Stephan M. Wagner

Negative Default Dependencies in Supplier Networks

Daniel Dumke — Mon, 18 Jun 2012 12:45:08 +0000

If people talk about disruptions and network effects within the supply chain, the associations are most often negative.

The picture of an automotive/just-in-time supply chain comes to mind, where a small screw from a distant supplier did not get delivered in time and all production processes within the whole network suddenly come to an involuntary halt.

But on the other hand there are companies profiting from these smaller and larger disruptions: competition.

To analyze these effects we have a look at the consequences of negative default dependence between suppliers. The full paper can be found here.

Default dependence and method

Empirical research on corporate defaults in the finance literature indicates that corporate defaults often cluster in time and that the default of a company is frequently affected by the defaults of other companies. […]

In the automotive industry there are several reasons why positive default correlation may exist in supplier networks. First, automotive suppliers face similar challenges, such as large and powerful customers who force suppliers constantly to cut costs and invest heavily in R&D or the volatile prices of raw materials. It is likely that the automotive suppliers have to suffer from the consequences of these challenges in a similar way or are reacting in a comparable manner to cope with them. Second, suppliers may maintain relationships with other suppli- ers and share “technical and explicit information as well as tacit information” and “work together closely, exchange ideas, and even engage in joint venture projects.” Being linked so closely may result in comparable strategic and operative actions and behavior of the supplier firms. The consequence is that decisions that lead to financial problems are likely to be taken by both suppliers that are linked through close supplier–supplier relationships.

However there are also reasons/situation in which default-events might be negatively correlated:

First, after a supplier default, customers might shift business to another supplier in the network.

Second, the default of a supplier can result in lay-offs and the competitor will be able to hire more and qualified staff.

Finally, due to the reduced number of alternative sources, the buying firm may become more dependent on the surviving supplier who, through the gained power, may be able to incur higher profit margins and, thus, gain in financial stability.

The authors use copula-functions.

Financial data to calculate default probabilities for a case study are derived from Datastream (Thomson Reuters). This data was used to calculate the individual default intensities.

For the worldwide 100 largest suppliers to the automotive OEMs in 2005 that were included in the Datastream database, we extracted the necessary data required for specifying and adjusting our model.

Figure 1 shows the default intensities for selected companies.

Figure 1: Company Characteristics (Wagner et. al, 2011)

The default dependencies were calculated using numerical results of a simulation.

Results and implications

The authors draw three conclusions from the results of their analysis.

First, our estimation of default intensities of selected first-tier suppliers in the automotive industry supports the concerns raised in the literature about the financial stability of automotive suppliers. Supplier default intensities above 5% are disquieting for the respective automotive OEMs.

Second, the simulation results depict that negative default dependence among suppliers in a supplier network has consequences for the survival probabilities of the entities in the network. The higher the individual default intensity of a supplier, the stronger the effect of negative default dependence on its survival probability after the default of the other supplier. […] for example, the portfolio with low default intensity suppliers demonstrated to increase the survival probability of the second supplier by 2.7% and the portfolio with high default intensity suppliers by 15.4% (in comparison to the independence case).

Third, in addition to the dependence level, the dependence structure, reflected in our model by the choice of copula, is an important factor for modeling default dependence in a supplier portfolio.

The following management implications can be given:

Purchasing managers should be aware that negative default dependence between suppliers may exist and take this into account for their sourcing decisions. A better understanding of the randomness and relatedness of supplier defaults internal to the supplier network can help firms to plan for uncertainty, take proactive measures to reduce risk (e.g., switch a supplier), and achieve better, less variable outcomes.

Firms should preferably establish relation- ships with suppliers that have low default intensities, and with suppliers that will benefit from the default of their competitors – given that the default of the competitor will not significantly shift the power in the buyer–supplier relationship

Conclusion

Not only surviving competitors can potentially profit from the default of its contestant, but also their clients may profit.

This research shows that interdependencies – no matter if positive or negative – have to be analyzed and should included in the decision making process.

