Paragraph 1(3) calls for a separate evaluation of downturns per jurisdiction. This can become very onerous for globally active companies and, if there are only a few exposures in this jurisdiction, it may also lead to less meaningful results in the subsequent evaluation of the loss data realised in this ju-risdiction. It should also be possible in this case to determine the downturn impact on the LGD glob-ally in the event of global crises that have affected (almost) all jurisdictions in at least one economic factor.
In our opinion, identifying the time series for externally provided default and loss rates required in Article 2(1)(a)(iii) and (iv) constitutes an excessive effort, considering that external data are actually available for this at the end. Furthermore, it is not clear how pool data should be viewed in this con-text.
Article 3(1)(b) only allows a shorter observation period than the preceding 20 years if the economic factor under consideration has been subject to significant changes due to the country’s accession to the EU. In our opinion, this can at best be an example, which should also be identified as such.
In principle, there may also be a significant change in the time series for individual economic factors in many other scenarios, leading to severe characteristics in the time series that are independent of the economic environment. In Germany, for example, the “Hartz IV” reforms in 2005 changed the way the jobless numbers are counted, and meant that former recipients of social welfare benefits who are unfit for work now also count as unemployed – in contrast to the previous years. This results in a peak in the German unemployment rate in 2005 that is not related in any way to economic down-turn conditions. Unemployment statistics for Germany adjusted for this effect are not available. In cases like this, it should be possible either to consider this economic factor only from the time of the significant change or to neglect the peak year 2005 as a downturn period. In general, with regard to the problem of breaks in time series, we would propose reverting to the more generous require-ments of the first RTS draft.