Cian O’Neill and Nicholas Vause.
Certain policy actions require a high level of precision to be successful. In a recent paper, we find that using margins on derivative trades as a macroprudential tool would require such precision. Such a policy could force derivative users to hold more liquid assets. This would help them to meet larger margin calls and avoid fire-selling their derivatives, which could affect other market participants by moving prices. We find that perfect calibration of such a policy would completely eliminate this fire-sale externality and achieve the best possible outcome, while simple rules are almost as effective. However, calibration errors in any rule could amplify fire-sales and leave the financial system worse off than if there had been no policy at all.
Margin requirements for derivative positions have greatly reduced risk in the financial system. They force these positions to be backed by collateral, which curtails risk by preventing losses at one investor spreading to others. However, a cost of this additional protection arises because investors can expect to face regular margin calls. These could be calls for ‘variation margin’ (VM), which cover current changes in the value of a derivative, or additional ‘initial margin’ (IM), which protects against potential future changes in value and thus depends on the volatility of the derivative. Either way, since margin calls can only be met with liquid assets, any shortfall in liquid asset holdings could force investors to liquidate their derivative positions. To reduce this risk, investors may scale back their initial derivative investments or increase their precautionary holdings of liquid assets at the expense of higher yielding alternatives.
Regulators can primarily influence IM requirements and the level at which they are set is really important. Setting them too high would reduce expected returns by overly constraining investors’ derivative positions and boosting their holdings of low-yielding liquid assets. Setting them too low would allow margins to increase sharply in times of stress, potentially leading to what Brunnermeier and Pedersen (2009) call a ‘liquidity spiral’. In such spirals, margin calls following an initial change in derivative prices force investors to liquidate positions, but these liquidations exacerbate the price moves, which leads to further margin calls. Moreover, this generates an externality, since the price volatility affects all market participants and not just those who are forced to liquidate positions. In addition, low margins in the upswing of the financial cycle can allow investors to excessively increase their leverage by helping them to take large derivative positions using only a relatively small amount of collateral. This increases their sensitivity to future price changes, which can help bring about the next downswing. Geanakoplos (2010) describes this ‘leverage cycle’ in more detail.
The tendency for margins to amplify changes in volatility over the financial cycle has led some, including European Systemic Risk Board (2017), to call for the introduction of macroprudential margins. Similar to the countercyclical capital buffer for banks, the idea is that an IM buffer could be increased during the upswing of the cycle when initial margins were low and leverage was high. It could then be released when the cycle turned, freeing up liquidity for investors and preventing the liquidation of derivative positions that could amplify initial price changes. Thus, such a policy could reach beyond regulated institutions to mitigate this kind of amplification.
In our paper, we test whether such a macroprudential buffer could be beneficial by creating a model of a derivative market. Figure 1 summarises this model. In our derivative market, optimistic and pessimistic investors establish long and short positions in a particular contract, hoping to benefit from future changes in the value of an asset to which the contract is linked (called its ‘underlying’). This requires them to immediately post some collateral as IM. In addition, during the life of the contract, they must meet any calls for VM or additional IM. If they have insufficient liquid assets to meet these calls, they must liquidate some of their derivative positions. This has a fire-sale impact on the price of the derivative. We show that this affects the returns of liquidating investors and all others in the market. Thus, we capture in our model the externality mentioned above.
Figure 1: Summary of derivative market model
We quantify the externality by solving the model for the derivative positions and liquid assets that investors hold. We compare solutions in which investors are left to their own devices (the ‘private optimum’) and where an omniscient social planner chooses on behalf of all of them (the ‘social optimum’). Investors hold larger derivative positions relative to liquid assets in the private optimum because they do not care about the potential costs that liquidations would impose on others. This creates a wedge between investment returns in the private and social optima, which measures the externality.
It is possible to introduce a macroprudential IM buffer that eliminates this wedge and achieves the best possible outcome at all points of the financial cycle. The buffer forces investors to hold more liquid assets against their derivative positions, and resizing it according to market volatility at each point in the financial cycle always replicates the social optimum. However, such discretionary resizing of the buffer would require, amongst other information, regular and comprehensive details about the potential future price movements of the underlying. Unfortunately, this forward-looking measure of price volatility is not observable, so it would have to be estimated, and any errors in these estimates would lead to errors in sizing the optimal buffer.
To be effective, such a macroprudential buffer would not only need to be set at just the right level, but also released under any type of liquidity stress, whether that reflected IM or VM calls. In short, buffers should be there to be used. This is a novel result as policy tools proposed in the literature have focussed on easing liquidity strains solely from IM calls, and not VM calls. We find that releasing the buffer only for IM calls would result in fire-sales that could have been avoided by freeing up liquid assets to help meet large VM calls. This would also undermine returns on a day-to-day basis, as investors would hold additional low-yielding liquid assets as a precaution to help meet VM calls, since they would know the buffer would not be released for these.
Since setting the optimal buffer could require an unrealistic level of information, we think it is more instructive to investigate the performance of buffers based on simpler rules. Firstly, we look at three tools inspired by those in European Market Infrastructure Regulation (EMIR). These are a margin floor, which puts a lower limit on IM; a stress-weighting mechanism, which always takes stressed periods into account in IM calculations, and a proportional margin buffer, which increases IM by a certain percentage but releases this amount when overall IM requirements would otherwise be increasing rapidly. While under EMIR, margin buffers can only be released in response to rising IM requirements, we release our versions of these buffers with VM or IM calls to facilitate comparison with other approaches to setting buffers.
Figure 2: Effectiveness of alternative macroprudential buffers
The blue bars in Figure 2 show the size of the externality (y-axis) at different points in the financial cycle (x-axis) under the alternative EMIR-based tools. These should be compared with the green diamonds, which show the size of the externality with no policy in place. A bar that is smaller than the level of the green diamond shows that a buffer is effective at reducing the externality. This reveals the EMIR-style tools are beneficial at many points of the cycle. However, in states of low volatility the floor and stress-weight tools demand much additional margin, pulling investment returns below those of the private optimum (since the two blue bars on the left of the chart are larger than the level of the green diamonds). So, by getting the calibration wrong and demanding too large a buffer, a macroprudential margins policy could end up making things worse.
A rule that varies the size of the buffer with the inverse of volatility does much better (orange bars). A constant buffer also performs very well (purple bars); and this might be preferable from an implementation perspective. However, even though setting an ideal constant buffer would not require updates through the cycle to potential future price movements of the underlying, it would still need comprehensive information about the structure of the market, including on the balance of long and short investors and their risk appetites.
In the absence of such information, calibration errors could easily offset the benefits of macroprudential margin buffers. This reflects our finding of only a narrow range of buffers that enhance welfare. Below this range, buffers do not bind, while above it they force investors to hold more low-yielding liquid assets than necessary to meet most potential margin calls, diverting resources from more-productive investments. There is a delicate balancing act at play, and lots more study might be needed before it is attempted given that implementation errors could be costly.
Cian O’Neill works in the Bank’s Macroprudential Strategy & Support Division and Nicholas Vause works in the Bank’s Capital Markets Division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.