Int. J.

Production Economics 169 (2015) 169–178

Contents lists available at ScienceDirect

Int. J. Production Economics

journal homepage: www.elsevier.com/locate/ijpe

A behavioral experiment on inventory management with supply

chain disruption
Sourish Sarkar a,n, Sanjay Kumar b
a
Sam and Irene Black School of Business, Pennsylvania State University—Erie, 5101 Jordan Road, Erie, PA 16563, USA
b
College of Business, Valparaiso University, Valparaiso, IN 46383, USA

art ic l e i nf o a b s t r a c t

Article history: Various operational strategies for mitigating supply chain disruption have been studied theoretically, but
Received 10 February 2015 few studies have investigated behavioral decision-making in multi-echelon supply chains experiencing
Accepted 27 July 2015 disruptions. We explore the effects of communicating disruption information in real-time to supply
Available online 5 August 2015
chain members using the beer distribution game in a controlled laboratory setting. Both upstream
Keywords: (manufacturer) and downstream (retailer) disruptions are independently considered, and in each of
Bullwhip effect these scenarios, the difference between sharing and not sharing the disruption information is
Behavioral operations investigated. We find that supply chain disruptions may cause higher order variability when compared
Multi-echelon supply chain to the base case (no disruption). For a disruption at an upstream echelon, sharing the disruption
Disruption
information is found beneficial in reducing order variability and supply chain cost—upstream echelons
experience more benefits from information sharing than downstream echelons. Therefore, we advocate
that manufacturers share supply disruption information in real-time in order to benefit from a reduced
bullwhip effect and its associated costs. For a disruption at a downstream echelon, sharing disruption
information does not appear to have a significant benefit. Past studies have shown the importance of
sharing downstream inventory information with upstream supply chain members. In the event of
disruptions, our results demonstrate that sharing upstream disruption information with downstream
members is beneficial.
& 2015 Elsevier B.V. All rights reserved.

1. Introduction
Supply chain disruptions are a growing concern for firms,
especially those firms that depend upon global networks of
high-efficiency/low-inventory suppliers (Knemeyer et al., 2009).
Announcements of supply chain disruptions have significant
impacts on both short- and long-term financial performances
(Hendricks and Singhal, 2003, 2005). While supply chain man-
agers are well aware of the consequences of disruptions, a majority
of them are reluctant to build resilient supply chains due to
operational cost increases (Chopra and Sodhi, 2014). A recent
survey (Deloitte, 2013) found that 45% of the executives admit that
their supply chain risk management programs are only somewhat
effective or not effective at all; a majority of the surveyed
executives identified organizational silos and lack of collaborations
as the main barrier for effective risk management programs.
A large amount of academic literature has been devoted to
studying risk mitigation strategies in general, and information shar-
ing and collaboration strategies in particular (cf. Christopher and
n
Corresponding author. Tel.: þ 1 814 898 6800; fax: þ 1 814 898 6223.
E-mail address: szs15@psu.edu (S. Sarkar).
Peck, 2004; Kleindorfer and Saad, 2005; Tang, 2006; Gurnani et al.,
2012; Heckmann et al., 2015). However, today's supply chains are
typically multi-echelon, and the disruption literature rarely addresses
the issues that arise from the dynamics of a multi-echelon serial
supply chain. The performance of such a supply chain is different
from that of a two-echelon system as the bullwhip effect (Lee et al.,
1997), which is an increase in order variability at the upstream
echelon, is a typical observation for the former, but not latter. In
addition to the operational reasons, behavioral reasons also exist for
the bullwhip effect (Croson and Donohue, 2006). The scant literature
available on multi-echelon supply chain disruption focuses only on
analytical and simulation models (e.g., He and Zhao, 2012; Baghalian
et al., 2013; Schmitt and Singh, 2012; Schmitt et al., 2015). Research
on the behavioral aspects of firms in the wake of disruption is limited
(Bode et al., 2011).
Hendricks and Singhal (2005) found empirical evidence that
firms typically postpone public announcements of disruptions.
Anticipated stock price declines and high equity risks may drive
the managers of the disrupted facilities to conceal the information
related to disruption (or at least underreport the information to the
supply chain partners). Section 409 of the Sarbanes–Oxley Act of
2002 mandates real-time disclosure of any event that can affect a
publicly-listed firm's finance or share price. However, the reporting

