Better anti-procyclicality? From a critical assessment of anti-procyclicality tools to regulatory recommendations

Thomas Siegl and Daniel Steinberg

Abstract

In the aftermath of the 2007–9 global financial crisis, the regulation of the financial industry was the focus of the regulators in many countries. Alongside the clearing mandates, the European Markets Infrastructure Regulation (EMIR) seeks to mitigate procyclical behavior of the central counterparty margins in times of crisis or stress. The regulations are known as anti-procyclicality (APC) tools. Such tools are used by most central counterparties worldwide. We conduct a critical assessment of the existing APC tools from both a qualitative perspective and a quantitative perspective. We calibrate and compare initial margin models and derive APC results for a representative Deutscher Aktienindex (DAX) portfolio in Germany. This quantitative assessment is complemented by qualitative policy recommendations toward more effective APC tools that extend beyond initial-margin-based approaches (such as improved transparency, collateral eligibility, clear definitions of procyclicality) and the reconsideration of the loss distribution within the clearing system.

Download PDF

Introduction

1 Introduction

Since the 2007–9 global financial crisis several efforts (including clearing mandates) have been made to regulate the financial industry more effectively in order to strengthen financial stability, with the intention of mitigating or even eliminating the potential procyclical effects of clearing. However, the actual implementation of anti-procyclicality (APC) regulation generally suffers from a very real and far-reaching lack of precision in the definition of APC; that is, there is no agreement in the literature or in similar central counterparty (CCP) regulations on a detailed definition of APC for CCP clearing. This leads to ambiguity with respect to the implementation of practical tools fostering APC ex ante and the measurement of the effectiveness of these tools ex post.

The rest of the paper is organized as follows. Section 2 gives an overview of APC approaches and their general limitations. Section 3 discusses the APC tools introduced under the European Markets Infrastructure Regulation (EMIR) (European Union 2012), which are now used by most CCPs globally (see International Swaps and Derivatives Association 2021). Section 4 gives the results of our empirical analysis of different initial margin (IM) models together with APC tools and their measurements in the context of a hypothetical Deutscher Aktienindex (DAX) position. Finally, based on our holistic understanding of APC gained from this analysis, Section 5 makes suggestions and derives indicative effectiveness assessments for alternative APC tools that extend beyond the narrow focus of the existing quantitative IM regulation. Section 6 states our conclusions.

2 The general interpretation of anti-procyclicality

2.1 The definition of APC

Regulatory sources do not provide a consistent and holistic definition of APC, and the following list details various interpretations of the purpose of the APC regulation.

1. (1)

   In their Principles for Financial Markets Infrastructures (PFMI), the CPSS and IOSCO define procyclicality as “changes in risk-management practices that are positively correlated with market, business, or credit cycle fluctuations and that may cause or exacerbate financial instability” (Committee on Payment and Settlement Systems–Technical Committee of the International Organization of Securities Commissions, 2012, Paragraph 3.6.10).

2. (2)

   The European Securities and Markets Authority (ESMA) confirms that “EMIR does not include an explicit definition of procyclicality”. Correspondingly, the requirements of their EMIR Regulatory Technical Standards (RTS) describe only the aims to “prevent and control possible procyclical effects, providing that margining requirements shall be forward-looking, stable and prudent” and to “avoid when possible disruptive or big step changes and establish transparent and predictable procedures for adjusting margin requirements in response to changing market conditions” (European Securities and Markets Authority, 2015, Paragraph 3).

3. (3)

   The ESMA also defines procyclicality as “the tendency of a financial variable to move with the cycle, which is undesirable where the variable exacerbates financial stress. For instance, margins often behave this way, as they tend to rise in times of crisis. It is this tendency of margin requirements to increase in times of market stress which is captured in the notion of procyclicality of margin requirements” (European Securities and Markets Authority, 2022a, Paragraph 5).

4. (4)

   The Bank of England also expects APC policies to help avoid large unexpected jumps in IM requirements (Bank of England, 2020, p. 16).

5. (5)

   The European Association of Clearing Houses (EACH) suggests the purpose of APC measures is to avoid margin requirements falling too low in good times, which would entail a potentially destabilizing correction in bad times (European Association of CCP Clearing Houses 2021).

6. (6)

   The Commodity Futures Trading Commission (CFTC) Market Risk Advisory Committee considers that “CCPs should seek to limit the reactivity of margins to volatility changes, which may exacerbate liquidity stress, and balance that with overall margin coverage levels” (Market Risk Advisory Committee 2021).

Note that the CPSS–IOSCO definition addresses APC in a rather broad way, including more than just margin changes. Rather, general “changes in risk-management practices” should be the focus of the regulation. These changes would need to be “positively correlated with market, business or credit cycle fluctuations” in order to fall under the definition, and further “cause or exacerbate financial instability”. However, it remains ambiguous how the positive correlation or causal relationship between financial instability and credit cycle fluctuations should be determined (in advance) in order to avoid such fluctuations.

In contrast, the EMIR RTS definition (European Securities and Markets Authority 2015) is even more ambiguous, since no connection with “market, business or credit cycle fluctuations” is required, and the specified goal to “prevent and control possible procyclical effects” is also a circular argument. This leaves us with the vague characteristics of “forward-looking, stable and prudent” margin requirements, which are not well suited for any practical implementation as none of these concepts is well defined or exactly specified by parameters. Moreover, the relationships or priorities between the various factors are not laid out when trade-offs are encountered. The term “disruptive” is also not specified or characterized in detail. The only roughly measurable and defined criterion is therefore to “avoid when possible big step changes”.

The ESMA statement (European Securities and Markets Authority 2022a) is more precise, as “increases in margin requirements” can be captured more directly in times of “market stress”. Unfortunately, this definition is neither rooted in the EMIR nor directly reconcilable with the CPSS–IOSCO definition. In fact, in Paragraph 11 the document addresses increases in margins that may have acted in a procyclical manner, but proceeds to state that this could have been “potentially diffusing or even amplifying liquidity stress to other parts of the financial system”, indicating that not only “market stress” but also “liquidity stress” would play a role.

The Bank of England description refers only to large unexpected jumps, without any reference to market stress (Bank of England 2020), and the EACH statement refers to the assessment of “good times” when margins should not fall too low. A CFTC remark sees the reactivity of margins to volatility changes in the center of procyclicality (Market Risk Advisory Committee 2021), with liquidity stress as a trigger that is not easy to define properly.

2.2 The effectiveness of APC tools

The ambiguity and inconsistency in the definition of APC is also reflected in the measurement of the effectiveness of APC tools.

1. (1)

   The ESMA guidelines do not provide a clear metric to measure (anti-)procyclicality. Instead, according to European Securities and Markets Authority (2019, Paragraph 10), the CCPs’ “competent authorities should ensure that any CCP supervised by them defines quantitative metrics to assess the margins … in the context of margin procyclicality”. Potential metrics are listed as “margin changes over a defined period”, the “standard deviation of margin” or a “peak-to-trough ratio over a defined period”,¹¹ 1 The peak-to-trough ratio is introduced formally in Section 4.4.1. but neither the priority nor the target levels are provided for the various metrics.

2. (2)

   European Securities and Markets Authority (2022a) highlights that during the Covid-19 stress episode, European Union (EU) CCP margin models “reacted differently, with some models performing in a more procyclical manner than others” (Paragraph 14). Rather than focusing on a clear definition of APC, the ESMA highlights the “divergent implementation of the APC tools” and the “need for higher granularity of the relevant provisions” (Paragraph 16), while noting that it is “difficult to define common targets” and referring to “policy discussions” at the international level (Paragraph 41).

3. (3)

   Its first recommendation, the CFTC market risk advisory committee proposes to implement “a standard set of metrics to measure procyclicality that are suitable to the specific CCP to enable the CCP to determine targets to be achieved” (Market Risk Advisory Committee 2021), implicitly noting therefore that the targets to be achieved are not sufficiently clear.

