The minimum balance at risk: a proposal to mitigate the systemic risks posed by money market funds.

McCabe, Patrick E. ; Cipriani, Marco ; Holscher, Michael 等

ABSTRACT This paper introduces a proposal for money market fund (MMF) reform to mitigate the systemic risk and externalities that arise from the funds' vulnerability to runs and to protect shareholders who do not redeem quickly when runs occur. Our proposal would require that a small fraction of each MMF shareholder's recent balances, called the "minimum balance at risk" (MBR), be available for redemption only with a delay of 30 days. Most regular transactions in the fund would be unaffected; the requirement would only affect redemptions of the shareholder's MBR. In addition, in the rare event that a fund suffers losses, the MBRs of investors who have recently made large redemptions would absorb losses before those of nonredeeming investors. This subordination of redeeming investors' MBRs would create a disincentive to redeem if the fund is likely to have losses, but would have little effect on incentives when the risk of loss is remote. We use empirical evidence, including a novel data set from the U.S. Treasury and the U.S. Securities and Exchange Commission on MMF losses in 2008, to calibrate an MBR rule that would reduce the vulnerability of MMFs to runs.

**********

By many measures, money market funds (MMFs, or "money funds") are a popular financial product. With $2.7 trillion in assets under management at the end of 2012, MMFs represented over a fifth of all U.S. mutual fund assets, according to the Investment Company Institute. MMFs are key intermediaries of short-term funding and hold large fractions of the short-term debt issued by financial institutions in capital markets. (1) The popularity of money funds largely reflects the attractiveness of their defining feature for many investors: MMFs, unlike other mutual funds, typically maintain a stable $1.00 share price (also known as the fund's "net asset value," or NAV). Their stability of principal allows MMFs to serve as an important cash management tool for individuals, firms, institutions, and governments.

However, MMFs are vulnerable to runs, and given the size of the money fund industry and its importance in allocating short-term funding to financial institutions, this vulnerability poses considerable risk to the financial system. The potentially dire consequences of a run on MMFs were evident in September 2008, when the Lehman Brothers bankruptcy caused one fund to "break the buck" (that is, its NAV fell below $1, so its investors suffered principal losses) and triggered massive and widespread redemptions from other MMFs. (2) These outflows were a key factor in the virtual shutdown of short-term funding markets and a broader curtailment of credit supply (see, for example, Federal Open Market Committee 2008, Board of Governors of the Federal Reserve System 2009, U.S. Securities and Exchange Commission 2009, President's Working Group on Financial Markets 2010). The severity of the damage to financial stability caused by the run in 2008 led to unprecedented government interventions to support MMFs in order to halt the run. Since then, the funds' vulnerability has continued to pose risks to the financial system. The heavy exposures of MMFs to European financial institutions, for example, have put the funds at risk of transmitting strains from Europe very rapidly to U.S. short-term funding markets (Financial Stability Oversight Council 2011, Chernenko and Sunderam 2013). Yet policymakers have fewer tools available now to address MMF runs than they did in 2008; in particular, the Treasury's Temporary Guarantee Program for Money Market Funds, which effectively halted the run in 2008, would no longer be possible under current law: the Emergency Economic Stabilization Act of 2008 specifically prohibits the Treasury from reusing that mechanism. In light of the systemic risk stemming from MMFs' susceptibility to runs, calls for reform have come from government agencies (for example, Schapiro 2010, 2011, 2012, President's Working Group on Financial Markets 2010, Financial Stability Oversight Council 2011, 2012a, 2012b), from academics (for example, Squam Lake Group 2011, Hanson, Scharfstein, and Sunderam 2012), and from the financial industry (for example, Mendelson and Hoerner 2011, Goebel, Dwyer, and Messman 2011).

This paper proposes a new approach to mitigating the vulnerability of MMFs to runs by introducing a "minimum balance at risk" (MBR) that could provide a disincentive to run from a troubled money fund. (3) The MBR would be a small fraction (for example, 5 percent) of each shareholder's recent balances that could be redeemed only with a delay. The delay would ensure that redeeming investors remain partially invested in the fund long enough (we suggest 30 days) to share in any imminent portfolio losses or costs of their redemptions. However, as long as an investor's balance exceeds her MBR, the rule would have no effect on her transactions, and no portion of any redemption would be delayed if her remaining shares exceed her minimum balance.

An MBR could be introduced in a manner that preserves the basic features of MMFs. Funds could, for example, maintain stable $1 NAVs and honor most transactions without any delay or restrictions. The loss allocation rules that are central to the MBR concept would affect shareholders' actual balances only in the event that a fund breaks the buck and closes. Implementation would require an MMF to track just two additional variables for each investor: her MBR and any portion of her MBR that she has requested to redeem. Importantly, the MBR concept could be introduced together with other reforms. In particular, we suggest that an MBR would work well in tandem with a capital buffer requirement for money funds.

An appropriately designed MBR would have several benefits. By discouraging investors from redeeming from a troubled MMF, an MBR would help prevent a run and its destabilizing and costly repercussions. This not only would benefit the fund and its investors but also would mitigate the externalities that result from an MMF's potential to propagate strains throughout the financial system. An MBR also would benefit investors who are not prone to redeem shares quickly at the first sign of trouble for an MMF. In particular, retail investors (small, typically individual investors, as distinct from institutional investors such as corporate treasurers), who historically have been less quick to run from distressed funds, would enjoy additional protections because the MBR would prevent non-redeeming shareholders from shouldering all losses in the event that a fund breaks the buck. Moreover, by clarifying that investors cannot avoid imminent losses by redeeming shares quickly in a crisis, an MBR should strengthen incentives for early market discipline for MMFs and motivate investors to identify potential problems in a fund long before any losses occur.

We begin in section I with a discussion of the structural vulnerability of MMFs to runs and recent proposals for reform. Section II outlines the MBR concept, and section III illustrates how an MBR would counter investors' incentives to redeem shares in a troubled MMF. Section IV reviews empirical evidence, including that from a novel data set from the U.S. Department of the Treasury and the U.S. Securities and Exchange Commission (SEC) on MMF losses during September and October 2008, to calibrate an MBR rule that would reduce the vulnerability of MMFs to runs. Section V discusses policy issues relevant to the introduction of an MBR. Section VI concludes the paper.

I. Background, Policy Context, and Literature

The vulnerability of MMFs to runs can, in large measure, be traced back to their stable $1 NAVs, to the characteristics of investors who are attracted to stable-value funds, and to the methods that MMFs use to maintain price stability. Like other mutual funds, MMFs provide maturity and liquidity transformation by holding a combination of highly liquid assets and less liquid, longer-dated securities while allowing shareholders to redeem shares on demand. But unlike other mutual funds, MMFs redeem shares at a price that almost never varies from $1 per share (or, in a few MMFs, from $10 per share). Indeed, the historical success of the funds in maintaining principal stability (only two money funds have "broken the buck" since 1983, when the SEC adopted rule 2a-7 to govern MMFs) has attracted a large, highly risk-averse shareholder base that includes many sophisticated institutional investors. (4) These shareholders reportedly view principal stability as the "hallmark" feature of MMFs (Investment Company Institute 2009, Stevens 2011) and hence are prone to pulling out of an MMF quickly at any sign of trouble.