http://dx.doi.org/10.1016/j.ijpe.2015.07.032
0925-5273/& 2015 Elsevier B.V. All rights reserved.
170 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
of a supply chain disruption is still discretionary. Schmidt and
Raman (2012) have shown that the managers disproportionately
underreport disruptions that are less damaging to the firm value.
The New York Times (Bunkley, 2011) reported that following the
Japanese earthquake and tsunami in March of 2011, automobile part
suppliers in Japan released only limited information to U.S. auto-
mobile companies on the extent of damage caused by disruption as
well as tentative timelines for resuming production.
Underreporting or hiding of disruption information may affect
supply chain performance. The disruption risk management lit-
erature has emphasized the importance of having “supply-chain
wide visibility of vulnerabilities” (Kleindorfer and Saad, 2005). A
multi-industry empirical study by Blackhurst et al. (2005) also
indicated the importance of “real-time sharing of correct informa-
tion” across supply chains in mitigating disruptions. To study the
effect of information sharing, we use laboratory experiments as
this approach offers tight control over decision environments (Falk
and Heckman, 2009). A number of past studies conducted labora-
tory experiments to investigate behavioral decision making and
the bullwhip effect in serial supply chains; nevertheless, none of
these studies, to our knowledge, examined the effect of real-time
disruption information sharing in a multi-echelon supply chain,
which is the central focus of our study.
A beer distribution game framework is followed in our experi-
ments. The beer game was first introduced by Forrester (1958) and has
been used in numerous studies on behavioral experiments. The basic
game mimics a serial supply chain using four players (i.e., a retailer,
wholesaler, distributor, and manufacturer), who individually manage
inventories and operating costs at four locations. Customer demand is
exogenous. The players are asked to minimize the supply chain costs
(i.e., sum of inventory holding and backorder costs incurred at all
echelons), which is a challenging task due to the presence of lead-
times in the supply chain. We incorporate supply chain disruption in
the traditional beer game. Although practically any echelon of a supply
chain can face disruptions, we consider only two locations of disrup-
tions in this research: the upstream (i.e., a manufacturer disruption)
and downstream ends (i.e., retailer disruption) of the supply chain.
These disruptions are independently considered.
For each disruption location, we consider two scenarios. In the
first scenario, when the disruption starts, the manager of the
disrupted facility communicates the details of the disruption (i.e.,
potential impact and anticipated downtime) to all members. We
term this as “sharing disruption information” scenario. In the
second scenario, no information transparency exists, and the
manager decides not to reveal the disruption to the other
members of the supply chain. This is described as “not sharing
disruption information” scenario. Regardless of whether the dis-
ruption information is shared, we assume that the disrupted
facility is unable to perform any operational activities for the
duration of the disruption. That means order placement for
inventory replenishment and fulfillment of demand from immedi-
ate customers do not happen. Disrupted echelons are ‘inactive’ for
the duration of a disruption. Supply chain partners could only
suspect from this abnormal inactivity that a part of their supply
chain is temporarily nonoperational. Yet, they are unable to find
any clarification for this inactivity as the manager of the disrupted
facility is not willing to disclose the information. Therefore, the
exact location, potential impact and duration of the disruption are
unknown to the echelons other than the disrupted echelon.
By using a 2  2 experimental design (i.e., sharing/not sharing
disruption information, upstream/downstream disruptions) along
with a base case (no disruption), this research investigates the
effect of sharing the disruption information on order quantities
and operating costs in a multi-echelon serial supply chain. Our
experimental data shows that sharing information about upstream
disruption provides significant benefit to the entire supply chain
with the upstream echelons gaining the larger share of the benefit.
In contrast, little benefit is observed in sharing information about
the downstream disruption.
The remainder of this article is organized as follows: a theoretical
background is presented in Section 2, followed by the hypotheses
development and description of experimental design in Section 3.
The analysis and results are outlined in Section 4, and a discussion of
the results is undertaken in Section 5. The paper ends with conclud-
ing remarks in Section 6.
2. Theoretical background
Analytical techniques and theories in operations management
often prescribe the optimal way in which a certain system or
process should function. However, practical supply chains and
processes are “complex social systems where human behavior is a
central driver” (Gino and Pisano, 2008). Analytical research often
ignores human behavioral aspects in modeling systems, which may
make the theories deviate from practice (Bendoly et al., 2006).
Therefore, it is important to compliment the theoretical analysis
with experimental observations (Tokar, 2010; Knemeyer and Naylor,
2011). This research investigates the benefit of information sharing
in the wake of disruption using a controlled laboratory experiment.
In the remaining part of this section, we review the prior work in
three distinct areas in the supply chain management literature:
(i) supply chain disruption, (ii) behavioral research on serial supply
chains, and (iii) information sharing in supply chain. Taken together,
these three areas provide us a basis for hypotheses development.
2.1. Supply chain disruptions
Supply disruptions are unforeseen events that prevent the
normal operations in a supply chain. The current trends of rising
outsourcing/offshoring and just-in-time initiatives are adopted to
increase supply chain efficiency. However, these initiatives can
also make the supply chains more vulnerable to disruptions. A
significant amount of work has been done in recent years on
various risk identification and mitigation strategies in the event of
disruptions. Broadly speaking, operational strategies for managing
disruptions can be proactive, such as inventory or sourcing
strategies, or reactive, such as rerouting and demand management
(Tomlin, 2006). For a recent review on the supply chain disruption
literature, see Snyder et al. (2012).
Information transparency is critical in managing supply chain
disruptions. Kleindorfer and Saad (2005) mentioned that achiev-
ing visibility of vulnerabilities throughout the entire supply chain
is a difficult task, as “a company with special vulnerabilities may
have every incentive to hide these from other supply chain
participants”. Risk sharing contracts may help in truthful reporting
of risk-sensitivity information (e.g., Xiao and Yang, 2009) and
thereby mitigating supply chain risk (e.g., Xia et al., 2011).
2.2. Behavioral research on serial supply chain
A number of studies used the beer distribution game in beha-
vioral research on serial supply chains. Sterman's (1989) experimen-
tal setup used a nonstationary customer demand function unknown
to the participants—the demand jumped only once during the
progress of the game. The analysis provided evidence of the
systematic underweight of pipeline inventories in ordering decisions
made by the participants. This underweight is a behavioral cause for
the oscillation and amplification of orders (i.e., the bullwhip effect).
Lee et al. (1997) identified four operational causes of the bullwhip
effect (i.e., demand signal processing, inventory rationing, order
batching and price fluctuation). Croson and Donohue (2006) found
 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
that the bullwhip effect remained even if these operational causes
were removed from the system. Their work showed that the
bullwhip effect is also caused by the behavioral reasons as the
decision-makers underweight the supply line. Sharing inventory
information cannot eliminate the underweighting bias. Sharing
point-of-sale (POS) data helps reduce the bullwhip effect even when
the demand distribution is known in advance to all members in a
supply chain (Croson and Donohue, 2003), which is again due to the
behavioral reasons. Regarding the inventory information sharing,
Croson and Donohue (2005) found that downstream information
sharing significantly reduced order oscillations, whereas upstream
information sharing was not beneficial.
Wu and Katok (2006) showed that only having experience with
the beer game might not be sufficient for improving supply chain
performance. They found that experience together with supply chain
collaboration helped to reduce the bullwhip effect. However, Tokar
et al. (2012) showed that training can improve the performance of
the participants. In this regard, a large number of past studies used
university students as the subjects in the beer game experiments (e.
g., Croson and Donohue, 2003; Steckel et al., 2004; Zhao and Zhao,
2015). Only a few studies used industry practitioners: Croson and
Donohue (2006) used members of a professional organization of
logistics managers, Tokar et al. (2011) used senior-level members of
consumer products manufacturer in the US, and Niranjan et al.
(2011) used an Asian supplier of automotive steering system. These
studies observed consistent outcomes regardless of the decision-
maker being a student or an experienced practitioner.
Coordination and information sharing problems do not arise in a
simple, single-echelon inventory system. Even for such a system,
non-optimal ordering decisions were made by the participants
(Bloomfield and Kulp, 2013). Ancarani et al. (2013) introduced a
demand uncertainty in the beer game setting and observed that the
participants held fewer inventories in response to the increased
uncertainty. Consumer demand uncertainty and forecasting issues
were completely eliminated in Croson et al. (2014) by using a
constant demand in each period that was known to every partici-
pant—the bullwhip effect still remained. This finding has been
explained by using the coordination risk, which acts as a buffer
against the uncertainty about the way other members of a supply
chain behave. Mitigating coordination risk factors and carrying
coordination stock were found to moderate the bullwhip effect.
Behavioral experiments on a multi-echelon system by Zhao and
Zhao (2015) showed that full information sharing (i.e., sharing every
echelon's orders, inventories and stockouts) might not improve
supply chain performance. Unit inventory holding cost was assumed
equal to the unit stockout penalty cost in their experiments.
Participants were observed to focus more on reducing stockout for
an information sharing scenario, and reducing on-hand inventory for
a no-information sharing scenario. Rong et al. (2008) incorporated
supply uncertainty due to a limited production capacity of the
manufacturer and introduced several capacity shocks during the
game. Participants were aware of the timing of these shocks and
responded with a reverse bullwhip type order variability.
Several studies focused on the effect of demand fluctuations:
Tokar et al. (2011) investigated the effect of knowing the timing
and magnitude of the demand spikes due to retail promotions in
both single-echelon and multi-echelon systems. Sharing of the
demand spike information was found helpful to the supply chain
members. Tokar et al. (2014) extended this analysis and suggested
that sharing information on the timing of store-level promotions
was critical for retailers to improve supply chain performance.
2.3. Information sharing in supply chain
The literature considers information asymmetry and a lack of
transparency as key reasons for supply chain inefficiency and
171
increased bullwhip effect (Fiala, 2005). The game theoretic models
address the information sharing vs. private information issues. For
example, Li and Zhang (2008) considered a supply network
consisting of one manufacturer and multiple retailers, and showed
that sharing of the retail demand information by all retailers
maximized the supply chain profit. A simulation model by Zhao
et al. (2002) on a similar supply network showed that information
sharing and ordering coordination significantly improved the total
cost and service level. Empirical research with firm-level survey
data also provided evidence of this improvement (e.g., Zhou and
Benton, 2007; Prajogo and Olhager, 2012).
The literature on information sharing in supply chains is exten-
sive. However, a majority of the studies focused on the benefit of
sharing demand or inventory information. Sharing this information
helps the supply chain entities to make better ordering decisions,
which can reduce order oscillation and amplification resulting in
lower supply chain costs. Various mechanisms developed for effec-
tive information sharing are discussed in the literature, such as a
vendor-managed inventory system (e.g., Angulo et al., 2004), ERP
tools (e.g., Kelle and Akbulut, 2005), or advance demand information
(e.g., Thonemann, 2002; Sarkar and Shewchuk, 2013). However, scant
attention is paid to the effect of sharing disruption-related informa-
tion. As discussed earlier, managers are often unwilling to reveal the
operational vulnerabilities to the supply chain partners. However,
sharing the information on vulnerabilities is an integral part of the
risk assessment process (Sodhi and Tang, 2012). Therefore, it is
important to understand the behavioral effect of sharing the infor-
mation on supply chain disruptions. Exploring the operational
benefits in real-time sharing of disruption information serves as
the motivation of the current paper.
3. Hypotheses development and experimental design
3.1. Hypotheses
Humans tend to underestimate the chance of having a negative
event and overestimate the chance of having a positive event
(Weinstein, 1980). When the decision makers have no prior
information on the possibility of disruption, we do not expect them
to systematically plan for the disruption before it actually starts.
Disruption contributes to the supply demand mismatch, which
aggravates the bullwhip effect (compared to the base case). How-
ever, the effect of communicating the disruption information is not
clear. An absence of information transparency creates ambiguity
over operational processes. Using natural-event experiments,
Abdellaoui et al. (2011) showed that people opt to take more
insurance as the uncertainty and ambiguity increases. In inventory
management context, accumulating stocks at present time can
provide insurance from future backordering. Sharing of disruption
information promotes transparency and reduces uncertainty, which
should result in lower order oscillations and mitigated bullwhip
