Bank risk models may eventually take market sentiment into account, using information from news feeds and social media sites to help predict future events, according to John Macdonald, executive vice-president at risk management technology vendor Algorithmics, an IBM company.
Speaking in a video interview with Risk on the future of risk management technology, Macdonald said early attempts to incorporate market sentiment into models have already been made, and further research is currently under way.
Pointing to the example of Stockton, a city in California that recently filed for bankruptcy, he suggested this kind of information could have added an extra dimension to risk analysis.
"Could that have been predicted?" Macdonald asked. "There were significant numbers of retail closures in Stockton. Forbes actually listed it as the most miserable city in the US. It had one of the highest crime rates, and it has one of the highest obesity rates in the whole of the US. So could you have got sentiment from this wealth of data?"
Collecting and making sense of this kind of unstructured data poses a number of challenges – not least, how to get the right balance between traditional market and credit risk data and other, less objective forms of information. "That's the really interesting bit, and where the research has to go," he added.
However, this is unlikely to be the primary focus for technology vendors and the banks using their products – at least in the short term, said Macdonald.
"Is there going to be massive investment in trying to define sentiment? I think there is going to be investment, but it is going to be in parallel to using the investments already made to run the business. What do I mean by that? The delivery of right-time information – and by right time, I mean when you need it. That could be daily, it could be once a year, or it could be instantaneously to help you make a business decision. That's where I think the real investment is going, because looking back and saying, ‘well, what happened?' – everyone can do that."
The week on Risk.net, November 25-December 1, 2016Receive this by email