4. (4)

   Murphy et al (2016) state that “both the term ‘procyclicality’ and the three EMIR procyclicality mitigation tools lack precise definitions”.

Since the effectiveness measures of APC tools are not specified precisely, as exemplified above, effective regulation and, above all, harmonized application of rules is not possible. As a consequence, Murphy et al (2016, p. 4) highlight that it “may be appropriate to move from tool-based procyclicality regulation to one based on the desired outcomes”. This is also mirrored by the CFTC, which recommends prioritizing the “desired outcome of reducing procyclicality, not the specific means of reducing it” (Market Risk Advisory Committee 2021).

The next section highlights the limitation of APC tools from a general perspective.

2.3 General limitations of the IM-based APC regulation

2.3.1 The IM has a smaller effect than a VM

The effectiveness of IM-based APC regulation is sometimes seen as quite limited in comparison with the (unlimited) requirements to pay variation margins (VMs). Accordingly, Committee on the Global Financial System (2010) notes that: “One interpretation of the market participants’ views is that raising the initial margins … to contain financial leverage in order to counteract these procyclical effects of leverage may not, however, dampen the large and disruptive VM calls that can arise in adverse market conditions.” The analysis of the Public Quantitative Disclosure of CCPs conducted by CCP12 (2021, p. 23) found that VMs based on a quarterly average were significantly larger than IM changes in the first quarter of the Covid-19 pandemic (ie, 2020 Q1). Gurrola-Perez (2020) also argues that margin calls are largely driven by the VMs rather than by the IM.

Other papers seem to argue that the VMs would not contribute to procyclicality. For example, European Systemic Risk Board (2020, p. 17) states that the “exchange in variation margin is a redistribution of liquidity based on profits and losses … and is neutral with regards to the overall liquidity in the system”, while European Association of CCP Clearing Houses (2021, p. 12) states that “variation margins are … not procyclical by definition” and cites the following statement by Bank of England (2020, p. 16): “Variation margin does not typically remove liquidity from the system, but rather redistributes it.” This interpretation of “not procyclical” only applies to firms that act as intermediaries and/or use cleared derivatives to hedge noncentrally cleared but bilaterally margined derivatives. However, even for intermediaries, during the Covid-19 market turmoil centrally cleared VMs ranked the highest of all the demands that forced draws on liquidity resources (Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions, 2022, p. 30). First, end users (ie, firms hedging real economic exposure) losing liquidity might need to close positions, and there is no indication that end users gaining liquidity would enter compensating positions. Second, there is also no indication that those firms gaining liquidity would make it available to the firms losing liquidity. This is also seen in the Covid-19 market turmoil analysis by Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2022, p. 32ff), which gives a diverse picture of asset increases and asset liquidation for different end users but overall highlights various vulnerabilities in funding markets and impaired market liquidity at that time. Third, a VM might reflect not only liquidity losses but also accounting losses. Thus, liquidity might be needed precisely at a time when the credit quality is negatively affected. The ESRB statement that, in times of stress, “counterparties receiving variation margin might hoard liquidity, reducing liquidity in the system and amplifying the market downturn” (European Systemic Risk Board, 2020, Table 2) thus oversimplifies the situation.

2.3.2 The IM is only part of a buffer reserve needed for future VMs

A further category of liquidity requirements to be considered is the reserves for potential future VM outflows (see, for example, European Systemic Risk Board 2020, p. 18). While a VM might be seen as a zero-sum game, this reserve drains liquidity from all market participants. The interpretation of the effectiveness of APC IM tools largely depends on the way this reserve is viewed for a future VM. If the end users see IM as liquidity that can be reclaimed as positions are downsized, then the pressure to downsize positions due to a lack of liquidity would likely actually result not from the IM but from the need to hold an even larger amount of liquidity reserve.

2.3.3 The VMs for cleared and uncleared derivatives contribute to the VM scope

It should be noted that this need to hold liquidity reserves for the VMs not only is present in cleared products but has also, due to regulation, been partly introduced into over-the-counter (OTC) uncleared markets. European Systemic Risk Board (2020, Section 4.5) addresses these reserves in its policy options.

2.3.4 The IM procyclicality may be caused more by rising underlying prices than by volatility

In stressed stock market conditions, prices often drop, offsetting IM increases. In commodity markets, just the opposite occurs, as evidenced by recent dramatic increases in power, gas and oil prices (see Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions 2023, pp. 8–12). In these cases, procyclicality may be driven much more by rising prices than by increasing volatilities, limiting the effectiveness of APC tools based on only volatility.

The next section is devoted to the APC tools proposed by EMIR in particular.

3 The European Markets Infrastructure Regulation anti-procyclicality tools

European Union (2013, Article 28) prescribes three possible tools to limit procyclicality:

1. (a)

   applying a margin buffer equal to at least 25% of the calculated margins, which allows it to be temporarily exhausted in periods when calculated margin requirements are rising significantly;

2. (b)

   assigning at least 25% weight to stressed observations in the retrospective period; and

3. (c)

   ensuring that margin requirements are not lower than those that would be calculated using the volatility estimated over a 10-year historical retrospective period.

The application of these three tools may include mixing multiple tools, as shown in Figure 1.

Figure 1: Use of procyclicality tools. Source: European Association of CCP Clearing Houses (2021, Figure 6).

ISDA observes that most CCPs globally use at least one of the three APC tools prescribed by the ESMA, while most CCPs use floors (International Swaps and Derivatives Association 2021).

3.1 All tools basically come down to the application of an add-on

All the APC tools discussed above can be expressed in the following form:

$$M=r(1+\pi ),$$

(3.1)

where $M$ represents the APC margin (ie, the actual margin charged), $r$ the risk-based margin before APC tools and $\pi$ the add-on as a percentage. Regarding the add-on, the following options are possible in general.

1. (a)

   The add-on is simply the buffer, $b$, which ranges from 0 to 25% (ie, $\pi =b\in [0;0.25]$).

2. (b)

   The add-on can be interpreted as implementing an APC margin of $M=0.25s+0.75r$, as

   $$\pi =0.25\left(\frac{s-r}{r}\right),$$

   (3.2)

   where $s$ is the margin based on the maximum volatility observed so far (stressed volatility).

3. (c)

   The APC margin corresponds to $M=\max (r(\text{10 years});r(\text{current}))$, where $r$ equals the margin using the volatility estimated over the stated history. The corresponding add-on is

   $$\pi =\max (\frac{r(\text{10 years})-r(\text{current})}{r(\text{current})};0).$$

   (3.3)

While different interpretations exist, the above interpretation will be implemented on the level of the volatility to improve comparability (see Section 4.3).

The next section provides a critical assessment of the APC tools with particular reference to the market stress induced by the Covid-19 pandemic.

3.2 Critical assessment of EMIR APC tools

3.2.1 EMIR 25% buffer (EMIR RTS, Article 28(a))

While a 25% buffer has some advantages, it is not clear when to release it and when to build it up again; that is, the significantly increasing margin requirements to which the regulation refers do not necessarily materialize only in periods of market stress. Accordingly, releasing the buffer when margins are significantly increasing may still correspond to normal market circumstances, possibly followed by market stress later on. Adhering to the regulation would therefore only allow part of the 25% buffer to be released in order to still have further capacity in case a subsequent significant market stress occurs. In addition, with regard to the Covid-19 pandemic, ISDA states that margin requirements for certain equities increased by more than 300% in March 2020 compared with January 2020 (International Swaps and Derivatives Association 2021), making a 25% buffer ineffective. It is difficult to calibrate this APC tool in a meaningful way. The calibration should be dynamic and aligned with the underlying contract, requiring an analysis of previous price movements in both normal and stressed periods. There is also insufficient guidance about when this buffer can be used and when it should be replenished. These problems are also reflected in the article by Cominetta et al (2019), where the buffer tool reduces the peak-to-trough (PTT) measure by only 4%. European Securities and Markets Authority (2015, Paragraph 36) notes that the 25% buffer is small when margins are small, which undermines the practical applicability. The decision to exhaust the buffer means that either current procyclical effects or future procyclical effects are addressed, forcing a choice that dilutes the effect of the 25% buffer.