Although the stable NAV is critical for many MMF investors, no capital buffer or insurance guarantees a money fund's $1 share value. Instead, MMFs have relied on a combination of strict SEC rules on portfolio composition, the ability to round their NAVs to the nearest cent (see the previous footnote), and when all else fails, financial support from their sponsors (investment management firms and their affiliates) when they have the wherewithal to provide it (Moody's Investors Service 2010, Rosengren 2012, Schapiro 2012, Brady, Anadu, and Cooper 2012, Financial Stability Oversight Council 2012b). Importantly, this support is voluntary and provided on a discretionary basis; sponsors are not required to support an ailing fund, either by regulation or by contract.

However, if shareholders begin to doubt the ability of these mechanisms to prevent losses, they have strong incentives to redeem shares before others do. Institutional shareholders appear to be particularly attuned to these incentives (President's Working Group on Financial Markets 2010, McCabe 2010, Chernenko and Sunderam 2013). The imperative to be the first to exit means that any sign of serious strains for an MMF--or concerns that other shareholders perceive such a problem--may be enough to trigger a run.

The incentive to redeem before others do arises largely because investors who redeem shares from a troubled MMF may benefit by imposing costs on other shareholders. This uneven allocation of risks and losses between redeeming and nonredeeming shareholders results from the same money fund practices and features that help maintain the stable NAV. Most importantly, because MMFs round their NAVs to the nearest cent, an investor who redeems shares from a fund that has incurred a loss of less than 0.5 percent can still obtain $1 per share. In effect, the fund transfers a redeeming shareholder's pro rata share of the loss to the fund's nonredeeming shareholders, as the loss is concentrated over a shrinking number of shares.

MMFs' liquidity management practices also contribute to principal stability while heightening the advantages for investors who redeem quickly from troubled money funds. MMFs typically meet redemptions by disposing of their more liquid assets, rather than by selling a cross section of their holdings. Redeemers who receive $1 per share thus bear none of the liquidity costs of their redemptions and leave nonredeeming investors with claims on a less liquid portfolio.

MMFs' imperative to maintain a stable NAV (and comply with rule 2a-7) also leads the funds to hold similar portfolios, a fact that contributes to contagion risk among MMFs. That is, redemptions from one MMF can hurt shareholders in other funds. Money funds generally can hold only assets with the highest short-term ratings, and given the relatively small number of private firms with such ratings, MMFs that provide funding to private firms tend to have exposures to similar sets of counterparties--mostly large financial institutions. As of September 30, 2012, for example, 50 private issuers accounted for 91 percent of the nongovernmental investments of prime MMFs, which largely invest in short-term debt instruments issued by private firms, and all but 4 of these 50 issuers were financial firms. (5) Redemptions that force one MMF to sell less liquid assets may put downward pressure on the prices of these assets (particularly given the thinness of secondary markets for many money market instruments), place other MMFs at risk of suffering losses, and prompt shareholders in those funds to redeem shares preemptively.

The severity of the run on MMFs in September 2008 and its broader consequences prompted calls for reforms to mitigate the systemic risks arising from MMFs' structural vulnerability to runs. Subsequent concerns that the funds might transmit strains from the European debt crisis to U.S. short-term funding markets provided further motivation for reform (Financial Stability Oversight Council 2011, Schapiro 2011, Rosengren 2012).

In 2010 the SEC adopted amendments to rule 2a-7 to make MMFs more resilient to market disruptions by, for example, tightening liquidity, maturity, credit quality, and disclosure requirements. Nonetheless, as these reforms were adopted, SEC Chairman Mary Schapiro recognized a need "to pursue more fundamental changes to the structure of money market funds to further protect them from the risk of runs" (Schapiro 2010). Later that year the President's Working Group on Financial Markets (PWG) "agree[d] with the SEC that more should be done to address MMFs' susceptibility to runs" and offered eight options for reducing this vulnerability (PWG 2010, p. 1). (6) In August 2012, however, Chairman Schapiro announced that the SEC would not vote to propose further MMF reforms. A month later, Treasury Secretary Timothy Geithner requested that the Financial Stability Oversight Council (FSOC) use its authority to recommend that the SEC address the systemic risks posed by MMFs. The FSOC issued proposed recommendations in November 2012, including three alternatives for MMF reform: the first would require MMFs to have a floating NAV, the second called for an MBR paired with a small capital buffer, and the third proposed a larger capital buffer, possibly combined with other measures (FSOC 2012b). In June 2013 the SEC proposed two options for MMF reform: a floating NAV for prime MMFs sold to institutional investors and a requirement that funds impose or consider fees and restrictions for redemptions if their liquid assets fall below a threshold (U.S. SEC 2013).

More generally, proposals for MMF reform have focused on three possible alternatives to the MBR for mitigating the funds' vulnerability to runs: a switch to floating NAVs, a mandate that funds maintain capital buffers, and requirements that MMFs impose fees or other (non-MBR) restrictions on redemptions. All three options have merits and drawbacks. For example, the floating NAV, which has been advocated by, among others, the Group of Thirty (2009), Jeffrey Lacker (2011), the Wall Street Journal, (7) and Paul Volcker (2011), could eliminate the destabilizing consequences of NAV rounding and remove the discontinuity in MMF pricing--and some of the resulting disruptions--when share values fall below the 99.5 cent threshold. But MMFs with floating NAVs probably would remain at some risk of runs, in part because the funds' liquidity management practices would continue to give an advantage to early redeemers from funds with limited liquid assets (see, for example, Hanson and others 2012, Gordon and Gandia 2013). A full analysis of the relative merits of other reform options is beyond the scope of this paper, but we discuss their pros and cons in some detail in section V.

II. The Minimum Balance at Risk

The minimum balance at risk that we propose in this paper is a new approach to MMF reform that would mitigate the risk of runs by limiting the ability of redeeming investors to benefit by imposing risks, costs, and losses on nonredeemers. The MBR would be a small fraction (for example, 5 percent) of some measure of each investor's recent balances. The investor could redeem her MBR only with a delay, which would need to be sufficient (we suggest 30 days) to ensure that she remains partially invested in the fund long enough to share in any imminent portfolio losses or costs of her redemptions. The MBR would have no effect on an investor's transactions in the fund as long as her remaining shares exceed the MBR.

The MBR rule that we describe would impose no losses on any investor unless that investor's MMF breaks the buck. In the rare event that an MMF does break the buck and liquidate, the MBR rule would allocate losses differently than under current rules, which divide losses solely in proportion to each investor's shares in the fund when it closes. Importantly, an MBR would make investors who otherwise would have redeemed all of their shares shoulder some of a fund's losses.

