LO 53.1: Describe ways that errors can be introduced into models.

LO 53.1: Describe ways that errors can be introduced into models.
Models are highly useful in simulating real-life scenarios; however, they can suffer from several risks. M o d el ris k is the risk of incorrect trading or risk management decisions due to errors in models and model applications, which can lead to trading losses and give rise to legal, reputational, accounting, and regulatory risk. Biases in models themselves do not necessarily cause model risk; however, inaccurate or inappropriate inputs can create distortions in the model.
There are several ways in which errors can be introduced into models. These include bugs in the programming of model algorithms, securities valuations or hedging, variability of value at risk (VaR) estimates, or inaccurate mapping of positions to risk factors.
For example, bugs in programming occurred in May 2008 when Moodys used flawed programming to incorrectly assign AAA ratings to certain structured credit products. It happened again in October 2011 when bugs in the quant programming used by AXA Rosenberg1 led to investor losses. For Moodys, model risk was related to reputational and liquidity risk because the model errors had been discovered prior to being made public and coincided with a change in ratings methodology that resulted in no change to the ratings of certain products. As a result, Moodys was suspected of tailoring its model to the desired ratings, which damaged the companys reputation. For AXA Rosenberg, the discovery of the model error had not been made public in a timely manner, leading to both regulatory fines and considerable reputational damage to the firm. 1
1. AXA Rosenberg Group, LLC is a division of the French insurance company AXA.
Page 206
2018 Kaplan, Inc.
Topic 53 Cross Reference to GARP Assigned Reading – Malz, Chapter 11
Model errors in securities valuations or in hedging can create losses within a firm and lead to market risk and operational risk. M arket risk is the risk of buying overvalued (or, at a minimum, fairly valued) securities in the market that are thought to be undervalued. O perational risk is the risk of recording unprofitable trades as profitable.
Relying on market prices rather than model prices through marking positions to market can theoretically avoid model errors and reduce valuation risk. A problem with this approach, however, is that certain positions, including long-term bank commercial loans, are difficult to mark-to-market due to infrequent trading and complexities in valuation.
V a r i a b i l
i t y o f V a R E s t
i m a t e s