Research
My research focuses on applying econometrics, machine learning/AI, and simulation to solve real-world data-driven business problems in the context of supply chain operations management (SCOM). The central theme of my dissertation is the vulnerability of global supply chain networks.
Additionally, at MIT, I am actively involved in two main areas of research. Firstly, I am studying human-AI collaboration in demand planning through a field experiment conducted in collaboration with a major multinational fast-moving consumer goods (FMCG) company and its customer. Using Mediation Analyses, I am investigating the conditions under which human-AI collaboration is beneficial and when the algorithm version appears. Secondly, I am empirically analyzing the impact of supply chain digitalization on firm productivity.
In conducting these studies, I utilize machine learning techniques, including text analytics, web scraping, and image processing, to leverage both structured and unstructured data sources such as the web, images, and social media.
Research Interests
Topics:
Methods:
Journal Publications: [Google Scholar]
Applied Econometrics
Using a large sample of approximately 210,000 supplier-firm relationships, we develop new measures to operationalize
the industrial and geographical concentrations of firms' upstream industries. The developed measures show an
upstream industry might be competitive and contain many suppliers (low industrial concentration) and yet be
concentrated in a few nations or small geographical areas (high geographical concentration), thus leaving the
downstream firm vulnerable to geographical or geopolitical disruptions. Using COVID-19 as an exogenous shock,
we document that the firms whose upstream suppliers are operating in highly concentrated industries, on average,
experienced a reduction of 250 million US dollars in quarterly sales. Thus, competitive industries with ample
suppliers may still impose significant risks on downstream firms if the upstream suppliers' industries are
concentrated mainly in a few nations or small geographical areas. We find evidence that the link between upstream
suppliers' industry-level concentrations and downstream firms' sales growth is heterogeneous. Specifically, firms
can moderate the negative relationship by adjusting their supply chain networks. The negative relationship is weaker
for firms with a geopolitically diversified supply base and firms with more domestic suppliers. Our estimates are
robust and consistently identified through several different empirical strategies. The results of this study may
serve to help policymakers and managers systematically identify locations in supply networks that have the potential
to create bottlenecks. These bottlenecks may cause severe shortages or even threaten national security. Additionally,
the measures give insights to policymakers into which industries to support for onshoring.
Keywords: Global Supply Chain Network, Firms Resilience, Industrial & Geopolitical Concentrations, COVID-19
This study examines how firms adjust their sourcing decisions due to variations in policy uncertainty and if such
adjustments result in performance effects. Specifically, we assess how policy uncertainty influences a
focal firm’s decisions to restructure its supply chain along two dimensions: (a) onshoring/offshoring and
(b) geopolitical diversification/concentration. The empirical evidence suggests that managers decrease the
the ratio of onshore suppliers and diversify their supply base geopolitically in response to heightened upstream
policy uncertainty affecting the suppliers. However, we do not find evidence of the influence of domestic
policy uncertainty on the focal firm’s decision to adjust the structure of its supply base. To provide
actionable advice for mitigating adverse effects resulting from policy uncertainty, we also conduct a
posthoc analysis linking supply chain structure to focal firms’ performance. Our results indicate that,
consistent with transaction cost economics, firms with more suppliers onshore and a geopolitically
concentrated supply base demonstrates superior performance when their suppliers are exposed to heightened
policy uncertainty. As such, our investigation reveals that the managerial response of increasing offshoring
and diversifying supply bases may be counterproductive. The estimates are robust and identified through several empirical methodologies..
Keywords: Global Supply Chain Network, Policy Uncertainty, Resilience
Supply networks are vulnerable to disruptions not only because of their immediate damages but also because
of their cascading effects. We posit that the structure of supply networks plays a significant role in cascading
disruption and that the ability to find nexus suppliers can help to lessen the impact of cascading disruptions.
Therefore, we examine the effect of three network structural characteristics: modularity, assortativity, and
structural holes. We find that the lack of ties among the supplier’s neighbors (structural holes), increases the
impact of a cascade emanating from that supplier. In addition, modularity and assortativity have heterogeneous
effects on cascading disruptions across the conditional distributions of the outcome variables of performance.
Modularity and assortativity generally have deteriorating effects on network resilience; however, they have
beneficial effects during extreme events. In addition, we develop a new predictive model to identify the nexus
suppliers whose disruptions would significantly impact focal firms’ operations. The model gives managers a
tool to effectively identify nexus suppliers located deep in a supply network..
Keywords: Supply network, Nexus supplier, Agent-based simulation, Quantile regression, Multi-level model
We use a large-scale panel dataset to investigate the relationship between suppliers' exposure to political instability and firms'
performance. The empirical evidence suggests a U-shaped relationship between the political instability of suppliers' countries
and buyer firms' performance, indicating that firms experience the highest benefits when suppliers are located at the extreme
ends of the political instability spectrum. Sourcing from the least politically unstable countries can reduce transaction costs
(costs incurred in conducting economic transactions) to improve firm performance by limiting information search and contract
negotiations, which align with the purchase cost minimization explanations of the transaction cost economics (TCE) framework.
On the other hand, sourcing from highly politically unstable countries allows firms to leverage institutional arbitrage and
bargaining power to decrease procurement prices, aligning with TCE explanations related to agreement enforcement and compliance
expenses. Furthermore, we discovered that the relationship varies depending on the industry category of the firms. The U-shaped
relationship is more pronounced for firms in industries with higher productivity, while it is inverted for firms in durable goods
industries. Interestingly, our post hoc analyses did not reveal any U-shaped effect between economic policy uncertainty (EPU)
of suppliers' countries and buyer firms' performance. In fact, our sample reveals a negative linear relationship between the
EPU of suppliers' countries and buyer firms' performance, highlighting the distinction between EPU and political instability.
