Session6A

SESSION 6A: PROBABILITIES &
DECISION TREES IN FINANCE

Session 6
There is a chance!
1. Stock Market as a Random Walk
¨ One of the simplest, albeit weakest, tests of randomness is to test to see

whether there is a 50:50 chance that markets will go up (or down). Thus,
if you have a hundred days of market changes, you would expect to see 50
up and 50 down days.
¨ To test this proposition, you can look at price changes for n trading days,
and compute the number of days that the market was up (and down).
However, even if markets are random, the actual number that you find
will deviate from 50:50. The standard error is a function of the number of
trading days and can be computed as follows:
!.#∗!.#
¤ Std Error in probability= where n = Number of days in your sample.
%
¤ This can be generalized more generally to any two-outcome experiment as
&(()&)
Std Error in probability= where p = Probability of one of the outcomes
%
¤ Thus, if you observe a hundred trading days, the range, with 95% probability, on up
(or down) days, even if markets have a 50:50 chance of going up and down, would
be 45-55. (Std Error = 2.5%)
2
Probability that the market will go up (or
down)…
¨ There were 1257 trading days between January 1, 2016 and December
31, 2020. Over that period, the S&P 500 was
¤ Up on 700 trading days
¤ Down on 557 trading days
¨ Based upon that data, the probability that the market was up during the
period was 55.69% and the probability that the market was down was
44.31%.
¨ Can you use this to accept or reject the hypothesis that there was a 50%
chance of up/down days., at least during this period?
!.#∗!.#
¤ Std Error in probability, if random = = 0.0141 or 1.41%
(+#,
¤ Range with 95% confidence, on up probability = 0.5569 ±2 .0141 : Range: 52.87%
to 58.51% probability of up (or down) days, during this period
¤ At least during this period, you can reject the 50:50 up/down hypothesis.
¨ Can you extrapolate from this that there will be more up than down days
in the market in the future?
3
Conditional Probabilities…
¨ Using the same data )returns on the S&P 500 on a

daily basis from January 1, 2016, to December 31,
2020), and breaking down into up and down days:
¨ Converting these numbers into probabilities, you

get:
4
Cumulative Probabilities…
¨ If the probability that the market will go up (down) on a

given day was 55.69% (44.31%), all through the time
period, you can estimate the probabilities of the market
going up or down two, three or even ten days in a row.
¨ Thus, to estimate the probability that the market was up
three or five days in a row:
¤ Probability of three up days in a row = (0.5569)3 = .1727
¤ Probability of five up days in a row = (0.5569)5 = .0536
¨ Similarly, to estimate the probability that the market was
down three or five days in a row:
¤ Probability of three up days in a row = (0.4431)3 = .0870
¤ Probability of five up days in a row =(0.4431)5 = .0171
5
2. Transition Probabilities
¨ In investing and finance, it is common to put portfolio managers

and companies into groupings, reflecting their standing on a metric
(returns, debt ratios, dividends) or rankings in performance.
¨ Those rankings are not only used to judge investors and companies,
but are also used as predictors for the future. While every mutual
or hedge fund touts the warning that past performance is not a
predictor of the future, it is undeniable that funds that have
performed/ranked well in the past market themselves on that
basis, and that investors redirect their money to these funds.
¨ The test of whether chasing past performance is a good strategy
can be converted into a test of whether there is stickiness in
rankings. Put simply, if you ranked funds into groups based upon
success, do these rankings persist?
6
Money Manager Performance
The performance persistence varies across

