§ 8 of the introduction states correctly that one of the issues behind the crisis of 2007-2009 was that “complex transactions have been assessed by external rating agencies using erroneous modelling assumptions”. More precisely, the main modelling error that led to the disastrous consequences that we know was the assumption that the various assets in the securitised vehicles had no, or a very low, correlation. This assumption made for the illusion that securitised vehicles were safe investments. The reality is that correlation is a very unstable variable in financial markets and that, in the case of a crisis, different assets that had shown until then no or little correlation can suddenly become highly correlated. To this day, this erroneous assumption of a low correlation between assets is still used by credit rating agencies in their models. Nothing has changed on this issue since the crisis of 2007-2009. As a matter of fact, this assumption is the root of the mathematical illusion that the slicing and dicing financial engineering at the heart of securitisation creates investable assets with an improved risk/reward ratio. As a consequence, there is no doubt that the same causes would provoke the same consequences as we saw in the 2007-2009 crisis: a strong external shock could trigger again the default of many highly rated securitised investment vehicles.
The principle of distinguishing between the regulatory treatment of ‘qualifying’ securitisations and that of other securitisations (§ 9) makes sense, everything else being equal. As the principle for distinguishing those two types of securitisations is based on their level of simplicity, standardisation and transparency, the heart of the question asked in this consultation is really to determine whether a synthetic securitisation can be deemed to be simple, transparent and standard. We will show throughout this response to the Discussion Paper that calling a synthetic securitisation simple, transparent and standard is a contradiction in terms. In a nutshell, STS synthetic securitisation is a misnomer.
§ 14 infers that one of the reasons for not applying STS criteria to synthetic securitisation in the past was the lack of available data on market developments, volume and historical performance. This has not changed in a significant manner. Remarkably, the data showing better default performances for lower rating grades referred to in the Discussion Paper has not been tested during crisis times. From an economic standpoint, using data that has the double characteristic of being both very recent and untested in crisis times (when the erroneous low correlation assumption degenerates systematically, as explained above) makes the interpretation of the results shown useless. Here again, it is a grave risk-management mistake to infer crisis times’ correlations from normal times’ correlations. All financial risk-managers know that this is a recipe for disaster.
The introduction (§ 16) is correct in asserting that “the market of synthetic securitisation... has traditionally been characterised by issuance of bespoke transactions, i.e. it has not been standardised” and “it has been implemented in accordance with a wide spectrum of practices and was perceived at the time to increase the structural complexity”. By contrast, § 17 is wrong to infer that this was due to the lack of systematic data or analysis available “at that time”: the bespoke and complex nature of synthetic securitisation has nothing to do with the availability of data or of analysis. It is inherent to its very nature. The reasons why the STS framework was considered as impossible to apply to synthetic securitisation when it was developed and adopted (i.e. as recently as May 2017) are still entirely valid only two and a half years later: synthetic securitisation transactions are by nature bespoke, complex and not transparent. The reasons given in 2017 for not applying the STS framework to synthetic securitisation are as valid as ever.
Interestingly, § 31 states that the reason why “banks applying standardised approaches to credit risk are rarely originators of synthetic securitisations...is mainly due to reluctance to enter a largely unstandardised/bilateral market”. Is not this simple sentence an admission by the consultation itself that the market for synthetic securitisations is, by nature, not standardised, not to mention simplicity and transparency? Believing that the introduction of an STS framework would change this situation is highly unrealistic.
§ 32 explains the main motivation behind synthetic securitisation, i.e. arbitraging capital requirement regulation coming, among others, from the Basel III framework, the introduction of the output floor, IFRS 9 and the fundamental review of the trading book. This means that circumventing regulation is an essential motive for promoting STS synthetic securitisation. By construction, if STS synthetic securitisation enables market participants to roll out transactions with a similar, and even an increased level of risk given their additional levels of complexity and credit risk, with less prudential constraints, it will lead to higher financial systemic risk.
Finance Watch finds surprising that policy-makers should propose to market participants to put in place an STS framework for synthetic securitisation with regulatory arbitrage as a rationale. This leaves no doubt that applying the STS regulation framework to synthetic securitisation will lead to more risk with less control of the risk, to the detriment of financial stability.
The analysis of the historical performance based on the data provided by S&P is unambiguously biased to give a favourable view of synthetic securitisation. This data is impossible to interpret with rigour given the following methodological problems: 1) the analysis is applied to rated transactions only when the previous section (§ 42) explained that since the crisis most synthetic securitisation transactions have been unrated; 2) the concept of “lifetime default rate” is too vague to be of any use to interpret meaningfully, in particular when no indication is given on the lifetimes; 3) no information is given on the volumes considered and only the number of transactions is counted. Those different flaws prevent us from giving any economic sense to the data provided.
