SCHOOL OF SOCIAL SCIENCES
Edinburgh Business School
Financial Markets Theory
C38FM
Prof. Mustafa Caglayan
Finance Group
Heriot-Watt University, Edinburgh
Investments and Portfolio Management
Efficient Markets Hypothesis
Market efficiency
Operational
Transaction costs kept to a minimum.
Allocational
Prices are determined by equating the marginal returns of
all producers and savers.
Informational
All relevant information is fully, and instantaneously, priced
in.
Fama’s 3 Levels
Weak form efficient
Security prices fully, and instantaneously, reflect
information contained in past prices.
Semi-strong form efficient
Security prices fully, and instantaneously, reflect all publicly
available information.
Strong form efficient
Security prices fully, and instantaneously, reflect all
available information, whether public or private.
Testing weak form efficiency
Past price changes cannot be used to predict future price changes
Today’s price change should be unrelated to yesterday’s, or the day
before...
Random walk
Do prices follow a `random walk’? This means that the sequence of
price changes is indistinguishable from a series of numbers generated
by a random variable with zero mean and known probability
distribution.
If prices follow a ’random walk’ then the correlation coefficient between
price changes on day t and price changes on day t + 1 will be zero ⇒
they are serially uncorrelated.
𝜎∆𝑃𝑡 𝑃𝑡+1
Serial correlation coefficient 2
𝜎∆𝑃
𝑡
Serial correlation coefficients,
Fama (1965)
Runs test
A run is a sequence of price changes of the same
sign. Suppose you have n1 positive price runs and
n2 negative runs, then statistical theory leads us to
predict average runs and variance
𝑛1×𝑛2 (−1)(−2)
= +1, 2 = , where N = n1+n2
(𝑛1+𝑛2)/2 N−1
𝑁−µ
Then one can test Z =
An Example
• For example, a series of 20 coin tosses might produce
the following sequence of heads (H) and tails (T).
HHTTHTHHHHTHHTTTTTHH
1 2 3 4 5 6 7 8 9
The number of runs for this series is 9. We have 11 Heads and 9 Tails.
n(heads): 5; n(tails): 4, (i.e., n1 = n(heads); n2 = n(tails). APPLY the
formulas to get
5×4 (5.44−1)(5.44−2)
= +1 = 5.44, 2 = = 1.913;
(5+4)/2 (9−1)
9−5.44
Z= = 1.85 (z5%=1.96; hence we cannot reject randomness.)
1.913
Runs test - Dryden,1969
Testing semi-strong form efficiency
Published information cannot be used to predict future
price changes
Carried out using an event study methodology. What are
relevant events?
• Receiving or making a takeover bid.
• Announcing a dividend change.
• Announcing a major investment project.
• Announcing a share buyback.
How do we know if it has an effect?
Measure relative to what would have occurred otherwise, if
it had responded normally.
Event study
• Event window
𝑡0 is the time of the event of interest - define an
event window during which all effects of the event
are anticipate to occur, both before and after 𝑡0 .
Event study
Residual return
Residual return = Actual return - Predicted normal return
where the predicted normal return is based on that of a
representative sample, such as a sectoral index, or an
estimate, such as CAPM.
• 𝑅𝑖𝑡 is the residual return of the i-th company in period t.
σ𝑁
𝑖=1 𝑅𝑖𝑡
• 𝑅𝑖𝑡 = is the average residual return in period t.
𝑁
𝐶𝐴𝑅𝑇 = σ𝑇𝑡=𝑡1 𝑅𝑖𝑡 is the cumulative average residual return.
Event study of a dividend increase
Event study of a dividend increase - effect of the
increased dividend
Event study of a dividend increase - highly efficient
market
Testing strong form efficiency
Private information cannot be used to predict future price
changes
We do not know who has private information, so we can
only answer this by searching and finding a group, or class,
of investors who systematically earn superior returns. This
could be professional fund managers, insiders...
Lorie and Niederhoffer, 1968
LN identified a sample of trades by insiders and examined
their returns over a six month period.
How efficient are capital markets?
These examples, and many subsequent tests,
suggest that the market is indeed semi-strongly
efficient (therefore also weakly efficient) such that
even fundamental analysis will be ineffective.
On the other hand, we can find investors who
consistently outperform the market, suggesting
that private information can be used to predict
future share prices.
