Hacker News Reader

Comment by abuteau | Hacker News Reader

There's ton of criticism on MPT. Would like to hear you on that ! Are they valid ? Should we use PMPT ?

You'll want to check out Steve Keen, of "Debunking Economics" fame, for the gory details. He has a blog (debtdeflation.com or something) with a couple of finance class videos on it, and he runs or used to run a talk all over, a version of which you'll find in the Google Talks channel.

In short though, the entire thing is based on the idea that finance is based on guassians, whereas the evidence overwhelmingly suggests it's based on power laws instead. (And intringuingly, the consequence of the latter would mean that portfolio risk increases when you're diversified in many asset classes.)

The primary question is this: do we truly have models that can predict future asset correlations. In other words, are our assumptions about distribution of returns valid. Behavioral economics suggests that individuals react differently than mathematically predicted.

Personally, I think the biggest problem with any ideal portfolio allocation tool are black swan events (Taleb).

> I think the biggest problem with any ideal portfolio allocation tool are black swan events (Taleb).

Yes. Black Swans are certainly a problem. It's insane to think that investment returns are anything resembling a Gaussian distribution. In the financial crisis of 2008-2009, people were using phrases such as "a 9 sigma event" to describe the markets.

Wrong! Basically THERE'S NO SUCH THING AS a 9 SIGMA EVENT. Wikipedia gives the probability of a 6 sigma event as 1 in 500 million. So it the epitome of arrogance to think that we're so "special" that we just happened to be alive during a 9 sigma market move.

What it all really means is that the markets do not behave the way lazy eggheads want them to. Everyone understands Gaussian distributions and standard deviation. So they are eager to use that math where they shouldn't.

I'm sure you know all this. But for those people who aren't familiar with Taleb and want to read more, he wrote three pop books about it:

2001 Fooled by Randomness http://en.wikipedia.org/wiki/Fooled_by_Randomness

2007 The Black Swan http://en.wikipedia.org/wiki/The_Black_Swan_%28Taleb_book%29

2012 Antifragile http://en.wikipedia.org/wiki/Antifragile

I'm in the middle of reading Antifragile, and I am thoroughly enjoying it so far.

George Soros' perspective is grounding. Markets are moved by people with incomplete knowledge. The entire field of economics is based on a false analogy with Newtonian physics. There is no equilibrium, no mystical balancing force underlying the economic universe. Just a whole bunch of relative idiots trying to get rich for free.

Are these works (esp the latter) accessible to people with no economics background, but a decent foundation in maths (calculus & algebra) and basic statistics (distributions etc)?

Yes. Well, with knowledge of basic statistics you might start getting angry at how much it's dumbed down.

How do you see value investing vs MPT then ?

I often refers to a good paperon What has worked in investing [1]

I think Taleb criticism is right though and he demonstrates it through Mandelbrot fractal.

[1] http://www8.gsb.columbia.edu/sites/valueinvesting/files/file...

If by Taleb you mean Mandelbrot, then maybe.

Taleb wrote a book together with Mandelbrot.

Yeah, this whole "a responsible asset manager has to use MPT" is completely not true. A responsible asset manager has to consider risk and return, of course, but how they measure and evaluate risk and return is an art as much as a science. MPT-driven portfolios have been shown to underperform more naive formulations, mainly due to model risk and mis-estimation of parameters.

I completely agree with you. MPT/PMPT systems are contingent upon accurate modeling of risk and return and this is not a science.

Nassim Taleb is a vocal critic of MPT.

Reply on news.ycombinator.com