Friday, November 22, 2013

URSP Student Oleg Titorchuk Evaluates Alternative Methods of Credit Valuation and Credit Rating Agency Institutional Structure

I am fascinated with the nature of systems and interconnectivity of the agents within them. Financial system is one of the most interconnected sectors in any economy and the slightest variance of one agent can compound to an immense bullwhip effect. The Big Three Credit Rating Agencies role in the subprime mortgage crisis is greatly overlooked and my goal is to show the CRAs contribution and with clear numbers explain why the tale of a more efficient and more stable financial sector is the tale of deregulation.

            My research looks at corporate structure of the CRAs that rate virtually all fixed income in the U.S. and internationally. The first part of the project is focused on understanding how did Washington’s policies shaped the CRAs business model, and finance industry as a consequence. The second part takes datasets with credit default rates and looks at what factors were the most effective at predicting default rates of corporate bonds, CDOs and individual loans. As data scientists in the making, I am delighted to combine my domain expertise of finance and economics towards fascinating machine learning problems of predicting default rates and understanding what variables are the most useful for doing so.

            My routine resembles reading a lot of peer reviewed papers and running machine learning algorithms on the risk valuation models. Brainstorming about incentive structures and how to channel human interest for the interest of the public is a very engaging and continuous part of my work. To have the highest odds of success, you must be able to sort through vast volumes of information and find exactly what is relevant to you.

            This week, I have been researching legal constructs of the CRAs. This led me to discover that many research papers leverage isolated data sets to prove their point, and at times leave out some very important factors that fit into the overall picture. For example, one banker stated that on average CRAs are excellent at predicting risk outside of countries and CDOs, however, what he failed to mention is that on average corporate bonds default at the rate of 2% per year over the past 150 years.