Margins of Conservatism
Margins of conservatism (MoC) is the concept where a model development team shall try to measure the model impact from any remaining deficiencies after having deployed methods and techniques for adjusting for inconsistencies- and lack of accuracy in historical data. Conservative margins has always been part of credit risk models and regulatory expectations, but with the detailing of MoC-expectations in the Guidelines on PD estimation, LGD estimation and the treatment of defaulted exposures (EBA/GL/2017/16) a more distinct and clear view has been presented by the European Banking Authority on what the focus and ambition shall be with respect to MoC.
Below are some reflections from Henrik Nilsson on the work that is being done in the Nordic banking sector by smaller as well as larger IRB- institutions with respect to the challenges and trade-offs that are embedded when establishing a MoC-framework:
Mixing quantitative and qualitative approaches: It is noticed that smaller institutions tend to strive towards more quantitative based MoC-frameworks in comparison to larger institutions. This is in turn a consequence of smaller institutions having more homogenous and retail focused portfolios where the data, specifically default data, is ample. For larger institutions with a mix of corporate and retail portfolios more qualitative based approaches can be observed. The trend is not without trade-offs and where the smaller institutions face less flexibility in the MoC-review process while the larger risks ending up with a too subjective framework not being able to manage expectations on objectivity. There is still a hesitation amongst larger institutions in establishing separate frameworks for retail based- and corporate portfolios.
Balancing institutional consistency and flexibility: With size and geographical reach there is an increasing need to ensure a consistent framework implementation where the resulting MoC would be the same irrespective of which model development team that is performing the assessment. The counterbalance to the requirement on consistency is, of course, the need for flexibility due to the idiosyncrasies in the historical evolution of the portfolios and the IT infrastructure under which risk data has been collected. Smaller institutions may in turn end up with a high degree of flexibility but where long term employee turnover may result in inconsistent MoC-assessments over time. Related to the methodological soundness (see below) is the risk of establishing to many measurement options running the risk of initiating a dialogue regarding cherry-picking.
Analytical depth and aggregation techniques: With the establishment of more subsets for each MoC-category the analytical depth of each deficiency increases. This may however come at the cost of adapting evaluation criteria’s which does not fit all application portfolios. In addition, if the evaluation criteria are overlapping other aggregation techniques than the straight summation may be required risking applying scientific methods where there is no science.
Methodological soundness: A successful MoC-framework need to provide the model development team with a range of tools that can be deployed depending on the specific nature of the portfolio. Applying variance-based methods on low default portfolios may cause the paradox to arise that these portfolios are viewed as more risky in comparison to the other (non-LDP) portfolios within the institution. Not defining specific measurement approaches may on the other hand expose an institution to regulatory risks and where post-application (applying for the approval to use the MoC-framework) corrections may result in unwanted surprises in terms of capital costs. I.e. it is critical to assess the soundness of the proposed methods, by category of portfolio, during the establishment of the MoC-framework.
If you are interested in discussing or benchmarking your MoC-framework please contact Henrik Nilsson at firstname.lastname@example.org or +46 766 35 19 37