Supply Chain Management

Chapter 7
Behavioral (SCM or OM) Management

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE0

Learning objectives
• Understand how and why laboratory experiments are used in supply
chain/operations management.
• Understand behavioral models that explain observed newsvendor
behavior in the lab.
• Understand why these models are important for supply chain
coordination.

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE1

Agenda
7.1. Laboratory experiments
7.2. Explaining newsvendor behavior in the lab: anchoring
7.3. Supply Chain Contracting: preference for ex-post inventory errors

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE2

7.1 Laboratory experiments

Newsvendor („Newsboy“, „Zeitungsjunge“, „Zeitungsverkäufer“)

• Imagine you are a newsvendor

• You have to decide in the morning how much newspapers to buy (in order
to sell them to your customer)

• You pay per newspaper: 1 €

• Your revenue: 4 €

• Every paper not sold at the end of the day: zero salvage value

• You know that demand is uniformly distributed between 0 and 300

• How much do you order?

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE3

7.1 Laboratory experiments

• Students - pay for performances

• Decisions with real consequences (vs. hypothetical decisions)
• E.g., better newsvendor orders  higher payoffs

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE4

 7.1 Laboratory experiments

profit margin

Schweitzer and Cachon (2000)

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE5

7.1 Laboratory experiments
Why laboratory experiments
• Laboratory studies complement other methods (see Katok 2012)
– bridge the gap between analytical models and real business problems
– Analytical models: parsimonious and general, primarily normative,
assumptions make the mathematics tractable, highly internally valid
• Internal validity: extent to which a causal conclusion is warranted (Safety
stock increases when demand uncertainty increases, ceteris paribus!)
• External validity: extent to which it is warranted to generalize results to
other contexts.
• All empirical methods involve a trade-off between internal and external
validity.
– Surveys and field studies (secondary data) have high external validity (close to
the real settings being studied), but may be low on internal validity.
– Lack of ability to establish the cause and effect relationship
• suffer from being confounded (several possible explanations)
• or not having all the data that would ideally be required.

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE6

7.1 Laboratory experiments

• “Why do we allow wild animals to cross exactly here?”

• „Deer Crossing“: [Link]

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE7

7.1 Laboratory experiments
Why laboratory experiments

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE8

7.1 Laboratory experiments
Why laboratory experiments
• Empirical methods (including e.g. Machine Learning, Big Data
techniques…) measure correlation, but this may be different from root
cause
– Cause-effect : one – and exactly one – factor needs to be varied, keeping all
other factors constant!
• Laboratory experiments can accomplish that by hypotheses testing

Hypotheses (exemplarily)
Cause Effect
Crossing sign Wild animals cross the street
Loss aversion People order less than optimal
(see later on)

Watch:
[Link]
[Link]

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE9

7.1 Laboratory experiments
Why laboratory experiments
“Pull-to-center” effect
Low profit margin High profit margin
(Revenue: 1.33€) (Revenue: 4€)

Cause Effect
Crossing sign Wild animals cross the street
Loss aversion People order less than optimal
Anchoring on mean demand People order less for high profit
margins and more for low profit
margins
B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE10
Agenda
7.1. Laboratory experiments
7.2. Explaining newsvendor behavior in the lab: anchoring
7.3. Supply Chain Contracting: preference for ex-post inventory errors

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE11

7.2. Explaining newsvendor behavior in the lab
Overview

Pull-to-center effect
• Newsvendor experiments show that subjects make sub-optimal orders
• „Too high – too low“ pattern
– Orders are too low under high profit margins
– Orders are too high under low profit margins
• Explanations:
– Bounded rationality
• Subjects would like to maximize expected profits but are not able to do so
• Biases and cognitive limitations
– Preference based
• Subjects pursue goals that are different from expected profit
maximization
• … but they are still optimizing (… just a different goal….)

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE12

7.2. Explaining newsvendor behavior in the lab
Anchoring heuristic

Anchor-heuristic
• Anchor-factor: 0 ≤ 𝛼 ≤ 1
– Newsvendor anchors with weight 𝛼 on mean demand
• Heuristical order size under anchor (a) heuristic
– 𝑄 = 𝛼 ⋅ 𝜇 + 1 − 𝛼 ⋅ 𝑄∗
Example
• 𝑈 0,100 → 𝜇 = 50 , 𝑝 = 1, 𝑤 = 0.25, 𝑠 = 0, 𝐶𝑅 = 0.75
– Optimal order size: 𝑄 ∗ = 𝐹 0.75 = 75

• Observed order size: 𝑄 = 60

– 𝑄 = 𝛼 ⋅ 𝜇 + 1 − 𝛼 ⋅ 𝑄∗ = 𝛼 ⋅ 50 + 1 − 𝛼 ⋅ 75 = 60
 𝛼 = = 0.6

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE13

7.2. Explaining newsvendor behavior in the lab
Anchoring heuristic

Expected cost of anchoring

• For the uniform distribution we get (see Chapter 4)
– 𝐸 𝑃(𝑄) = 𝑝 − 𝑤 ⋅ 𝑄 − 𝑝 − 𝑠 and for 𝑎 = 0 and 𝑠 = 0 follows
⋅
– 𝐸 𝑃(𝑄) = 𝑝 − 𝑤 ⋅ 𝑄 − 𝑝 = 𝑝−𝑤 ⋅𝑄−
• Expected profits under anchoring
⋅
– 𝐸 𝑃(𝑄 = 𝑄 ) = 𝑝 − 𝑤 ⋅ 𝑄 − where
• 𝑄 = 𝛼 ⋅ 𝜇 + 1 − 𝛼 ⋅ 𝑄 ∗ and
• 𝑄∗ = ⋅ 𝑏 (for uniform distribution!)

