisis. Without coordination, uneven implementation of insurance will exert a destabilizing effect on nonguaranteed institutions as investor funds flow elsewhere or into risk-free instruments backed by the government. This danger is illustrated by the effect on deposit flows of the Irish government’s public guarantee of all deposits and debt instruments at six major Irish financial institutions, including Allied Irish Banks, Bank of Ireland, and Anglo Irish Bank in September 2008.54 The Irish guarantee caused deposit outflows from banking institutions elsewhere in Europe, including the United Kingdom, into Ireland as investors sought to shield themselves from rising credit risk.55 One straightforward response to this problem is to coordinate to exclude creditors who transfer into a jurisdiction during a crisis from the protection of that jurisdiction’s insurance. This would deter outflows seeking to take advantage of a more favorable insurance regime located elsewhere in a moment of panic. But such coordination would be difficult to achieve. In the financial crisis of 2007 to 2009, the US Treasury and FDIC limited foreign access to its stabilization programs to some extent (see table 12.3): foreign subsidiaries and branches were ineligible for the Capital Purchase Program, and foreign branches also were restricted from accessing the FDIC’s TLGP debt guarantee. Other Treasury and FDIC protections, including the PPIF, Transaction Account Guarantee, and various Federal Reserve facilities including the TALF and CPFF were, however, made available to foreign subsidiaries and branches.