effects. We find some support for our conjecture from the supply
chain management literature: lack of information transparency and
double-guessing by players contribute to the bullwhip effect (Geary
et al., 2006); an uncertainty of demand and supply also contributes
to it (Ancarani et al., 2013). The coordination risk increases with an
increase in the ambiguity and unpredictability of the partners
(Croson et al., 2014). In our experiments, not sharing information
creates more ambiguity, thus providing rationale for accumulation
of a larger inventory. This leads to our first set of hypotheses
regarding information sharing in the wake of retailer disruptions.
H1A. In the event of a disruption at downstream echelon (retailer),
sharing of disruption information with supply chain members will
reduce the level of order oscillations throughout the supply chain.
172 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
H1B. In the event of a disruption at downstream echelon (retailer),
sharing of disruption information with supply chain members will
reduce the magnitude of amplification of order oscillations through-
out the supply chain.
H1C. In the event of a disruption at downstream echelon (retailer),
sharing of disruption information with supply chain members will
result in lower supply chain cost.
In a serial supply chain, the order information flows upstream
and the material flows downstream. In addition, the upstream
order variability is more than the downstream variability due to
the bullwhip effect. The opposing direction of flows, together with
the difference in order variabilities could produce different out-
comes for the upstream disruptions, which led us to test a similar
set of hypotheses for manufacturer disruptions.
H2A. In the event of a disruption at upstream echelon (manufac-
turer), sharing of disruption information with supply chain members
will reduce the level of order oscillations throughout the supply chain.
H2B. In the event of a disruption at upstream echelon (manufac-
turer), sharing of disruption information with supply chain members
will reduce the magnitude of amplification of order oscillations
throughout the supply chain.
H2C. In the event of a disruption at upstream echelon (manufac-
turer), sharing of disruption information with supply chain members
will result in lower supply chain cost.
In case there is an impact of information sharing on decision-
making, we would like to investigate the benefits for various supply
chain echelons. Since the manufacturer receives the most distorted
demand information among all echelons, the bullwhip effect should
have the most severe impact on the manufacturer (i.e., the
manufacturer's operating cost is typically the highest in a supply
chain). Information sharing promotes transparency, which should
provide the largest benefit to the echelon most affected by the
distorted information. For example, Croson and Donohue (2003)
showed that the manufacturer is the primary beneficiary of sharing
POS data. It can be similarly argued that regardless of the location of
a disruption, sharing disruption information should be most ben-
eficial to the manufacturer. Therefore, we have the following set of
hypotheses, one each for downstream and upstream disruptions.
Downstream disruption:
H3A1. Sharing downstream disruption information with all supply
chain members will result in a larger reduction in order oscillation for
the upstream members (i.e., manufacturer and distributor) than that
for the downstream members (i.e., wholesaler and retailer).
H3B1. Sharing downstream disruption information with all supply
chain members will result in a larger reduction in supply chain cost
for the upstream members (i.e., manufacturer and distributor) than
that for the downstream members (i.e., wholesaler and retailer).
Upstream disruption:
H3A2. Sharing upstream disruption information with all supply
chain members will result in a larger reduction in order oscillation
for the upstream members (i.e., manufacturer and distributor) than
that for the downstream members (i.e., wholesaler and retailer).
H3B2. Sharing upstream disruption information with all supply
chain members will result in a larger reduction in supply chain cost
for the upstream members (i.e., manufacturer and distributor) than
that for the downstream members (i.e., wholesaler and retailer).
Finally, we compare the impact of upstream and downstream
disruptions on supply chains. In a serial supply chain, decision-makers
tend to underweight the supply line while making ordering decision.
This underweighting issue contributes to the bullwhip effect by
increasing the order variability at the upstream echelons (Croson
and Donohue, 2006). Implications of a disruption to the upstream and
the downstream sides can be different (Schmitt and Singh, 2012).
Therefore, disruption impact may depend on the location of disrup-
tion. We suspect that sharing of the disruption information may also
play a role here. Particularly when disruption information is not
shared, undisrupted echelons have little to no indication that the
disrupted echelon is temporarily nonoperational. Specifically during
upstream disruptions, the upstream echelons cannot fulfill the orders
placed by the downstream ones. The downstream echelons do not
know the exact reason for this irregularity, which may create a
nervousness in the downstream echelons. These echelons may
provide even less weight to the supply line after experiencing the
irregularities, resulting in large order sizes and increased bullwhip
effect. Therefore, we expect to see a high level of order oscillation for
the upstream disruptions. For downstream disruptions, the upstream
echelons have no specific information regarding the disruptions.
However, these echelons are able to fulfill arriving orders as they
remain operational. Since this normal behavior of upstream echelons
do not create irregularities, there is no additional cause for further
underweighting of the supply line. Therefore, when the disruption
information is not shared, we may expect to see a lower level of order
oscillation for downstream disruptions (compared to the upstream
disruptions).
When the upstream disruption information is shared, the exact
reason for the irregularity is known to the downstream echelons.
Therefore the additional cause for underweighting the supply line
should disappear. We may even expect the downstream echelons
to help in reducing the overall cost of the supply chain by not
placing large orders immediately following the disruption. In pre-
disruption periods, supply chain echelons may attempt to accu-
mulate inventory as a protection from the demand variance.
Carrying this additional inventory should be beneficial as it
provides protection from the impact of an upstream disruption.
For a downstream disruption, carrying such additional inventory
might not be very useful as the consumer orders cannot be
fulfilled due to the disruption. Immediately following the disrup-
tion, the entire supply chain should experience a spike in the order
level, as the retailer places a large order in an attempt to establish
the normal condition. The spike in order level should aggravate
the bullwhip effect (as compared to an upstream disruption).
Therefore we have the following hypotheses.
H4A. When disruption information is not shared, disruption at an
upstream echelon (manufacturer) results in a higher level of order
oscillations than a disruption at a downstream echelon (retailer).
H4B. When disruption information is shared, disruption at an
upstream echelon (manufacturer) results in a lower level of order
oscillations than a disruption at a downstream echelon (retailer).
3.2. Configurations
We conduct a laboratory experiment with a setup similar to the
traditional beer distribution game, but with the addition that we
introduce supply chain disruptions. Therefore our experimental
design mimics a decentralized, periodic review, serial inventory
systems with four echelons—retailer (R), wholesaler (W), distri-
butor (D) and manufacturer (M). Each echelon is managed by a
human participant, who places replenishment orders with
upstream echelon and fulfills the demand from the downstream
echelon over a set duration of the game. Manufacturer is uncapa-
citated. Lead times cause ordering delays and shipping delays
between each echelon. We assume two periods of ordering delays
 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
and another two periods of shipping delays for every echelon
except the manufacturer, who takes three periods to produce the
items (Fig. 1 shows a schematic of the supply chain). The number
of echelons and the corresponding delays in our experiments are
consistent with those assumed by the past researchers (e.g.,
Croson and Donohue, 2003, 2006).
During the experiments, the following sequence of activities
occurs in each echelon: (i) delivery of shipment from the upstream
echelon, (ii) demand arrival from the downstream customer, (iii)
shipment to fulfill the demand (items will be backordered in case
of stockout), (iv) placement of new order with the upstream
echelon for inventory replenishment. The first three activities are
carried out automatically; each participant only decides the order
amount in each period. Consumers' demand in each period is
uniformly distributed between 0 and 8 units, and it is indepen-
dently drawn between periods. The demand distribution is known
to all participants. At the start of the experiments, every echelon
has an initial inventory of 12 units and incoming shipments of
4 units in each of the next two periods. The outstanding order at
each echelon for the previous two periods is also 4 units.
Supply chain disruptions are unpredictable events. Therefore,
the participants have no prior knowledge that disruption may
happen during the experiments. We consider five treatments as
shown in Fig. 2. Treatment 1 is the base case, where no disruption
takes place. In the absence of disruptions, this setup is the same as
that for a traditional beer game. In each of the remaining
treatments, a disruption is introduced at either the retailer's or
the manufacturer's location starting in period 10. Each disruption
is five periods long—during this time, the disrupted echelon (i.e.,
retailer or manufacturer) cannot perform normal operations.
Though the disrupted echelon can observe the incoming demand
in every period, fulfillment of downstream demand, order place-
ment with upstream suppliers, and delivery of replenishment
items are prevented throughout the disruption. As soon as the
normal operations resume (i.e., at the start of period 15), (i) all
incoming shipments are received, (ii) accumulated incoming
demand during disruption are shipped (subject to inventory
availability), and (iii) disrupted echelon can place replenishment
order of any quantity. Similar to the pre-disruption periods,
information flow and material flow lead times are also applicable
for post-disruption ordering and delivery in the disrupted echelon.
The retailer disruptions are introduced in treatments 2 and 3.
When the disruption begins in a game (i.e., in period 10), the
retailer learns about the duration of disruption (i.e., five periods)
and its impact (i.e., no order placement, delivery and demand
fulfillment during the disruption; accumulated demands need to
be filled after the disruption, etc.). In treatment 3, the disruption
information is shared with the other echelons as soon as the
disruption starts. In a similar manner, the manufacturer disruptions
are introduced in treatments 4 and 5. Similar to the case of retailers
in treatments 2 and 3, the manufacturers in treatments 4 and
5 receive the disruption information at period 10—but the disrup-
tion information is immediately shared with other echelons only in
treatment 5. Treatments 2 and 4 (no information sharing) replicate
a practical scenario that managers of the disrupted facilities may
consider hiding the disruption information beneficial. The rationale
behind such an intention of the managers is discussed earlier with
real-life examples.
Fig. 1. Schematic of the beer distribution game model.
173
3.3. Implementation
We developed a web-based program using ASP.NET and uploaded
the program in a server, which can be accessed through internet. The
participants used web browsers in individual computer terminals to
place the order with the upstream echelons after observing the
incoming demands, current inventory levels, current shipment quan-
tities and the order quantities in the last 10 periods. Ten simultaneous
replications for each of the five treatments can be conducted by the
program. Common random demand streams [ U(0,8)] are used across
all treatments for a fair comparison of the performance differences. We
recruited 200 undergraduate business students from a major public
university in the US as the participants (subjects) in our experiments.
These subjects were randomly assigned into the five treatments, which
enabled us to have one student to manage every echelon for all ten
replications in each treatment (i.e., four subjects formed a team and
completed one replication). Every replication was conducted for 55
periods; however, the subjects had no prior knowledge on the total
number of periods in each experiment, which could have induced an
end-of-period behavior otherwise. Before beginning the experiments,
the subjects were briefed on the rules and objectives of the game. The
meaning of “uniform demand distribution” had been clearly explained
during the briefing. The subjects were told the amount of holding and
backorder costs (which were $0.50 and $1.00 per unit per period)—
these cost values are consistent with prior studies. The objective of each
game was to minimize the total supply chain cost. After the briefing, the
subjects practiced the games for five periods before the experiments
began. Any kind of communication between subjects was prohibited for
the duration of the game. We did not mention anything on disruption
during our briefing; although considering the business background of
all students, we can assume that they were familiar with the con-
sequence of a supply chain disruption. At period 10, some subjects (i.e.,
disrupted echelons in treatments 2 and 4, and all echelons in treatments
3 and 5) saw the disruption information screen in their web browsers.
Before beginning the games, participants were informed of the
incentives: $100 for the winning team and $50 for the second winner in
each treatment. Each student also received a bonus credit in the course-
work for participation in the game. While various monetary incentive
structures exist in the literature, ours is similar to the fixed prize schemes
used in Sterman (1989) and Ancarani et al. (2013). We believe that our
simple structure is readily interpretable, which allowed the participants to
focus on the ordering decisions rather than the incentive mechanism.
4. Results
Our primary units of analysis are the order quantities placed by
each subject and the corresponding inventory levels, which are
Fig. 2. Classification of the treatments used in the experiments.
174 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178

