Assignment 2

1. Portfolio math (10 points). Let Rπ denote the return on a portfolio located on the
minimum-variance frontier for risky assets only and suppose that π is different from the
global minimum-variance portfolio. Show that there is a portfolio z(π) also located on the
minimum-variance frontier for risky assets only, which has the property that Cov(Rπ , Rπ(z) ) =
0. Show that E[Rz(π) ] = (C − BE[Rπ ])/(B − AE[Rπ ]), where A, B, and C are the
constants defined in the slides. Hint: First show that the covariance between the re-
turn on a minimum-variance portfolio with mean m1 and one with mean m2 is equal to
1
∆ (Am1 m2 − B(m1 + m2 ) + C).

2. Portfolio math (10 points). Let Rmin denote the return on the global minimum-variance
portfolio of risky assets. Let R be the return on any risky asset or portfolio of risky
assets, efficient or not. Show that Cov(R, Rmin ) = V ar(Rmin ). Hint: Consider a portfolio
consisting of a fraction w in this risky asset. and a fraction (1 − w) in the global minimum-
variance portfolio. Compute the variance of the return on this portfolio and realize that
the variance has to be minimized for w = 0.

3. Alternative portfolio strategies (30 points). The optimal mean-variance portfolio is a

In this exercise you will run a horse-race between four different portfolio strategies, each of
which is a special case of the full mean-variance strategy. Diversification is common to all
the strategies, but they build a diversified portfolio in different ways. This leads to very
different performance.

The four portfolios strategies are:

• The tangency portfolio.

• The minimum variance portfolio. This is a special case of full mean-variance analysis
that does not estimate means; it implicitly assumes that all assets have the same
mean.
1
In the situation considered here with four asset classes, there are just 14 parameters to estimate. With 100
assets, there are 5,510 parameters to estimate. With 5,000 stocks (approximately the number listed in U.S.
markets) the number of parameters to estimate is over 12,000. The potential for errors is enormous.

1
 • The risk parity portfolio with weights equal to the inverse of standard deviations of
returns. This is also a special case of mean-variance analysis that does not estimate
means or correlations; it implicitly assumes that all assets have the same mean and
all assets are uncorrelated.

• The equally weighted portfolio. This has nothing to estimate. It is also a special case
of mean-variance analysis; it implicitly assumes that all assets are identical.

As you move from full-blown mean-variance analysis to equal weights, you estimate fewer
parameters and thus there are fewer things that can go wrong with the mean-variance
optimization. The extreme case is the equal weights, which require no analysis of data.2

The Excel file totalReturns.xlsx has monthly total return indices for four asset classes:
US Treasuries (from Bank of America Merrill Lynch bond indices), US corporate bonds
(from Bank of America Merrill Lynch bond indices), US stocks (from MSCI), global stocks
(from MSCI). It also contains 1-month T-bill rates, which is the relevant risk-free rate over
the sample period. The sample period starts in 1978.

For the four different portfolio strategies, track the performance from January 1988 to the
end of the sample period. Implement the strategies at time t using data for a ten-year
window. Therefore, the first portfolios are formed at the end of December 1987 using
returns from January 1978 to December 1987. The portfolios are held for one month (i.e,
until end of January 1988). The next portfolios are formed at the end of January 1988
using returns from February 1978 to January 1988. The portfolios are held for one month
(i.e, until end of February 1988), and so on. Use one-month T-bills as the risk-free rate.

(a) For each of the four portfolio strategies, plot the time series of the portfolio weights.
2
Long-run means are very tricky to estimate. Sampling at higher frequencies does not allow you to more
accurately estimate means – only extending the sample allows you to estimate the mean more precisely (this
is shown in a seminal paper by Merton (1980)). For any asset, the only way to gauge the long-run return is to
look at the index level at the beginning and at the end of the sample and divide it by time. It does not matter
how it got to that final level; all that matters is the ending index value. Hence, we can only be more certain
of the mean return if we lengthen time. This makes forecasting returns very difficult. Volatilities are much
more predictable than means. High frequency sampling allows you to estimate variances more accurately even
though it does nothing for improving estimates of means. Higher frequency data also allows you to produce
better estimates of correlations. But correlations can switch signs while variances can only be positive. Thus,
variances tend to be easier to estimate than correlations.

2
 (b) Compute mean and standard deviation of portfolio excess returns as well as the Sharpe
√
ratio (annualize by multiplying by 12)

(c) Interpret the differences in performance

4. Mean-variance optimization (20 points). In spite of the evidence above, consider an

(a) Draw the risky-asset-only minimum-variance frontier for

i. US stocks and global stocks

(In the (σ, µ) diagram, limit µ to the interval between 0.002 and 0.012 and σ to the
interval between 0 and 0.07.)

(b) Suppose you target a monthly expected return of 0.0100. In the case with all four asset
classes, compute the minimum-variance portfolio that achieves this expected return.
What is its return standard deviation on a monthly basis?

(c) Consider an investor who has mean-variance utility U = µp − a2 σp2 and a risk aversion
coefficient of 10. In the case with all four asset classes, what is the optimal portfolio?
What is its expected return and return standard deviation on a monthly basis? Draw
the corresponding indifference curve.

(d) Suppose now that a riskless asset is also available for investment. Assume that it has
a monthly return of 0.0042 (approximately the average T-bill rate over the sample
period). Draw the new mean-variance efficient frontier.

(e) What is now the optimal portfolio of the investor? What is its expected return and
return standard deviation on a monthly basis? Does the investor lend or borrow?
How much? Draw the corresponding indifference curve.