Second meeting of the InvestEU Risk Expert Group - Minutes

Brussels, 3rd December, 2019

Participants: Adam Matanyi (MFB), Agnieszka Falkowska and Michał Łukaszewski (BGK), Jose Miguel Molina Foncea (ICO), Jaakko Louekari (NIB), Ray Mangan (SBCI), Soline Dumontier (CDC), Ondrej Mates (EIB), Nicola Baggio (EIB), José Abos (EIB), Steven De Gruyter (EIB), Denis Weerts (CEB), Michel Käsmacher (KfW), Lipo Jokinen (Finnvera), Barry McGrath (EIF), Edwards Olding (EIF)

Commission: CHIARION CASONI Giorgio (ECFIN), FISCHER Andreas (ECFIN), TANASA Alina (ECFIN), QUERO MUSSOT Antoine (BUDG), DE RYCKE Frederic (BUDG), STRIUNGYTE Egle (BUDG)

COM (Andreas Fischer) recalled that the primary objective of the exercise of the WG is to allow the EC to quantify and monitor the aggregate risk-profile of the InvestEU guarantee to ensure that the assets set aside in the Common Provisioning Fund with a provisioning rate of 40% will not be fully depleted during the lifetime of the guarantee, applying a confidence level of 95%. This requires in the first place aggregating risk information from the different implementing Partners that use different rating metrics and systems and secondly mapping the individual IPs rating scales to a common scale to ensure that rating outputs are comparable and commonly understood.

The mapping exercise is however not a straightforward process as a direct translation across rating categories is rarely possible. Andreas Fischer presented the mapping methodology described in the GESY Discussion paper as an example - notably in light of the clear description of the mathematical properties of the transition matrix between two rating scales, which were discussed. Andreas Fischer also pointed out that the likely absence of a pool of entities for which ratings on the IPs and the master scale are available is a clear limitation for the application to InvestEU of the approach presented. Finding a “common language” for the quantitative comparison of the individual risk assessments would require the IPs to provide COM with the upper/lower bounds for the Probability of Default (PD) on their scale for a given time-horizon. The majority of the potential IPs present confirmed they already reconcile their own scale with the Moody’s scale, which would therefore qualify as a master scale. Some of the IPs (BGK, MFB) however voiced concerns about disclosing detailed information on PD curves for confidentiality reasons. This issue could be overcome, if the IPs concerned directly provided Moody’s ratings, which would eliminate the need for mapping their scale to master scale.

It was agreed that Andreas Fischer will send a questionnaire to potential IPs by 20 December to collect inputs on the main features of the PD over 1 year (number and description of rating classes etc, including upper/lower boundaries of each rating class to avoid the necessity for a double-rated pool of entities by IP). Potential IP will then have the choice between mapping their internal system to the master scale according to the defined methodology or asking COM to proceed based on their inputs, possibly complemented by expert judgement in case of less advanced rating systems. In all cases, a bilateral discussion will follow to ensure that the IP’s internal ratings were appropriately mapped to the common master scale. In case PD information cannot be disclosed for confidentiality reasons, the Moody’s rating scale will be applied.

Regarding the Loss-Given-Default (LGD) values, there is neither an obvious common scale nor a commonly accepted standard LGD rate. Andreas Fischer presented the two available methodological approaches: (1) using the IP’s internally validated and reported LGD at face value, thus fully relying on the expert judgement and internal validation of the IP in line with the delegation framework or (2) using one-size fits-all LGD (based on a benchmark, for example, Basel’s 55% for senior debt and 25% for subordinated debt) to ensure a consistent and comparable treatment. Some IPs highlighted potential risks stemming from the delegation framework where there might be an incentive for the IPs to better rate their portfolio in order to benefit from a lower price of the guarantee, in absence of any quality control of the different (and non-comparable) LGD outputs. The EIB supported the delegation model and emphasised that as long as an alignment of interest between the COM and the IP is ensured, the latter will rate entities as if the guarantee was not there as they will have no incentive to “cheat” on a systematic basis. Furthermore, any other LGD model departing from the one in use by the single IPs is prone to raise additional costs to participate in InvestEU. Other potential IPs recalled the Risk department of any bank is first liable to its Management Board, so “cheating” is not a real option. It was agreed that the delegation model (Option 1) is to be preferred.

Regarding the note on the Risk Methodological Framework circulated to potential IPs for comments, the EC (Giorgio Chiarion Casoni) clarified that:

• The quota of revenues pertaining to the Commission as risk related remuneration for risks taken under InvestEU, together with recoveries, will feed the InvestEU Guarantee Fund in addition to the 40% budgetary contribution, .
• Thematic products should be considered the exception rather than the rule under InvestEU as they are very expensive in terms of capital consumption,
• The COM risk appetite is to be understood in relation to the provisioning rate (40%) and the applied confidence level (95%). The assets in the InvestEU compartment of the Common Provisioning Fund are set in relation to the Value-at-Risk of the portfolio loss distribution at a confidence level of 95%,
• Portfolio correlation (typically assumed in the range of 20-40% for a normal loan portfolio) will be assessed prudently based on a simple approach.
• The maximum guaranteed amount for each Guarantee Agreement under Invest EU will be set in euros. Some potential IPs (BGK, MFB, CEB) pointed to the currency risk related to non-Euro operations as an additional risk/uncertainty and the potential costs associated to hedging activities applied to mitigate currency risk. This may need to be addressed to ensure a true level playing field between different geographies (EA vs non-EA) of InvestEU operation. Giorgio Chiarion Casoni recalled that the InvestEU Regulation foresees to cover such a risk within the limit of art 15.2 for the losses arising from currency fluctuations. The Commission will analyse this issue in further detail in the coming months.

Concerning the note on Remuneration of the EU guarantee, which was also circulated for comments at the previous meeting, the majority of potential IPs expressed a preference for a fee type remuneration model, which is easier to manage operationally than a revenue sharing model . The EIB proposal to apply the fee model to layered structures and a revenue sharing type model in case of pari-passu investments received limited support in view of the additional costs that would inevitably result for IPs who do not already have the required IT systems in place. A separate discussion should be held for equity-type investments. Regarding the timing for the payment of the fee, and considering that the exposure at risk before disbursement is zero, COM clarified that this point can be indeed addressed in the Guarantee agreement when defining the start date of the guarantee fee.

The next meeting will tentatively take place in the first half of February 2020.