• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Why Einstein Wouldn't Play Dice With Monte Carlo Risk Simulations?

As it is widely known, two of Einstein's most famous quotes are, God doesn't play dice with the universe and The definition of insanity is doing the same thing over and over again, but expecting different results.

I wonder what he would have thought of the continued use of Monte Carlo risk simulation models (MC) by financial institutions, credit reporting agencies, the Fed and other regulatory agencies today?

Recent research points to a flaw within MC based simulations due to the lack of parameters necessary to be more in-tune with todays modern economy. Notably, among others, value inputs for future credit defaults caused by the present number of borrowers under financial distress, such as, Housing Cost To Income ratios (HCTI). As defined here: https://economicgenome.blogspot.com

Ironically with regard to MC risk simulation models, not a lot has changed since 2008. Presently most of our risk models continue to be largely based on rear view mirror data mining.

An example of the methodology behind MC models: Imagine driving to and from work in your investment vehicle on the same road every day for the next 10 years without any concern of hitting potholes (credit defaults). If we were to run a MC simulation on a new road for next ten years, assuming no potholes were created within the first seven years; the results of the regression analysis would be: 365 days x twice a day x 7 years = 4,984 variables. With a correlation coefficient of 100%.

In a MC universe, the first 7 years would indicate the chances of your investment vehicle ever hitting a pothole for the remaining 3 years would be 0%. Those of us who have driven investment vehicles for more than ten years, know just how unrealistic these results are. In the real world, the effects of variables such as snow and ice (financial stress ) on the road (the borrower) for the remaining 3 years would be increased by such events as: rising interest rates, HCTI ratios and unemployment rates etc... All of which are not fully accounted for in past or current MC simulation based risk models.

The most recent example of the inadequacies of MC risk models was clearly on display in 2005. None of the Feds or investment firms MC risk simulations picked up on the impending financial crisis to come. Was this a one time Black Swan event?

Considering MC risk models didn't predict what was to come in 2005 or 3 years before any of the last 3 major recessions, one would tend to believe a clear pattern is beginning to formulate which excludes Black Swans altogether. This begs the question, should we continue to use the same risk models so heavily dependent on MC simulations that have failed us in the past; wouldn't Einstein view this as being the very definition of insanity?

Although MC risk simulation models have been and should continue to be an important tool against credit defaults, we should look towards improving our present models. I would like to encourage the financial risk community to have a friendly debate on this issue. I will start the discussion off by taking the opposing view by stating the current MC based risk simulations that have not worked sufficiently in the past, will not work in the future without a significant update to their parameters which are more in tuned to todays modern economy.

Can anyone post a current risk model being used by the Fed, or any of the major financial institutions today, which takes into consideration current borrower behavioral changes, such as a HCTI ratios?