Transcript Experimental Approach to Business Strategy 45-922

Strategic Corporate
Management
45-870
Professor Robert A. Miller
Fourth Mini 2017
Teaching assistant:
Chi (Faith) Feng: [email protected]
Preamble
Being “strategic” means intelligently seeking your own
goals in situations that involve other parties who do not
share your goals, a theme we emphasize in this course.
In business “corporate” typically refers to a publicly
traded company with limited liability, a corporation
owned by shareholders. We will focus more broadly on
business entities, and management goals.
And “management” refers to organizing people, when
you lack the absolute powers of a dictator, but can wield
some incentives and help set the rules.
Course objectives
1. Recognize strategic situations and opportunities
2. Summarize the essential elements in order to
undertake an analysis
3. Predict the outcomes from strategic play
4. Conduct experiments, that is “human simulations”,
to verify and revise your predictions
5. Analyze the experimental data to increase your
knowledge and familiarity using simple statistics
. . . to help you make better strategic decisions.
Course materials
The course website is:
http://www.comlabgames.com/45-870
At the website you can find:
 the course syllabus
 power point lecture notes
 experiments you can download
 project assignments
 the on line (draft) textbook
 other reading material
Week 1
Introduction to Strategy:
Best Responses and Game Design
We begin this course by laying out the four
basic questions of every strategic situation.
Then we define the extensive form, and
explain: what we mean by the empirical
distribution of moves, and we mean by the
best response to that distribution. The last
part of the lecture turns to game design.
A cola war
After struggling through the Great Depression
of the 1930s Pepsi finds its soft drink sales are
stalled in the 1940s.
Coke is the industry leader, and its products
command a premium price over Pepsi’s.
The country is at war, but remains segregated
along racial lines, with blacks economically
and socially disadvantaged.
Who are the main players in
this episode?
Pepsi shareholders
Coke shareholders
Management at Pepsi
White cola demanders
Black cola demanders
What options or choices
face the main players?
Pepsi could target its product line to African
American consumers, Pepsi could target a new
product line to African American consumers, or
Pepsi could pursue another strategy, such as
expanding its operations in Canada.
Coke could respond aggressively or passively
to any marketing initiative taken by Pepsi.
White consumers might be alienated by a
marketing campaign that targets African
American consumers.
How do the players evaluate the
consequences of their choices?
If Coke responds to an advertising campaign
both firms will sell more cola in return for
lower profits.
If Coke does not respond to Pepsi, how much
value will be added or lost to each company?
If the white community is alienated by both
companies targeting the African American
community, would Coke be hurt more than
Pepsi?
Where are the sources of
uncertainty in this unfolding drama?
Will white cola drinkers be alienated by the
introduction of a marketing campaign that
targets the African American community?
If both companies target blacks, the
probability of alienating whites is higher than
if only Pepsi does.
Moreover as the company with the bigger
white market share, Coke has more to lose in
this case.
Illustrating the four critical questions
in an extensive form game
The extensive form
The game we just played was represented by its
extensive form.
The extensive form representation answers the four
critical questions in strategy:
Who are the players?
What are their potential moves?
What is their information?
How do they value the outcomes?
These are the ingredients used to design all games in
extensive form.
Who is involved?
How many major players are there, and whose
decisions we should model explicitly?
Can we consolidate some of the players into a
team because they pool their information and have
common goals?
Should we model the behavior of the minor players
should be modeled directly as nature, using
probabilities to capture their effects on the game?
Does nature play any other role in resolving
uncertainty, for example through a new technology
that has chance of working?
What can they do?
Each node designates whose turn it is. It could be a
player or nature. The initial node shows how the
game starts, while terminal nodes end the game.
A branch join two nodes to each other. Branches
display the possible choices for the player who
should move, and also the possible random
outcomes of nature’s moves.
Tracing a path from the initial node to a terminal
node is called a history. A history is uniquely
identified by its terminal node.
What are the payoffs?
Payoffs capture the consequences of
playing a game.
They represent the utility or net
benefit to each player from a game
ending at any given terminal node.
Payoffs show how resources are
allocated to all the players contingent
on a terminal node being reached.
What do they know?
Each non-terminal decision node is associated with an
information set.
