Please use this identifier to cite or link to this item:
Rommer, Anne Dyrberg
Rommer, Anne Dyrberg
Year of Publication:
Danmarks Nationalbank Working Papers No. 27
The Basel Committee's Revised Framework for Capital Measurement and Capital internal models to estimate probability of default (PD) when calculating the minimum capital requirement using the internal ratings-based approaches. Valid estimates of the PDs require a considerable amount of data and default observations. Basel II allows for banks to pool their data to overcome their data shortcomings and a number of international data pooling projects have emerged. Thus even international banks need more data to fulfil the requirements of Basel II. To the best of our knowledge, so far no study has compared the banks' capital requirements calculated on the basis of PDs estimated from single-country creditscoring models and multi-country credit-scoring models and accordingly no study has discussed the incentive structure this might create for banks pooling data. The purpose of this paper is to illustrate the consequences on the calculated capital requirements of pooling data for estimation of PD from several countries. We construct a hypothetical portfolio of loans to small and medium sized enterprises for a hypothetical bank operating in France, Italy and Spain. For this purpose we use real world data extracted from the pan-European Amadeus database provided by Bureau van Dijk. Using this data, the PDs are estimated on the basis of single-country creditscoring models and on the basis of multi-country credit-scoring models with pooled data from the three countries. The estimated PDs are then used to calculate the minimum capital requirements. The result shows that there might be incentives for cherry-picking, i.e. that banks are motivated to choose a certain method because it results in a lower capital requirement. The calculated capital requirements vary with up to 18 percent depending on the choice of method for the hypothetical bank. Calculated for the individual countries it varies up to 47 percent. The results are of particular interest for banks operating in several countries, which plan to pool data from the various countries in order to estimate PDs, maybe due to lack of a sufficient single-country database. They are equally interesting for banks planning to pool data with banks from other countries to make up for an insufficient database. Though our default definition is the same for the three countries and we have controlled for variables such as age, size, legal form and sector of each firm, we find quite large differences in terms of the resulting minimum capital requirements for the portfolio in each of the three countries, when the PDs are estimated using a singlecountry credit-scoring model compared to using multi-country credit-scoring models. We show that it is not enough for banks to apply similar definitions of default and similar accounting regimes in the countries. Banks and regulators should also have a careful look into the models, especially the factors that drive financial distress, when banks pool data.
Appears in Collections:
Items in EconStor are protected by copyright, with all rights reserved, unless otherwise indicated.