Conceptually, the MBR straddles the two main proposed approaches to shoring up MMFs' stable NAVs: capital buffers and redemption restrictions. By identifying a minimum portion of each investor's balance that would be at risk for absorbing losses--whether or not the investor redeems shares--the MBR essentially would serve a function similar to that of capital. As a form of redemption restriction, the MBR rule would always be in place, so investors could not redeem preemptively to avoid the restriction. (In contrast, redemption restrictions that are imposed only in times of stress may increase the risk of MMF runs by creating an incentive to run before the restrictions are imposed.) But even though the MBR would be in place at all times, it would have no effect on most transactions in a fund, particularly during normal times. Only when an MMF appears to be at risk of losses would the MBR materially affect investors' incentives to redeem shares.

II.A. Basic Features

The MBR would be determined individually for each investor based on his "reference amount" R, which might be his maximum balance ("high-water mark") in the fund over the previous 30 days or his 30-day-average balance, or some other function of his recent balances. Any reference amount should be calculated excluding any shares that have been held back (that is, made subject to redemption with a delay).

The MBR for investor i would be a fraction m of the investor's reference amount. That is, [MBR.sub.i] = [mR.sub.i]. For illustration, we discuss a rule with m = 5 percent; in any case, m should be large enough to give investors confidence that, in a crisis, their MMF's losses would not exceed its aggregate MBR, that is, the sum of all its shareholders' MBRs.

An investor with a total balance B would have an available balance equal to B-MBR. As long as the investor's requested redemption would not bring his balance below his MBR, his transactions would be unaffected by the MBR rule. However, when the investor places a redemption request that would bring his balance below the MBR, any shortfall relative to the MBR would be redeemed with a delay. We propose a delay of 30 days.

Both the redemption delay and the period over which R is calculated should be long enough to minimize the opportunity for preemptive runs. If the delay is short, informed shareholders may continue to see redemptions at the first sign of a problem as an effective way to obtain full payment for all their shares before any losses are realized. The MBR in that case might even be destabilizing, since such investors might redeem all shares as fast as possible in the hope that the delay period passes before any material losses are realized. The 30-day delay that we propose should be long enough to minimize the effectiveness of such a strategy. (8)

II.B. Creating a Disincentive for Shareholders to Redeem

To be effective in braking redemptions from a troubled MMF and mitigating the externalities associated with runs, an MBR rule must create a disincentive for redemptions that is strong enough to counter investors' powerful reasons to redeem when an MMF is under strain. Since the MBR works by affecting the allocation of losses only in the event that a fund breaks the buck, a sensible way to create such a disincentive is by stipulating that redeeming investors absorb losses in the fund before other investors. The MBR rule that we propose would cause some or all of a redeemer's MBR to be subordinated relative to nonredeemers' MBRs. That is, if an MMF breaks the buck, redeeming shareholders' MBRs would absorb losses before nonredeemers' MBRs do. Thus, as long as losses do not exceed a fund's aggregate MBR, redeeming shareholders would shoulder a larger share of the losses than nonredeemers.

Specifically, the fraction of a redeeming investor's MBR that is subordinated would be the fraction of his available shares that he has redeemed. That is,

subordinated balance = MBR x (cumulative net redemptions/potential redemptions).

Cumulative net redemptions are the investor's reference amount less his current balance (zero if his balance exceeds his reference amount). (9) For example, if the reference amount is defined to be the investor's 30-day high-water mark, his cumulative net redemptions would be the high-water mark less his current balance. Potential redemptions are his reference amount less his MBR. Thus,

subordinated balance = MBR x (reference amount - current balance/reference amount - MBR),

as long as the reference amount exceeds the current balance (otherwise, the subordinated balance is zero). Hence, all that is needed to compute the subordinated balance is the investor's reference amount, his MBR (which is just a fixed proportion of his reference amount), and his current balance. For example, if an investor redeems all his available shares, his current balance is his MBR, the ratio of cumulative net redemptions to potential redemptions is 1, and the investor's entire MBR would be subordinated. If his net redemptions total only half of his potential redemptions, then half of his MBR would be subordinated.

Subordination of redeemers' MBRs has several advantages. First, as mentioned above, creating a disincentive to redeem is essential to achieving substantial reduction of the risk of runs on MMFs. An MBR without subordination of redeemers' balances only partly mitigates the incentive to run, and given the intensity of incentives to redeem from a troubled MMF, investors without a disincentive to run still should be expected to run. Second, the automatic strengthening of the disincentive when an MMF encounters trouble would allow the subordination rule to be in effect at all times without imposing an undue burden on a fund and its investors. Only in the event that a fund experiences problems would the disincentive become large enough to offset incentives to redeem. Even then, the rule still would allow redemptions, but investors would face a trade-off between redeeming shares to preserve liquidity and remaining invested to safeguard principal. Notably, the strengthening of the disincentive for redemptions would not require intervention by a fund's board or by government officials. Third, a disincentive based on subordination would allocate losses among investors only when an MMF breaks the buck--that is, in circumstances in which investors would lose money under current rules, too--but losses would be allocated first to those who have recently redeemed and who thus contributed most to strains on the fund. Fourth, the MBR would strengthen incentives for early market discipline for MMFs by clarifying that investors cannot quickly redeem all shares from a fund during a crisis. Since investors would have strong incentives to identify potential problems well before any losses are realized, the market discipline encouraged by the MBR likely would be based more on investors' assessments of the riskiness of a fund's strategy or operations, rather than on headlines that trigger runs. Thus, redemptions that result from incentives created by the MBR probably would be diffuse rather than concentrated and destabilizing.

II.C. A Retail Exemption

The MBR concept is quite flexible, and a variety of adjustments could be made to accommodate normative concerns. For example, some adjustments to the MBR rule may be desirable to protect investors who make small, routine redemptions from triggering subordination of their MBRs. One adjustment would exempt the first $50,000 of an investor's redemptions from subordination. An investor who redeems less than $50,000 still would be subject to the MBR, but none of her MBR would be subordinated. (10) Specifically, the rule would be

subordinated balance = MBR x (cumulative net redemptions - $50,000/potential redemptions),

for cumulative net redemptions in excess of $50,000, and zero otherwise. (11)

Thus, any investor with a balance of under $50,000 would never have any portion of her MBR subordinated, so the adjustment would exempt many retail MMF accounts from subordination. This may be appropriate, given that retail investors are generally much less prone to run from distressed MMFs than are institutional investors. (12) Of course, a $50,000 exemption for redemptions would reduce protections for nonredeeming investors, especially in retail MMFs. However, investors would continue to be protected by the MBR itself, which would ensure that redeeming investors share proportionally in any losses of principal as long as losses do not exceed the fund's aggregate MBR.