If a decision node is not connected to a dotted line, the
player assigned to the node knows the partial history.
If two nodes are joined by a dotted line, they belong to
the same information set, and the two sets of branches
emanating from them, which define the player’s choice
set, must be identical.
A player cannot distinguish between partial histories
leading to nodes that belong to the same information
set.
Changing the information
available to Coke
Suppose we draw
a dotted line
connecting the
two decision
nodes for Coke.
Then we prevent
Coke from seeing
which market
Pepsi enters,
before it chooses
whether to
acquiesce or not.
The empirical distribution
The empirical distribution is the probability
distribution of choices made by all the
players in the game.
It is formed from the relative frequencies of
choices made at each information set
observed in the experiment.
For example, the empirical distribution
characterizing Coke comprise two relative
frequencies, showing how likely Coke is to
cut price if Pepsi advertises in the:
1. Quebec market
2. African American market
Best response to empirical distribution
For any given player (Pepsi), treat the moves by all
other players (Coke) as nature.
Form the choice probabilities for the other players
from the relative frequencies of the choices observed
in the experiment (by experimental subjects playing
Coke).
This transforms the game into a dynamic
programming problem for the given player (Pepsi).
Use dynamic programming methods, such as
backwards induction, to find the (Pepsi’s )best
response to the empirical distribution.
To compute the best response for Pepsi you need to
know Pepsi’s payoffs but not Coke’s.
The Ware case
10 years ago Ware received a patent for Dentosite that
has since captured 60 percent share in the market.
National had been the largest supplier of material for
dental prosthetics before Dentosite was introduced.
A new material FR 8420 was recently developed by NASA.
If Ware develops a new composite with FR 8420 it will be
a perfect substitute for Dentosite.
If the technique is feasible then Ware would have just as
good a chance as National of proving it first.
If Ware develops it first they could extend the patent
protection to this technique and prevent any competitors.
Strategic considerations
Ware’s problem is bound to National’s.
Ware does not want to develop a technology that
would not be used if the competitor does not
develop it.
If National develops the technology Ware cannot
afford to drop out of the race.
It all depends how people at National see this
situation. Are Ware and National equally as well
informed?
Some facts
The Ware Case
10% Discount/year
$1.9091 Value of $1 in years 1+2
$3.4462 Value of $1 in years 3+4+5+6+7
$0.500 Ware entry cost/yr
$1.000 National entry cost/yr
$15 Range of possible
$20 future annual sales (millions)
50% Probability process feasible
50% Ware chance of winning race
50% Ware market share if Nat'l enters mkt
20% Ware profit margin
Ware case in the extensive form
Using the facts we can present the case
in the following diagram:
Simplifying the extensive form
Folding back the moves of chance that are
related to developing a new technology we
obtain the following simplification.
What should National do?
National is indifferent between the two choices if the
expected profits are equal.
If Ware chooses “in” with probability p, the value to
National from choosing “out” is 0, and the expected
profits to National from choosing “in” are:
-0.401*p + 1.106*(1 – p)
Solving for p we obtain:
-0.401*p + 1.106*(1 – p) = 0
=>
p = 0.734
Thus if Ware enters with a higher probability than
0.734, then National should stay out, but if Ware
enters with a lower probability than 0.734, National
should enter itself.
What should Ware do?
If National chooses “in” with probability q, then the
expected value to Ware from choosing “in” is:
-2.462*q - 0.955*(1 – q)
Also the expected value to Ware from choosing “out” is:
-3.015*q
Solving for q we obtain:
2.462*q + 0.955*(1 – q) = 3.015*q
=>
q = .633
Thus if National enters with probability higher than
0.633, then Ware should enter too, but if National
enters with probability lower than 0.633, then Ware
should stay out.
Rule 1
Play a best response to the empirical
distribution as best you understand it.
Summary
The extensive form summarizes the four dimensions
of every strategic situation (players, actions,
information and consequences).
The empirical distribution describes how all the
players choose their moves. If you know the
probabilities of the moves for everyone else, you can
calculate your own expected value from making the
different choices that you have.
Playing the best response to the empirical
distribution mimics optimal decision making. It only
requires you to know the other players, and their
choice probabilities.