3.2.2 EMIR 25% stress (EMIR RTS, Article 28(b))

One issue with adding 25% stressed observations is that the actual weight of stress on entering a stress period increases above the 25%. In addition, market condition may be individually stressed but potentially not quite as stressed as the “stressed observations” used. In both cases IM would be more expensive than necessary during stressed periods. Both effects potentially increase the amount of deleverage needed by the system in comparison with a situation where no APC IM tool is applied. Further, in a portfolio context, adding stressed observations increases (or decreases) the correlation and likewise causes potential increased margin offsets (or excess diversification benefits). Thus, a continuous recalibration and justification of stressed scenarios is needed, making the application complex. Correspondingly, ISDA believes that further guidance is needed on how these stressed scenarios are chosen and maintained, as well as on the framework governing this process (International Swaps and Derivatives Association 2021).

3.2.3 EMIR 10-year floor (EMIR RTS, Article 28(c))

With respect to the 10-year floor, ISDA highlights that (International Swaps and Derivatives Association 2021): “This floor is easy to calculate and apply to an existing margin model. However, once the floor is exceeded, this tool has no further impact on procyclical IM increases. During the COVID-19 crisis, 10-year floors mostly had little impact, as 2008 stress periods had rolled off the lookback period. It was regular margin models with a shorter lookback period that drove higher margin requirements.” Moreover, European Securities and Markets Authority (2015, Paragraph 50) notes that under stressed conditions, extraordinary margin calls become inevitable and no buffer might be available to absorb the rapid increase of margin requirements. It would therefore be pure coincidence (ie, a volatility above or below the 10-year floor) that the tool generates any anti-procyclical effect limiting the tendency of margin requirements to increase in times of market stress (see European Securities and Markets Authority 2022a).

4 Empirical analysis

4.1 Data

We make use of DAX closing price data ranging from $t=T_{\max }$ (November 19, 1982) to $t=0$ (November 27, 2022) as a proxy for a futures position. We assume a long DAX index position and use the models in Section 4.2 in the two variants ${\mathrm{CF}}_1$ and ${\mathrm{CF}}_2$ for the confidence factors ${\lambda }_1=0.94$ and ${\lambda }_2=0.99$.

The analysis of APC measures only goes back to the end of November 1987 ($T_{\mathrm{hist}}$); the model for the maximum volatility is in a more or less stable setting after Black Monday (October 19, 1987).

4.2 Models

We test the procyclicality tools based on a value-at-risk (VaR) model with an exponentially weighted moving average (EWMA) measure of volatility based on relative returns of the prices, $P$, of the respective assets, for convenience. Thus, the risk-based IM is defined as

$$r(t)=\mathrm{CF}(t)\times P(t)\times \underset{ ={\sigma }_{\lambda }(t)}{\underbrace{\sqrt{\frac{1}{{\sum }_{\tau =1}^{250}{\lambda }^{\tau }}\left(\sum _{\tau =1}^{250}{\lambda }^{\tau }{\left(\frac{P(t+\tau -1)-P(t+\tau )}{P(t+\tau )}\right)}^2\right)}}}\times \sqrt{H}.$$

(4.1)

Accordingly, CF represents the confidence factor used to determine the VaR at the desired 99% confidence level. The square root is the EWMA daily volatility based on the stated time series and $t$ is the point in time under consideration, while $H$ is the holding period (one day or more to measure margins; typically, one day for backtesting versus one day return). It is known that EWMA measures become more cyclical (react more strongly to changed volatility) if the decay factor, $\lambda$, is significantly less than $1$. We use ${\lambda }_1=0.94$ for this case, derived from the RiskMetrics standard (JP Morgan-Reuters 1996), which is a rather cyclical choice. If the parameter is closer to $1$, the volatility measure is less cyclical. We set ${\lambda }_2=0.97$, closer to the level of 0.99, which was proposed, for example, by Murphy et al (2014, p. 9). We generally use $H=1$, for simplicity.

Typically, the confidence factor applied to calibrate the IM estimate is more stable when ${\lambda }_1=0.94$ and, for example, the confidence factor ${\mathrm{CF}}_1(t)=2.6495$ from a Student $t$ distribution with four degrees of freedom usually produces an adequate backtesting quality for many financial time series.

For ${\lambda }_2=0.97$, the confidence factor typically requires the assumption of an even more heavy tailed distribution, making it worthwhile to adapt the confidence factor when needed. This continuous recalibration can be done by using a fixed factor that is adjusted “by hand” from time to time (eg, based on model validation and backtesting results) to ensure adequate risk coverage,²² 2 For more details we refer the reader to Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2023, p. 11). or can, in our case, be proxied by the implementation of an automatic recalibration based on a filtered historical simulation approach. This sets a daily adaptable confidence factor based on quantiles $q_{\alpha }$ of the standardized time series of returns:

	${\mathrm{CF}}_2(t)$	$={\displaystyle \frac{1}{2}}\left(\right|q{(1\%)}_{\tau =t}^{t+250}\left({\displaystyle \frac{P(\tau )-P(\tau +1)}{P(\tau +1){\sigma }_{\lambda }(\tau +1)}}\right)|$
		$\mathrm{\hspace{1em}\hspace{1em}}\mathit{\hspace{1em}\hspace{1em}}+\left|q{(99\%)}_{\tau =t}^{t+250}\left({\displaystyle \frac{P(\tau )-P(\tau +1)}{P(\tau +1){\sigma }_{\lambda }(\tau +1)}}\right)\right|).$		(4.2)

This calculation is stabilized by a floor (here ${\mathrm{CF}}_1$) and a cap (here set at 3.5) inspired by Commodity Clearing (2022).

Pure historical VaR or equally weighted volatility estimators were not used, as they would be as affected by a stressed period leaving the observation period as they would by stressed periods entering the observation period. The effect on margin changes would therefore not be a pure reaction to current market stress.

An interesting exercise is to contrast the two models (with ${\lambda }_1$ and ${\lambda }_2$) as there is an obvious trade-off between the cyclicality of EWMA volatility and the cyclicality of the confidence factor.

4.3 Tools

The following tools (see also Section 3) are implemented.

4.3.1 The zero-add-on tool

For reference purposes, one implementation does not use any add-on: this is the pure risk-based margin calculated in (4.1).

4.3.2 The basic buffer tool

The basis buffer (see Commodity Clearing 2022) has the add-on

$$\pi (t)=\min \left\{1-\max \left[(\frac{{\sigma }_{\lambda }(t)-{\min }_{\tau =t}^{T_{\max }}({\sigma }_{\lambda }(\tau ))}{{\max }_{\tau =t}^{T_{\max }}({\sigma }_{\lambda }(\tau ))-{\min }_{\tau =t}^{T_{\max }}({\sigma }_{\lambda }(\tau ))}-0.2,0)\right]\frac{1}{0.8}\right\}\times 0.25.$$

(4.3)

The minimum and maximum volatility of the time series so far are used to calibrate the point where the volatility buffer is released, initially with 25% for low volatilities, where the buffer is gradually released starting from 20% of the difference between maximum and minimum volatility until the buffer reaches zero at maximum volatility.

