Unformatted text preview: ins stock returns and believe the factor
will rise, then they can (1) long stocks that are
more sensitive to the factor, (2) short the stocks
that are less sensitive to the factor, and (3) make
profits from the rise of factor
profits
If you do not believe efficient markets, you make
If
profits from speculations
profits
If you believe efficient markets, profits will quickly
If
disappear, or they are related with “true” risk. If the
latter is the case, then you just earn risk premium560
latter Speculate using factor models
Suppose we want to speculate using the FX
Suppose
exposure (For example, we want to make profits
by betting on the depreciation of US dollar)
by
For Excel, see “speculation” in
For
“Lecture2_StockPricing”
“Lecture2_StockPricing”
As you can see, by longing Dell and shorting GE,
As
we achieve: (1) a “zeroinvestment” portfolio (2)
“neutralize” the risk exposure to market risk, so
that we can eliminate other risks; (3) So, if we
believe that the USD will depreciate against Euro,
we will profit over 9% when USD depreciates by
1%.
561 The implementation of multifactor
models (1)
Basically speaking, you can throw in any
Basically
economic variables in the multifactor model
You can throw in any quantitative economic
quantitative
variables in the multifactor model
variables
For qualitative variables (e.g., country risk), you
For qualitative
can assign number to each status (e.g, 1least
risky, 5 most risky) and thus “quantify” these
variables
variables
Or, you can use binary distribution to present the
Or,
variable in regression (1 = risky countries; 0 =
riskless countries)
riskless
562 The implementation of multifactor
models (2)
So, how many variables should we consider?
There are two stepwise approaches:
There
stepwise
Narrowdown: First, throw in all variables in the
Narrowdown First,
regression; then, drop the insignificant ones and
rerun the regression. Keep doing so until you
find all variables appear to be significant.
find
Expansion: First, only include one variable that
First,
should be the most important one (e.g., market
premium); then, throw in another variable to see
if this new one is significant – (a) if not, then
drop it and try another – (b) if yes, then keep it
and throw in another variable until all variables
have been tried. 563 Linking return models to dividend models 564 General dividend model for stocks
∞ Dt
Po = ∑
t
t = 1 (1+ rt)
P0 = Current stock price
Dt = Future cash dividends (can be fixed)
Future
rt = expected stock return from CAPM and
other linear models (can be fixed)
other
The key concept here is: your future
The
dividend is “discounted” by expected return
dividend 565 Example
Let’s use GE as the example because it is
Let’s
financially solid and provides constant dividends
financially
We go to check GE’s dividend process since
We
2000
2000
Now, we set the follows: (1) fixed r = 5% and (2)
Now,
dividends after 2011 remains the same
dividends
Let’s try to used dividend data since 2000 to
Let’s
estimate GE’s stock price in 2000
estimate
See “GE dividend” in “Lecture2_StockPricing”
The estimated price is 11.67 but the market price
The
was 50 at the end of December 2000, so GE
was overpriced back in 2000
was
566 Exercise 7
See “Exercise 7” tab in “Lecture2_StockPricing”
Try to compute GE stock price using data table
Try
function for various expected stock returns,
given (1) presumed dividend process and (2)
expected returns ranging from 1% to 10%
Let’s try to used dividend data since 2000 to
estimate GE’s stock price in 2000
estimate
Plot the expected returns (in Xaxis) and stock
Plot
price (Yaxis)
price 567...
View
Full Document
 Spring '09
 returns

Click to edit the document details