Professional Documents
Culture Documents
Erik Kole
September 6, 2010
In the course Asset Pricing we will often differentiate functions of vectors and matrices.
Typically these functions have a linear or a quadratic form. An example of a linear function
is the return of a portfolio of n assets. For the portfolio return we write rp , while we use r
to denote the vector with the returns of the n assets and w for the portfolio weights. This
yields
n
X
p 0
r =wr= w i ri , (1)
i=1
For the quadratic case of σ p differentiation is bit more advanced, as we need the product
1
rule:
n n
∂σ p X X
= bkj wj + wi bik
∂wk j=1 i=1
n n n
X X X (4)
= bkj wj + bki wi = 2 bkj wj
j=1 i=1 j=1
=2Bk w.
In the second step, I used the fact that a variance matrix is symmetric (so bik = bki ). In
the third step I used vector notation again, with Bk being the k th row of B.
Finally, we need to find the derivative of a whole vector instead of one of its elements.
The derivative of a function g(w) is defined as
∂ g(w)
1 ∂w
∂ g(w)
∂g(w) ∂w
≡ .
2
(5)
∂w ..
∂ g(w)
∂wn