One has to keep in mind though, that Wagner et al.‘s results heavily rely on their method to estimate the default dependencies within the supplier portfolio.

This might induce additional uncertainty in the form of model risk.

Reference:

Wagner, S., Bode, C., & Koziol, P. (2011). Negative default dependence in supplier networks International Journal of Production Economics, 134 (2), 398-406 DOI: 10.1016/j.ijpe.2009.11.013

Rate This:

Tags:

dependence

correlation

supplier

default

An Empirical Investigation into Supply Chain Vulnerability

Daniel Dumke — Tue, 05 Jun 2012 06:30:56 +0000

One basic assumption in risk-aware supply chain design is the notion that the design of the supply chain actually has an impact on the vulnerability of the supply chain.
This question has been analyzed about six years ago in a broad empirical study by Wagner and Bode.

Method

The authors use a rather large sample of companies in Germany. Overall nearly 5000 supply chain professionals were asked to participate and 760 actually took part in the study.
Most sample companies had an industrial focus (72% versus service (20%) and trade (9%)).
This study is founded on a similar sample as this other study by Wagner and Bode analyzing the impact of risks on supply chain performance.

Model

The author focus on some supply chain design variables, which supposedly increase supply chain vulnerabilities. Figure 1 shows the assumed relationship between those drivers of supply chain vulnerability and three supply chain risk categories.

Figure 1: Concept: Relationship between Design and Risk

The authors propose the following hypothesis, which are then tested using the empirical data:

H1: The higher the drivers of supply chain vulnerability, the higher the level of demand side risk a firm faces.

H2: The higher the drivers of supply chain vulnerability, the higher the level of supply side risk a firm faces.

H3: The higher the drivers of supply chain vulnerability, the higher the level of catastrophic side risk a firm faces.

Results

The results show that all hypothesis are supported by the findings of the authors. However the design factors/vulnerabilities only explain part of the observed supply chain risks (7% for H1, 13% for H2, 3% for H3).
Demand side risk was influenced by strong customer dependence and strong supplier dependence.
Supply side risk was influenced by supplier dependence, single sourcing and global sourcing.
Lastly, catastrophic risk was impacted by the degree of global sourcing.

The authors draw the following conclusions:

First, the supply chain vulnerability variables in our model explain a rather small portion of the variance in the risk arising from demand side risk sources. It is a low but not astonishing value since the majority of the vulnerability variables focuses on the upstream supply chain. However, the results reveal that customer dependence increases demand side risk. This finding indicates that firms that are dependent on some customers are exposed to a higher risk of suffering from the detrimental effects of demand volatility and poor downstream information. This could be because of order batching or limited possibilities of demand pooling. […] This leads to the hypothesis that, beyond the investigated variables, there are several additional aspects both internal and external to the supply chain that determine a firm’s exposure to supply chain risk.

Second, risk derived from supply side risk sources is elevated by supplier dependence, single sourcing and global sourcing. Supplier dependence obviously amplifies the threat from poor quality, supply shortages, sudden demise of one of these suppliers, and poor logistics performance. Although this argumentation also applies to single sourcing, the single sourcing approach seems to be less hazardous than general dependence on some suppliers. This is because single sourcing is usually aligned with a closer relationship that might absorb some of the supply side risk.

Third, when it comes to risk from catastrophic risk sources it has to be taken into consideration that the sample data was collected in Germany which has been a very “calm” place with regard to disasters. Here, it is solely global sourcing that is a significant factor that exposes firms to higher risk from catastrophes. The implementation of a global sourcing strategy stretches the supply chain geographically which ultimately means more peril points for the information and material flow. The robustness and resilience of regional or national supply chains is usually higher. Surprisingly, the study shows that supplier dependence decreases the risk exposure to catastrophes. Again we would argue that this is because of lack in supply flexibility. Firefighting against the consequences of catastrophic events might be more successful with the ability to quickly adjust the supply.

