Incident: A rushed model, conditions nobody checked

Module 2 · Incident file · About 35 minutes

What happened

On 16 January 2013 JPMorgan Chase published the report of its Management Task Force on the 2012 losses in its Chief Investment Office (CIO), the trades the press called the "London Whale". Appendix A of that report tells the story of one risk model. The US Senate Permanent Subcommittee on Investigations reprinted the whole report as Exhibit 98 of its hearing of 15 March 2013 (S. Hrg. 113-96, Volume 1, pages 1086 to 1094), and the quotations below come from that reprint.

The model calculated value at risk (VaR), an estimate of how much a portfolio could lose in one day. A London-based quantitative expert, "the modeler", began a new VaR model in or around August 2011, while keeping his existing duties. He "asked the trader to whom he reported for additional resources", but "did not receive any". He submitted it for approval on 25 November 2011. The bank's Model Review Group "performed only limited back-testing", comparing results over "a subset of trading days during a two-month period".

In January 2012 the CIO exceeded its VaR limit. The new model "was about to come online that would show a substantially reduced VaR". On 23 January the trader wrote to "keep the pressure on our friends in Model Validation". The review group authorised the model on 30 January, for use from 27 January. At the hearing, Senator Levin said the CIO's VaR fell overnight from $132 million to $66 million.

In May reviews found the model's spreadsheet error, described in the next section. By the end of 2012, Senator McCain said at the same hearing, the trading losses "stood at a staggering $6.2 billion". The task force did not blame the model alone. The losses were "the result of a number of acts and omissions, some large and some seemingly small".