II.D. Loss Allocation Rules with an MBR

In the event that an MMF with an MBR rule suffers losses, we assume that those losses would be allocated in the following order:

--Losses would be allocated first to a capital buffer (if any). The MBR rule that we propose would work well in tandem with a capital buffer, since a well-calibrated MBR rule would make liquidity-related losses less likely for MMFs and hence make a buffer of any size more effective. Moreover, a capital buffer would augment an MBR in mitigating investors' incentives to redeem when a fund suffers a loss. We assume that if losses exceed the capital buffer, the fund would be liquidated and remaining losses would be distributed according to the MBR rule. (13)

--Any losses in excess of the buffer would be absorbed on a pro rata basis by the subordinated portions (if any) of shareholders' MBRs.

--Any additional losses would be absorbed on a pro rata basis by the remaining portions of shareholders' MBRs.

--Any remaining losses would be divided on a pro rata basis over all other shares in the fund.

III. A Closer Look: Incentives to Redeem under Different MBR Rules

Here we use a stylized example to illustrate the MBR concept and the relative effectiveness of different MBR rules in discouraging runs on money funds. Specifically, we examine how an MBR would affect the incentives of an MMF shareholder when she learns that her MMF is in distress. We show that investors' decisions to redeem depend not only on the principal losses that the fund might suffer, but also on the liquidity costs associated with losing access to cash invested in an MMF that closes. (14) We also explore how different MBR rules affect the linkages between one investor's losses and other investors' redemptions.

We consider an MMF with an MBR of 5 percent of investors' recent high-water marks (their reference amounts), and we focus on outcomes for small investors who each own 0.1 percent of the fund's assets. To illustrate how an MBR might complement a small, NAV-stabilizing capital buffer, we assume that the fund has a 0.5 percent (50 basis point) buffer, so that in normal times it maintains a mark-to-market ("shadow") NAV of $1.005. (15) Losses are absorbed first by this buffer, and as long as losses are smaller than 0.5 percent, the fund can remain open and no losses are imposed on shareholders. Any losses exceeding the buffer force the closure of the fund, however, and are allocated first to subordinated MBRs, then to non-subordinated MBRs, and finally to other MMF shares, as described in section II.D. The assumption that the fund must close if its shadow NAV falls below $1 rules out dilutive redemptions (that is, investors receiving $1 for shares when the shadow NAV falls below $1), which are costly for nonredeeming shareholders. (16) Indeed, in part because the NAV-stabilizing buffer can eliminate the rationale for allowing dilutive redemptions, we suggest that an MBR be introduced with such a buffer and that dilutive redemptions be banned.

[FIGURE 1 OMITTED]

III.A. No MBB (Status Quo but with a 0.5 Percent Capital Buffer)

Figure 1 shows how individual shareholders' losses would vary with the losses in an MMF without an MBR, a situation equivalent to the status quo except that we assume that the fund also has a 0.5 percent capital buffer. (17) We consider three investors: one who redeems all shares just before the losses are realized (the "fully redeeming investor"), one who redeems 25 percent of her shares, and one who redeems no shares. Each investor's losses are plotted as a fraction of her preredemption assets (which are assumed to be her reference amount). (18) In showing losses for each investor, we also assume that others' redemptions from the fund are trivial.

As long as the fund's losses do not exceed 0.5 percent, the fund stays open and no investor incurs a loss. However, once losses exceed 0.5 percent, stark differences emerge. The fully redeeming investor bears no loss under any circumstance. The investor who redeems nothing suffers losses in proportion to the fund's losses in excess of 0.5 percent. For example, in the case illustrated by the dashed lines, if the fund loses 5.5 percent, the nonredeeming investor loses 5 percent. The investor who redeems 25 percent of her available balance just before the fund's loss experiences only 75 percent of the losses that the nonredeeming shareholder suffers.

Figure 1 illustrates only part of investors' strong incentive to run under current rules, even with the small, NAV-stabilizing buffer. Those who redeem immediately before a loss is reflected in a fund's share price are able to shield themselves from any loss of principal. A second reason to run is that other shareholders' redemptions at $1 per share concentrate losses on the remaining shareholders. The losses incurred by the nonredeeming investor thus depend on the behavior of other investors. Figure 2 illustrates this point by relaxing the assumption in figure 1 that redemptions by other investors are trivial. In this case the losses for a nonredeeming investor grow (the upward-sloping region of the loss function rotates to the left) as the fraction of shares that others redeem rises from zero to 25 percent and 50 percent. In contrast, the fully redeeming shareholder's losses are still unaffected by others' redemptions: that shareholder's losses are zero in all cases. This sharp disparity in outcomes for redeeming and nonredeeming investors highlights the problem that, under current rules, when a fund is perceived to be in trouble, investors have a strong incentive to rush for the exits before others do.

A third reason to redeem quickly from an MMF in trouble is that others' redemptions can force the fund to sell assets, which may be costly for nonredeeming shareholders (this point is not illustrated in our charts). For example, if a fund suffers losses because it must dispose of illiquid assets to raise cash, nonredeeming investors shoulder at least a portion of those losses and thus effectively subsidize redeeming investors. Even if a fund has enough liquid assets to raise cash without suffering losses, heavy redemptions leave remaining investors with claims on a less liquid portfolio.

[FIGURE 2 OMITTED]

A fourth reason for an MMF shareholder to redeem quickly is to preserve her own liquidity if the fund closes. We examine the incentive to preserve liquidity in section III.B.

III.B. The "Simple" MBR Rule: No Subordination

Under what we call the simple MBR rule, redeeming shareholders' MBRs are not subordinated. Instead, as figure 3 shows, losses in excess of a fund's NAV-stabilizing buffer are allocated in proportion to shareholders' total MBRs, as long as the fund's losses do not exceed its buffer plus its aggregate MBR (for a 5 percent MBR rule, that loss threshold falls at 5.5 percent of the fund's preredemption assets). The figure depicts the losses of investors who redeem none, 25 percent, and all of their available shares in an MMF with a simple MBR rule. Whereas all of an investor's shares are available for redemption in the absence of an MBR, only 95 percent of each investor's reference amount is available with a 5 percent MBR.

[FIGURE 3 OMITTED]

The simple MBR rule is a notable improvement on the "status quo" (that is, a fund with just an NAV-stabilizing buffer). Because investors' varying redemptions do not immediately reduce their MBRs, the simple MBR rule ensures that all investors' loss functions lie atop one another--at least as long as the fund's losses are less than its buffer plus its aggregate MBR. Hence, the simple rule ameliorates the allocation of losses among investors and reduces incentives to redeem.

However, the simple rule has a very serious limitation: it only reduces the incentive to redeem shares in a troubled MMF; it does not eliminate that incentive. Even with the simple MBR rule, investors in a troubled MMF would have several good reasons to redeem. First, the fully redeeming investor limits his losses to his MBR; any losses in excess of a fund's buffer plus its aggregate MBR are shouldered exclusively by nonredeeming shareholders. In the stylized example in figure 3, once the fund's losses exceed 5.5 percent, nonredeeming shareholders' losses exceed those of redeeming shareholders. Hence, redeeming all MMF shares is beneficial if losses are large, and it does not make the investor worse off if losses are small. Second, redeeming investors still may be able to shift liquidity-related costs arising from their own redemptions to other investors in the fund. These costs may be especially important during episodes of financial strain, when liquidity is likely to be at a premium.