Conclusion

I would argue that it is always hard to measure risk consistently in a qualitative study. People are likely to evaluate the same risk quite differently, which might lead to unclear results.
Furthermore the low impact of these specific design variables emphasizes the view that there are many more factors (internal and external) that impact the exposure to supply chain risk.

Reference:

Wagner, S.M., & Bode, C. (2006). An Empirical Investigation into Supply Chain Vulnerability Experienced by German Firms Erich Schmidt Verlag, 79-96

Rate This:

Tags:

design

vulnerability

exposure

Modeling Defaults of Companies in Multi-Stage SC Networks

Daniel Dumke — Mon, 28 May 2012 14:07:06 +0000

Agent-based supply chain models are build using small entities (agents), which might represent a single company.
Each of the agents has its own goals and rules of operation programmed into a computer. The interaction between several agents of this kind leads to a more realistic and complex behavior of the system.

There are several different schools for the quantitative analysis of supply chain risks. Simulation is one of them and agent-based models show several distinct advantages: They are both easier to understand and allow for a more complex system behavior, than other quantitative methods.

After this introduction I would like to have a look at a current agent-based supply chain model, which analyzes the effect of bankruptcies on supply chains. The full paper can be downloaded here.

Model

The authors focus on a comprehensive view on the supply chain: several stages (horizontally and vertically) are modeled. Figure 1 shows an exemplary supply chain.

Figure 1: Exemplary Supply Chain Structure (Mizgier et al. 2012)

Each circle represents one node or agent, the connections are drawn as lines.
The model features five additional characteristics worth mentioning:

Price dispersion (prices can vary between companies)
Evolution of supply chain topology (links between companies can be changed, by the agents themselves)
Network reconfiguration (the reconfiguration is based on the price)
Production dynamics (output is determined by the invested working capital)
Dynamics of costs of production (random changes to the environment every five periods, lead to changes in the cost function of the companies)

Last but not least, companies may go bankrupt if they are not able to perform their short therm debts. There are no loans.

Results

First, the network performs as expected. During the first period turbulences can be observed. Figure 2 shows the utilization of working capital during the simulation (1 equals 100%).

Figure 2: Capacity Utilization (Mizgier et al. 2012)

After 200 iterations a typical start scenario might look like figure 3.

Figure 3: Network Structure after Initialization Period (Mizgier et al. 2012)

Starting from this state the

The firms with the best profit/cost ratio are growing and adding new suppliers, whereas the working capital of the firms whose sales price is higher than the mean price of the given stage is slowly decaying and results in the defaults of firms.

A stable state might look like figure 4.

Figure 3: Network Structure after Stabilization (Mizgier et al. 2012)

The authors deduce three implications from those results:

The first implication is that during the process of assessment of the company’s risk exposure, managers should keep their focus on the global structure of the supply chain network instead of being restricted to the own portfolio of suppliers and customers.

Secondly, as a result of the dynamics of the topology of the supply chain network, strong competition in prices and fast changing technology, even the most reliable firms should be monitored and constantly re-evaluated in terms of their production capacity and risks associated with their structure of connections.

Third and most important, managers should find ways to cut costs and reinvest the free cash flows in new technology of production, which will allow further cost reductions and the development of new innovative and cheaper products.

Conclusion

Being stuck in the supply chain can lead to negative consequences, when ripple effects cause multiple suppliers and customers to default. Furthermore, these supply chain partners might also not be the most efficient ones, and these prices affect each participating company.
All in all, a great paper, but I would have liked to read more about the authors’ efforts to validate and verify the model’s integrity.

Reference:

Mizgier, K., Wagner, S., & Holyst, J. (2012). Modeling defaults of companies in multi-stage supply chain networks International Journal of Production Economics, 135 (1), 14-23 DOI: 10.1016/j.ijpe.2010.09.022

Rate This:

Tags:

agent-based

Simulation

analysis

Assessing Vulnerability of a Supply Chain

Daniel Dumke — Mon, 10 Oct 2011 14:42:00 +0000

At this year’s HICL conference in Hamburg, I was able to present some of my own research. In the follow-up discussions several points were highlighted, especially focussing on the viability of supply chain wide cooperation and collaboration efforts and on the difficulties of doing a realistic quantification of supply chain risks.