4.3.3 The 10-year floor tool

This tool uses the add-on

$$\pi (t)=\max (\frac{1}{{\sigma }_{\lambda }(t)}\frac{{\sum }_{\tau =t}^{\min (t+3650,T_{\max })}({\sigma }_{\lambda }(\tau ))}{\min (T_{\max }-t,3650)}-{\sigma }_{\lambda }(t);0),$$

(4.4)

where the sum represents the average EWMA volatility of the time series estimated using a 10-year historical period.

4.3.4 The 25% stress tool

This tool uses the add-on

$$\pi (t)=25\%\times \frac{{\max }_{\tau =t}^{T_{\max }}({\sigma }_{\lambda }(\tau ))-{\sigma }_{\lambda }(t)}{{\sigma }_{\lambda }(t)},$$

(4.5)

where “$\max$” is the maximum EWMA volatility of the time series so far when implementing the stressed observations tool.

4.3.5 The Murphy (2016) buffer

This tool uses the add-on

$$\pi (t)=\min (\frac{{\max }_{\tau =t}^{T_{\max }}({\sigma }_{\lambda }(\tau ))-{\sigma }_{\lambda }(t)}{{\sigma }_{\lambda }(t)};25\%),$$

(4.6)

where the max is the maximum EWMA volatility of the time series so far. The tool essentially combines the basic buffer tool with a release mechanism that would reduce the buffer only if the applied volatility after the buffer were to exceed the maximum volatility so far (we refer the reader to Murphy et al (2016, Section 2.6) for more details).

4.3.6 The Murphy (adapted) tool

This tool is similar to the one above but uses the add-on

	$\pi (t)$	$=\min \{{\displaystyle \frac{1}{2}}[\max ({\displaystyle \frac{1}{{\sigma }_{\lambda }(t)}}{\displaystyle \frac{{\sum }_{\tau =t}^{\min (t+3650,T_{\max })}({\sigma }_{\lambda }(\tau ))-{\sigma }_{\lambda }(t)}{\min (T_{\max }-t,3650)}},0)$
		$\mathrm{\hspace{1em}\hspace{1em}}\mathit{\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}}+{\displaystyle \frac{{\max }_{\tau =t}^T({\sigma }_{\lambda }(\tau ))-{\sigma }_{\lambda }(t)}{{\sigma }_{\lambda }(t)}}];25\%\},$		(4.7)

where the max and sum in the square bracket respectively represent the maximum and mean EWMA volatility of the time series. This tool essentially combines the basic buffer tool with a release mechanism similar to that of Murphy, which has the problem that the buffer add-on is charged even if the volatility is near (but not yet at) the historical maximum level. By combining this with the 10-year floor tool, the add-on is released earlier.

4.4 Measurement

Given the size of the cleared derivative markets, both unintended consequences and costs need to be considered as well as the intended consequences when adding non-risk-based IM regulatory requirements to risk-based regulatory requirements (see, for example, Murphy and Vause 2021).

Accordingly, the unintended consequences and costs could be any of the following.

1. (1)

   The APC measures could be ineffective, which, in the literature, tends to be associated with

   1. (a)

      the PTT ratio (Murphy et al 2014, 2016; Cominetta et al 2019; European Securities and Markets Authority 2019), or

   2. (b)

      the $n$-day increase in IM (Murphy et al 2014, 2016; Cominetta et al 2019; European Securities and Markets Authority 2015, 2019; Szanyi et al 2018).

2. (2)

   The APC measures could hide an ineffective risk-based margin in the sense that the add-on would help to cover an incorrectly calibrated IM model, and in times of stress the reduced add-on could therefore lead to miscalibration.

3. (3)

   The additional APC-compliant IM add-ons increase the overall costs to the economy in the following ways.

   1. (a)

      Capital is blocked as IM and the economy incurs the following capital costs:

      1. i.

         credit line and utilization costs, depending on the collateralization and credit quality of the borrowing market participant;

      2. ii.

         CCP and clearing member fees, based on the amount of collateral;

      3. iii.

         if the IM is passed through to a CCP via a clearing member (not individually segregated), the IM is also exposed to risk (credit risk or liquidity risk).

   2. (b)

      The additional costs of hedging may make it less attractive and therefore expose market participants to more risk than necessary. Such unhedged risks could force hedging if volatility increases to such an extent that the economic exposure would no longer be tolerable, which would allow passive procyclicality to materialize from non-existent hedging positions.

   3. (c)

      The costs of hedging may be different for different market participants (eg, producers or consumers) and therefore affect prices through supply and demand.

4. (4)

   Stabilizing IM in times of unstressed market conditions may create a false sense of security. Being exposed to larger margin movements in normal times could make liquidity buffers larger, and therefore make market participants better prepared for stressed IM movements.

Variants of the first three classes of measurement criteria for the abovementioned effects are proposed in the literature. For example, European Securities and Markets Authority (2022a) suggests that: “When setting its APC policy, the CCP attempts to consider these three dimensions: stability, conservativeness and overcollateralization.” To the best of our knowledge, the fourth category has not been considered in past analyses.

4.4.1 Effectiveness

As highlighted above, there is no agreement in the literature or in regulations about how APC effectiveness should actually be measured; this may be due to the lack of agreement on the definition itself, as discussed in Section 2.1. Regulators therefore usually leave the problem to CCPs and see themselves as reviewers of the approach. See, for example, European Securities and Markets Authority (2019, p. 4): “Competent authorities should ensure that any CCP supervised by them defines quantitative metrics to assess the margins, including margin add-ons, in the context of margin procyclicality.” In the EU, multiple such measures are to be developed by CCPs to address two typical but different interpretations of the purpose of APC measures (ie, a long-term measure and a short-term measure). Again, regulations and practices often use a PTT measure and an $n$-day increase-in-margin measure. However, these measures quantify cyclicality instead of procyclicality (see Murphy et al 2014, p. 7) because they are completely unrelated to the consequences of causing or exacerbating stress. While this problem has been addressed in the literature (see, for example, Wong and Zhang (2021), who highlight the possibility of applying a procyclicality measure restricted to subperiods in which volatility is elevated), the solutions therein are rarely implemented. Another problem in applying and implementing these measures as guiding principles is that they can only be determined ex post but are used to make on-the-spot decisions requiring the anticipation of future scenarios. The optimal strategy will only become clear in retrospect.

Long-term measure: peak-to-trough.

Proponents of the PTT measure may argue that positions in derivatives markets are held for a long time, and the leverage that is built up in times of low margin rates is the cause of later necessary deleveraging trades, which in turn create additional volatility and exacerbate market stress.

To specify the PTT measure precisely, a time period $T_{\mathrm{PTT}}$ in which both the peak and trough are determined needs to be defined (in our analysis this is set to one year). Usually, the PTT measure is considered as a long-term measure. We will take the average of the measure denoted by $\mathrm{avPPT}$:

$$\mathrm{avPPT}=\frac{1}{T_{\mathrm{hist}}+1}\sum _{\tau =0}^{T_{\mathrm{hist}}}\mathrm{PPT}(\tau )$$

(4.8)

with

$$\mathrm{PTT}(t)=\frac{{\max }_{\tau =t}^{t+T_{\mathrm{PTT}}}(M(\tau ))-{\min }_{\tau =t}^{t+T_{\mathrm{PTT}}}(M(\tau ))}{{\min }_{\tau =t}^{t+T_{\mathrm{PTT}}}(M(\tau ))}.$$

(4.9)

Being a backward-looking measure, it cannot guide day-to-day decisions on the application of APC measures, and it introduces biases depending on the current volatility situation relative to a particular time period, or relative to mean or stress volatility. In addition, there is only a very indirect link between this measure and the partially articulated APC goal of avoiding unanticipated big-step increases in margin.

Short-term measure: maximum margin increase.

