Skip navigation

Daily Archives: August 18th, 2009

Well done Emanual Derman. At last someone is asking the right questions. I broke EMH, Portfolio Risk/Diversification and CAPM in my1995 University College Dublin 500-page Master’s thesis. I showed that,
1. Markets are only as efficient as the data reported.
2. Portfolios do not fully diversify away unsystematic risk.
3. CAPM is only valid within a 90-day range.

I published my thesis in 2002 as a 278-page book, titled “A Rational Approach to Unsystematic Risk, Re-Thinking Modern Finance” . You can still purchase it from http://www.quantumrisk.com/books.html while stocks lasts.

I too had a quick look at Andrew Lo’s AMH. It could be a good approach. However, being an engineer by training, I’m not keen to rush into more theory. In 1995 I had proposed how we could measure market efficiency (see table below). While I am not a whole hearted fan of EMH, I believe we should use metrics to measure market efficiency to monitor and improve markets. My 1995 results suggest that market volatility plays a part in market efficiency.

Stock Exchange Index Prob. of
Weak Form
Prob. of
Strong Form
Bangkok SET Index 100.00% 81.59%
Frankfurt DAX 87.29% 77.42%
Gold Gold Spot 100.00% 94.18%
Hong Kong Hang Seng 100.00% 90.49%
Kuala Lumpur Composite 83.40% 76.09%
Kuala Lumpur Emas 81.46% 66.15%
London FTSE 81.33% 71.33%
New York Dow Jones 84.13% 59.05%
New York S&P 500 84.20% 80.68%
Singapore All Share 76.26% 74.83%
Sydney All Share 100.00% 98.87%
Tokyo Nikkei 82.64% 76.79%

___________________________________________________________________
Disclosure: I’m a capitalist too, and my musings & opinions on this blog are for informational/educational purposes and part of my efforts to learn from the mistakes of other people. Hope you do, too. These musings are not to be taken as financial advise, and are based on data that is assumed to be correct. Therefore, my opinions are subject to change without notice. This blog is not intended to either negate or advocate any persons, entity, product, services or political position.
___________________________________________________________________

Returns are almost always long tailed distributions. Depending on the circumstances, they can have negative long tail (the ‘famous’ lognormal fat tails in risk analysis), positive long tails or both.

If you don’t have the experience or the expertise, handling all these variations can be quite a challenge. For small samples I have observed many times people ignore the tails – i.e. the justification is that these outliers are ‘infrequent’ and due to some other unknown process (a nice way to say “I don’t want to trouble myself”) – and then fit a reasonable normal distribution.

Another problem I have observed with small samples  is that the distribution can change markedly from sample to sample. This can be a severe problem. The concern here is, if you don’t know the true population distribution, how do you model it correctly? I don’t know how others treat this if they do or if they just ignore the whole thing and assume normality.

Note: depending on the data even 100 data points can be considered small.

One important reason why normal can be used. If you are measuring the average of many small samples, than normal gives a good fit. You have to make the distinction that you are working with averages and not the return statistic itself but I find that when this happens many times people don’t. 

___________________________________________________________________
Disclosure: I’m a capitalist too, and my musings & opinions on this blog are for informational/educational purposes and part of my efforts to learn from the mistakes of other people. Hope you do, too. These musings are not to be taken as financial advise, and are based on data that is assumed to be correct. Therefore, my opinions are subject to change without notice. This blog is not intended to either negate or advocate any persons, entity, product, services or political position.
___________________________________________________________________