Third, even with the MBR, shareholders have an incentive to redeem to safeguard their own liquidity, because those who do not redeem from an MMF that closes may lose access to their cash during a prolonged liquidation phase. Figure 4 illustrates this point under the simple assumption that shareholders would be indifferent between having their investments locked up in a closed MMF during a liquidation phase of indefinite duration and having to incur a 0.5 percent fee to obtain all of their assets immediately. (19) The figure shows losses for investors both excluding (reproduced from figure 3) and including this opportunity cost of lost liquidity. The magnitude of this cost depends on whether the fund must close and the value of the investor's remaining shares in the fund if it does close. (If losses erode the value of an investor's shares in a closed fund, the value of his lost liquidity declines proportionally.)

The nonredeeming investor's loss function inclusive of liquidity cost jumps up at the point when losses exceed 0.5 percent, since at that point the fund must close and the investor's entire balance is locked up for the duration of the fund's liquidation process. Importantly, with a simple MBR rule, when liquidity costs are taken into account, any loss large enough to cause fund closure causes greater losses for nonredeeming than for redeeming investors. (20) In this example the fully redeeming investor's liquidity loss is at most 2.5 basis points (0.5 percent of 5 percent of assets), and it declines as the investor's MBR is absorbed. Hence, for this investor the line representing losses inclusive of liquidity cost lies just a bit above the line for losses exclusive of that cost when the MMF itself experiences losses of less than 5.5 percent.

[FIGURE 4 OMITTED]

This example illustrates some of the pros and cons of the simple MBR rule. This rule would reduce redeemers' first-mover advantage and allocate any losses that a fund might suffer more equally. Hence, it would have some benefit, for example, in protecting retail investors, who have proved to be much less likely to run than institutional investors. However, the simple rule leaves investors with strong incentives to run from distressed MMFs, since a redeeming investor incurs no additional cost but limits her losses to the size of her MBR, preserves her liquidity, and possibly imposes some liquidity costs on others. Nonredeeming shareholders would still be at greatest risk of suffering losses. Hence, the simple MBR rule does not eliminate the destabilizing allocation of risks and costs between redeeming and nonredeeming shareholders. Under such a rule, MMFs--particularly those with large institutional investors that may be highly motivated to preserve liquidity--still would be vulnerable to runs and constitute a source of systemic risk.

III.C The "Strong" MBR Rule: Adding Subordination

The strong MBR rule addresses the shortcomings of the simple rule by subordinating the MBRs of redeeming investors to create a disincentive to redeem. As outlined in section II.B, the fraction of an investor's MBR that is subordinated would be equal to the fraction of her available balance that she has redeemed. Figure 5 shows the allocation of losses in a fund under the strong rule, again with the assumptions that all other redemptions from the fund are trivial and that liquidity costs due to losing access to shares in a closed fund are zero. (For simplicity, we do not incorporate the retail exemption discussed in section II.C.)

Under the strong rule, if the fund's losses are smaller than its buffer plus its aggregate MBR, the fully redeeming investor fares worse than the investor who redeems less or not at all. In this stark example, the fully redeeming investor loses her entire MBR for MMF losses just slightly larger than 0.5 percent, because other investors do not redeem anything. By redeeming her entire available balance, she has put her full MBR in a subordinated position, but without other redemptions, her MBR is the fund's total subordinated MBR. (21) Importantly, however, the fully redeeming investor still is better off than others if the fund's losses exceed its buffer plus its aggregate MBR.

Figure 6 adds consideration of the potential liquidity costs for investors who may have their money locked up in a closed MMF through a prolonged liquidation process. The figure shows that even when these costs are taken into account, the strong rule can create a disincentive to redeem.

[FIGURE 5 OMITTED]

Figure 7 shows how a fully redeeming investor's losses depend on the behavior of other investors under a strong MBR rule (for simplicity, we again exclude the costs of lost liquidity). The fully redeeming investor's losses are less extreme for small losses to the fund when other investors redeem as well. That is, when other investors redeem shares, they also contribute to the fund's subordinated MBR, so each investor owns a smaller share of the aggregate subordinated MBR. (22) At the same time, a nonredeeming investor's losses are also smaller when other investors redeem, because the subordinated MBRs of redeeming investors provide those who do not redeem some protection from losses. (23)

[FIGURE 6 OMITTED]

[FIGURE 7 OMITTED]

By making redemptions from a distressed MMF potentially costly, the strong MBR rule can help offset the incentives to redeem that arise from investors' preference for liquidity and the concern that losses might exceed a fund's capital buffer plus its aggregate MBR. The cost of redemptions also may offset the liquidation costs that redeemers impose on other investors. In addition, under this rule, redemptions protect nonredeeming investors by providing a subordinated buffer to absorb losses. In sum, the strong rule may offset and even reverse the incentive to redeem from a troubled fund, and thus may help stabilize a distressed MMF, allow it to weather a difficult period, and mitigate the externalities associated with a run.

The benefits of the strong rule are evident in a comparison of figures 2 and 7. Figure 2 shows that under current rules, an investor's losses depend on other investors' behavior in a destabilizing way. The more other investors are expected to redeem, the greater the expected losses for non-redeeming investors. This dynamic provides a strong incentive to redeem at the first sign of trouble, before others do. In contrast, figure 7 shows that with a strong MBR rule, an investor's losses depend on other investors' behavior in a stabilizing way. Redeeming investors contribute to the fund's subordinated MBR and provide more protection for nonredeeming investors. This dynamic stabilizes the fund and benefits retail investors and others who may not redeem quickly when MMFs encounter problems. Even so, it is important to recognize that if expected losses are large enough (or if the MBR is too small), investors may be better off if they redeem shares.

IV. Calibrating an MBR

In this section we use empirical evidence to calibrate how large an MBR would need to be to reduce the vulnerability of MMFs to runs, and we provide some guidance on the appropriate delay for redeeming shareholders' MBRs. When an MMF encounters strains and the likelihood of losses increases, investors' decisions to redeem or not depend on how the choice affects their expected losses (including the value of lost liquidity) if the fund breaks the buck and closes. That is, for a fund in distress, conditional expectations are critical, and investors will redeem when doing so reduces their expected principal and liquidity losses in a break-the-buck scenario. (24) Thus, as input for our calibration, we examine historical evidence on the size of MMF losses when they have occurred, including evidence from a novel and important data set on MMF losses in 2008. We also study the value of preserving liquidity in a crisis, which strengthens investors' incentive to redeem from a fund in distress. Finally, we examine evidence that can be helpful in setting the delay period for redemptions of investors' MBRs.