Proponents of the maximum margin increase (MMI) measure would argue that this measure is most closely aligned with the partly articulated requirement to avoid big-step margin increases or the “stability of margins” (see also European Securities and Markets Authority 2022a, Paragraph 132). We calculate this measure as the maximum IM increase ($\mathrm{MIMI}$) and the maximum VM increase ($\mathrm{MVMI}$) over a period of $n$ days ($n=3,30$). As part of our calculation, we apply the maximum increase within the $n$-day period (rather than the start-to-end increase), as this drives potential liquidity shortfalls. We first calculate the maximum absolute amount of the IM increase, $\mathrm{IMI}(t)$ and the VM increase, $\mathrm{VMI}(t)$, respectively,³³ 3 VMI is defined as the VM for the period from $t$ to $t+1$. within the 3- or 30-day period. This increase is then related to the IM, $M$, at the beginning of the $n$-day period and a maximum is applied up to a historical time, $T_{\mathrm{hist}}$:

	${\mathrm{IMI}}_{n\text{days}}(t)$	$=\underset{\tau =t-n}{\overset{t-1}{\max }}(M(\tau )-M(t)),$		(4.10)
	${\mathrm{VMI}}_{n\text{days}}(t)$	$=\underset{m=1}{\overset{n}{\max }}\left|\left({\displaystyle \sum _{\tau =t-m}^{t-1}}(VMs(\tau ))\right)\right|,$		(4.11)
	${\mathrm{MIMI}}_{n\text{days}}$	$=\underset{\tau =n}{\overset{T_{\mathrm{hist}}}{\max }}{\displaystyle \frac{{\mathrm{IMI}}_{n\text{days}}(\tau )}{M(\tau )}},$		(4.12)
	${\mathrm{MVMI}}_{n\text{days}}$	$=\underset{\tau =n}{\overset{T_{\mathrm{hist}}}{\max }}{\displaystyle \frac{{\mathrm{VMI}}_{n\text{days}}(\tau )}{M(\tau )}}.$		(4.13)

4.4.2 Conservativity

We introduce two other measures: the average outlier frequency and the proportion of days for which the daily backtest shows a red or yellow traffic light (Basel Committee on Banking Supervision 1996). Outliers are counted by considering the returns, $R$, of the portfolio in relation to the required margins, $M$, using the indicator function

The averages $\overline{\mathrm{outlier}}$, $\overline{\mathrm{yellow}}$ and $\overline{\mathrm{red}}$ are determined as follows:

	$\overline{\mathrm{outlier}}$	$={\displaystyle \frac{{\sum }_{t=0}^{T_{\max }}\mathrm{OC}(\tau )}{T_{\max }+1}},$		(4.14)
	$\overline{\mathrm{yellow}}$			(4.15)
	$\overline{\mathrm{red}}$	$={\displaystyle \frac{{\sum }_{t=0}^{T_{\max }}{\mathrm{𝟏}}_{\mathrm{OC}(\tau )\ge 10}}{T_{\max }+1}},$		(4.16)

where $\mathrm{OC}$ denotes the outlier count,

4.4.3 Cost of overcollateralization

It seems natural that costs of APC measures should be considered as a selection criterion when developing APC measures. European Securities and Markets Authority (2022a, Paragraph 49) notes that excessive overcollateralization (especially during stress periods) should be avoided. The following three measures are considered as the average,

$$\mathrm{avg}(Q,w)=\frac{{\sum }_{t=0}^{T_{\mathrm{hist}}}Q(t)w(t)}{{\sum }_{t=0}^{T_{\mathrm{hist}}}w(t)},$$

of the ratio $Q(\cdot )$ with given weights $w(\cdot )$ over the period to $T_{\mathrm{hist}}$.

1. (1)

   The average required additional IM, $\overline{m}$, which utilizes the required margin, $M$, and the risk-based margin, $r$, based on the following formula:

   $$\overline{m}=\mathrm{avg}(\frac{M(t)-r(t)}{r(t)};1).$$

   (4.17)

2. (2)

   The equally weighted average overcollateralization, $\overline{o}$, which utilizes the measure of required margin, $M$, and absolute returns, $|R|$, for the portfolio,

   $$\overline{o}=\mathrm{avg}(\frac{M(t+1)-|R(t,t+1)|}{M(t+1)};1).$$

   (4.18)

3. (3)

   The stress-weighted average overcollateralization, $\overline{s}$, utilizes a stress-focused measure of the required margin, $M$, versus absolute returns $|R|$:

   $$\overline{s}=\mathrm{avg}(\frac{M(t+1)-|R(t,t+1)|}{M(t+1)};\frac{\sigma (t+1)-{\min }_{\tau =t}^{T_{\max }}\{r(\tau )\}}{{\max }_{\tau =t}^{T_{\max }}\{r(\tau )\}-{\min }_{\tau =t}^{T_{\max }}\{r(\tau )\}}).$$

   (4.19)

4.4.4 Surprise measure

The surprise measure addresses the maximum relation between a 30- or 3-day IM increase in comparison with the 30-day volatility of IM at the beginning of the period. A high ratio means that market participants may have been underexposed to volatility in the IM in unstressed times, and therefore the IM change came as more of a surprise to them than if the volatility of IM had previously been higher. We measure the surprise factor, $\mathrm{sur}(t)$, using an equally weighted estimator of the standard deviation over 30 days, and we use the maximum surprise factor, $\mathrm{Msur}$, in our analysis:

	${\mathrm{sur}}_{n\text{days}}(t)$	$={\displaystyle \frac{{\max }_{\tau =t-n}^{t-1}(M(\tau ))-M(t)}{{\mathrm{SD}}_{\tau =t+1}^{t+30}(M(\tau )-M(\tau +1))}},$		(4.20)
	${\mathrm{Msur}}_{n\text{days}}$	$=\underset{\tau =0}{\overset{T_{\mathrm{hist}}}{\max }}{\mathrm{sur}}_{n\text{days}}(\tau ),$		(4.21)

where $\mathrm{SD}$ denotes the standard deviation.

4.5 Analysis

Figure 2: The peak-to-trough measure.

The average PTT ratio as shown in Figure 2 suggests that the 10-year floor and 25% stress tool are most effective. The adapted Murphy tool fares better than the original tool proposed by Murphy et al (2016); all 25% buffer tools fare roughly the same as the risk-based IM, with the basic buffer tool faring the best. The EWMA factor (PTT from 1.97 to 1.33) is actually found to have an even greater effect than the APC tools, indicating that regulators should look more closely at risk-based IM. While vague wording (forward-looking, stable and prudent margin requirements that limit procyclicality to the extent that the soundness and financial security of the CCP is not negatively affected) is found in EMIR RTS 153 (European Union, 2012, Article 28), Figure 2 indicates that ex-post regulation (APC tools) is out of proportion compared with the regulation of the anti-procylical nature of the risk-based margin itself.

Figure 3: Maximum (a) 3- and (b) 30-day increases in IM and in VM in relation to IM.

The maximum $n$-day increases in the margin depicted in Figure 3 show that, for the end users of the derivatives, a change in IM only plays a minor role in the overall liquidity management on the longer of the two timescales. On the shorter timescale (three-day measure) the liquidity to be held against VM payments still dominates changes in IM, but IM changes also play a sizable role. The 25% stress APC tool is most effective in reducing the relative impact of the VMs, as it increases IM significantly. The 10-year floor tool is not particularly effective in mitigating IM increases. This indicates that large increases in VaR occur when the volatility exceeds mean volatility. Moreover, the 10-year floor tool is roughly on par with the 25% buffer tools: the earlier a buffer tool releases the buffer, the better it is at managing the maximum IM increases.

Figure 4: Average time for which VaR is in the yellow and red zones (average outlier ratio).

