[R-SIG-Finance] Joint estimation of APT in R

Hebbertt Soares hsoares85 at gmail.com
Wed Nov 12 13:44:43 CET 2008


Hello UseRs,

I am working with Arbitrage Pricing Theory (APT), an extension of CAPM
proposed by Roll and Ross. One of the most troubling issues of this
model is the estimation of its two components: the factor
sensitivities and risk premia..

Most of the literature estimate this equation in two steps, using Fama
and Macbeth (1973) approach.

However, as Burmesteir and McElroy (1988) shows this approach is
inefficient. It would be more efficient doing a joint estimation of
factor sensitivities  and risk premia as proposed in as Burmesteir and
McElroy (1988) paper.

This estimation would be done trough a system of nonlinear
multivariate equations with the following structural form (suppose two
assets and two factors):

return of asset 1 - risk free rate = constant*coefficient(2) +
coefficient(2)*factor 1 +
coefficient(3)*coefficient(4)+coefficient(4)*factor2

return of asset 2 - risk free rate = constant*coefficient(5) +
coefficient(5)*factor 1 +
coefficient(3)*coefficient(5)+coefficient(5)*factor2

This system has N equation with K parameter, where N is the number of
assets and K is the number of factors and the number of interaction
among factors.

The estimation can be done using SUR method.

I've been looking systemfit (specially nlsystemfit function), and I
have not been able to find a way of implementing this method in R.

Does anyone in the list implemented APT joint estimation in R? How?

Thank you very much,

Hebbertt



More information about the R-SIG-Finance mailing list