Regards
Brad

If your models calc’s different capital amounts depending on the current economic position, this would be a PIT (point in time) system. If the models are immune to the economic environment, they are part of a through-the-cycle (TTC) system. This relates to the pro-cyclicality issue.

Most retail models react to economic inputs to some extent (hybrids of PIT/TTC are possible). This is because the models must be sensitive to changes in risk due to non-economic factors and splitting these out completly (from economic factors) is very difficult.

First off, the stress test should take the models as they stand and stress any inputs that are economically sensitive. This should demonstrate the variation in pillar 1 capital during a downturn scenario. In theory, with the same risk profile of the portfolio, this should give the same answer regardless of the current economic position. However, it is probably not that straight forward. If the model inputs consider relative shifts in economic factors rather than absolute positions then it gets more tricky. (ie model changes interest rates by +5% rather than setting a fixed 10% rate).

The thought process should be how your current models (within the live process and governance restrictions) will perform in a downturn and what capital figures will be calculated. If any changes to the model would be required during an actual downturn, then these need to be fully documented, approved and set up with governance controls. You can’t assume a (management) action might happen in the future because it seems sensible now.

That would be the pillar 1 stress test. Pillar 2 is another story.

e.g. Scenerios like terrorist attacks affecting the credit market, there’s not much of a correlation of the probability of it happening and the credit cycle from the start. Thus, this stressed scenario shouldn’t change where you are in the cycle.

e.g. For the likes of Subprime, it matters much as consumer’s behaviour in taking credit becomes too much for the market to bear.

In short, it all depends on the scenarios you are considering.

I am currently working on stress testing economic capital for a retail bank. What we are confused about at the moment is whether a stressed scenario changes according to where you are in the cycle. In other words given that our current exposure remains the same, and the macroeconomic factors going into the model change (as we move through the cycle) should the value of capital at the 1 in 100 event stay the same or should this point move.

To put this into context, currently we are sitting above our long run average, and when we run run our model, stressing our data according to where we currently sit in the cycle, we achieve a higher capital number and a higher VaR for a 1 in 100 event.

So just to confirm, our confusion lies in whether you apply the stress to the long run average or whether you apply the stress to where you are in the cycle.

Any comments would be greatly appreciated