According to Figure 4, the backtesting quality in this example is already adequate for the zero-add-on tools, with zero outliers in the red zone in all cases, and a yellow zone that is, on average, less frequently hit than the permitted type-1 error (10.8%) (Basel Committee on Banking Supervision, 1996, Table 1). The average outlier frequencies (1.01% and 0.85%) are slightly above and below 1%, respectively. Therefore, the risk-based margin is well calibrated. The add-ons increase the conservativity further, especially in the 25% stress tool. Deciding whether to include the add-ons in IM backtesting is therefore important.

Figure 5: Maximum change in IM over 30 days and 3 days in relation to the volatility in IM changes.

The 10-year floor tool (Figure 5) produces the most surprising changes in IM since it stabilizes the IM too much in unstressed times. If the actual volatility is below this 10-year floor, almost all market participants will feel only the effects the price change of the underlying asset and only small changes in volatility. Interestingly, changes in the risk factor associated with the $\lambda =0.97$ volatility measure produce more surprises than the more cyclical shorter-term $\lambda =0.94$ volatility measure, due to the proxied recalibration effects.

Figure 6: Average cost of APC tools measures.

The average cost of the 25% stress method is the highest, as shown in Figure 6. We have already anticipated this in the previous graphs. What is surprising is this method also has the most overmargining when weighted with a stress focus. The 10-year floor tool produced the second highest relative cost of add-on but ranks similarly to the buffer tool in terms of overmargining. The Murphy and adapted Murphy tools are inferior to the basic buffer tool in essentially all cost components, but the adapted tool fares slightly better. The average relative cost being close to 25% in the Murphy (2016) and adapted Murphy tools indicates that the buffer is only rarely released.

4.6 Overall comparison

Based on the abovementioned figures, it becomes apparent that the 25% stress tool is the most expensive, even though it shows broad benefits in the $n$-day increase measures and the PTT. The 10-year floor tool in turn has its greatest benefit in the PTT measure, and does not perform particularly well in the $n$-day increase measure, and comes at a significant cost in terms of the surprise measure. The Murphy (2016) tool could be improved by the proposed adjustment, but the basic buffer tool produces even better overall results, indicating that a gradual release is preferable to the more stress-focused release of the Murphy (2016) tool, which retains too much buffer in cases when the volatility is already close to (but not yet at the level of) the stress volatility. All measures, models and tools produce plausible effects but exhibit very different side effects, making clear communication of the objective even more important in order for CCPs to follow the intent and not (only) the vague letter of the regulations. The effect of the VMs on procyclicality is significant and should be reflected more significantly in the regulatory analysis to put in perspective the efficiency of the measures proposed. While Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2022, p. 38) addresses the streamlining of VM processes between the centralized and decentralized markets as a next step, the concerns of the end users of the derivatives hedging the underlying economic risk still do not appear to have been addressed. Further, the regulatory focus on returns and volatility may be justified for equities (where prices tend to drop in times of stress), but should be reconsidered separately in the case of commodities, whose prices may rise significantly in a stressed environment. This effect is not yet adequately reflected in existing APC regulations and may require more fundamental changes, which we will consider in the next section.

5 Discussion and directions for future research

Table 1: Indicative assessment of policy recommendations. [The table depicts indicative assessments of policy recommendations based on the measures effectiveness, conservativity, costs and surprise. “$+$” indicates corresponding improvements in a dimension; “0” indicates no clear effect; “$-$” indicates a negative effect.]

Measure	Effectiveness	Conservativity	Cost	Surprise
Sizing liquidity reserves (5.1)	0	0	0	$+$
Better recognition of hedged positions in collateral (5.2)	$+$	$+$	$+$	$+$
Releasing the buffer (5.3)	$+$	$+$	$-$	0
Reconsidering safety standards (5.4)	0	0	$+$	0

In this section we address the criteria effectiveness, conservativity, cost and surprises for each of our recommendations, which are summarized in Table 1. It is often implicitly assumed that the downside of the margin tools during “unstressed times” only affects counterparty risk protection (see, for example, Cominetta et al 2019). However, hedging is required for practical purposes (in particular, for commodities) and impeded by additional costs. For example, citing energy companies, European Securities and Markets Authority (2021, Paragraph 64, p. 23) highlights that: “The current level of the clearing thresholds for commodity derivatives and the methodology to calculate positions … can limit their ability to enter into the derivative transactions they would need to do in the context of activities contributing to the energy transition.” Instead of fine-tuning the tools (see, for example, Wong and Zhang 2021; Murphy and Vause 2021; Goldman and Shen 2020) to mitigate procyclicality, it might be preferable to consider alternatives to IM-based margin tools. These are discussed in the following subsections.

5.1 Sizing liquidity reserves

As stated above, a reserve of readily available liquidity needs to be anticipated. Thus, the IM could be seen as part of the reserve that is returned, to the degree that positions are closed out, and it would mostly compensate overall VM payments. European Systemic Risk Board (2020) highlights that: “Liquidity demands from variation margin calls are typically larger than those from initial margin calls and can be more sudden.” Changing IM therefore has a limited effect on the overall reserve calculation. Gurrola-Perez (2020) cites Jon Cunliffe, the Bank of England’s deputy governor for financial stability, noting that “the answers may lie more in ensuring that financial market participants … understand how margin call can evolve in a stress and have the resilience to manage the consequent liquidity pressures” (Cunliffe 2020). To improve this understanding, the stress tests mandated to size the default fund according to EMIR Article 43 could also be reused to give market participants information that allows them to prepare for increased liquidity requirements (ie, distributing stress test results for the VMs and change in IM to market participants calculated by the CCP by assuming these extreme but plausible scenarios materialize in their portfolios).

Evidence of effectiveness.

Recent regulatory reviews highlighted the need for tools to anticipate and to mitigate surprises originating from margin changes linked to “(margin) stress testing tools” (Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions, 2022, p. 27ff) and “what-if functionality” (Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions, 2023, p. 19). Referring to our assessment criteria, the surprise dimension would be positively affected.

5.2 Better recognition of hedged positions in collateral positions

For the hedgers, any nonhedged risk may improve the liquidity position but would impact the profit-and-loss and equity risks. These risks could, like liquidity risks, trigger deleveraging that would then affect the (un)hedged positions. Hedgers may therefore actually be harmed by the margin add-ons in unstressed times. The procyclical nature of increased IM could be mitigated by better recognition of the hedged positions as eligible collateral. The hedged position for a bank can sometimes easily be converted to liquidity by using central bank eligible collateral. This conversion would be harder for nonbanks (eg, liquid securities of a mutual fund could be used as collateral for a loan from a bank that has access to central bank liquidity), and harder still for non-central-bank eligible collateral. This often applies to foreign exchange or commodity exposures. As not all hedged positions can be used as collateral for CCPs, better recognition might be achieved by improving the way in which end users can provide margin coverage by using bank guarantees. Banks in turn could use the hedged positions as collateral for the bank guarantees provided, as they may find it easier to accept such nonliquid collateral. Limitations in application originate from EMIR Article 46, which references EMIR RTS 153 Article 39 in combination with Annex I, Sections 2 and 2a and from the PFMI standards, which prohibit the use of bank guarantees unless they are fully backed by acceptable collateral that is realizable on the same day, such as “an explicit guarantee from the relevant central bank of issue” (Committee on Payment and Settlement Systems–Technical Committee of the International Organization of Securities Commissions, 2012, Paragraph 3.5.2, footnote 63). Thus, a solution would be to provide such explicit guarantees, as foreseen by PFMI, for (parts of) bank guarantees provided as IM coverage (eg, to the degree otherwise provided for nonmarketable assets eligible for Eurosystem credit operations, and to the degree that coverage by hedged positions exists). This allows a fast and efficient solution within the acceptable risk framework of central banks.

Evidence of effectiveness.