IV.A. Previously Available Data on MMF Losses

Although hundreds of MMFs have suffered material losses at various times over the last 30 years, information on the magnitude of those losses is not readily available. The historical record of losses in MMF portfolios has been obscured by fund sponsors' long-standing practice of providing discretionary financial support to MMFs that were in danger of breaking the buck, even though such support is not required by either regulation or contract. SEC Chairman Schapiro reported in 2012 that sponsors had intervened more than 300 times to support MMFs since they were introduced in the 1970s (Schapiro 2012). Moody's found 144 cases from 1989 to 2003 in which U.S. MMFs received such support (Moody's Investors Service 2010). Steffanie Brady and others (2012) report 123 instances of support for 78 different money funds between 2007 and 2011. MMF sponsors reportedly intervened as recently as November 2011 to support their MMFs. (25)

We have no comprehensive data on the MMF losses that would have occurred in the absence of sponsor support. Sponsors are not required to disclose the value of their support for a distressed MMF at the time of an intervention, and available data generally do not provide enough information to estimate the magnitude of NAV declines that would have occurred without interventions. (26)

Still, the historical record and recent research provide some useful evidence. The two U.S. MMFs that have broken the buck since the adoption of rule 2a-7 in 1983 provide a couple of observations: the Community Bankers U.S. Government Money Market Fund lost 3.9 percent of its value in 1994 (U.S. SEC 1999), and the Reserve Primary Fund lost approximately 1.6 percent in 2008. (27) Brady and others (2012) document 31 MMFs that between 2007 and 2011 received sponsor support that exceeded 0.5 percent of the fund's assets--enough, that is, that the support was probably required to prevent those funds from breaking the buck. Support was more than 2 percent of assets for 10 of these funds and exceeded 3 percent for 4 funds. (28) Analysis by Moody's showed that at the time of Lehman's bankruptcy, 15 MMFs held Lehman obligations that ranged from 0.25 percent to 5.6 percent of fund assets and averaged 1.9 percent--hence, the Reserve Primary Fund's position (1.3 percent) was less than the average among funds that held Lehman obligations at that time (Moody's Investors Service 2012c).

IV.B. New Data on MMF Losses

We analyze here a new data set that provides additional evidence on the scale of losses that have occurred in MMFs. The data come from the U.S. Department of the Treasury and the SEC, which collected information about certain MMFs that participated in the Treasury's Temporary Guarantee Program for Money Market Funds in 2008. MMFs with shadow NAVs below $0.9975 were required to report information about their portfolios, including what their NAVs would have been without sponsor-provided guarantees, such as capital support agreements (U.S. Department of the Treasury 2008). Even so, the NAV data do not reflect the full extent of losses that might have occurred without sponsor interventions, since the effects of some types of sponsor support, such as direct cash infusions to a fund and outright purchases of securities from a fund at above-market prices, are not excluded from reported shadow NAVs. Of course, the data also do not reflect portfolio losses that might have occurred in the absence of extensive government support in 2008 for MMFs, short-term funding markets, and some financial institutions.

The Treasury-SEC data include the shadow NAVs of reporting MMFs (but not the funds' identities) from September 5 to October 17, 2008. (29) As the first row of table 1 shows, 72 funds reported their shadow NAVs at some point over this period. The average of the minimum shadow NAVs reported by each of these funds during this period was $0.989; that is, the average loss was $0.011 per share (1.1 percent). The second row indicates that 29 funds reported shadow NAVs that would have fallen below $0.995--enough to break the buck--at some point during this episode. (30) On average, these funds' shadow NAVs would have dropped to $0.978 (a 2.2 percent loss) without sponsor support. Even when three outliers (funds that reported minimum shadow NAVs of $0.935, $0.929, and $0.903) are excluded, the average shadow NAV of funds that would have broken the buck was $0.985 (third row of table 1).

IV.C. Liquidity Costs of Losing Access to Shares in a Closed MMF

As discussed above, when an MMF breaks the buck and closes, its shareholders suddenly lose access to their cash for what may be a prolonged liquidation process. Indeed, the Reserve Primary Fund, which broke the buck on September 16, 2008, still had not completed the distribution of all assets to shareholders more than 3 years later (Schapiro 2012). Clearly, a fund's closure imposes a liquidity cost on its shareholders.

There are many challenges in estimating this cost and thus in assessing the strength of shareholders' incentives to redeem shares in a distressed fund to avoid lost liquidity. The opportunity cost to shareholders of an unplanned loss of access to cash could depend on a number of factors, including shareholders' individual circumstances, broader financial conditions, and the length of time over which access is lost.

We use several approaches to estimate the value of shareholders' lost liquidity (these are discussed in more detail in our working paper). One is to examine the average net cost to an investor who must take out a business loan from a bank to replace cash in a closed MMF. We show that this net cost (the rate on the bank loan less the yield on the investor's MMF shares until they are liquidated) averaged about 2 percent at an annual rate during the financial crisis. (31) Liquidity premiums for financial instruments, particularly during periods of financial strain when investors most likely would be motivated to run from MMFs, provide another perspective on the costs of lost liquidity. Although these premiums are difficult to estimate, various methods discussed in our working paper suggest that premiums might range from 1 to 2 percent at an annual rate.

The cost of lost liquidity in a closed MMF also depends on the length of time over which investors are likely to lose access to their cash. Ideally, shareholders might expect to receive cash as the securities held by their fund mature, although payments of subordinated claims would have to wait until other claims have been paid. Among prime MMFs, which largely invest in the debt securities of private firms, the weighted-average life of portfolio holdings was 71 days in 2012, and the longest-dated nongovernment security in each fund's portfolio matured, on average, in 314 days. However, in practice, investors' wait might extend for years, as the Reserve Primary Fund example shows.

The evidence we have reviewed indicates that the opportunity costs of lost liquidity to investors in a closed MMF are probably significant. Assuming that investors expect to lose access to shares in a closed MMF for 6 months, we believe that an MBR should be designed to withstand redemption pressures that incorporate investors' incentives to avoid lost-liquidity costs of at least 0.5 percent of the value of their MMF shares.

IV.D. How Large Should an MBR Be?

To calibrate the size of an MBR sufficient to create a disincentive for investors to run from a distressed MMF, we draw on the empirical evidence on historical MMF losses described in sections IV.A and IV.B and assume that MMF losses are distributed exponentially, with an unconditional mean of 1.5 percent, so that losses conditional on a fund breaking the buck (that is, when losses exceed 0.5 percent of assets) average 2 percent. (32) On the basis of the evidence discussed in section IV.C, we also assume that the liquidity costs to an investor in an MMF that closes are 0.5 percent of the value of the shares that the investor cannot access. Finally, we assume that the fund has a small (0.5 percent) capital buffer and that when it encounters distress, other investors do not redeem. (33)

Figure 8 illustrates the calibration under these assumptions. Each line in the chart shows how an investor's expected losses vary with her own redemptions for an MBR of a given size. When the MBR is small--1 or 2 percent under our assumptions--these expected-loss curves slope downward, so investors reduce their expected losses by redeeming more shares. That is, if the MBR is too small, investors still have an incentive to run from a fund in distress. With an MBR of 3 percent, the expected-loss function slopes slightly upward, indicating that the investor would be better off not redeeming. MBRs of 4 or 5 percent provide more convincing disincentives to redeem. This exercise suggests that an MBR of 3 percent, combined with a small, NAV-stabilizing buffer, might be enough to protect MMFs from runs during a crisis. (In our working paper, we examine the calibration issue in more detail and conclude that an MBR of 3 to 4 percent might be needed.)