Our estimates are robust, consistently identified through several empirical methodologies, and align with theoretical predictions..
Keywords: Supply Chain, Global Sourcing, Political Instability, U-Shaped Effect, Empirical Research
Applied Optimization
Supply chains are prone to several operational and disruption risks. In order to design a resilient supply chain network
capable of responding to such potential risks suitably, this paper proposes a novel framework for the business
continuity-inspired resilient supply chain network design (BCRSCND) problem, which includes three steps. First,
four resilience dimensions including Anticipation, Preparation, Robustness, and Recovery are considered to quantify
the resilience score of each facility using a multi-criteria decision-making technique and considering a comprehensive
set of resilience strategies. In the second step, the critical processes and their business continuity metrics
(which are vital for supply chain continuity), are identified. The outputs of the first two steps provide the inputs
of a novel two-stage mixed possibilistic-stochastic programing (TSMPSP) model. The model aims to design a multi-echelon,
multi-product resilient supply chain network under both operational and disruption risks. The proposed TSMPSP model allows
decision makers to incorporate their risk attitudes into the design process. After converting the original TSMPSP model
into the crisp counterpart, several sensitivity analyses are conducted on different features of hypothetical disruptions
(i.e. their severity, likelihood and location) and DM’s risk attitudes from which useful managerial insights are provided.
Keywords: Resilient Management, Supply Chain Design, Business Continuity Management, Two-Stage Stochastic Programming, Pssibilistic Programming
This paper investigates the use of sourcing strategies to achieve supply chain resilience under disruptions.
The coping strategies considered are single and multiple sourcing, backup supplier contracts, spot purchasing,
and collaboration and visibility. Collaboration and visibility, which affect suppliers’ recovery capabilities
and a buyer’s warning capability, have not been similarly modelled in the past. A scenario-based mathematical
model is developed such that it considers objectives under uncertainties including disruption risks and operational
risks. A broad numerical study examines its output for various risk attitudes in a decision-maker, ranging from risk
neutral to risk averse. The sensitivity of procurement strategies to other key parameters such as recovery and warning
capabilities is examined. One of the major findings is that buyer’s warning capability plays a vital role in enhancing
supply chain resilience. We seek to build on these efforts to further support disruption planning and mitigation and
to obtain a deeper understanding of the relationship between supply chain characteristics and resilience.
Keywords: Resilient Supply Chain Design, Disruption Risks, Collaboration Two-Stage Stochastic Programming, Mixed-Integer Programming
Most of current logistics network design models in the literature typically assume that facilities are always available
and absolutely reliable while in practice, they are always subject to several operational and disruption risks.
This paper proposes a reliable closed-loop supply chain network design model, which accounts for both partial
and complete facility disruptions as well as the uncertainty in the critical input data. The proposed model is
of mixed integer possibilistic linear programming type that aims to minimise simultaneously the total cost of
opening new facilities and the expected cost of disruption scenarios. An enhanced possibilistic programming
approach is proposed to deal with the epistemic uncertainty in input data. Furthermore, the p-robustness criterion
is used to limit the cost of disruption scenarios and protect the designed network against random facility disruptions.
Several numerical experiments along with sensitivity analyses on uncertain parameters are conducted to illustrate
the significance and applicability of the developed model as well as the effectiveness of the proposed solution
approach. Our results demonstrate that operational and disruption risks considerably affect the whole structure
of the designed network and they must be taken into account when designing a reliable closed-loop logistics network.
Keywords: Disruptions Risks, Fuzzy Possibilistic Programming, p-Robustness, Miexd-Integer Programming
This paper designs a reliability distribution network with limited capacity under partial and complete
facility disruptions. To increase the reliability of the distribution network, a new mixed integer linear
programing model is developed by considering multiple mitigation strategies including diversification,
fortification, and transshipment. The distribution network constitutes of reliable distribution centers
are more expensive, always available and not affected by disruption, and unreliable distribution centers.
Thus, they might be fortified at any level of reliability. Several numerical examples with sensitivity
analyses are conducted to illustrate the usefulness of the proposed model. Results demonstrate that the
transshipment strategy is more effective than the other mitigation strategies on distribution network
reliability and cost.
Keywords: Distribution Network Design, Disruption Risks, Mitigation Strategies, Pssibilistic Programming
Simulation
Supply chain literature has amply explored the effect of different resilience strategies in the face of supply chain disruptions.
Firms often apply a multitude of resilience strategies in tandem. Such strategies can vary from slack inventory to volume
flexibility and responsiveness to backup capacity. Yet, there is a lack of empirical or analytical evidence in how the
combination of resilience strategies affects firm capabilities in the face of supply chain disruptions. In this paper,
we use simulation modeling techniques to determine the effect of different combinations of resilience strategies in a
systematic and stepwise manner. Our modeling considers different disruption attributes (capacity or delay), their effect
(severity and likelihood), and their origin (i.e. upstream or downstream). A number of interesting observations are made.
First, combining resilience strategies is not always beneficial and can occasionally have detrimental effects. Moreover,
resilience strategies that are beneficial at the node (firm) level may prove ineffective at the system level. We also
find that inventory and volume flexibility are strategies that combine well with others. The study offers several
contributions to research and management in supply chain disruption and the study of resilience.
Keywords: Supply Chain Resilience, Supply Chain Disruption, Propagation, Layered Defense Theory, Simulation
Conference Presentations
Academic Service
Workshop
Journal Referee
Session Chair
INFORMS Student Chapter