The most recent version of EMIR RTS 153 foresees a similar solution, but with a temporary application limited to EU Regulation on wholesale Energy Market Integrity and Transparency (REMIT) derivatives and to nonfinancial counterparties that are eligible as clearing members or limited to public or central bank guarantors. These limits and the lag in implementation means that for future situations, the full backing essentially still negates the benefits of bank guarantees as APC tools in many cases (see also European Securities and Markets Authority 2022b, Paragraph 30; Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions 2023, p. 12ff). Our proposal provides a permanent solution to the problem, which would be PFMI compatible. The cost dimension is improved by the reduced cost of collateral. As the collateral values would be linked with VMs, the effectiveness and surprise dimensions would benefit. Excess collateral would be more likely, improving the conservativity dimension.

5.3 Releasing the buffer/add-on

Only the buffer tool directly allows the release of the add-on to be linked to an observation of stressed market conditions that could be positively correlated with market, business or credit cycle fluctuations, as defined by Committee on Payment and Settlement Systems–Technical Committee of the International Organization of Securities Commissions (2012). Reducing or filling up the required add-on could be linked to either

1. (1)

   an observation of predefined market based criteria, such as credit spreads, volatilities or benchmark prices nearing historical maximum levels or quantiles thereof; or

2. (2)

   macroprudential settings similar to the regulatory-prescribed macroprudential capital buffer.

Such a mechanism would make it more likely that the full buffer amount would be at a CCP’s disposal when needed, and buffer releases would not be diluted. Such buffer releases would be superior in terms of the effectiveness dimension, however, as margin releases would be more targeted, and additional margin buffer would be held most of the time, improving the conservativity dimension and worsening the cost dimension.

5.3.1 Concentrated buffer releases

CCPs could be required to implement the buffer tool on a portfolio basis rather than on an individual risk factor basis, in combination with a policy to detect end-of-day possible stressed market conditions. In such cases, a decision process could be triggered to release (parts of) the portfolio buffer. Thus, increased IM requirements could be mitigated before they are published to the market and clearing participants.

Evidence of effectiveness.

O’Neil and Vause (2018) address the case of macroprudential buffers and argue that the buffers should be released to help meet margin calls that would lead to fire sales. This argument also applies to other, regulatory-prescribed buffers such as existing EMIR APC buffers. An application in everyday situations to just mitigate margin calls is wasteful; the buffer should rather only be released to help avoid fire sales, regardless of whether the source is IM or VM.

5.3.2 Regulatory decision to release buffers

Regulatory prescribed buffers already exist for banks and are also designed to address systemic stress, albeit on a very different timescale.⁴⁴ 4 Behn et al (2020) states that such capital buffers are “meant to mitigate procyclicality by acting as shock absorbers in times of stress” and that “buffers mitigate negative externalities related to excessive deleveraging or fire sales that could otherwise harm the economy”. The EMIR does not foresee a supervisory buffer, even though it has been considered by European Systemic Risk Board (2020). We do not argue in favor of such a supervisory additional buffer but note that, through APC tools, there exists a rule-based buffer that is implemented and released by CCPs. However, due to the intraday time frame (ie, only hours for end-of-day processing of CCPs), the setting of the EMIR APC buffer release requires formula-based decisions. In the past, formula-based decisions have not been ideal for macroprudential purposes (see Wiessmann 2020). Having only regulation-based tools for regulators (necessitating changes in law) is also not ideal, and the time lag for changes in law would be more significant than in a situation in which a regulator could simply choose to mandate CCPs to release their existing APC buffers.

Evidence of effectiveness.

The massive changes in liquidity requirements from clearing gas and power in 2022 resulted in multiple recent state- or central-bank-sponsored decisions, such as credit lines for margin financing from the German state-owned KfW or the Bank of England’s 2022 Energy Financing Scheme, and a mandated release of existing APC buffer might serve as a timely (additional) remedy.

5.4 Reconsidering safety standards

The APC tools may increase non-IM-based procyclicality. Alternatively, IM procyclicality might be limited by, for example, speed limits or upper limits, as discussed by Goldman and Shen (2020). What the latter methods have in common is that risk coverage from the CCP perspective is shifted from IM to the common resources of the default fund, thus ensuring prefinanced and efficient risk coverage. However, in times of market stress (from the perspective of the clearing member), all these methods (add-on and IM (increase) limits) effectively lead to a lower safety standard (than usual), including: increased default risk of the CCP itself to the clearing member; increased risk of default fund usage for the clearing member; or increased risk that clearing members could suffer from a client default due to insufficient collateral. Clearing members might deal with that increased risk versus their clients through

1. (1)

   increased fees,

2. (2)

   tighter IM limits or

3. (3)

   putting additional margin requirements on top of the IM of their clients

(all of which may be procyclical). This problem is addressed, for example, by European Systemic Risk Board (2020). Raykov (2018) studied this effect in a simplified model and linked it to insufficient risk aversion by clearing members. It should be noted that a high degree of risk aversion could remove the need for any treatment of the increased risk but would probably also limit the access of weaker counterparties to clearing, which in our view would not be a desirable outcome.

Limiting the risk impact on clearing members would improve the cost dimension. Additional elements to the default waterfall might be allowed, such as guarantees, catastrophe bond structures, or insurance or reinsurance for the default fund, all of which distribute the risk and liquidity requirements to areas of the economy possibly not directly affected by the stressed market conditions. Increased default fund risk due to lower than usual IM coverage could be recognized and allocated explicitly. In the case of a clearing member default that exceeds a predefined safety level of IM, the CCP could distribute its uncovered loss to end-user clients through VM haircutting without increasing its own risk or affecting the risk of clearing members. This compromise between CCP and OTC safety levels may be more appropriate than the general increase of margin or default fund levels in times of stress for a risk that may never materialize or that is significantly reduced by the existing IM.

Evidence of effectiveness.

Possible evidence of this reactivity in the Covid-19 market turmoil is presented by Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2022, p. 29), which states that 17% of intermediaries indicated that they made material changes to the credit limits applied to counterparty positions or the type of credit limits imposed on those positions. Also Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2023, p. 20) provides evidence of additional margin requirements from clearing members that are discretionary and hard to understand and predict. Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2023, p. 18) also suggests that market participants are already balancing funding and liquidity risk against counterparty and market risk. Some have actively migrated positions from centrally cleared to OTC markets and/or reduced overall hedging. Allowing market participants to transact within the centrally cleared system while keeping most of the advantages might be better than pushing market participants away from central clearing.

5.5 Summary of qualitative measures

Consistent with the evidence of effectiveness highlighted in the previous subsections, Table 1 summarizes the indicative qualitative assessment of the policy recommendations based on the effectiveness, conservativity, costs and surprise dimensions, respectively.

6 Conclusions

Based on a comparison of the effects of APC tools with effects of VMs cyclicality, we argued that regulators should move away from a tools-based approach to a required effect regulation, and we proposed different non-IM-based approaches to limit procyclicality. These approaches have the potential to be much more effective than the existing APC IM tools or could at least be important additions to the tool kit.

Declaration of interest

The authors report no conflicts of interest but note that Thomas Siegl previously worked as a board member at a central counterparty. The authors alone are responsible for the content and writing of the paper. All possible remaining errors are their own.

Acknowledgements

We thank the editor and two anonymous reviewers for their valuable input.

