You are on page 1of 4

Measuring Return

• Return is a measure of how well a security performs as


an investment.

Lecture 10
Pt Dt+1 Pt+1
Measuring Return
Time t Time t+1

• Gross return = (Dt+1+Pt+1)/Pt = 1+Rt+1


• Net return = (Dt+1+Pt+1-Pt)/Pt = Rt+1
• Net return = income yield + capital gain
= Dt+1/Pt +(Pt+1-Pt)/Pt .

Return: Example Compounding Return


• Buy a stock at the beginning of the year for $50 • By convention, we often report returns on an annual
basis (at an annual rate), even if the period over which
• At the end of the year it pays $2 dividend we calculate return is different from a year
• At the end of the year, the ex-dividend price is • This convention facilitates return comparisons
$55 • The correct way to do this is to raise the gross return to a
power
• Gross return = (2+55)/50=1.14 • Example: this quarter a stock has a return of 8%
• The gross return this quarter is 1.08
• Net return = (2+55-50)/50 = 0.14 = 14%
• If this same gross return were repeated for four quarters,
• Income yield = 2/50 = 0.04 = 4% the return after a year would be (1.08)4 =1.36
• Capital gain = (55-50)/50 = 0.10 = 10% • Thus the annualized return is 36% (NOT 32% = 4 x 8%).

Compounding Return Lower Bound on Return


• Similarly, if an investment doubles your money • For assets with limited liability, the worst that can
over a decade, the gross return over the decade happen is that you lose your initial investment
is 2 • In this case, the gross return is zero and the net
• The annualized return is 21/10 = 1.072 return is -1 = -100%
• An annual return of 7.2% will double your money • For such assets, we can define the log return as
if it is repeated each year for ten years. the natural log of the gross return
• But some assets do not have limited liability:
– Insurance participation at Lloyds of London
– Companies with legal liabilities taken over by new
owners with deep pockets
– Short positions

1
Log return and compounding Arithmetic Average Return
Consider a compounded return
1+R= If this same gross return were repeated for
four quarters, the return after a year would be
(1.08)4
Log return=ln((1+R)4)=ln((1.08)4)=4*ln(1.08)

Geometric Average Return What’s the Difference?

Which Average Is Better? Which Average Is Better?


It depends on your purpose. It depends on your purpose.
• G is a good summary of past performance, since it • A is the expected return if you hold the asset for a single
depends only on the cumulative return earned over the randomly selected period.
sample period used to compute the average • If each period’s return is a new draw from the same
• Under the reasonable assumption that log returns are return distribution, then the expected gross return if you
symmetric, G gives the median of the return distribution. hold the asset for K periods (the K-period “buy and hold”
Half the time you will do better, half the time worse than return) is (1+A)K. This is greater than (1+G)K.
G. • Relative to A, G penalizes returns for risk. The penalty is
• Under the reasonable assumption that log returns are exactly the right one for an investor with log utility of
symmetric with a single peak, G gives the mode of the wealth. Such an investor chooses the portfolio with
return distribution, the single most likely outcome. maximum G.

2
Portfolio Return
• The return on a portfolio is a weighted average
of the returns on the individual assets in the
portfolio
Portfolios and Indexes • The weights in this average are the fractions of
the value of the portfolio that is invested in each
asset. These are called portfolio weights.

Index Construction Index Construction


• An index is a portfolio that is thought to be representative • Equal-weighted indexes place most of their
of the general market.
weight on small companies. Equal-weighted
• Each index is described by the included assets and the
weights placed on each asset. portfolios require frequent trading (why?)
• Alternative weighting schemes: • Value-weighted indexes summarize the fortunes
– Equal-weighted, wi = 1/n. of all the dollars invested in the index
– Price-weighted, wi = Pi/(P1 + … + Pn). Example: Dow Jones
Industrial Average. constituents. Value-weighted portfolios never
– Value-weighted, wi = Vi/(V1 + … + Vn), where Vi = market value = need to be rebalanced (why?)
price per share x number of shares outstanding. Example: S&P
500. • Price-weighted indexes are bizarre and exist
– Free-float-weighted. Same as value-weighted except that only by historical accident.
market value calculation excludes shares that are not traded.
Example: Morgan Stanley Capital International (MSCI) indexes
of foreign stocks.

Stocks, Bonds, Bills, and Gold Stocks, Bonds, Bills, and Gold
Real Cumulative Returns
• The return history fits the basic story of 1000
S&P500
equilibrium with risk aversion 10-year T-bonds
• Stocks do well when the economy does well, 100
1-month T-bills
Gold
and have the highest average return
Value ($)

• Gold does well when the economy does poorly, 10


and has the lowest average return
• Bonds and bills are intermediate
1
1925

1930

1935

1940

1945

1950

1955

1960

1965

1970

1975

1980

1985

1990

1995

2000

0.1
Time

3
Does the Measurement Interval Does the Measurement Interval
Matter? Matter?
• If returns in each period are uncorrelated with returns in
• By convention, we report returns on an annualized basis other periods, then the measurement interval does not
• Does the measurement interval matter? For example, matter
does it matter whether we calculate the average using • To see this, look at g, the average of log returns
monthly data and then annualize, or using annual data • The log return over two periods is the sum of two
successive log returns, rt + rt+1
directly?
• Thus the mean of the two-period return is twice the
mean of the one-period return
• The variance of the two-period return is twice the
variance of the one-period return:

Var(rt + rt+1) = Var(rt) + Var(rt+1) + 2Cov(rt, rt+1) = 2Var(rt)

because the covariance term is zero by assumption

Does the Measurement Interval


Normality
Matter?
• For most asset returns, log returns are approximately
• So mean and variance scale in proportion to the normal over periods of a month or more
measurement interval, implying that g and a both scale
in proportion as well • Why? Because monthly returns are generated by
• When you annualize g and a you undo this effect and get multiplying daily returns within the month
numbers that do not depend on the measurement • Thus monthly log returns are generated by adding daily
interval returns within the month
• The measurement interval does not matter!
• These daily returns are approximately independent
• But note the assumption we had to make to get this
result: returns in each period are uncorrelated with • The Central Limit Theorem says that sums of
returns in other periods. We will come back to this later independent random variables approach the normal
in the course distribution, no matter what the distribution of the
underlying daily shocks

Which Average Is Better? Which Average Is Better?


It depends on your purpose. It depends on your purpose.
• G is a good summary of past performance, since it • A is the expected return if you hold the asset for a single
depends only on the cumulative return earned over the randomly selected period.
sample period used to compute the average • If each period’s return is a new draw from the same
• Under the reasonable assumption that log returns are return distribution, then the expected gross return if you
symmetric, G gives the median of the return distribution. hold the asset for K periods (the K-period “buy and hold”
Half the time you will do better, half the time worse than return) is (1+A)K. This is greater than (1+G)K.
G. • Relative to A, G penalizes returns for risk. The penalty is
• Under the reasonable assumption that log returns are exactly the right one for an investor with log utility of
symmetric with a single peak, G gives the mode of the wealth. Such an investor chooses the portfolio with
return distribution, the single most likely outcome. maximum G.

You might also like