APT vs Multifactor CAPM

[wrongsaidfred]

Hi David,

I was wondering if there was a way to easily describe the difference between the two models. The general idea behond them seems exactly the same (multiple factors, multiple sensitivities) so I was wondering exactly what the difference is.

Thanks in advance for any help you could provide.

Mike

 

[David Harper CFA FRM]

Hi Mike,

Grinold Chapter 7 (previously in the curriculum) goes into some technical depth, and Elton & Gruber contains quite a bit (piecemeal) of the theory, showing how the difference amounts to assumption flexibility in the APT. I'm not a deep expert, as apparently, understanding the difference requires understanding the theory of each; in practice, IMO, the APT ends up looking like the multi-factor CAPM. Although the APT is completely "creative:" factors can be implicit or statistical (ie, unlike the CAPM where, I think, you would specify risk factors that have an economic "narrative" such as the small company premium).

Here is Amenc's summary (Chapter 6), this is the best short version i have in my library:
"In 1976, Ross proposed a model based on the principle of valuing assets through arbitrage theory (see also Roll and Ross, 1980). This model, called the Arbitrage Pricing Theory (APT) model, is based on less restrictive assumptions than the CAPM. While the CAPM assumes that asset returns are normally distributed, the APT does not hypothesise on the nature of the distribution. The APT model does not include any assumptions on individuals’ utility functions either, but simply assumes that individuals are risk averse. This simplification of the assumptions allows the model to be validated empirically." - Amenc Chapter 6, page 148

I attached the Roll & Ross paper that Grinold recommends on the topic (difference btwn CAPM and APT), in case it's useful.

I hope that's a start, thanks, David

 

[wrongsaidfred]

Works for me.

Thanks!

Mike