compiled for periods 10–50. Data before period 10 is not considered, manufacturer disruptions) suggests a potential role of information
as all treatments are expected to be similar before the start of sharing in improving the overall cost.
disruptions. The impact of a disruption is often longer than the
disruption duration (Schmitt and Singh, 2012), therefore we consider
4.1. Hypothesis test
up to 50th period in our analysis. Following Croson et al. (2014) and
Zhao and Zhao (2015), we use Grubbs (1969) procedure to identify
To further explore the benefit of information sharing and particu-
the outliers in all treatments. One team from each of the treatments 1,
larly to verify our hypotheses, we perform the Mann–Whitney U test
3 and 5 are found to be outliers and eliminated from our analysis.
(also called Wilcoxon–Mann–Whitney test, see Siegel and Castellan
Fig. 3(a)–(e) shows the order quantities and Fig. 4(a)–(e) show the
(1988), p. 128), which is a nonparametric test to determine whether
order variances for each echelon in all five treatments. By definition,
two independent groups were drawn from the same population. This
the bullwhip effect is a phenomenon of increased order variance at a
test, widely used in similar behavioral studies, has the advantage of
given echelon relative to the order variance at the downstream
not relying on the normal distribution assumption. First we compare
echelon. Figs. 3 and 4 confirm the presence of the bullwhip effect
the difference in order variances between the base case (treatment 1)
in all treatments. A nonparametric sign test (Siegel and Castellan,
and each of the treatments 2–5.We report the p-values from the
1988, p. 80) considering “success” as an increase in the variance of
Mann–Whitney U test as: 0.07 (treatment 2), 0.05 (treatment 3), 0.02
orders placed in two successive echelons confirms the amplification
(treatment 4) and 0.64 (treatment 5). This test shows a significant
of order variance to the upstream of these supply chains (p-value
increase in the order variability due to disruptions; treatment 5 (i.e.,
o0.01 or better). The median supply chain costs incurred during
periods 10–50 are $1671.50, $2566.00, $2565.00, $3312.50, and
$1731.50 for treatments 1–5, respectively. The corresponding per-
period costs incurred by the individual roles are shown in Fig. 5.
These values depict similar supply chain costs in treatments 2 and 3
(i.e., no information sharing vs. information sharing in the retailer
disruptions), but a large difference between the costs in treatments
4 and 5 (i.e., no information sharing vs. information sharing in the

