Quant models. Forgotten variables?
September 14, 2009
Here is a widespread claim: The problem with quant models is that they forgot a few variables. This position on the causes of crisis plays into two dominant arguments: The first is that the methodology of modeling is fundamentally sound; what was wrong was the models themselves which need tinkering. The second, also discussed in this article in today’s NYTimes, is that social network analysis is part of the solution because it can track the contagious spread of ideas about markets that can lead to panic that exacerbates market malfunction.
My question is: what is the methodological basis of social network analysis? It is perhaps ironic that some of the biggest disciplinary critics of the economists are now working on an algorithmic platform that allows them to be folded into the logic of economics and quantitative finance. The Times article cites an NSF grant awarded to at team at Cornell to work on collective behavior. The team is populated by computer scientists, economists and finance experts… with narry a sociologist in sight.
This article by Bruno Latour explaining how new quantitative techniques that allow individuals to be tracked, rather than treated in aggregates, are fundamentally different than traditional statistical tools, may be useful starting point for addressing these issues. Latour’s insight is that the history of quantification is not just a story about doing better statistics by adding factors to combat incompleteness; rather the progress of models has involved the rise of novel type of quantification methods that have changed the nature of the world that these techniques presume to model.
Will quant models survive? Yes. But the simple narrative that financiers will continue to tinker with their toys might not be enough to explain their resilience…