PART 2 - Behavioral Economics

Economic experiments (method) =/  behavioral economics (concepts)

Effect from A on B

Problem: How to establish a single cause?

• often note easy indetifiable (Great Chicago fire)

Role of manipulation:

• Holland/Rubin wrote: no causation without manipulation

• Granger said: many causal questions cannot be tackled within Holland’s Framework, because they cannot be treated (Race)

• a specific question that could theoret. be answered by literature

• Usually: What is the effect of A on B?” // What is the effect of race on earnings?

Most RQ are sucht that:

• there is a way to measure the implicit claim

• the implicit claim is:

  • generalizable to multiple units

  • falsifable

  • causal

Stable Unit Treatment Value Assumption

• no interference: potential outcomes for any unit do not vary with treatment of other units

• no versions of treatment: for each unit, no differnet versions of each treatment level (which lead to different potential outcomes)

• individuals

• two possible conditions: treated or not treated

—> man kann nur den expected value vom treatment-outcome beobachten, wenn auch getraetet wurde, nicht wenn T=0

• impossible to observe treated and not treated of the individual

• when we measure each individual either was treated or not

• aims to solve the fundamental problem of causal inference

1. ABA Design

2. Full / Complete randomization

3. Factorial design

4. Block design

5. Within-subject design

6. Crossover design

•       Experiment starts with condition A

•       …it continues with condition B…

•       …and it ends with condition A

•       It allows to observe behavior before treatment, during treatment, and after treatment is removed

-               Simplest form

-               Random sample from entire pool

-               No possible correlation with observable characteristics

-               Drawback: sample size in each treatment is random

• Solves random sample size problem

• Pre-determined no. of subjects assigned to each treatment/ contribution of treatments

• Subjects randomly assigned to each condition

• Subjects are divided according to one or more observable characteristics / blocks:

  • Age group

  • Gender

  • Nationality

• Minimizes effects of systematic error

• Treatments are randomly allocated to experimental units inside each block

• multiple conditions repeated over the same block

  • impact of subject-specific effect is eliminated

• Order effects: we need to randomize or balance the order in which every conditions is presented, so effects of the order are ruled out

• smaller sample size required

• for longitudinal or observationsal studies

• solves the Order effect

• Example: half subjects assigned to sequence “condition 1 then 2”, half assigned to sequence “condition 2 then 1”

  —> if there is a difference btw. Groups, the Order effect  is confirmed

• Controlled experiments (trials)

• Natural experiments

Popular experiments:

• Ultimatum game

• Dictator game

• Trust game

• Public good game

• Traditional economic theory predicts full rationality in individual or group decisions -> not actually the case

• acceptance that humance are not perfectly rational decision makers arrived late 1980s

• laboratory experiments are the tool to isolate causes behind apparently irrational behavior

Popular findings:

• reciprocity (Wechselseitigkeit)

• altruistic motives

• emotions

• Rational inattentions

  • human focus on subsets of their informations set

We need data:

• several choices over time

• Cost and benefits associated with each choice

• Each guess gives us an inequality involving delta

  —> we obtain a range for delta

• (impatience)

• Short-run vs. long-run impatience

• preference reversal (dynamic inconsistencies)

• Demand for commitment

deltas are not stable over time

—> one marshmallow now or in 1h is not the same as one marshmallow in 1 year or 1 year and 1h

—> preferences for both options are not the same as suggested by exponential discounting model (with exponential discounting, the same level of impatience for any delay of the same length is implied)

—> short-run impatience (e.g. payday loans) = tendency for immediate gratification / discount heavily on short-term decisions

BUT:

people are in fact quite patient in the long run, e.g.:

• save for retirement

• invest in education

• exercise often

• do problem sets

(preference reversal=dynamic incinsistencies)

• Dynamic consistency: action a person thinks she should take in the future always coincides with action that they actually prefer to take once time comes

• i.e. person’s preferences at different points in time are consistent with each other —> no “intra-personal conflicts”

Analytically:

—> in the real world plenty of examples of intra-personal conflicts

Solution to express both short-term impatience and long-run patience:

LEISURE GOOD:

INVESTMENT GOOD:

Student has some beta, delta and instantaneous (dis-)utilities and must do a problem set within a certain time frame.

• If future misbehavior raises cost of current misbehavior, sophistication helps in overcoming short-run impatience (tends to be true for investment goods with immediate costs)

• If future misbehavior lowers cost of current misbehavior, sophistication hurts in overcoming short-run impatience (tends to be true for leisure goods with immediate reward)

Urn with 300 balls, 100 red 200 blue or green (unkown prob.)

Q1: 100€ if right color guessed

Q2: 100€ if right color not guessed

Outcome:

Q1: Red —> implies Pr > Pg, Pb

Q2: Red —> implies Pr < Pg, Pb

—> Ambiguity aversion

Results of a maximin strategy:

• choosing the action that maximizws utility in the worst-case-scenario

(Prospect ≙ Lottery)

Alternative descriptive model of decision making under risk, aiming at accounting for violations of EUT (expected utility theory)

Two phases:

• Editing phase: preliminary analysis of offered prospects

• Evaluation phase:edited prospects are evaluated

Heuristic rules & operations to narrow down options to be considered (3-2-1-operations: CCC SS D):

• Coding outcomes as gains and losses relative to a reference point (Distance from a reference point X0)

  • Utility with reference point U(X/X0) = v(X-X0)

• Combination of probabilities associated with identical outcomes

• Cancellation: discarding components that are shared by offered prospects (isolation effect)

• Segregation of the risk-less from the risky components of the prospect

• Simplification: rounding probabilities and outcomes

• Detection of dominance: some prospects may dominate others

• some operations either permit or prevent others

• editing phase may generate anomalies

  • cancellation of common components may generate inconsistencies

  • simplifications that eliminate small differences may induce intransitivities

• same prospect in different framing could be edited differently

—> Individuals are assumed to choose prospect with highest value

• Defined on deviations from a reference point: carriers of value are changes in wealth/welfare, rather than final states

• Concave for gains & convex for losses: diminishing sensitivity: impact of change diminishes with distance to reference point

• Steeper for losses than for gains: loss aversion: losses loom larger than corresponding gains

• level of probability weight is a way of expressing risk tastes

• curvature captures how sensitive people are to differences in probability

• the level of proability weight is a way of expressing risk tastes (if you hate to gamble, you place low weight on any chance of winning anything)

  
  

• Kahneman & Tversky’s final weighting function shape is the inverted-s

• reflect principle of diminishing sensitivity: impact of change in probability diminishes with distance from boundary —> concave near 0, and convex near 1

• principle yields subadditivity for very unlikely events and superadditivity near certainty

• Not well-behaved near endpoints: very small probabilities can be either over-weighted or neglected altogether (rounding or certainty effect)

• individuals have a strong tendency to remain at the status quo

• Results of Samuelson & Zeckhauser:

  • an alternative becam more popular when it was designated as the status quo

  • advantage of status quo increases with the number of alternative

• people demand much more to give up an object than they would be willing to pay to acquire it

• reversibility of indifference cuves not given due to loss aversion

Example:

• Three subject groups:

  • sellers were given a coffee mug and were asked to indicate whether or not they would sell the mug at a series of prices ranging from $0.00 to $9.25

  • buyers indicated whether they were willing to buy a mug at each of these prices

  • choosers were asked to choose, for each of the possible prices, between a mug and cash

Median valuations were $7.12 for sellers, $3.12 for choosers, and $2.87 for buyers