Fig. 4. Variance of orders by role and team for each treatment.

Fig. 3. Average ordering quantity for each echelon across all treatments.
Fig. 5. Role-specific cost break-up for median supply chain cost in each treatment.
R¼ retailer, W ¼ wholesaler, D¼ distributor, M ¼ manufacturer. Treatment 1 is base
case; treatment 2 is retailer disruption with no information sharing; treatment 3 is
retailer disruption with information sharing; treatment 4 is manufacturer disrup-
tion with no information sharing, treatment 5 is manufacturer disruption with
information sharing.
 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178 175

the manufacturer disruption with information sharing) is the only
exception. When we compare the variances of the orders across
treatments 2 and 3 (i.e., disruption at retailer), none of the differences
are found to be significant at the individual echelon levels. We also
compare the order variances in the supply chain level combining all
echelons: test statistics is still not significant (n2 ¼ 40, n3 ¼ 36,
mean variance2 ¼ 37:5, mean variance3 ¼ 39:6, p ¼ 0:67; subscript
2 and 3 represent treatments 2 and 3, respectively), therefore we do
not find support for H1A. However, a similar comparison test for the
order variances in manufacturer disruptions demonstrates a signifi-
cant difference between treatments 4 and 5 (results shown in Table 1),
therefore H2A is supported. We conclude that real-time sharing of the
disruption information by manufacturers reduces order oscillations.
To verify the impact of information sharing on order amplifica-
tion, we again use the Mann–Whitney U test for a comparison of
amplification ratios (i.e., the ratio of order variances at adjacent
echelons) across treatments: p-values for neither the echelon level
comparisons nor the overall comparisons indicate a significant
difference between sharing and not sharing the disruption informa-
tion. For the retailer disruptions, test result for the overall compar-
ison is the following: n2 ¼ 40, n3 ¼ 36, mean amplification2 ¼ 1:430,
mean amplification3 ¼ 1:480, p ¼ 0:54. For the manufacturer disrup-
tions: n4 ¼ 40, n5 ¼ 36, mean amplification2 ¼ 1:711, mean
amplification3 ¼ 1:432, p ¼ 0:32. Therefore H1B and H2B are not
supported. We found evidence that the disruption information
sharing by the manufacturers does help to reduce the order
oscillations; however, no such improvement is observed for order
amplifications. It should be noted here that the unchanged order
amplification ratios do not subdue the benefits of information
sharing, since oscillation reduction (as opposed to amplification
reduction) is a direct measure of supply chain performance improve-
ment. Croson and Donohue (2003) showed an example: if the order
oscillations are reduced by a constant factor for each echelon, then
the amplification ratios will remain unchanged, but the overall
supply chain will have a reduced oscillation.
We further investigate the effect on supply chain cost at each
echelon across treatments. The p-values for the Mann–Whitney U
test are not significant for the comparison between treatments in
case of retailer disruptions, but they are significant in case of
manufacturer disruptions (results shown in Table 2). The data
provides no support to H1C, but it does support H2C.
Since H1A–H1C are not supported, our analysis indicates no
significant benefit in sharing downstream disruption information.
As there is no benefit, H3A1 and H3B1 are not applicable.
As shown in Tables 1 and 2, the benefit of disruption information
sharing appears to be larger for the upstream echelons (i.e., distributor
and manufacturer) than the downstream counterparts (i.e., retailer
and wholesaler). We verify the significance of this difference by
grouping the order variances of distributor and manufacturer together
and compare those values across the treatments 4 and 5
(i.e., disruption at manufacturer) using the Mann–Whitney U test:
the difference is found to be significant (n4 ¼ 20, n5 ¼ 18,
mean variance 4 ¼ 197:6, mean variance 5 ¼ 60:9, p ¼ 0:003). A
similar grouping of the downstream echelons (i.e., retailer and whole-
saler) also shows a significant difference across the treatments
(n4 ¼ 20, n5 ¼ 18, mean variance4 ¼ 19:1, mean variance5 ¼ 13:7,
p ¼ 0:048). Both the upstream and downstream groups have lower
order oscillations because of the information sharing; however, the
upstream group achieves a reduction of ð197:6  60:9Þ
197:6  100 ¼ 69:2% on
average, while the downstream group achieves a reduction of only
ð19:1  13:7Þ
19:1 100 ¼ 28:1%, which supports H3A2.
We use a similar analysis for comparing the operating costs of the
upstream and downstream echelons across treatments 4 and 5.
Operating cost reduction achieved by information sharing is signifi-
cant for the upstream members (n4 ¼ 20, n5 ¼ 18, mean cost4
¼ 1302:2, mean cost5 ¼ 692:9, p o 0:002, reduction¼ 46.8% on
average), but not significant for the downstream counterparts
(n4 ¼ 20, n5 ¼ 18, mean cost4 ¼ 383:3, mean cost5 ¼ 366:5, U ¼
326, p ¼ 0:47). Therefore, H3B2 is also supported.
We compare the order oscillations in supply chains (i.e., by
grouping order variances in all four echelons together) across
treatments 2 and 4. Recall that these two treatments incorporate
the retailer disruption and the manufacturer disruption, respec-
tively—in none of them the disruption information is shared with
the other echelons. The difference in order oscillations is not
significant (n2 ¼ 40, n4 ¼ 40, mean variance2 ¼ 61:3, mean
variance4 ¼ 108:3, p ¼ 0:49), therefore H4A is not supported. Next,
we compare treatments 3 and 5. These treatments incorporate
information sharing for the retailer disruption and the manufac-
turer disruption, respectively. The difference in order oscillations
across these two treatments is significant (n3 ¼ 36, n5 ¼ 36,
mean variance3 ¼ 67:1, mean variance5 ¼ 37:3, p ¼ 0:005), which
supports H4B.
4.2. Weight on inventory and supply line
Following Sterman (1989), and Croson and Donohue (2006), we
next test the underweighting of supply line issue by the following
regression for each subject in our experiments:
n o
i  1;g
Oig ig ig ig
t ¼ max 0; α0 þ αI I t  1 þ αR Ot  2 þ αS St þ αN N t þ αt t þϵ ð1Þ
where i, g and t indicate the echelon (i.e., i ¼ 1, 2, 3 and 4 for
retailer, wholesaler, distributor and manufacturer), team (i.e.,
g ¼ 1, 2, …, 10) and time period (t ¼ 10, 11, …, 50), respectively.
Order placed by echelon i in current period t is Oig t . The explana-
tory
 variables are the inventory level in the previous period

I ig
t  1 , the orders arrived from the downstream echelon in
i  1;g
current period
  (Ot  2 ), the shipments received from upstream
ig
echelon
  S t , the outstanding order quantity in the supply line
Nig t , and the time period t. The regression provides the best-fit
model that predicts the ordering decisions of each subject. How-
ever, as noted in Croson and Donohue (2003), there is no

Table 1 Table 2
Difference in order variances across treatments 4 and 5 (manufacturer disruption). Difference in operating cost across treatments 4 and 5 (manufacturer disruption).

Echelon Retailer Wholesaler Distributor Manufacturer Overall Echelon Retailer Wholesaler Distributor Manufacturer Overall

n4 10 10 10 10 40 n4 10 10 10 10 10
n5 9 9 9 9 36 n5 9 9 9 9 9
mean variance4 11.9 26.2 143.8 251.5 108.4 mean cost4 296 471 1263 1341 3371
mean variance5 9.3 18.1 34.4 87.5 37.3 mean cost5 291 442 724 662 2119
Ustatistics 74 73 63 61 1124 U statistics 91 78 66 59 62
p-Value 0.21 0.18 0.03 0.02 0.006 p-Value 0.97 0.36 0.05 0.02 0.04

