To discourage the “originate-to-distribute” business model (which created the US subprime mortgage securitisation industry and caused the recent financial crisis) policymakers require that originators retain sufficient risk to have so-called “skin-in-the-game”. In defining retention rules, European policymakers settled on the wording of “not less than 5%”. The 5% value was selected, instead of a higher percentage such as 10% or 20%, because of constraints imposed by the international accounting standard, IAS 39. IFRS 9, which will replace IAS 39 in 2018, will similarly constrain what is feasible. This short note explains how accounting constraints drove the choice of 5% and explains why these constraints will apply under IFRS 9.
1. Introduction
Current proposals for a high securitisation retention such as 20% are not new in banking regulation. Such proposals were discussed in 2009-2010 when retention was central to policymaker preoccupations.
At the time, in discussing retention rules, policymakers were trying to decide on an appropriate degree of “skin-in-the-game” that would reinforce appropriate incentives for securitisation originators. A range of possibilities was discussed, including percentages around 4-8% and higher figures such as 10% or 20%.
In finalising retention rules, however, policymakers were ultimately obliged to take into account what was feasible under accounting standards and, in particular, the International Accounting Standards (IAS) that are applied in Europe (although not in the US).
Bank securitisation of balance sheet assets is governed by IAS 39 dealing with “financial assets”. Under IAS 39, the accounting decision tree (see Appendix) for a securitisation is as follows:
1) [§ 17(a)] Have the rights to the cash flows from the asset expired?
The answer is normally “No”, as the securitised assets generate interest and principal cash flows. If “No”, go to § 18(a).
2) [§ 18(a)] Has the entity transferred its rights to receive the cash flows from the asset?
The answer is normally “Yes”, as the entity, i.e. the bank originator, will not receive the cash flows from the securitised assets because that right now belongs to the securitisation SPV. If “Yes”, go to § 20(a).
3) [§ 20(a)] Has the entity transferred substantially all risks and rewards?
This is where things are complicated by accounting rules.
The need to be consistent with those rules was the main technical factor that led to the adoption of the “not less than 5%” retention rule. IAS 39 does not define what “risks” or “rewards” mean[2]. Nor does it state what “substantially all” means. This is because the accounting culture in Europe is “principlebased” (unlike the US approach which is “rule-based”).
So, what is “substantially all”? The question has been widely debated by academics, policymakers and accountants. Under US GAAP practices, the term “substantially all” is generally viewed by US accountants as meaning “90% or more”. Over the years, this threshold was also accepted by European accountants when they interpreted the IAS norms, in particular IAS 39 § 20(a).
Being able to answer “Yes” to the question § 20(a) would enable a bank originator to derecognise the securitised assets (see Appendix), i.e. to remove them from the bank’s balance sheet. However, if the skin-in-the-game retention is “10% or more”, a bank originator would have to answer “No” to this question. A “20% retention” of each tranche, which was mentioned in some circles in 2009-2010, would clearly fall into that category, making the accounting derecognition highly unlikely.
2. Does it matter if banks cannot derecognise the assets from the balance sheet?
In 2009-2010, policymakers viewed their “skin-in-the-game” objectives as a) keeping in check the US originate-to-distribute business model and b) ensuring that the interests of bank originators were aligned with those of investors.
With a 20% retention (as an example of “more than 10%”), a bank could still derecognise the securitised assets in two ways (see Appendix): i. A bank could avoid going through IAS 39, § 20(a), by avoiding IAS 39 altogether. This could be achieved by not recognising the assets in the first place! Specifically, a bank could create a series of funds which would acquire the assets directly at the time of origination. These assets would, therefore, never touch the bank’s balance sheet. Such practices would push the banking system towards an accelerated, at arm’s length, originate-to-distribute business model, exactly what policymakers have wished to avoid both in 2009-2010 and more recently. ii. A bank might be able to answer “No” to the question IAS 39, § 20(c): “Has the entity retained control of the asset?” For this, the bank originator would have to lose control of the assets. So, with a 20% retention constraint, a bank could only derecognise its balance sheet assets if it could adopt a “fire-and-forget” origination model. This seems highly inadvisable and again represents a business model that policymakers would not wish to encourage.
