The concept of an economic downturn was designed to address potential adverse macroeconomic conditions within estimations of the Internal Ratings-Based (IRB) risk parameter models. The regulatory definition of a downturn was first introduced in the earliest Basel II regulation, however, there were no strict rules and no detailed guidance on how the downturn period should be identified, and how a downturn adjustment should be estimated. The whole regulatory and risk management community within the European Central Bank (ECB) regulated single market spent significant time developing and updating the downturn methodology of their respective financial institutions. In this blog article we will talk about the latest regulatory requirements on this topic and will discuss the impact of a downturn period on the main components of the LGD Model.The publications by European Banking Authority (EBA) prescribing more detailed guidance on how a downturn period should be incorporated into Credit risk models is originated from two main references:
- Final Draft Regulatory Technical Standards, EBA/RTS/2018/04, as of 16 November 2018: contains mainly the guidance on the downturn period definition
- Guidelines for the estimation of LGD appropriate for an economic downturn, EBA/GL/2019/03, 6 March 2019: describes main approaches for downturn adjustment quantification
However, in the recent history of some countries, based on the approach described in these regulatory guidance, the downturn period has never happened or when it happened there was no real impact on the observed risk parameters. Similarly, when such a downturn hit the economy of the respective territory it was such a long time ago that the bank did not have such loan books by that time. Therefore, the downturn adjustments for such portfolios were estimated to be either very small or were practically not existing. However, taking into account the current turbulent times and considering various macroeconomic forecasts, we are currently expecting a serious downturn period with some grave consequences for almost all industries. For some countries, such a downturn period might be extremely severe compared to the crisis they have experienced in the past.
Therefore, it would be useful to conduct a benchmarking study, and to get at least a high-level understanding to what level the risk parameters can move during the expected next downturn period. According to the regulation, downturn adjustments are usually performed for the estimation of the LGD and EAD risk parameters. In this blog article we will focus on the LGD risk parameter and its components as it is generally considered to be of a more complex nature in its estimation methodology than that of the EAD risk parameter. Or in other words, the calculation methods are demanding a high granularity of data and include several steps. For the PD risk parameter, an explicit representation of the downturn as an add-on has not been foreseen, though at least one downturn period is expected to be covered by the default observation period to reflect its impact on the observed through the cycle default average.
In one of the previous articles on this Regulatory Blog (IRB Loss Given Default Modelling: Risk Differentiation Function, as of 4 May, 2020) it was already discussed that according to the best practices of LGD modelling, we usually assume several components as structure of an LGD model:And, these main LGD components driving the final LGD estimation are:
- Probability of cure event – P (Cure) or PC. It is also useful to consider PC at different points in time in default vintages because the behavior of cure rates on short defaults and long default is expected to be different during a downturn period. By short defaults it is meant, that a client was in default not for a significant amount of time (usually from a couple of months till a couple of years). However, this definition is not very strict and very much depending on the portfolio under the analysis;
- Loss given the fact that the default cured – Loss (Cure) or LGC. Loss in case of cure event is usually not modelled, because the value is rather deterministic, and usually explained by business processes rather than by factors used for modelling. In most of the cases, the value is estimated as a value slightly higher than 0;
- Loss given the non-cure event – Loss (Non-Cure) or LGNC. LGNC is also worth considering at different points in time in default vintages.
These listed components are explicit components. However, there are also some implicit drivers that might have significant indirect impact on the overall estimation of the LGD value. For example, to make the sensitivity assessment more detailed and sounder, it is useful to consider such factor as Time in Default as it is driving indirectly the LGD parameter through the application of the maximum work out period and the estimation of expected recoveries.
Probability of cure event
It is worth distinguishing cure rates related to the short times in default (1, 2 years) and the longer ones (3, 4, …, till default lifetime). The behavior of such types of cure rates might be different should adverse macroeconomic conditions occur. Usually, the biggest impact (the ratio of this parameter’s value in downturn period to the parameter’s long-run average) is coming from the short-term cure rates. The reason for that is that the defaults occurring during a downturn period are very unlikely to cure in the first two years upon the start of the default event as borrowers are still affected by the on-going crisis and the sources of income are still not revived. On the longer time horizons as soon as the adverse conditions are softening (for example, for retail portfolios in the Northern European market, it has been observed that the recoveries were improving significantly from around the 3rd year in default onwards) the impact of downturn might turn to become benevolent, though still being visible.
For many portfolios, the short-term cure rate reflects the highest impact during the adverse macroeconomic conditions among all LGD components.
Loss given cure event
By and large, the loss given cure event is a deterministic value and it is usually not modelled. However, it does not mean that there is no change for this parameter in times of crisis. For example, depending on internal credit policies of a bank, bank might soften its collection processes and might allow more insignificant write-offs to occur during a downturn period. In other words, if a bank previously was writing-off outstanding dues below 50 EUR, the bank can increase this threshold, for example, up to 100 EUR. In this way, the bank will support with such actions its clients, and it will not use up collection resources on small exposures dues and focus on the larger exposures due. This will lead to an increase of the observed LGC parameter in downturn periods.
Loss given non-cure event
Similar to the Probability of cure event, for the Loss given non-cure event it is useful to distinguish between the short and the long term LGNCs. Recoveries (1 – LGNC) are very unlikely to happen at the start of the default, however, they should expectedly increase as soon as the adverse conditions ease off. Based on the various approaches, it is useful to assess LGNC parameter not based only on resolved cases but also to analyze the dynamics of loss caused through the incomplete defaults.
Time in default
Time in default is not included explicitly in the LGD formula above, however, it has been still having indirect impact on the final loss given default estimation. Time in default is closely interrelated with the time through when recoveries are collected. Moreover, it influences the estimation of maximum workout period and as the result the process of augmentation of incomplete recovery cases.
It is also considered to be a widely used practice not to jeopardize the process of secured loans collection in times of adverse economic conditions. Most of the banks are waiting with the liquidation of the underlying collateral because selling the underlying collateral during the downturn period will not yield a favorable liquidation value. The key reason for this is the common observation that during the recession most of the collateral prices are usually decreasing. And sometimes the loss of collateral could be very dramatic. Consequently, banks can tend to hold on for a couple of years until the collateral values are improving at least partially comparable to the prices before the start of the adverse macroeconomic conditions.
What does that mean for banks using the IRB approach?
Summarizing our discussion, the expected impact on the LGD estimation of the downturn conditions are varying by the LGD components. The final impact depending on the time horizon is presented in the table below:
The macroeconomic crises are exogeneous shocks and cannot be forecasted. Downturn effects are significantly depending on the various factors such as location of the country (Southern Europe or Northern Europe), the type of the portfolios and the structure of the collection processes. Therefore, in most of the cases the value projections of the LGD parameters during the course of the adverse macroeconomic conditions should follow the previously described dynamics, though in some specific cases the LGD parameter value projection might deviate from what has been observed in practice.
For further enquiries
PwC can support you with Downturn identification and quantification under the latest regulatory framework. Our experts have many years of experience in designing IRBA rating systems and implementing IRB 2.0 requirements. Should you have further questions please contact the Quant team by the email addresses below.
Phone: +49 69 9585 5874
Phone: +49 69 9585 5406