Subscripts 4 and 5 indicate treatments 4 and 5, respectively. Disruption information Subscripts 4 and 5 indicate treatments 4 and 5, respectively. Disruption information
is not shared in treatment 4, but is shared with other echelons in treatment 5. is not shared in treatment 4, but is shared with other echelons in treatment 5.
p-Values shown are for two-tailed comparison using Mann–Whitney U test. p-Values shown are for two-tailed comparison using Mann–Whitney U test.
176 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
guarantee that the subjects actually ordered according to the
estimated parameter values. The average adjusted R2 value ranges
between 0.63–0.69 across the treatments. The optimal values of αI
(i.e., the coefficient of inventory level) and αN (i.e., the coefficient
of outstanding order quantity) are –1. If the subjects neither
underweight nor overweight the supply line, the values of αI and
αN will be the same. However, we find the average value of
αI ¼ –0:216 and αN ¼ –0:072 in treatment 1 (base case). The
difference between αI and αN is statistically significant
ðαI o αN ; p  value o 0:01Þ, which indicates underweighting of sup-
ply line by the subjects. In the absence of disruption, Sterman
(1989) obtained: αI ¼ –0:261, αN ¼ –0:088, and Croson and
Donohue (2006) obtained: αI ¼ –0:237, αN ¼ –0:030. These values
are similar to ours, which provides a validation of our
experimental setup.
We further calculate the values of (αI þ αN ) for the upstream
members (i.e., the average of manufacturer and distributor) and
the downstream members (i.e., the average of wholesaler and
retailer) of each game in the treatments 4 and 5. The value of
(αI þαN ) implies the weight of inventory position (i.e., inventory
levelþ on-order inventory) in the ordering decisions. The p-value
for Mann–Whitney U test is weakly significant for the difference
(mean αI þ αN for the upstream treatment 4 ¼  0.116, mean αI þ αN
for the upstream treatment 5 ¼  0.192, p-value ¼0.07), indicating
the effect of information sharing on the weightage of inventory
position for the manufacturer disruptions. The analysis suggests
that when disruption information is not shared, the decision-
makers give even less weight on the inventory position. The
corresponding p-value for the difference in downstream members'
weightages is not significant. The coefficient values obtained
through regression analysis may help to explain the findings in
the previous subsection: when the manufacturer's disruption
information is not shared, an upstream decision-maker gives less
weight on the inventory position that results in more severe
underweighting issue compared to the information sharing case.
Sharing disruption information helps to ameliorate the under-
weighting, hence the order variability reduces significantly.
5. Discussion
The collective results of the treatments indicate that in the case
of a disruption at the upstream echelon (i.e., manufacturer), it is
beneficial to share the disruption information (such as the
expected duration and significance) with the other echelons in
the supply chain. Sharing of the upstream disruption information
helps to mitigate the order variance at both the upstream and
downstream echelons, which results in lower operating cost for
the supply chain. Information sharing reduces ambiguity over
supply chain operations, which may help improve trust and
long-term reliability in the buyer–supplier relationship. The entire
supply chain may also get the assurance of timely warning for any
future disruptions. Disrupted upstream echelons gain the larger
share of the overall benefit: both order variability and operating
costs drop significantly. In the case of a disruption at the down-
stream echelon, we do not find any significant improvement in
supply chain performance by sharing disruption information with
supply chain members. From the perspective of the bullwhip
effect, a disruption at the manufacturer's location is found to be
less damaging than a disruption at the retailer's location if the
disruption information is shared. In the absence of information
sharing, no such conclusion can be made.
A vast amount of literature on supply chain risk management
has investigated supply chain incentives and contract design,
where an underlying assumption is that the manufacturers are
unwilling to reveal disruption information and need external
motivations to comply with the information transparency require-
ments. This concern is legitimate, as external mandates, such as
the Sarbanes–Oxley Act, have been devised to encourage man-
agers to disclose disruption information in real-time. However, our
findings show that the managers of upstream echelons may be
better off sharing the disruption information for their own benefit
in regard to reduced bullwhip effects and associated supply chain
costs. The main hindrance for building resilient supply chains is
often the associated operating cost increases (Chopra and Sodhi,
2014). However, maintaining information transparency may not
increase the operating costs, but still reduces the impact of a
disruption at the manufacturer's location.
It is interesting to compare our findings with Croson and
Donohue (2005), who using a beer game framework in their
behavioral experiments, showed that sharing inventory informa-
tion of the downstream echelons is effective in reducing the
bullwhip effect, and sharing upstream inventory information
provides little benefit. Their experiment considered the same
retail demand distribution as our experiments, but did not con-
sider any disruptions in supply chains. We find that sharing
information on upstream disruptions reduce the bullwhip effects
and associated costs, while sharing information on downstream
disruptions did not create much of a difference. Here a comparison
can also be drawn with the findings of two other studies. In the
first study, Schmitt and Singh (2012) suggested that addressing
downstream disruption should be prioritized over addressing
upstream disruptions under resource constraints, as the down-
stream disruptions can have a faster effect on consumers. When
risk mitigation initiatives are directed toward increasing informa-
tion transparency throughout the supply chain, our study shows
that increasing transparency for an upstream disruption is more
important than that for a downstream disruption. In the second
study, Croson and Donohue (2003) found that sharing POS data
was more beneficial to the echelon furthest from the source of the
POS data. We find that sharing upstream disruption information is
more beneficial to the upstream echelon itself.
In our analysis, we did not find support for H4A as the order
oscillation levels are similar for the upstream and downstream
disruptions in the absence of information sharing. In the case of an
upstream disruption, postponing the fulfillment of large orders due to
the disruption can cause high order oscillation levels. However, for
downstream disruptions, a spike in the retailer's order quantity
immediately after the disruption may cause high order oscillation as
well. In information sharing cases, downstream disruptions have
more impact, which supports the finding of Schmitt and Singh (2012).
We offer an explanation for our findings on the impact of
downstream vs. upstream disruptions. In a demand-driven supply
chain, the retailer has more control than the manufacturer over
the aggregate order quantities flowing through all echelons: even
a small change in a retailer's order can result in a larger order
volume for the manufacturer; however, increased production
volume at the manufacturer's end does not necessarily increase
outbound order shipment from the manufacturer. In both cases of
downstream disruptions (i.e., information sharing and no informa-
tion sharing), retailers place larger orders after the disruption to
clear backorders that occurred during the disruptions and also in
anticipation of future disruptions. However, in both cases, the
retailer has the same information on the disruptions regardless of
whether she is sharing the information. Therefore, the order
quantity from the retailer should not be significantly different
across the two treatments. Other echelons only react to the
retailer's demand, and knowing about the disruptions may not
create a large difference in their ordering decisions. On the other
hand, due to upstream disruptions, the shipments may take an
