MScFE 610 Econometrics – Video Transcript (Lecture 1) Module 6: Unit 1
Unit 1: Portfolio Choice and Empirical Modeling of Expected
Portfolio Return and Risk – Basic Static Results
In this module, we combine the results we developed in the first five modules of this course into a
holistic approach to the modeling of the expected returns and risk to a portfolio of assets, the
central tool that a financial engineer or investor uses to manage investment returns and the
related risks.
Portfolio optimization
In this video we will be looking at basic static results of empirical modeling of expected portfolio
risk and return.
To focus on the core ideas of portfolio choice, which you will study in detail in a later course, we
only consider long positions in two risky assets and one safe asset. All the results we will use extend
readily to more assets (which would usually be the case in a balanced, well diversified portfolio),
but this highlights the most important relationships and results. If you need more detail, consult
Ruppert and Matteson (2015).
Consider two risky assets with the following characteristics:
Asset 1 has expected return 𝜇! with variance 𝜎!" (or standard deviation 𝜎! ). Asset 2 has expected
return 𝜇" with variance 𝜎"" . The correlation between returns is given by 𝜌!" .
Since the expectation operator is linear, this means a long portfolio 𝑝 with weight 𝑤 on asset 1 will
have expected return:
𝜇# = 𝑤𝜇! + (1 − 𝑤)𝜇"
Due to the simple formula for the variance of the sum of two random variables, the variance of the
portfolio will be:
𝜎#" = 𝑤 " 𝜎!" + (1 − 𝑤)" 𝜎"" + 2𝑤(1 − 𝑤)𝜌!" 𝜎! 𝜎"
© 2019 - WorldQuant University – All rights reserved.
3
MScFE 610 Econometrics – Video Transcript (Lecture 1) Module 6: Unit 1
With two assets, the investment problem is not very interesting: given the parameters of the
returns, there is only one way to get a given expected return, which fixes the variance, and one
way to get a specific variance, which fixes the expected return.
Once we add a safe or risk-free asset, however, we can ask questions about the “optimal
portfolio”.
To keep things simple, we can consider the optimal portfolio from two angles:
1. If the investor has a specific desired expected return from a portfolio, the optimal
portfolio is the portfolio that attains that expected return with the lowest possible
variance.
2. If the investor wants a specific risk or portfolio variance, then the optimal portfolio is the
one that achieves that portfolio variance with the largest possible expected return.
Let’s define the safe asset as having a fixed return 𝜇$ and, by definition, zero variance.
The two questions we need to ask are:
1. How much of our portfolio must be invested in the safe asset? Let’s denote this with
(1 − 𝑣), so that we invest 𝑣 in some portfolio of the two risky assets.
2. Given 𝑣 to be invested in the two risky assets, how must we allocate this across the two
assets?
The second question has a fixed answer: we always invest in the tangency portfolio that
represents the optimal trade-off between risk and return across the two risky assets. We only
state the result here. See Ruppert and Matteson (2015) for the derivation. The tangency portfolio
is given by the relative weight 𝑤% on asset 1:
.𝜇! − 𝜇$ /𝜎"" − .𝜇" − 𝜇$ /𝜌!" 𝜎! 𝜎"
𝑤% = .
.𝜇! − 𝜇$ /𝜎"" + .𝜇" − 𝜇$ /𝜎!" + .𝜇! + 𝜇" − 2𝜇$ /𝜌!" 𝜎! 𝜎"
© 2019 - WorldQuant University – All rights reserved.
4
MScFE 610 Econometrics – Video Transcript (Lecture 1) Module 6: Unit 1
Yielding risky portfolio with expected return and variance:
𝜇 % = 𝑤% 𝜇! + (1 − 𝑤% )𝜇"
𝜎%" = 𝑤%" 𝜎!" + (1 − 𝑤% )" 𝜎"" + 2𝑤% (1 − 𝑤% )𝜌!" 𝜎! 𝜎"
The answer to the first question depends on the goal of the investor.
The expected return 𝜇& and variance 𝜎&" on an arbitrary portfolio with (1 − 𝑣) in the safe asset is
given by:
𝜇& = 𝜇$ + 𝑣.𝜇 % − 𝜇$ /
𝜎&" = 𝑣 " 𝜎%"
These equations then describe the optimal trade-offs between the final expected return and
variance of all portfolios in this model: the lowest risk. The final choice of 𝑣 depends only on the
risk appetite of the investor.
Suppose our investor requires return equal to 𝜇∗ . Then, for any given, and usually data dependent,
values of 𝜇$ and 𝜇 % , the optimal fraction of wealth 𝑣 ∗ to invest in the tangency portfolio is given by:
𝜇∗ − 𝜇$
𝑣∗ =
𝜇 % − 𝜇$
You can easily derive a value for 𝑣 if the investor is targeting a specific portfolio variance instead.
This video looked at basic static results of empirical modeling of expected portfolio return and
risk.
© 2019 - WorldQuant University – All rights reserved.
5