The analysis on the historical performance based on the data provided by IACPM brings no useful information to assess the performance or the resilience of the market as it considers only transactions originated in 2017 and 2018. Not only assessing the performance of any asset class with so little data and on such a short period of time is meaningless, but testing such complex structures outside of times of crisis is an exercise that brings no useful information whatsoever.
On the question of the changing regulatory environment (§ 62), Finance Watch would like to point out that the Securitisation Regulation has been in place since 1 January 2019 only (less than one year), and that it is far too early to draw any significant lesson from its implementation, notwithstanding the idea of amending it already.
Finance Watch also opposes the assertion of the consultation (§ 68) that “the developments in the last few years have strengthened...the relevance of an STS regulatory framework”. This affirmation relies on no strong evidence or demonstration.
§ 70, 71, 72, 73 and 74 offer a good description of the reality of the complexity of synthetic securitisation transactions and of the fact that those transactions can under no circumstance be considered as simple, standardised or transparent. A particularly important point is the fact that synthetic securitisation gives rise to two levels of credit risk, as counterparty risk with the originator arises on top of assuming an already complex credit risk in which a correlation risk is embedded in the underlying credit exposures.
Assertion of § 78 that “there is no evidence that would suggest that synthetic securitisation structure inherently results in higher losses than traditional securitisation structure” relies on no reliable data covering crisis times or low credit cycle phases.
Assessment of pros and cons:
Finance Watch regrets the argument of § 81 in favour of synthetic securitisation (Increasing relevance of the product in the context of regulatory developments): as stated above (answer to question 2), we are witnessing the situation of a regulator offering regulated entities a route to circumvent regulation. This creates a conundrum on the consistency of pros and cons arguments, in particular in the light of the statement in § 113 of the Consultation Paper that the STS framework contemplated for synthetic securitisation would not be Basel compliant.
The argument of § 85. b. that synthetic securitisation offers “an instrument for hedging of tail risk in economic downturn” is in contradiction with the assertions made higher of a good performance of synthetic securitisation transactions. By definition, if synthetic securitisation enables originators to hedge their tail risk, it will result in losses for investors in extreme conditions (it is the very definition of “tail risk”). The (correct) insistence on tail risk shows that the risk considered is not the risk incurred in normal times but in times of crisis. This reinforces the arguments developed earlier in our response to this consultation that synthetic securitisation transactions convey with them a particularly high risk that reveals itself in times of crisis. As for the argument that “it can enable the risk transfer from bank to non bank entities and hence facilitate allocation of risk more widely within the Union financial system”, it sounds like a good description of what happened during the 2007-2009 crisis and seems, surprisingly, to be calling for a repetition of the worst financial crisis the world has ever experienced. We dare hope that this is not the objective. All in all, synthetic securitisation is a factor of acceleration and amplification of financial crises.
The argument of § 85. c. in favour of the “notable potential to free up originators’ balance sheets to allow further lending to the economy” is doubly flawed: the “freeing balance sheets” part takes us back to the “viva regulatory arbitrage” argument already pointed out, and the “further lending to the economy” is unfounded as there is no lack of lending capacity to the economy in the European Union. If anything, the European Union economy is over-banked, not under-banked. Sound economic projects can find the bank lending that they need. Importantly, given the costs and the margins inherent to synthetic securitisation transactions, the yield given to synthetic securitisation investors is, by construction, lower than the yield perceived by a bank lending to a corporation, large or small, and keeping the loan and its risk on its balance sheet. The last argument of § 85. c. “This may be relevant for the revival of SME lending” reveals also an internal contradiction of the Discussion Paper: on top of the fact just mentioned that there is no shortage of lending capacity in the EU, SME lending in the EU is essentially made by small and medium size banks (as opposed to large banks who devote often as little as 3% to 5% of their balance sheet to SME lending among a paltry 12% to 15% devoted to lending to non-financial corporations). As pointed out in § 31 of the Discussion Paper, synthetic securitisation is not being done by banks applying Standardised Approaches to credit risk (i.e. small and medium size banks), but by those applying Internal Rating-Based models (i.e. large banks). In other words, the Discussion Paper shows that synthetic securitisation is not made by the banks that lend mostly to SMEs (small and medium size banks) and suggests at the same time that STS synthetic securitisation may be a factor of revival of SME lending. This is not coherent.