This exercise estimates only a minimum effective size for an MBR; the appropriate size in practice would depend on broader criteria. Most notably, the goal of internalizing the significant social costs of redemptions in a crisis suggests that the optimal slope for the investor's expected-loss function should be positive--that is, the investor should face expected redemption costs that incorporate the large externalities associated with runs from MMFs. (34)

[FIGURE 8 OMITTED]

IV.E. How Long a Delay for MBR Redemptions?

To be effective in mitigating MMFs' vulnerability to runs, the delay period for redemptions of investors' MBRs must be long enough to inhibit "preemptive" runs. With very short delays, investors may expect that redeeming at the first sign of a problem will allow them to recover their MBRs before any losses are realized. Because this strategy would require an investor to exit a distressed fund earlier than he would under current rules, a short delay might make MMFs even more vulnerable to preemptive runs.

At the same time, delays should not be so long as to unnecessarily impose liquidity costs on shareholders or impede market discipline for MMFs. In particular, very long delays might inhibit investors from pulling away from a fund with poor management, even long before the emergence of any specific strains in its portfolio.

The events of September 2008 suggest that a very short delay, such as 1 week, would not be enough to prevent preemptive runs on MMFs. Even amid the run triggered by Lehman's bankruptcy early on Monday, September 15, every MMF except the Reserve Primary Fund managed to survive without breaking the buck until the end of the week, when the Treasury introduced its temporary guarantee program and removed concerns about losses. (As noted above, many funds avoided breaking the buck only because they received considerable discretionary sponsor support.) With an MBR delay of just a few days, an MMF shareholder who redeemed from a fund even after Lehman failed might have avoided any loss and shifted potential risks and losses to nonredeemers.

The maturities of MMF assets suggest an upper bound for the delay period. Because any problems in a fund's portfolio at the time of an investor's redemption should be resolved by the time the fund's longest-lived asset matures, the delay period might be as long as 397 days, the maturity limit for privately issued assets that MMFs can hold. On average, as noted above, a prime MMF's longest-dated security held in 2012 (excluding government securities) matured in 314 days. Although the 30-day delay that we suggest is relatively short in comparison, about half of prime MMFs' assets (on average) mature within 30 days, so a delay of that length should provide a reasonable amount of time to ensure that any problems with assets at the time an investor redeems are resolved before disbursement of her MBR.

Since MMFs must report on their websites detailed portfolio holdings data on a monthly basis, a 30-day delay means that an MMF usually would publish updated data at some point between an investor's request for redemption of her MBR and her receipt of payment. Hence, a 30-day delay would help ensure that investors who request redemptions based on information about a problem that has already occurred in an MMF's portfolio would not be paid before updated information about the portfolio is publicly released.

V. Policy Considerations

This section first explores how the MBR rule would compare with other proposals for MMF reform and then examines a number of practical issues that might affect the implementation, operation, and effectiveness of an MBR rule.

V.A. How Would an MBR Rule Compare with Other Options for MMF Reform?

As described in section I, advocates of MMF reform have focused on four basic approaches to reduce the funds' vulnerability to runs. Besides the MBR, the alternatives include requiring the funds to have floating NAVs, requiring them to have a capital buffer, and imposing fees or (non-MBR) restrictions on redemptions.

The risks that MMFs pose to the financial system are externalities: substantial social costs that redeeming MMF investors and other market participants currently do not bear. Hence, any reform should aim at internalizing these costs. Doing so, however, likely would cause a reduction in the equilibrium size of the industry, might drive up required returns for MMF investments, and could increase borrowing costs in short-term funding markets. (35) For these reasons, all serious MMF reform options are unpopular with MMF shareholders, with the firms that offer MMFs, and with those that borrow from MMFs. Comments from market participants on the FSOC's proposed recommendations on MMF reform have overwhelmingly opposed all three of the alternatives put forth by the council. (36)

MMFs offer market-based yields that reflect the risk of the securities they hold, while at the same time offering principal stability and redemption on demand. This combination is inherently unstable, and all effective reform alternatives would diminish the attractiveness of MMFs by taking something off the table: capital buffers would diminish yields, a floating NAV would reduce principal stability, and any redemption restriction would limit liquidity (FSOC 2012b). The net benefits of reform will depend not only on the extent to which it succeeds in stabilizing MMFs, but also on how market participants respond. Reforms that drive institutional investors into lightly regulated or unregulated MMF substitutes may be less effective at improving systemic stability than alternatives that preserve some of the attractive features of today's MMFs. (37) Although the focus in this paper is on the potential benefits of an MBR for money funds, the most effective reform strategy might be one that creates different types of MMFs with different risk mitigation features, so investors can choose the MMFs that best suit their preferences rather than move money to unregulated vehicles. For example, reforms might allow for funds with MBRs as well as funds that are exempt from MBR rules because they have substantial capital buffers or hold only Treasury securities.

A FLOATING-NAV REQUIREMENT A requirement that all MMFs have floating NAVs has considerable appeal among policymakers, academics, and others for its simplicity, elimination of NAV rounding, and potential to improve investors' understanding of MMF risks. A floating NAV would partially internalize the cost of redemptions by ensuring that investors who exit a fund that has incurred a loss do not receive $1 per share. It would also make fluctuations in the value of an MMF portfolio more transparent in share prices, could help investors become more cognizant of MMF risks, and might reduce the ex ante incentives of portfolio managers to take risks, since those risks would result in greater (and more observable) share price volatility. (38)

However, unlike investors in MMFs with an MBR, investors in floating-NAV money funds would still have a strong incentive to redeem quickly during crises (FSOC 2012b, Hanson and others 2012, Gordon and Gandia 2013). As mentioned above, MMFs typically meet net redemptions by disposing of their more liquid assets, rather than by selling a cross section of all of their holdings. Such liquidity management practices help funds avoid realizing losses from sales of less liquid investments. But these practices essentially subsidize redeeming investors by leaving the remaining investors with claims on a less liquid portfolio. The subsidy may be immaterial in normal times, but during periods of market strain, when liquidity is at a premium, the incentive to redeem grows. Indeed, some empirical research suggests that floating-NAV money funds in Europe were as vulnerable to runs as stable-NAV funds during the financial crisis (Gordon and Gandia 2013), although other work finds that stable-NAV funds were indeed more vulnerable (Witmer 2012).

