Beware The Model All Logic

On returning from a 2-week “electronic silence” vacation in the Swiss Alps and in Milan, I saw The New York Times magazine (Aug 17) article, “Dr. Doom.” It profiled an unconventional economist, Professor Nouriel Roubini, who on Sept 7, 2006, predicted for an audience of peers at the International Monetary Fund that the US financial markets would be rocked by a major crisis. The attendees snubbed him for basing his case on comparisons to other recent economic crises instead of mathematical models. Dr. Roubini, who reveres the great John Maynard Keynes for being “the most brilliant economist who never wrote down an equation,” went on to co-write a 400-plus page book on the crisis without using a single equation,

I don’t know Dr. Roubini or his work at al. Even so, I immediately granted him “street cred” because I had written an 800-page (double spaced! Double spaced!!) equation-less doctoral thesis and had experienced significant difficulty in convincing many academics of its value. Several had proudly proclaimed that though they were professors of manufacturing, they had never been inside a single factory. Yet, they were sure that without math models to back me, my 4 years of research in labs and factories, my work as an engineering intern and my “asset review” visits to numerous plants (as a junior banker) could not teach the world anything interesting about manufacturing and technology. Ironically, while most theses lie unread, I received royalty checks for about 5 years for mine and a major consulting company built a successful business around its ideas.

Don’t get me wrong: math models are very important. As an engineer and a doctoral candidate, I chose to study math/applied math for over 8 years. I am very thankful that my computer and the aircrafts that bore me to Europe and back are backed by math models. Finally, at Ishan Advisors, I have built complex statistical models that a blue-chip client is using to drive strategy decisions worth hundreds of millions of dollars.

However, executives and business academics should never forget that a model is a simplification of reality, it is not reality! A model includes a handful of factors that govern the real world, but excludes many, many more. Hopefully, the choices made are correct, but that is not always true. For example, in 1988, the top-tier peer-reviewed journal Management Science published the its first math model of Just-in-Time – 30 years after Toyota started implementing the capability. The model’s complexity challenged the understanding of several top-notch mathematicians at the Harvard Business School. Yet, in the interest of “mathematical tractability,” it assumed away the only thing that makes JIT truly valuable – uncertainty.

Another major problem with math models is that for most part, they are “steady state” models: they compare the world at state B to the world as it had been in state A. What they ignore – the path (the “transient states”) from A to B – is often as important (if not more so) because the “real world” spends more time transitioning than staying steady.

The networked world will make both these shortcoming serious impediments to the building of good models: It will be increasingly hard to isolate a situation from all the influences that could affect it. Executives should be wary of any “expert” who bases his/her views solely on mathematical models, just as they should be wary of anyone who ignores the insights of models. They should work hard to combine “gut feel” and qualitative insights with models. In other words, they should turn more to people like Dr. Roubini (who said that his comparions had been based on his deep understanding of standard economic models) than the model builders who criticized him.

Rabindranath Tagore, the philosopher who won the Nobel Prize in literature, had once written, “A mind all logic is like a knife all blade. It makes the hand bleed that uses it.” Truly words to live by.

Category: Business Environment, Business Tools | Tags: , Comment »

Leave a Reply

Back to top