So, the retention threshold of 5% (expressed in the wording as “no less than 5%”) has enabled banks both in the US (with US GAAP) and European banks (with IAS norms) to size their retention in a way that would avoid failing the derecognition requirements.
3. Retention options
We now turn to the forms that retention can take. The European Regulation 648/2012, Article 405, defines several different retention options. (This was later clarified by the Commission Delegated Regulation 625/2014.) These retention options may be summarised as follows:
(a) retention of “no less than 5%” pro rata in each of the tranches sold or transferred to investors (or vertical tranche equivalence or vertical asset allocation equivalence);
(b) retention of the originator's interest of “no less than 5%” of the securitised revolving exposures;
(c) retention of randomly selected exposures in a pool of at least 100 exposures, equivalent to “no less than 5%” of the nominal value of the securitised exposures;
(d) retention of the first loss tranche and, if necessary, other tranches having the same or a more severe risk profile than those transferred or sold to investors and not maturing any earlier than those transferred or sold to investors, so that the retention equals in total “no less than 5%” of the nominal value of the securitised exposures;
(e) retention of a first loss exposure “not less than 5%” in every securitised exposure.
Under IAS 39, Option (a) is clearly constrained by the accounting derecognition threshold described above in that the “no less than 5%” is capped by the “substantially all” interpretation to be “no more than 10%”.
Under Option (b), the way the securitisation is structured makes a difference, as the assets backing the originator’s interest and the investors’ interest can be segregated or not. If they are not segregated, the “substantially all” derecognition requirements will be relevant. If they are segregated, the retention level in itself is not an accounting issue, but a level-playing-field issue, with other forms of retention or other forms of securitisation structures.
The randomly retained assets under Option (c) do not follow the IAS 39 treatment at all, as they have not been transferred into the SPV in the first place. Under this form of retention, it is conceivable to ask for more than 10%, even 20%! But, lawmakers should make sure that such retained assets can be transferred into fully consolidated structures, so as not to restrict the bank’s way of funding itself. Without such flexibility, a bank might switch the funding mechanism of the assets from securitisations to covered bonds, with all the balance sheet problems and asset encumbrance issues that this creates.
Option (d) does not make sense for a bank that wants to derecognise the securitised assets and reduce the size of its balance sheet. On the other hand, it could make sense for an asset manager that originates managed securitisations and is not constrained by asset consolidation issues.[3] Finally, at the other end of the balance sheet management spectrum, some might like to claim the entire portfolio on their balance sheet, when their [only] asset (i.e. their investment) is only 5% of it[4]…
Option (e) is too complex to be handled in this paper. As long as it is not the main retention mechanism used by banks, it does not create a systemic risk.
4. Impact of IFRS 9 on retention
IFRS 9 will replace IAS 39 as of January 2018 and the final text has already been published. Under it, derecognition thresholds are unchanged compared to IAS 39. The 90% value for having “substantially” transferred “all risks and rewards” will continue to apply. It is explicitly stated in paragraph IFRS 9 B3.3.6 of the section Derecognition of financial liabilities.
Nevertheless, IFRS 9 does introduce one major difference compared to IAS 39 that will have an impact on retention. This is the accounting treatment of marked-to-market versus amortised cost for the retention piece. For “contractually linked instruments” which includes securitisation tranches, the section B4.1.21.C of IFRS 9 specifies the rules under which the tranche can be booked at amortised cost. This is possible only provided that the expected loss of the securitisation tranche is not higher than the expected loss of the underlying pool.
The consequence of this rule is that retention in the form of first loss tranches (see Option (d) in Section 3) will automatically be accounted for under mark-to-market accounting. Increasing the level of retention will therefore increase exposure of bank originators to market volatility in their financial results. An unintended consequence is that this will contribute to further instability in the financial system.
5. Conclusion
Proposals of high securitisation retention levels, such as 20%, are not new in in banking regulation. Policymakers examined this issue in 2009-2010 but were constrained in their choice by the then international accounting norms, in particular IAS 39. Lawmakers in Europe cannot influence[5] these norms as they are decided by an independent board (IASB). Obliged to follow an approach consistent with accounting rules, European policymakers in the past settled on a minimum 5% threshold. The room left between the derecognition cap of 10% and the retention floor of 5% enables bank originators to provide additional financial products or ancillary services (such as servicing or collateral administration) to securitisation vehicles.