unusually long time to be delivered. When disruption information
is not shared, other echelons do not get a clear explanation for this
 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
delay, and they assume it is because of the manufacturer's
incompetency. They (especially the distributor, who is immedi-
ately downstream) react by attempting to pile up stocks to buffer a
higher coordination risk in coping with an incompetent manufac-
turer. The attribution theory (Kelley and Michela, 1980; Coombs
and Holladay, 1996) can help explaining this behavior as
outlined below.
Supply chain disruption is an organizational crisis for the
disrupted echelon. Once the supply chain members detect the
inactivity of the disrupted echelon due to a crisis, they evaluate the
cause and make attributions. Communication with the stake-
holders (i.e., information sharing with the supply chain members)
can influence the evaluation and judgment made, which in turn
influences the supply chain members' feelings and behavior
toward the actor (i.e., the disrupted echelon). The attribution is
influenced by the received information and perceiver's beliefs (e.g.,
suppositions about the causes, and expectations of the effects).
The locus of causality (i.e., internal or external) and controllability
of cause are two important aspects of attribution (Ford, 1985).
When ambiguity exists, the causal analysis is biased toward
controllable factors (Kelley and Michela, 1980). Past research on
helping behavior demonstrated that the presence of intentional
factors created a stronger attribution than the presence of unin-
tentional factors behind an incident (Piliavin et al., 1969). In the
information sharing case of an upstream disruption, downstream
echelons may have weaker attributions of the actor's responsibility
(which is indicated by a relatively smaller order size compared to
that for the no information sharing case). However, a lack of
information creates an ambiguity, resulting in a strong attribution
of the actor's operational incompetency, which makes it more
challenging to coordinate supply chain operations when disrup-
tion information is not shared.
A majority of the subjects, who had no precise information
regarding the cause of supply irregularity, actually indicated in a
post-game questionnaire that a disruption could have happened
during the game. These subjects still responded to their attribu-
tions during the game and not to the supply irregularity itself. A
potential difference in the attributions may have contributed to
the difference in the outcomes.
6. Conclusions
This study is perhaps the first to report the findings of a
behavioral experiment on multi-echelon supply chains with dis-
ruptions. For upstream disruptions, we find that complete infor-
mation transparency among the supply chain members improves
supply chain performance. Prior literature on supply chain disrup-
tions focused more on mitigation tactics (i.e., actions in advance of
a disruption, such as supplier selection, dual sourcing, contract
design, etc.), and less on contingency tactics (i.e., actions only in
the event of a disruption). This paper contributes to the literature
by discussing a contingency tactic: building resilient supply chains
by real-time sharing of disruption information. Our findings can be
used by managers to make informed decisions on real-time
sharing of disruption information with their supply chain partners.
This work creates a number of avenues for future research.
First, we use a specific duration for the disruptions in our
experiments. One extension may consider the effect of information
sharing for disruptions that are longer or shorter in duration. The
effect of disruptions occurring at the intermediate echelons of a
serial supply chain can also be investigated. A second potential
extension is to study the effect of sharing distorted information.
We assume that at the beginning of disruptions, the disrupted
echelons know the exact durations of the disruptions. Although it
may not always be possible to know the exact duration in advance,
177
the durations are often predicted with reasonable accuracy. How-
ever, when the disruption duration is significantly misestimated,
the supply chain members receive distorted information, which
may have an impact on order decisions. Another potential direc-
tion for future research is investigating the causal schemata in the
subjects' behavior. A possible explanation for the difference in
ordering behavior of the subjects is due to the attributional
difference. However, attributional complexity scale varies with
the perceptions of individual subjects (Fletcher et al., 1986),
therefore an in-depth study on the causal inference process may
be conducted.
In reality, a supply chain manager may be reluctant to share the
disruption information due to a fear that the firm's goodwill or
reputations will be damaged. A limitation of this study could be
the absence of this goodwill factor consideration. However, we
have investigated the immediate impact of sharing disruption
information on the operational efficiency rather than the long-
term effect on goodwill.
References
Abdellaoui, M., Baillon, A., Placido, L., Wakker, P.P., 2011. The rich domain of
uncertainty: source functions and their experimental implementation. Am.
Econ. Rev. 101 (2), 695–723. http://dx.doi.org/10.1257/aer.101.2.695.
Ancarani, A., Di Mauro, C., D'Urso, D., 2013. A human experiment on inventory
decisions under supply uncertainty. Int. J. Prod. Econ. 142 (1), 61–73. http://dx.
doi.org/10.1016/j.ijpe.2012.09.001.
Angulo, A., Nachtmann, H., Waller, M.A., 2004. Supply chain information sharing in
a vendor manager inventory partnership. J. Bus. Logist. 25 (1), 101–120.
Baghalian, A., Rezapour, S., Farahani, R.Z., 2013. Robust supply chain network design
with service level against disruptions and demand uncertainties: a real-life
case. Eur. J. Oper. Res. 227 (1), 199–215. http://dx.doi.org/10.1016/j.
ejor.2012.12.017.
Bendoly, E., Donohue, K., Schultz, K.L., 2006. Behavior in operations management:
assessing recent findings and revisiting old assumptions. J. Oper. Manag. 24 (6),
737–752. http://dx.doi.org/10.1016/j.jom.2005.10.001.
Blackhurst, J., Craighead, C.W., Elkins, D., Handfield, R.B., 2005. An empirically
derived agenda of critical research issues for managing supply-chain disrup-
tions. Int. J. Prod. Res. 43 (19), 4067–4081. http://dx.doi.org/10.1080/
00207540500151549.
Bloomfield, R.J., Kulp, S.L., 2013. Durability, transit lags, and optimality of inventory
management decisions. Prod. Oper. Manag. 22 (4), 826–842. http://dx.doi.org/
10.1111/poms.12017.
Bode, C., Wagner, S.M., Petersen, K.J., Ellram, L.M., 2011. Understanding responses to
supply chain disruptions: insights from information processing and resource
dependence perspectives. Acad. Manag. J. 54 (4), 833–856. http://dx.doi.org/
10.5465/amj.2011.64870145.
Bunkley, N., 2011. Piecing together a supply chain. The New York Times. Retrieved
from 〈http://www.nytimes.com/2011/05/13/business/global/13auto.html?_r=0〉.
Chopra, S., Sodhi, M.S., 2014. Reducing the risk of supply chain disruptions. MIT
Sloan Manag. Rev. 55 (3), 73–80.
Christopher, M., Peck, H., 2004. Building the resilient supply chain. Int. J. Logist.
Manag. 15 (2), 1–14. http://dx.doi.org/10.1108/09574090410700275.
Coombs, W.T., Holladay, S.J., 1996. Communication and attributions in a crisis: an
experimental study in crisis communication. J. Public Relat. Res. 8 (4), 279–295.
http://dx.doi.org/10.1207/s1532754xjprr0804_04.
Croson, R., Donohue, K., 2003. Impact of POS data sharing on supply chain
management: an experimental study. Prod. Oper. Manag. 12 (1), 1–11. http:
//dx.doi.org/10.1111/j.1937-5956.2003.tb00194.x.
Croson, R., Donohue, K., 2005. Upstream versus downstream information and its
impact on the bullwhip effect. Syst. Dyn. Rev. 21 (3), 249–260. http://dx.doi.org/
10.1002/sdr.320.
Croson, R., Donohue, K., 2006. Behavioral causes of the bullwhip effect and the
observed value of inventory information. Manag. Sci. 52 (3), 323–336. http://dx.
doi.org/10.1287/mnsc.1050.0436.
Croson, R., Donohue, K., Katok, E., Sterman, J., 2014. Order stability in supply chains:
coordination risk and the role of coordination stock. Prod. Oper. Manag. 23 (2),
176–196. http://dx.doi.org/10.1111/j.1937-5956.2012.01422.x.
Deloitte, 2013. Executives see growing supply chain risks, costlier disruptions.
Retrieved from 〈http://deloitte.wsj.com/cfo/2013/02/27/executives-see-growing-