Among the cons of the development of STS synthetic securitisation, the argument is developed rightly in § 79 that the STS framework might be perceived as a label of quality, and that it could therefore mislead investors. This would be inevitable and it would create a high responsibility for policy-makers and regulators that made this framework possible and thus amplified systemic risk.
Another cons argument, not mentioned in the Discussion Paper, is the impossibility for NCAs to monitor in real time the transactions and therefore to perform their supervisory role in an effective manner on those structures. In the best of cases, they will have an ex-post annual view of the transactions realised in their respective markets, but under no circumstance with the granularity that would enable them to understand thoroughly and ex-ante the risks piling up in the market. In other words, supervisors are running blind on this issue, and embedding synthetic securitisation in an STS framework would only make matters worse.
§ 92 takes again as a pro argument a supposedly increased banks’ lending capacity whereas, as already stated, there is no shortage of such capacity in the EU. It affirms also without any demonstration, and contrary to the obvious conclusion that even a superficial analysis of the market would lead to, that regulatory recognition of the STS product would increase stability. Nothing could be further from the truth.
No. The “originate to distribute” model has demonstrated the danger it represents during the 2007-2009 crisis and nothing can justify a preferential capital treatment of such a model.
Notwithstanding the fact that the criteria on simplicity are, once again, a contradiction in terms given the level of complexity described (credit risk at two different levels, tranching of credit risk at the first level, interest rate risk, currency risk, legal risk in the definition of an event of default, etc.…), these criteria are characterised by 1) wishful thinking (e.g. criterion 1 “robustness of credit protection”), 2) vagueness of concept (e.g. criterion 4 “homogeneity in terms of asset type”) and, most importantly, 3) an impossibility for NCAs or, as the case may be, ESAs to perform their supervisory duty (e.g. criterion 8 “in the case of securitisations where the underlying exposures are residential loans, the pool of loans should not include any loan that was marketed and underwritten on the premise that the loan applicant was made aware that the information provided might not be verified by the lender”, or criterion 13 “The underlying exposures should have been underwritten on the basis that their repayment was not intended to be predominantly reliant on the refinancing of such underlying exposures or on the resale value of the assets that are being financed by those underlying exposures”). There are two reasons for this impossibility: supervisors have neither the data nor the human resources to perform that duty on the synthetic securitisation market. This is most problematic as a regulation that cannot be enforced at supervisory level is, by nature, useless.
The reasoning given in the answer to question 7 on simplicity applies also to question 8 on standardisation.
One more example can be given to illustrate the impossibility of supervisors to perform their duties when it come to synthetic securitisation: in the rationale for criterion 15, it is said that “Derivatives should be allowed as underlying exposures of a synthetic STS securitisation only where those derivatives are used for the single purpose of hedging the currency and interest rate risk arising from the underlying exposures that are not derivatives”. This rule is not only a perfect example of wishful thinking, but also a school case of a situation that supervisors will never have the possibility to control and enforce. Many other examples of such situations could be given.
The reasoning given in the answer to question 7 on simplicity and to question 8 on standardisation applies also to question 9 on transparency.
One additional point can be added however:
The rationale for criterion 25 mentions that “the audit prior to issuance… should be carried out with a confidence level of a least 99%”. This wording infers that the underlying statistical distribution of synthetic securitised assets is either normal or lognormal, when, as experience has shown, such statistical distributions do not describe properly the behaviour of financial assets during crises, not to mention the behaviour of highly complex securitised assets. Moreover, using a reasoning founded on confidence levels is not conceptually coherent with the tail risk argument developed in § 85. It is essential that policy-makers, if and when they rely on mathematical concepts, be coherent and rigorous, and avoid giving to themselves a false sense of security by the mere fact that they refer to statistical concepts that become meaningless if and when not applied properly.
We encourage policy-makers to read again the specific criteria for synthetic securitisation and ask themselves whether they really believe that those criteria can be considered under any stretch of the imagination as simple (cf. discussion on credit events, the description of interim credit protection payment or the fact that “the credit protection agreement establishing the synthetic securitisation should be structured as contingent premiums” (criterion 32), the recognition that “full work out of losses can be a lengthy process” (criterion 31 / several years), or the extreme complexity and uncertainty of the interaction with BRRD in case of bankruptcy of the originator (criterion 34)).
A STS synthetic securitisation framework should not be introduced altogether, as it would increase significantly financial systemic risk without any benefit to the real economy. This question is therefore not relevant in our view.
If and when a regulation has negative consequences for the public good, the question of the level playing field is irrelevant. The STS framework would be such a regulation; therefore level playing field arguments should not be part of the discussion.