- Markets (Geographical, asset class) Each of these probability
- Time (Some periods have more estimates has a standard
persistence than others) error that will decrease
- Style classes (value versus growth, as the sample size
small vs large cap)
It is also difficult for investors to convert
increases.
this statistical persistence into returns.
7
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
3.00%
3.50%
4.00%
4.50%
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
Global Corporate Default Rate (as % of Loans Outstanding)
2011
3. Probability of Corporate Default
2012
2013
2014
2015
2016
2017
2018
2019
2020
8
Conditional Probabilities: Bond Ratings
and Default Rates
9
A Multiple Discriminant Model of Default:
The Altman Z Score
Altman analyzed 66
manufacturing companies, of
which 33 became bankrupt
within the years 1946- 1965
and the other half were
existing companies in 1966.
10
A Probit Model: Hostile Acquisitions
¨ While there are no easy pathways to making money, it seems

clear that investors in companies that are targeted in
acquisitions (especially hostile ones) earn high returns, but
only if they invest before the event.
¨ There are probit models for predicting companies that will be
targeted, and they involve:
¤ You start with all firms that publicly traded at the start of a period
¤ The dependent variable becomes the stand-in for whether a firm is
targeted in a hostile acquisition
¤ The independent variables reflect what you believe are key drivers of
hostile acquisitions, including poor stock price performance, lagging
accounting returns and managers with little or no shareholdings.
¤ You build a probit model that will yield as output an equation that
resembles a regression, but will yield a probability of a hostile
acquisition.
11
Source: Hostility in Takeovers: In the Eyes of the Beholder, G.W. Schwert, JF 2000
3. Decision Tree: An Example
¨ Consider a pharmaceutical drug for treating Type 1 diabetes that has gone
through preclinical testing and is about to enter phase 1 of the FDA approval
process.
¤ Phase 1 is expected to cost $ 50 million and will involve 100 volunteers to determine safety
and dosage; it is expected to last 1 year. There is a 70% chance that the drug will successfully
complete the first phase.
¤ In phase 2, the drug will be tested on 250 volunteers for effectiveness in treating diabetes over
a two-year period. This phase will cost $ 100 million and the drug will have to show a
statistically significant impact on the disease to move on to the next phase. There is only a 30%
chance that the drug will prove successful in treating type 1 diabetes but there is a 10% chance
that it will be successful in treating both type 1 and type 2 diabetes and a 10% chance that it
will succeed only in treating type 2 diabetes.
¤ In phase 3, the testing will expand to 4,000 volunteers to determine the long-term
consequences of taking the drug. If the drug is tested on only type 1 or type 2 diabetes
patients, this phase will last 4 years and cost $ 250 million; there is an 80% chance of success.
If it is tested on both types, the phase will last 4 years and cost $ 300 million; there is a 75%
chance of success.
¨ If the drug passes through all 3 phases, the costs and annual cash flows are below:
Disease treatment Cost of Development Annual Cash Flow
Type 1 diabetes only $ 500 million $ 300 million for 15 years
Type 2 diabetes only $ 500 million $ 125 million for 15 years
Type 1 and 2 diabetes $ 600 million $ 400 million for 15 years
13
The Tree…
14
The Fold Back
15
4. Scenario Analysis: easyJet and Brexit in
2019
16
No Deal Brexit Bad Deal Brexit Soft or No Brexit

Restructuring cost £500 million £300 million $0
(up front)
Revenue growth 3.00% 5.00% 5.00%
Operating Margin 6.00% 7.00% 8.00%
Sales to Capital 1.73 1.73 1.73
Ratio
No Deal Brexit Delayed & Messy Soft or No Brexit

Brexit
Probability 25% 50% 25%
Value Per Share £12.02 £15.70 £19.38
Expected Value per share = .25 (£12.02) + .50 (£15.70) + .25 (£19.38) = £15.70
Aswath Damodaran
16

Session6A

Uploaded by

Copyright:

Available Formats

Session6A

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Session6A

Uploaded by

Copyright:

Available Formats

SESSION 6A: PROBABILITIES &

DECISION TREES IN FINANCE

¨ One of the simplest, albeit weakest, tests of randomness is to test to see

¨ Using the same data )returns on the S&P 500 on a

¨ Converting these numbers into probabilities, you

¨ If the probability that the market will go up (down) on a

¨ In investing and finance, it is common to put portfolio managers

The performance persistence varies across

¨ While there are no easy pathways to making money, it seems

No Deal Brexit Bad Deal Brexit Soft or No Brexit

No Deal Brexit Delayed & Messy Soft or No Brexit

You might also like