An additional concern is that implementing the floating-NAV option would be tantamount to "eviscerating" MMFs (Investment Company Institute 2009). Opponents of a floating NAV have cited several concerns about its potential impacts on MMF investors, including tax and accounting complications that might substantially diminish the funds' appeal (see, for example, Investment Company Institute 2013). Because lightly regulated substitutes for MMFs would continue to offer stable NAVs, any reform that forced all MMFs to have floating NAVs might lead to a large migration of assets to these less regulated vehicles.

The floating-NAV option would offer additional flexibility to MMFs in an environment of very low interest rates. Some commentators have suggested that the easing of monetary policy could be constrained by concerns that very low interest rates might put MMFs out of business (see, for example, Sellon 2003, Keister 2011, Coeure 2012). (39) A money fund with a floating NAV might prove less vulnerable to low rates because it could charge expenses that exceed its earnings on portfolio securities without opening a gap between the share price and the underlying value of its shares. (40)

A CAPITAL BUFFER REQUIREMENT Like the MBR rule, capital creates a buffer to absorb potential MMF losses. Each approach has advantages and disadvantages.

A capital buffer would give MMFs some capacity to absorb losses without breaking the buck. This could allow the funds to operate without disruption even after suffering material losses and could eliminate the possibility of dilutive redemptions. An MBR, in contrast, would not prevent funds that suffer losses from breaking the buck.

A substantial capital buffer--one comparable in magnitude to the MBR that we recommend--could have two additional advantages over an MBR rule. First, capital might allow a more efficient allocation of risk among investors. Presumably the owners of the capital would have greater tolerance for risk than investors in a stable-NAV money fund (Hanson and others 2012). An MBR primarily would clarify the allocation of risks among MMF investors, although it is possible that MMF investors would contract with more risk-tolerant third parties to bear the risk of an MBR. Second, a substantial capital buffer could help curtail the MMF industry's historical reliance on voluntary sponsor support, which poses considerable systemic risk (McCabe 2010, PWG 2010, FSOC 2012b). However, a capital buffer that is too small to absorb reasonably foreseeable losses would not offer these advantages.

Some disadvantages of a capital buffer arise from the challenges of raising the large sums that would be needed. As of the end of 2012, each percentage point of capital as a share of industry-wide MMF assets would have required more than $25 billion in funding, depending on the scope of the capital requirement (for example, whether it applied to MMFs investing only in Treasury securities). At least three possible funding sources have been suggested. First, MMFs might retain income that normally would be distributed to shareholders (see, for example, Goebel and others 2011). However, this process would require many years to build a substantial buffer unless the net yields paid to investors are reduced very sharply, and the incentives to protect the buffer would be unclear. Second, a buffer might be raised from third-party investors in capital markets (see, for example, BlackRock 2011, Hanson and others 2012). This would facilitate an efficient allocation of MMF risks by shifting them to investors who are more willing and able to take risks. However, it would require the creation of a market for a new, untested type of security and might entail significant underwriting and other costs. Third, a buffer could be financed by the MMF sponsors themselves (see, for example, BlackRock 2010). Although this approach would be simpler than obtaining funding in the capital markets, it could lead to further consolidation of the MMF industry among the affiliates of large, systemically important financial institutions. Hence, sponsor-funded capital could further concentrate systemic risks.

One clear advantage of a strong MBR over a capital buffer is that subordination creates a disincentive to run. Indeed, with such an MBR, investors can be better off not redeeming, even if they recognize that a fund's losses might exceed the size of its aggregate MBR. In contrast, capital only reduces the incentive to run from a fund at risk: investors still would have good reason to exit if the danger exists that losses might exceed the capital buffer. Thus, an MBR might do more to mitigate the vulnerability of MMFs to runs than a capital buffer of the same size.

A capital buffer and an MBR rule also would have different implications for investor incentives. By shielding MMF investors from some losses, a meaningful capital buffer would shift the incentives for ex ante risk management from the MMF's shareholders to the owners of the capital buffer, who might be well suited to this task (Hanson and others 2012). (41) But an MBR should improve ex ante risk management as well, because an MBR would be, in effect, a buffer provided by the MMF shareholders themselves. By removing investors' option to redeem shares quickly when losses appear imminent, an MBR rule would strengthen shareholders' incentives to monitor the funds' risk taking well before problems materialize.

Both the MBR and a capital buffer might reduce shareholder demand for MMFs, but for different reasons. The costs of the capital buffer presumably would be funded in the same way as other MMF expenses and hence would reduce the net yields that MMFs pay to investors. The MBR, in contrast, would reduce the liquidity of MMF shares.

Because the strengths of a capital buffer and an MBR rule complement one another, we suggest that an MBR would be particularly effective if paired with a capital buffer. The MBR could bolster a capital buffer by creating a disincentive to redeem from an MMF in distress; this would lower the likelihood that capital might be eroded by redemption-driven fire sales of assets. A more substantial capital buffer could help reduce the importance of voluntary sponsor support for MMFs and make funds more resilient to losses. But even a small buffer could provide a means to eliminate dilutive redemptions. Notably, Alternative Two of the FSOC's proposed recommendations on MMF reform includes an MBR in tandem with a small capital buffer (FSOC 2012b).

REDEMPTION FEES AND NON-MBR RESTRICTIONS An MBR, if implemented, would be in place at all times. Proposals for other types of restrictions or fees for redemptions that are always in place appear less promising. For example, although a delay in disbursement of all redemptions would provide some extra time for an MMF to raise cash to meet heavy redemptions in a crisis, it would not reduce investors' incentive to redeem, nor would it halt the dynamics of a run. Moreover, imposition of a significant delay on all redemptions would eliminate much of the utility of MMFs for cash management. A fee on all redemptions would be an even more substantial departure from principal stability than a floating NAV, since investors would lose money on every redemption.

Conditional fees or restrictions for redemptions--that is, fees or restrictions that are not always in place--have garnered significant support from the MMF industry as an alternative to other options (see, for example, HSBC Global Asset Management 2011, BlackRock 2012, Investment Company Institute 2013). Such "standby" arrangements would allow MMFs to function much as they currently do in normal times but would halt or penalize redemptions when MMFs are under strain. For example, recent proposals would "gate" (that is, halt) redemptions if a fund's liquid assets fell below a threshold and then would charge a fee for any redemptions until the fund's liquidity was restored.

Conditional fees and restrictions might actually increase the risk of preemptive runs on MMFs in distress and speed up the spread of runs to other funds. The possibility that fees or gates might be imposed would heighten investors' incentives to redeem quickly from troubled MMFs, especially if they suspect that other investors will do so (FSOC 2012b, Hanson and others 2012). In contrast to running from a distressed MMF with a strong MBR rule, exiting an MMF with conditional fees or restrictions immediately before they are imposed would be essentially costless to the redeeming investor. In addition, given the similarity of MMF portfolios, news that one fund has halted redemptions could spark runs on other funds (Rosengren 2013).