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE14

7.2. Explaining newsvendor behavior in the lab
Anchoring heuristic

Order size 𝑸𝜶

𝜶
Expected proftis
𝐸 𝑃(𝑄 = 𝑄 )

𝜶
B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE15
7.2. Explaining newsvendor behavior in the lab
Alternative behavioral models

Other well known behavioral models (e.g. for bachelor thesis?)

• Demand chasing
• Probabilistic choices
• Preference for minimizing ex-post inventory error
• Loss aversion
• ….

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE16

Agenda
7.1. Laboratory experiments
7.2. Explaining newsvendor behavior in the lab: anchoring
7.3. Supply Chain Contracting: preference for ex-post inventory errors

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE17

7.3. Supply Chain Contracting
Preference for minimizing ex-post inventory error

• So far: buyer is expected profit maximizing

• Now: newsvendor suffers more from underages/overages than the actual
(opportunity cost)
– For every unit of the difference between the demand realization 𝑑 and the
order size 𝑄 the newsvendor suffers a (psychological) cost of 𝛿 (on top of the
actual (opportunity) cost)
– We call it a preference for “minimizing ex-post inventory differences”

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE18

7.3. Supply Chain Contracting
Preference for minimizing ex-post inventory error

Example: Psychological cost

• Supplier sells at wholesale price 𝑤
• Critical fractile with psychological cost:

• Supply chain optimal critical fractile:

𝑝−𝑤+𝛿 𝑝−𝑐
=
𝑝 − 𝑠 + 2𝛿 𝑝 − 𝑠

Rearranging gives
⋅ ⋅ ⋅( )
𝑤=

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE19

7.3. Supply Chain Contracting
Preference for minimizing ex-post inventory error

Example: 𝑈[0, 𝟏𝟎𝟎], 𝑝 = 1, 𝑠 = 0, 𝜹 = 𝟐

𝒘 Supplier makes
profits
even under
coordinating
wholesale price
contracts!

 Works only for

critical ratio smaller
w=c
than 0.5 (low profit
margin): buyer
𝒄 orders too much
with 𝛿 > 0

0.5

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE20

Summary & Outlook

• Understand how and why laboratory experiments are used in supply

chain/operations management.
– To test theory and bridge the gap between theory and practice.
– Highly internally valid conclusions regarding root-cause effects.
– Important: only one factor is allowed to be varied in order to establish root-
cause effect relationship.

• Understand behavioral models that explain observed newsvendor

behavior in the lab.
– Behavioral models capture observed behavior
– They are useful to derive hypothesis (ceteris paribus analysis is possible) and
may be used for theoretic analysis of other situations (e.g., supply chain
context)
– Two broad categories: Bounded rationality and preference based explanations

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE21

Summary & Outlook

• Understand why these models are important for supply chain

coordination.
– Our coordinating contracts (quantity discounts, risk sharing contracts) assume
that all parties are fully rational and expected profit maximizing
– Laboratory experiments show that this does certainly not hold in the lab
– It seems reasonable that it does not hold in practice
– Outlook: The behavioral models introduced
• may serve as input for designing behaviorally robust contracts ( active
research field, see Becker-Peth et al. (2013))
• may provide insights why mathematically equivalent contracts (e.g., buy-
back and revenue sharing contracts) perform differently in the lab (see
Katok and Wu 2009)

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE22

References

• Schweitzer and Cachon (2000) Decision Bias in the Newsvendor Problem with a Known
Demand Distribution: Experimental Evidence. Management Science 46 (3), 404–420.
• Becker-Peth, Katok and Thonemann (2013) Designing Buyback Contracts for Irrational But
Predictable Newsvendors. Management Science 59 (8), 1800–1816.
• Bolton, Ockenfels and Thonemann (2012) Managers and students as newsvendors.
Management Science, 58(12), 2224-2233.
• Kahneman and Tversky (1979) Prospect theory: An analysis of decision under risk. In:
Econometrica: Journal of the econometric society, S. 263–291.
• Katok and Wu (2009) Contracting in Supply Chains: A Laboratory Investigation. Management
Science 55 (12), 1953–1968.
• Katok (2012) Using laboratory experiments to build better operations management models.
Foundations and trends in technology, information and operations management 5(1) , 1–88.
• Long and Nasiry (2014) Prospect Theory Explains Newsvendor Behavior: The Role of
Reference Points. Management Science 61 (12), 3009–3012.
• Luce (2005) Individual choice behavior: A theoretical analysis. Courier Corporation

• German: Thonemann (Operations Management), 3 edt., Chapter 10

B EHAVIORAL P ROF. D R . GvUIDO VOIGT S LIDE23