References

• Basel Committee on Banking Supervision (1996). Supervisory framework for the use of “backtesting” in conjunction with the internal models approach to market risk capital requirements. Standards Document, January, Bank for International Settlements, Basel. URL: http://www.bis.org/publ/bcbs22.htm. 
• Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2022). Review of margin practices. Consultative Document, September, Bank for International Settlements, Basel. URL: http://www.bis.org/bcbs/publ/d537.htm. 
• Basel Committee on Banking Supervision–Committee on Payments and Market Infrastructures–International Organization of Securities Commissions (2023). Margin dynamics in centrally cleared commodity markets in 2022. Report, May, Bank for International Settlements, Basel. URL: http://www.bis.org/bcbs/publ/d550.htm. 
• Behn, M., Rancoita, E., and Rodriguez d’Acri, C. (2020). Macroprudential capital buffers: objectives and usability. ECB Macroprudential Bulletin, Volume 11. URL: https://econpapers.repec.org/article/ecbecbmbu/2020_3a0011_3a1.htm. 
• Bank of England (2020). The Bank of England’s supervision of financial market infrastructures: 15 February 2019–3 December 2020. Annual Report, December, Bank of England. URL: http://www.bankofengland.co.uk/-/media/boe/files/annual-report/2020/supervision-of-financial-market-infrastructures-annual-report-2020.pdf. 
• CCP12 (2021). Annual markets review in central counterparty clearing. Report, July, Global Association of Central Counterparties, Shanghai. URL: https://ccp-global.org/wp-content/uploads/2023/07/CCP12_AMR_2022_20230713-final.pdf. 
• Cominetta, M., Grill, M., and Jukonis, A. (2019). Investigating initial margin procyclicality and corrective tools using EMIR data. ECB Macroprudential Bulletin, Volume 9. URL: http://www.ecb.europa.eu/pub/financial-stability/macroprudential-bulletin/html/ecb.mpbu201910_5{~}6c579ba94e.en.html. 
• Committee on Payment and Settlement Systems–Technical Committee of the International Organization of Securities Commissions (2012). Principles for financial market infrastructures. Final Report, April, Bank for International Settlements, Basel. URL: http://www.bis.org/cpmi/publ/d101a.pdf. 
• Committee on the Global Financial System (2010). The role of margin requirements and haircuts in procyclicality. CGFS Paper 36, Bank for International Settlements, Basel. URL: http://www.bis.org/publ/cgfs36.htm. 
• Cunliffe, J. (2020). Financial system resilience: lessons from a real stress. Speech at the Investment Association Webinar, London, June 9. URL: http://www.bankofengland.co.uk/speech/2020/jon-cunliffe-speech-at-investment-association. 
• European Association of CCP Clearing Houses (2021). CCP resilience during the Covid-19 market stress. Position Paper, June, EACH, Brussels. URL: https://bit.ly/47VGlax. 
• European Commodity Clearing (2022). ECC derivatives market margining. Version 1.6.1. ECC, Leipzig. URL: http://www.ecc.de/fileadmin/ECC/Downloads/Risk_Management/Margining/ECC_Derivative_Market_Margining_V1.6.1.pdf. 
• European Securities and Markets Authority (2015). Review on the efficiency of margining requirements to limit procyclicality. EMIR Review Report 2, August, ESMA, Paris. URL: https://bit.ly/48NOQpg. 
• European Securities and Markets Authority (2019). Guidelines on EMIR anti-procyclicality margin measures for central counterparties. Guidelines, April, ESMA, Paris. URL: https://bit.ly/3SMJYvf. 
• European Securities and Markets Authority (2021). Review of the clearing thresholds under EMIR. Discussion Paper ESMA70-156-5010. ESMA, Paris. URL: https://bit.ly/48NOQpg. 
• European Securities and Markets Authority (2022a). Review of RTS No 153/2013 with respect to the procyclicality of margin. Consultation Paper, January, ESMA, Paris. URL: https://bit.ly/3ucotdx. 
• European Securities and Markets Authority (2022b). Emergency measures on collateral requirements: draft Regulatory Technical Standards amending Commission Delegated Regulation (RTS) 153/2013. Report ESMA91-372-2466, ESMA, Paris. URL: https://bit.ly/3HFGRii. 
• European Systemic Risk Board (2020). Mitigating the Procyclicality of Margins and Haircuts in Derivatives Markets and Securities Financing Transactions. ESRB, Frankfurt (https://doi.org/10.2849/752255). 
• European Union (2012). Regulation (EU) No 648/2012 of the European Parliament and of the Council of 4 July 2012 on OTC derivatives, central counterparties and trade repositories. Official Journal of the European Union 55(L201), 1–59. URL: https://data.europa.eu/eli/reg/2012/648/oj. 
• European Union (2013). Commission Delegated Regulation (EU) No 153/2013 of 19 December 2012 supplementing Regulation (EU) No 648/2012 of the European Parliament and of the Council with regard to regulatory technical standards on requirements for central counterparties. Official Journal of the European Union 56(L52), 41–74. URL: http://data.europa.eu/eli/reg_del/2013/153/oj. 
• Goldman, E., and Shen, X. (2020). Procyclicality mitigation for initial margin models with asymmetric volatility. The Journal of Risk 22(5), 1–41 (https://doi.org/10.21314/JOR.2020.435). 
• Gurrola-Perez, P. (2020). Procyclicality of CCP margin models: systematic problems need systematic approaches. Research Working Paper, World Federation of Exchanges, London (https://doi.org/10.2139/ssrn.3779896). 
• International Swaps and Derivatives Association (2021). Covid-19 and CCP risk management frameworks. Report, January, ISDA Clearing Member Committee, New York. URL: http://www.isda.org/a/3jjTE/COVID-19-and-CCP-Risk-Managament-Frameworks-January-2021.pdf. 
• JP Morgan-Reuters (1996). RiskMetrics: Technical Document, 4th edn. Reuters Limited and Morgan Guaranty Trust Company of New York. URL: http://www.msci.com/documents/10199/5915b101-4206-4ba0-aee2-3449d5c7e95a. 
• Market Risk Advisory Committee (2021). Recommendations regarding CCP margin methodologies. Report, February, MRAC, Commodity Futures Trading Commission, Washington, DC. URL: http://www.cftc.gov/media/5706/. 
• Murphy, D., and Vause, N. (2021). A cost–benefit analysis of anti-procyclicality: analysing approaches in central counterparty initial margin models. The Journal of Financial Market Infrastructures 9(4), 27–50 (https://doi.org/10.21314/JFMI.2021.013). 
• Murphy, D., Vasios, M., and Vause, N. (2014). An investigation into the procyclicality of risk-based initial margin models. Financial Stability Paper 29, Bank of England, London (https://doi.org/10.2139/ssrn.2437916). 
• Murphy, D., Vasios, M., and Vause, N. (2016). A comparative analysis of tools to limit the procyclicality of initial margin requirements. Staff Working Paper 597, Bank of England, London (https://doi.org/10.2139/ssrn.2772569). 
• O’Neil, C., and Vause, N. (2018). Macroprudential margins: a new countercyclical tool? Staff Working Paper 765, Bank of England, London (https://doi.org/10.2139/ssrn.3284418). 
• Raykov, R. S. (2018). Reducing margin procyclicality at central counterparties. The Journal of Financial Market Infrastructures 7(2), 43–59 (https://doi.org/10.21314/JFMI.2018.106). 
• Szanyi, C., Szorodai, M., and Váradi, K. (2018). Supplementation of the regulation of anti-cyclical margin measures. In Proceedings of the 32nd European Conference on Modelling and Simulation, Tholen, C., Werner, J., and Wellhausen, J. (eds). European Council for Modeling and Simulation/Curran Associates (https://doi.org/10.2139/ssrn.3242078). 
• Wiessmann, P. (2020). Antizyklischer Kapitalpuffer zur Erhöhung der Widerstandsfähigkeit der Banken: Eine theoretische und empirische Analyse. GRIN Verlag, Munich. 
• Wong, L. W., and Zhang, Y. (2021). Procyclicality control in risk-based margin models. The Journal of Risk 35(5), 79–102 (https://doi.org/10.21314/JOR.2021.010). 

Download PDF

• Keywords
• CCP
• Initial margin
• Margin models
• Procyclicality
• Risk management
• Original research