Decision Bias in the Newsvendor Problem

Decision Bias in the Newsvendor Problem

with a Known Demand Distribution:

Experimental Evidence

Schweitzer and Cachon, 2000

Introduction

• Newsvendor problem: short selling season with stochastic demand, one order opportunity, balance between underage & overage costs.

• Problem is well-known, but there is little engagement with how managers make decisions.

• Order Q ex-ante is rarely the best order Q ex-post.

Introduction

• Managers’ decisions don’t correspond to expected profit-maximizing order Q (Fisher and Raman 1996), but no explanation provided.

• Anchoring and insufficient adjustment bias (Sterman 1989).

• Influence of Supply chain design (Croson and Donohue 1998).

Introduction

• Preferences other than profit maximization – risk-averse behavior (Eechhouldt et al. 1995).

• Inventory level heuristic.• Biased forecast – done away with

in this paper.

Descriptive Models

• Inefficient order Q not consistent with:- Risk aversion- Risk-seeking preferences- Prospect theory preferences- Loss aversion- Stockout aversion.- Undervaluing opportunity costs• Consistent with: - Preference to reduce ex-post inventory error- Anchoring and insufficient adjustment.

The Newsvendor Problem

• q, D, μ, F, f, c, p, s.

• High- and low-profit products:

Utility Maximizing Orders

Risk Neutral Preferences

• un(w) = w -> max un(w) = max E[π(q,D)]

• qn is the optimal risk-neutral order Q

Risk-averse and Risk-seeking Preferences

Prospect Theory Preferences

• Risk-averse over domain of gains and risk-seeking over domain of losses (Kahneman and Tversky 1979);

• reference point = current wealth.• with all gains = less than qn.• with all losses = more than qn.• with both = depends.

Loss-averse Preferences

q1<qn δq1/δλ<0

Waste-Averse Preferences

qt<qn

Stockout-Averse Preferences

qm>qn

Underestimating Opportunity Costs

qo<qn

Preference for Minimizing Ex-Post Inventory Level

Anchoring and Insufficient Adjustment

• mean anchor heuristic – on mean demand, adjusting towards qn.

• chasing demand – on prior order Q, adjusting towards prior D.

Experiment 1: Uniform Demand

• 34 MBA students• selling “wodgets”• not informed about total # of

rounds, or future price and cost.• 30 inventory decisions• critical fractiles: 25 and 75%-> 75

and 225 qn.

Experiment 1

Experiment 1

Experiment 1: Procedure

• Initial inventory decisions – unconfounded

• Average high-profit order Q = 176.68<225;

• Average low-profit order Q = 134.06>75.

• Double repeated measures generalized linear model.

• Mean and chasing demand heuristics.

Adjustment Scores

• mean anchor heuristic:

• chasing demand heuristic:

Experiment 2: High Demand Distribution

• 44 2nd year MBA students• losses are not possible <-

increased D range: low [1,300] and high [901,1200]

• qn low: 225 or 75• qn high: 1125 or 975

Results

• High-D: 186.88 or 1092.55• Low-D: 142.17 or 1021.81