Normal and power law distributions

About a week back, I had sent a long email to my client espousing the virtues of the power law distribution, and how it was relevant to the project that I was working on. To be honest, I didn’t really expect a reply to that mail. It was fairly long, and parts of it were technical, while most of my clients are primarily business people. In the best case, I decided after sending the mail, it would be ignored, and in worse cases, the client would think of me as a freak and start wondering why I’ve been engaged.

The next morning I got a rather long reply from my client. It appeared that he had taken great pains to understand my model and its relevance to his problem, and he responded with the model that he had built for similar analysis. While the model he built was simplistic (during the course of his email he repeatedly pointed out his non-statistical background), I found it to be amazingly accurate, and that it modeled the behaviour we were looking to model rather accurately. The only drawback of the model was that it wasn’t a particularly scalable model, and it wasn’t easy to modify.

Given my experience at a rather large and successful investment bank that used to work for (documented¬†here), this was indeed a pleasant surprise. The modeling department of that bank had been virtually taken over by people skilled in continuous math and linear algebra, and the people there had little patience for any model that didn’t follow some variation of Brownian motion (of course, most of that firm’s systems were set up assuming one such distribution). Despite having been through the lows of the 2008 Financial Crisis, the modelers there steadfastly refused to let go of their supposedly time-tested models, and any comments of distributions not being normal were seen with suspicion.

Over a year and half back, when I decided that I should move out of the investment banking industry (where I found quant having saturated to some extent) and into other sectors where there wasn’t much utilization of quantitative techniques, one of my apprehensions had been that I would frequently run into business people who simply wouldn’t touch modeling with a barge pole. I had been apprehensive of working with people with absolutely no quantitative background.

After my recent experience, though, I’m happy to have made the transition. I do hope that other clients are also as receptive to new ideas in quantitative techniques.

Put Comment