The accounting norms are evolving again with the introduction of IFRS 9. IFRS 9 will increase financial instability through the requirement of mark-to-market assessment of retained first loss tranches (see Option (d) in Section 3 above). Any increase from today’s 5% retention level in this form of retention will compound the problem mechanically increasing financial instability further.
Not all forms of retention are affected equally by the accounting rules. Theoretically, different retention thresholds could be set for different options of retention, therefore, the same logic means that an across-the-board threshold will have different accounting impacts depending on the form of retention. For example, a unique threshold of 20% will disable the balance sheet management of securitisations using the retention form that consists of holding a percentage of each tranche (see Option (a) in Section 3).
The consequence of piling yet more regulatory constraints on securitisation instruments is likely to drive banks to other instruments. For example, if retention levels in the case of randomly selected assets (see Option (c) in Section 3) are increased too high, covered bonds will dominate securitisations, complicating monetary policy and weakening banks’ ability to issue unsecured funding because of balance sheet encumbrance.
In any case, in the re-opened debate on retention levels, we strongly advise policymakers to design rules that are consistent both with the intended consequences and, current and future accounting practices.
Appendix
Figure: IAS 39 Accounting Decision Tree for a Securitisation
[§ 15] Consolidate all subsidiaries (including any SPE)
[§ 16] Determine whether the derecognition principles below are applied to a part or all of an asset (or group of similar assets)
Yes [§ 17(a)] Derecognise the asset Have the rights to the cash flows from the asset expired?
No
[§ 18(a)] Has the entity transferred its rights to receive the cash flows from the asset?
Yes No No [§ 18(b)] Has the entity assumed an obligation to pay the cash Continue to recognise the flows from the asset? asset
Yes Yes [§ 20(a)] Has the entity transferred Derecognise the asset substantially all risks and rewards?
No Yes [§ 20(b)] Has the entity retained Continue to recognise the substantially all risks and rewards? asset
No No [§ 20(c)] Has the entity retained control of the asset? Derecognise the asset
Yes Continue to recognise the asset to the extent of the entity’s continuing involvement
Notes
[1] Duc Dam Hieu, Dr. Georges Duponcheele and Alexandre Linden work in the Asset Finance and Securitisation team at BNP Paribas. Dr. William Perraudin is Director of RCL and Adjunct Professor of Imperial College, London. Correspondence should be addressed to the authors at duc.damhieu@bnpparibas.com, georges.duponcheele@bnpparibas.com, alexandre.linden@bnpparibas.com, william.perraudin@riskcontrollimited.com. Any errors are ours. The views expressed are the authors’ own and not necessarily those of BNP Paribas.
[2] One of the roles of the securitisation structurer (who puts together a transaction for an originator) is to calculate the accounting “risks” and “rewards”. The accounting “rewards” are normally understood to be the interest cash flows (excluding expected losses and excluding funding costs) derived from the securitisation tranches, plus accounting “rewards” associated with other embedded instruments. The accounting “risks” are normally understood to be the distribution of the expected losses among the securitisation tranches, plus the accounting “risks” associated with other embedded instruments. Accounting “rewards” from ancillary services, such as servicing or collateral administration, when not priced at arm’s length, are also taken into account. Once all accounting “risks” and “rewards” are determined, the structurer determines the proportion that has been transferred, and whether this proportion satisfies the “substantially all” test in § 20(a).
[3] This form of retention enables managers to take the most risky, most yielding, portion of the capital structure. Whether this provides a true alignment of interest with investors is open to debate, as it can produce counter-beneficial effect of yield maximisation via risk optimisation, leaving investors exposed when the portfolio is under stress with no one whose interest is aligned. Increasing the retention to above 5% will lead to yield reduction, and managers will push for yield optimisation via risk maximisation… Not such good idea. In other words, a wiped-out asset manager is in nobody’s interest.
[4] The going concern accounting rules can be misleading for creditors, but this is another debate…
[5] We do not think this is necessarily a good thing as the damage of “mark-to-market” so much favoured by the accounting rules (and reinforced in IFRS 9) is one of the few lessons of the crisis that has not been fully addressed by policymakers.