Granular Data Offer New Opportunities for Stress Testing

Abstract

Nowadays, more and more granular data is being collected. The 2008-2010 crisis has shown that authorities were missing crucial information for accurately identifying risks in the financial system. This realization has led to a significant increase in the depth and scope of information being reported across the system. At the same time, the cost of reporting has gone down due to further digitization.

In this paper we will examine how increased granularity allows authorities to put the data to better use and we will focus in particular on stress testing. Stress testing involves postulating severe scenario’s that often have yet to materialize. More granular data combined flexibly from different sources allow for more accurate assessment of such adverse hypothetical states of the world. For instance, detailed information on inter-bank exposures allows for understanding bank to bank contagion much better.

We will briefly discuss the developments in the last decade and discuss what possibilities arise. The new granular data sets allow us to implement stress tests on various levels of aggregations using the very same data sets. Then we will turn to prerequisites in terms of for instance meta data allowing to connect data sets and data governance specifying how to treat the data. Finally, we will discuss remaining challenges, ranging from data quality up to high performance IT infrastructure and a company culture that embraces data.

Keywords: Granular Data, Stress Testing

JEL Classification: G20