Jul 26

Five new worksheets (operational risk)

by David Harper, CFA, FRM, CIPM

FRM |

image

As a precursor to Monday's screencast (Operational Risk A), I uploaded five worksheets to the member page. Please note: I tagged none of them with yellow highlight. That means I do not consider them essential to an exam-passing strategy. But they may still give a helpful boost, especially if you are struggling with a concept.

You might think operational risk will finally give relief in regard to quantitative methods; alas, it is not so!

I uploaded the following today:

  • Deutsche bank's business line/event type (BL/ET) matrix. Real simple replication of the cells used in the "LDA at Work" reading
  • Illustration of loss distribution approach (LDA) using parametric distributions (i.e., compounds a Poisson frequency distribution with a lognormal severity distribution)
  • Illustration of loss distribution approach (LDA) using empirical distributions. This also compounds a frequency with a severity distribution, but uses tabulation.
  • Illustration of economies of scale (Saunders' technology reading)
  • Illustration of a mixture distribution where extreme value theory (EVT) is used to model the loss tail (i.e., GPD in the tail). This is sort of advanced so don't bother unless you really want to look.

I will show the charts in the screencast. The key thing I would say about the distributions (based on the Deutsche Bank LDA reading) is:

  • It builds on the foundation laid in Gujarati, you are never wasting time to get solid in the distributional terms reviewed in Gujarati (e.g., PDF/PMF vs CDF, empirical vs. parametric). The LDA approach is less difficult if you are grounded in the more basic ideas around distributions
  • The first "innovation" in the DB reading is: to get a loss distribution, they are compounding a frequency distribution with a severity distribution. Operational losses occur in two dimensions: first they either happen or they do not happen with some frequency, then if they do happen, the loss occurs with some severity or magnitude. (hence we talk about high frequency, low severity losses or low frequency, high severity losses)
  • The second innovation, for DB,  is to "slice" the severity distribution into pieces: an empirical distribution for the body, then a parametric distribution for the tail (illustrated at top).

Comments

  1. Be the first to leave a comment!

Leave a Comment