supply-chain-risks-costlier-disruptions/〉.
Falk, A., Heckman, J.J., 2009. Lab experiments are a major source of knowledge in
the social sciences. Science 326 (5952), 535–538. http://dx.doi.org/10.1126/
science.1168244.
Fiala, P., 2005. Information sharing in supply chains. Omega 33, 419–423. http://dx.
doi.org/10.1016/j.omega.2004.07.006.
Fletcher, G.J.O., Danilovics, P., Fernandez, G., Peterson, D., Reeder, G.D., 1986.
Attributional complexity: an individual differences measure. J. Personal. Soc.
Psychol. 51 (4), 875–884. http://dx.doi.org/10.1037/0022-3514.51.4.875.
178 S. Sarkar, S. Kumar / Int. J. Production Economics 169 (2015) 169–178
Ford, J.D., 1985. The effects of causal attributions on decision makers' responses to
performance downturns. Acad. Manag. Rev. 10 (4), 770–786. http://dx.doi.org/
10.5465/AMR.1985.4279100.
Forrester, J., 1958. Industrial dynamics: a major breakthrough for decision makers.
Harv. Bus. Rev. 36 (4), 37–66. http://dx.doi.org/10.1007/978-3-642-27922-5_13.
Geary, S., Disney, S.M., Towill, D.R., 2006. On bullwhip in supply chains—historical
review, present practice and expected future impact. Int. J. Prod. Econ. 101 (1),
2–18. http://dx.doi.org/10.1016/j.ijpe.2005.05.009.
Gino, F., Pisano, G., 2008. Toward a theory of behavioral operations. Manuf. Serv.
Oper. Manag. 10 (4), 676–691. http://dx.doi.org/10.1287/msom.1070.0205.
Grubbs, F.E., 1969. Procedures for detecting outlying observations in samples.
Technometrics 11 (1), 1–21.
Gurnani, H., Mehrotra, A., Ray, S., 2012. Supply Chain Disruptions:Theory and
Practice of Managing Risk. Springer, London.
He, Y., Zhao, X., 2012. Coordination in multi-echelon supply chain under supply and
demand uncertainty. Int. J. Prod. Econ. 139, 106–115. http://dx.doi.org/10.1016/j.
ijpe.2011.04.021.
Heckmann, I., Comes, T., Nickel, S., 2015. A critical review on supply chain risk –
Definition, measure and modeling. Omega 52, 119–132. http://dx.doi.org/
10.1016/j.omega.2014.10.004.
Hendricks, K.B., Singhal, V.R., 2003. The effect of supply chain glitches on share-
holder wealth. J. Oper. Manag. 21 (5), 501–522. http://dx.doi.org/10.1016/j.
jom.2003.02.003.
Hendricks, K.B., Singhal, V.R., 2005. An empirical analysis of the effect of supply chain
disruptions on long-run stock price performance and equity risk of the firm. Prod.
Oper. Manag. 14 (1), 35–52. http://dx.doi.org/10.1111/j.1937-5956.2005.tb00008.x.
Kelle, P., Akbulut, A., 2005. The role of ERP tools in supply chain information
sharing, cooperation, and cost optimization. Int. J. Prod. Econ. 93–94, 41–52.
http://dx.doi.org/10.1016/j.ijpe.2004.06.004.
Kelley, H.H., Michela, J.L., 1980. Attribution theory and research. Annu. Rev. Psychol.
31 (1), 457–501. http://dx.doi.org/10.1146/annurev.ps.31.020180.002325.
Kleindorfer, P.R., Saad, G.H., 2005. Managing disruption risks in supply chains. Prod.
Oper. Manag. 14 (1), 53–68.
Knemeyer, A.M., Zinn, W., Eroglu, C., 2009. Proactive planning for catastrophic
events in supply chains. J. Oper. Manag. 27 (2), 141–153. http://dx.doi.org/
10.1016/j.jom.2008.06.002.
Knemeyer, M.A., Naylor, R.W., 2011. Using behavioral experiments to expand our
horizon and deepen our understanding of logistics and supply chain decision
making. J. Bus. Logist. 32 (4), 296–302.
Lee, H.L., Padmanabhan, V., Whang, S., 1997. The bullwhip effect in supply chains.
Sloan Manag. Rev. 38 (3), 93–102.
Li, L., Zhang, H., 2008. Confidentiality and information sharing in supply chain
coordination. Manag. Sci. 54 (8), 1467–1481. http://dx.doi.org/10.1287/
mnsc.1070.0851.
Niranjan, T.T., Wagner, S.M., Bode, C., 2011. An alternative theoretical explanation
and empirical insights into overordering behavior in supply chains. Decis. Sci.
42 (4), 859–888.
Piliavin, I.M., Rodin, J., Allyn, P.J., 1969. Good samaritanism: an underground
phenomenon? J. Personal. Soc. Psychol. 13 (4), 289–299. http://dx.doi.org/
10.1037/h0028433.
Prajogo, D., Olhager, J., 2012. Supply chain integration and performance: the effects
of long-term relationships, information technology and sharing, and logistics
integration. Int. J. Prod. Econ. 135, 514–522. http://dx.doi.org/10.1016/j.
ijpe.2011.09.001.
Rong, Y., Shen, Z.J.M., Snyder, L.V., 2008. The impact of ordering behavior on order-
quantity variability: a study of forward and reverse bullwhip effects. Flex. Serv.
Manuf. J. 20, 95–124. http://dx.doi.org/10.1007/s10696-009-9054-3.
Sarkar, S., Shewchuk, J.P., 2013. Use of advance demand information in multi-stage
production-inventory systems with multiple demand classes. Int. J. Prod. Res.
51 (1), 57–58. http://dx.doi.org/10.1080/00207543.2011.648279.
Schmidt, W., Raman, A., 2012. When supply-chain disruptions matter (No. 13-006).
Harvard Business School Working Paper (Revised January 2013).
Schmitt, A.J., Singh, M., 2012. A quantitative analysis of disruption risk in a multi-
echelon supply chain. Int. J. Prod. Econ. 139 (1), 22–32. http://dx.doi.org/
10.1016/j.ijpe.2012.01.004.
Schmitt, A.J., Sun, S.A., Snyder, L.V., Shen, Z.J.M., 2015. Centralization versus
decentralization: risk pooling, risk diversification, and supply chain disruptions.
Omega 52, 201–212. http://dx.doi.org/10.1016/j.omega.2014.06.002.
Siegel, S., Castellan, J.N.J., 1988. Nonparametric Statistics for the Behavioral Sciences.
McGraw-Hill, Inc, New York.
Snyder, LV., Atan, Z., Peng, P., Rong, Y., Schmitt, AJ., Sinsoysal, B., (2012). OR/MS
Models for Supply Chain Disruptions: A Review. Working Paper.
Sodhi, M.S., Tang, C.S., 2012. Managing Supply Chain Risk. vol. 172. Springer Science
& Business Media, New York.
Steckel, J.H., Gupta, S., Banerji, A., 2004. Supply chain decision making: will shorter
cycle times and shared point-of-sale information necessarily help? Manag. Sci.
50 (4), 458–464. http://dx.doi.org/10.1287/mnsc.1030.0169.
Sterman, J.D., 1989. Modeling managerial behavior: misperceptions of feedback in a
dynamic decision making experiment. Manag. Sci. 35 (3), 321–339.
Tang, C.S., 2006. Perspectives in supply chain risk management. Int. J. Prod. Econ.
103 (2), 451–488. http://dx.doi.org/10.1016/j.ijpe.2005.12.006.
Thonemann, U.W., 2002. Improving supply-chain performance by sharing advance
demand information. Eur. J. Oper. Res. 142, 81–107. http://dx.doi.org/10.1016/
S0377-2217(01)00281-8.
Tokar, T., 2010. Behavioural research in logistics and supply chain management. Int.
J. Logist. Manag. 21 (1), 89–103. http://dx.doi.org/10.1108/09574091011042197.
Tokar, T., Aloysius, J.A., Waller, M.A., 2012. Supply chain inventory replenishment:
the debiasing effect of declarative knowledge. Decis. Sci. 43 (3), 525–546. http:
//dx.doi.org/10.1111/j.1540-5915.2012.00355.x.
Tokar, T., Aloysius, J.A., Waller, M.A., Williams, B.D., 2011. Retail promotions and
information sharing in the supply chain: a controlled experiment. Int. J. Logist.
Manag. 22 (1), 5–25. http://dx.doi.org/10.1108/09574091111127534.
Tokar, T., Aloysius, J., Williams, B., Waller, M., 2014. Bracing for demand shocks: an
experimental investigation. J. Oper. Manag. 32 (4), 205–216. http://dx.doi.org/
10.1016/j.jom.2013.08.001.
Tomlin, B., 2006. On the value of mitigation and contingency strategies for
managing supply chain disruption risks. Manag. Sci. 52 (5), 639–657.
Weinstein, N.D., 1980. Unrealistic optimism about future life events. J. Personal. Soc.
Psychol. 39 (5), 806–820. http://dx.doi.org/10.1037/a0020997.
Wu, D.Y., Katok, E., 2006. Learning, communication, and the bullwhip effect. J. Oper.
Manag. 24 (6), 839–850. http://dx.doi.org/10.1016/j.jom.2005.08.006.
Xia, Y., Ramachandran, K., Gurnani, H., 2011. Sharing demand and supply risk in a
supply chain. IIE Trans. 43 (6), 451–469. http://dx.doi.org/10.1080/
0740817X.2010.541415.
Xiao, T., Yang, D., 2009. Risk sharing and information revelation mechanism of a
one-manufacturer and one-retailer supply chain facing an integrated compe-
titor. Eur. J. Oper. Res. 196 (3), 1076–1085. http://dx.doi.org/10.1016/j.
ejor.2008.05.004.
Zhao, X., Xie, J., Zhang, W.J., 2002. The impact of information sharing and ordering
co-ordination on supply chain performance. Supply Chain Manag.: Int. J. 7 (1),
24–40. http://dx.doi.org/10.1108/13598540210414364.
Zhao, Y., Zhao, X., 2015. On human decision behavior in multi-echelon inventory
management. Int. J. Prod. Econ. 161, 116–128. http://dx.doi.org/10.1016/j.
ijpe.2014.12.005.
Zhou, H., Benton, W., 2007. Supply chain practice and information sharing. J. Oper.
Manag. 25 (6), 1348–1365. http://dx.doi.org/10.1016/j.jom.2007.01.009.