I already read a great paper on this topic some time ago: “Assessing the vulnerability of supply chains using graph theory” by Stephan M. Wagner and Nikrouz Neshat (2010), which I present you today.

Disruptions and vulnerabilities

Several authors argue that several factors help increase the vulnerabilities of today’s supply chains. When supply chain complexity increases (e.g. supply chain length, higher division of labor, …), the vulnerabilities also rise. Furthermore there is evidence that natural and man-made disasters are on the rise as well (figure 1).

Figure 1: History of Disasters (Wagner and Neshat, 2010)

Figure 2 shows the links between supply chain vulnerability drivers and disruptions.
So, since natural- and man-made-disaster most often cannot be influenced directly, the authors argue that the focus has to be on reducing the vulnerabilities themselves.

Figure 2: Connections between Vulnerabilities, Disruptions and Risks (Wagner and Neshat, 2010)

Assessment of vulnerabilities

Wagner and Neshat suggest a four step algorithm based on graph theory, which is used to calculate a Supply Chain Vulnerability Index (SCVI). The algorithm is based on graph theory (which has been part of another study here). Key to the understanding of the SCVI is the risk driver mentioned above. Figure 3 shows an abstract example with three vulnerability drivers (D1 to D3) and their links.

Figure 3: Vulnerability Drivers and their Links with the corresponding Matrix (Wagner and Neshat, 2010)

To calculate the SCVI the supply chain risk manager has to create a map / graph of the current risk drivers (step 1), find the corresponding relations between the risk drivers (step 2), calculate the “influence matrix” (step 3) and deduce the SCVI. As a forth step SCVIs of different companies can be compared to gain further insights for optimization and risk reduction.

Figure 4 shows the main drivers of supply chain vulnerability.

Figure 4: Selection of Vulnerability Drivers in Structure, Demand and Supply (Wagner and Neshat, 2010)

Using graph theory makes it possible for the risk manager to get an overview of the influencing factors of supply chain risks. It also enables him to act on this knowledge. Figure 5 shows a possible simplification / vulnerability reduction.

Figure 5: Reduction of Supply Chain Vulnerability Drivers leads to a Simplification of the respective Graphs (Wagner and Neshat, 2010)

D3 in figure 5 “is called a ‘sink’ – which means that D3 can only be influenced by three other drivers and cannot influence others. Considering the graph nodes and edges, supply chain managers can apply risk management methods and implement mitigation strategies to omit or alleviate some of the vulnerability drivers. Figure 5 shows the graph after D3 has been omitted. As one can see, the resulting graph contains less vulnerability than it did prior to implementing the measures.”

Empirical data for the SCVI

The authors also conduct a survey with over 700 participants from different industries. Based on their feedback the authors assess the Supply Chain Vulnerability Index for eight different industries. The results are summarized in figure 5. And give support to the hypothesis that the automotive industry has one of the highest risk levels.

Figure 6: Supply Chain Vulnerability within several Industries compared (Wagner and Neshat, 2010)

Conclusion

I really like the graph approach to assessing supply chain vulnerabilities. And I think it is a great method to support the understanding of a complex system like the supply chain. The article combines two very interesting aspects of it: the practical implementation and the assessment of supply chain vulnerability and a survey to compare different vulnerability levels across industries. The complete article, where you find more details on the survey results and the algorithm for calculating the index, can be downloaded here.
From a business and research point of view this article should direct the supply chain risk management efforts especially in the industries with the highest risk levels, Automotive and ICT.

