Financial Instruments: Education Session — BdE Provisioning Model

Date recorded:

Two representatives from the Bank of Spain (regulator of Spanish banks) presented the statistical provisioning approach which the BdE requires from entities regulated by them. The representatives explained that, in their opinion, the model incorporates losses incurred due to under-pricing of credit in times of boom in the lending cycle due to market over-optimism. The lending cycle per the model is closely correlated with the economic cycle as a whole.

The model is based on a statistical formula which incorporates an element of collective impairment relating to the point in the lending cycle ('alpha') and incurred losses relating to individual assets ('beta'). The alpha component is a collective assessment and applied to the change in the portfolio of assets at each date of assessment. Thus in times of boom in the lending cycle, the alpha component is high relative to the beta component, whereas this trend is reversed in times of slump as incurred losses relating to the general cycle are in effect transferred to individual assets; the alpha element can be negative, reflecting over-conservative pricing of credit. The representatives explained that they believe the advantage of the model is the early detection of credit losses.

The Bank has around 6 asset classes which it views as homogeneous, for each of which an alpha (effect on asset class of stage of the lending cycle) and beta (historical incurred losses relating to individual assets) are kept. In order to assess the lending cycle, the BdE holds data from the Spanish national credit register dating back to 1988 for each of the asset classes, which equates roughly to 2 full lending cycles. The representatives stressed that this is an incurred loss model as the inputs to the model are derived only from historic experience.

Board members asked questions around various aspects of the model. One member asked what approach the BdE takes for new products where there is little historic data. The representatives replied that this was not an issue that had caused much difficulty in Spanish banking given the relative absence of new products - for example, credit cards represented only around 1% of total lending in Spain. Another member pointed out that, for loans given out during periods of boom, the model effectively results in a large 'day 1'-type loss for lenders. The representatives replied that this was a necessary reflection of credit pricing in these times.

Another member asked whether BdE was aware of how much credit data other central banks held, and thus how practicable a system such as this would be for banks from other countries. The representatives replied that they were aware of some central banks holding extensive credit data; however it was unlikely that many would hold data in sufficient detail going back as far as 1988. Another board member asked whether, if an expected loss model were incorporated within IFRS, the bank would continue to use its model. The representatives replied that, in their opinion, it was possible to use the model to estimate expected losses, however it would be a relatively simple approach. The representatives agreed with board members' views that, were this model adopted more widely, more active involvement would be required from banking supervisors in assessing provisioning than is currently the case given the complex nature of the model and underlying data. The representatives put forward the view that, in times of economic difficulty such as the present time, banks would be less likely to voice the view that an approach such as their model would lead to competitive disadvantage. This had been their experience in Spain.

The chairman thanked the representatives for their presentation.

Correction list for hyphenation

These words serve as exceptions. Once entered, they are only hyphenated at the specified hyphenation points. Each word should be on a separate line.