Models and Crises: Do Financial Models Affect the Incidence and Severity of Financial Crises?


November 28th 2013

Time: 6.00pm - 8.00pm
Venue: Hong Kong Theatre, ground floor Clement House, LSE. For directions click here.
Chair: David Webb (LSE)
Speakers: Jon Danielsson (SRC, LSE), Robert Macrae (Arcus Investment), Jonathan Martin (Markham Rae), Mark Salmon (University of Cambridge), Wolf Wagner (Tilburg University)

Some financial models have been singled out as contributing to the crisis, for example Gaussian copula models for CDO pricing. Other models, such as those used for risk forecasting, are widely seen as having failed before 2007, but still are a central element in the new financial regulations. The panel of experts, drawn from industry and academia, will debate the usefulness of financial models, disagreeing on the question of whether they should be sent to the bin or used even more intensively.

Photographs from this event are available to view here.

The support of the Economic and Social Research Council (ESRC) is gratefully received [grant number ES/K002309/1].
Full Report

In coordination with the SRC conference ‘Frontiers of Research in Systemic Risk Forecasting’, we organised a follow-up evening session where distinguished members of both academia and industry were invited to debate the subject of whether or not ‘Financial models affect the incidence and severity of financial crisis?’ The panel consisted of Robert MacRae a Managing Director of Arcus Investments; Wolf Wagner Chairman of the European Banking Centre Tilburg and Professor of Economics at Tilburg University; Jonathan Martin the Chief Executive Officer and Chief Risk Officer at Markham Rae hedge fund; Mark Salmon Research Director of High Frequency Finance at Imperial College; and Jon Danielsson co-Director of the SRC. The Head of LSE's Finance Department, David Webb, moderated the lively conversation that ensued.

The idea of hosting a panel discussion on the intrinsic role of risk models in the formation and dissemination of risk is inherently paradoxical.  However, if we look back on how some of the aspects of the crisis evolved we can see that posing such a question should not be controversial.  Danielsson made this point clear from the outset when he pointed to the design fault of the majority of financial models.  Namely, that before the crisis the risk signaled by the models was at historic lows.  This articulation calls us to recognise that models have the potential to be wrong at exactly the moment we need accuracy the most.  He summarised this concept by separating the notion of risk into two categories: perceived and actual risk.  The former being the risk that financial models track and the latter being the risk that arises endogenously from the actions of agents.  By design, these two forms of risk have the potential to be negatively correlated, and hence, the source of risk in financial modeling.

MacRae and Wagner reiterated the same concerns over risk modeling.  They added that models lead to overconfidence, which in turn leads to excessive risk taking as market participants feel safer. This may lead to an increase in the incidence and severity of financial crisis.  They both point to many possible reasons for this, but one particular point MacRae made is that models often assume that the period over which you have estimated the model is similar to the upcoming one.  This clearly has the potential of being a dangerous assumption.

The tone of the discussion took a turn towards an environment of debate when Salmon aptly proclaimed, ‘all models are wrong, but they still be useful’.  He says that it is the job of risk managers to be more aggressive in their communication with management over the appropriateness of risk modeling under different situations.    

Martin however objected to some of the attacks being waged on the viability of risk models.  He claimed that risk models are inevitable and useful, and as such, we are absolutely and profoundly reliant on them. ‘Without a model you cannot get a crisis,’ he maintained, ‘Instead, you only get a mess.’  The crisis, he argued, serves to illuminate the shortcomings of our methodologies and models.  Crisis therefore highlights the need for improvement.  If we improve the models, we stand to lesson the severity of the crisis, but never eliminate it. 

As the debate came to a close opinions remained divided as to the role of risk modeling moving forward.  Perhaps the discussion was best summarised by the following comment, ‘There are ways to make the financial system more crisis prone, and ways to make it less crisis prone.  Making the system more heterogeneous is one way towards more stability.’ The heterogeneity of opinion throughout the debate may be a signal we are on the right track.