Reference:

Wagner, S.M., & Neshat, N. (2010). Assessing the vulnerability of supply chains using graph theory International Journal of Production Economics, 126, 121-129

Rate This:

Tags:

Impact of Risks on Supply Chain Performance

Daniel Dumke — Mon, 25 Apr 2011 14:15:00 +0000

I am often astounded by the fact how many great articles I haven’t read yet. A good scientific paper contains an comprehensive description of the methodologies used, a theoretical foundation and literature review from which hypothesis are drawn, which are then confirmed or rejected in the course of the paper. And of course, it is always a plus to actually find some results in the course of the analysis.

From this category I would like to introduce an article by Wagner and Bode (2008) on an empirical examination of supply chain performance along several dimensions of risk. The complete article is defnitly worth a read and can be found here (free of charge).

Introduction & Hypothesis

The authors start by highlighting the major factors involved in the importance of supply chain risk management nowadays: There has been an upward trend in crisis and catastrophes, at least on the level of media coverage and awareness. Furthermore modern supply chains have been optimized for leanness in recent years, leaving them more exposed towards risks.
To decide on the the possible investments on mitigation strategies, a supply chain manager needs to know the probability of an adverse event and the effect of the disruption on supply chain performance.
Hendricks and Singhal (reviewed here 1, 2) already concluded, that SC risks can have severe consequences on performance and the stock price.The goal of this article was to establish a more direct link between what risks affect the performance most. The authors build their hypothesis building on contingency and strategic choice theory, arguing that the environment is a major determinant of business performance, nonetheless managers still can make “strategic choices” to improve performance. Of course this sounds like common knowledge, and it is, but it is essential from a scientific point of view to root his own ideas in a foundational theory.

Using a literature review the authors decide on the following risks to include into their research: (1) demand side; (2) supply side; (3) regulatory, legal and bureaucratic; (4) infrastructure; and (5) catastrophic.
From these risks five hypothesis emerge, following this structure:

H₁: The higher the demand side risk, the lower the supply chain performance. …

Methods

In contrast to Hendricks and Singhal, who analyzed the impact on the level of stock prices and public company performance data, Wagner and Bode conducted a survey with a final sample size of 790 supply chain professionals from companies located in Germany. Non-response bias has not been found. The questionnaire has been carefully designed, for each hypothesis multiple measures were included. E.g. demand side risk measures consisted of two items that capture the risk deriving from the interaction (or lack thereof) with customers and volatility of the market. The results were analyzed using correlation analysis and reliability evaluations

Results

The results show that in the view of the experts demand and supply risks have an impact on supply chain performance, whereas there is no significant relationship between regulatory, infrastructure and catastrophic risks and supply chain performance.

Results of Correlation Analysis (Wagner and Bode, 2008)

On first sight these results stand in contrast to the results of Hendricks and Singhal, where a huge impact of any disruption on performance has been encountered. This can be explained by the different focus of the analysis, where Wagner and Bode took a sample from all companies, which naturally only seldom experience severe disruptions whereas Hendricks and Singhal especially picked the larger (and worthy of a public announcement) disruptions. So a catastrophic event still has a negative impact on the sample group, but since the probability is so low it does not pose a significant impact on performance.

The managerial implications therefore are:

Importance of a supply chain risk management framework
Scarce resource should be deployed to mitigate the “every day” risks from supply and demand

Conclusion

This article also concludes as a text book example of “how to write a scientific paper”, it finishes with recommendations for the implementation of the results and further research gaps and still leaves something to say for critics like me:

It is very interesting to see the findings by Hendricks and Singhal in a new light, from the point of view of a supply chain professional.
But this study also points the perception of risks (have a look Part 1 and Part 2) as a major driving factor in these results.
Wagner and Bode also discuss the factual risk situation in the sample companies (all from Germany), which probably skews the results. All the risks mentioned might be (perceived or real) quite different in other countries.

Reference:

Wagner, S.M., & Bode, C. (2008). An empirical Examination of Supply Chain Performance along Several Dimensions of Risk Journal of Business Logistics, 29 (1), 307-325

Rate This:

Tags:

performance

risk

disruptions