Strategies in Dynamic Decision Making

1
Strategies in Dynamic Decision Making

• An Experimental Investigation on the Rationality
  of Decision Behaviour

John D. Hey LUISS University of York Julia A.
Knoll University of Düsseldorf
2
Content

• Introduction
• Experimental Design
• The Decision Strategies
• Summary

3
Introduction

• Dynamic decision problems under risk

end

• sequence of interdependent decisions
• external forces - Nature
• Nature moves with a known probability
• very complex

payoff nodes
decision node
chance node
start
4
Introduction

• How do people solve dynamic decision problems?
• Economic Theory vs. Experimental Evidence
• Theoryrational dynamic decision making implies
  to plan ahead ? optimal decision on every
  decision level
• Experiments subjects fail to anticipate or plan
  their future decisions (Carbone Hey, 2001 Bone
  et al. 2003)
• ignore the fact that they will take decisions in
  the future
• simplify the decision problem.

5
Experimental DesignThe decision tree

• 3 decision levels
• Nature moves
• with equal probability
• independent of previous moves
• special features
• notepads
• replay function
• dominance property of the payoffs

6
Experimental DesignImplementation of the
Experiment

• 4 attempts with the same set of payoffs
• payment
• payoff on each attempt ? 4 payoffs
• final payoff randomly chosen out of those 4
  payoffs

• EXEC, University of York
• 92 subjects
• individually at their own speed
• instructions
• written
• power point
• questions

7
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

8
The Decision StrategiesEffort Minimizer /
Ignorants

• do not check payoffs at all or
• checked payoffs arbitrarily but ignored this
  information
• few subjects wrong decision on DL3 although they
  have checked the remaining two payoffs before
• very fast
• almost no cognitive effort

9
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

10
The Decision StrategiesBackward Inducters
Overview

• in common tackle the decision problem backwards
• three subgroups with different degrees of
  rational decision behaviour
• Rationalists completely rational
• Quasi-Rationalists almost rational
• Simplifier partially rational
• different degrees of Backward Induction

11
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

12
The Decision StrategiesBackward Inducters The
Rationalists

• subjects who behave according to the assumptions
  of economic theory
• top-down
• starting point DL3
• infer decisions for DL 3
• from decisions on DL3 decisions for DL2 are
  backward inducted
• last step inference of decision for DL1 from
  decision on higher decision levels

13
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

14
The Decision StrategiesBackward Inducters The
Quasi-Rationalists

• top-down
• almost the same decision behaviour as
  Rationalists
• BUT mistake on DL2
• take irrelevant information into account when
  inferring the decision for DL1
• only 3 subjects (3)

15
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

16
The Decision StrategiesBackward Inducters
Simplifier the desperates

• top-down
• backward induct the optimal decision until DL2
• get stuck on DL2
• apparently do not know how to further reduce the
  information
• bottom-up take a random decision on DL1

17
The Decision StrategiesBackward Inducters
Simplifier Effort Time Savers

• first bottom-up
• first decision on DL1
• number of decisions reduced to 6
• then top-down
• backward induct the optimal decisions for the
  remaining branches

18
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

19
The Decision StrategiesForward Worker

• bottom-up
• check (most of the) 64 payoffs
• make a move on DL1
• check the remaining 16 payoffs
• make a move on DL2
• check the remaining 4 payoffs
• make a move on DL3
• no decision is taken in advance
• decisions will only be made when subjects get to
  a decision node

20
The Decision StrategiesOverview

• Effort Minimizer / Ignorants
• Backward Inducters
• Rationalists
• Quasi- Rationalists
• Simplifier
• Desperates
• Effort Time Savers
• Forward Worker
• Strategy Mixers

21
The Decision StrategiesStrategy Mixers

• subjects use different strategies
• could not be assigned unequivocally to one of the
  groups
• guessing
• take in new information
• changes the way they tackle the decision problem

22
Summary

• there are subjects who do behave according to
  economic theory
• Rationalists
• only 12
• few subjects seem to have an idea how to solve
  the decision problem but fail when implementing
  their strategy
• Quasi-Rationalists and Desperates
• majority of subjects does not decide rationally
• decision strategies are fast and do not require a
  lot of cognitive effort
• cost-benefit analysis

23

• Thank you very much for your attention!