Vertical collaboration for multi-modal and multi-carrier transportation chains bears great synergy potential, enabling the integration of otherwise isolated proprietary transport networks, and thus facilitating the flow of goods and creating a more efficient and flexible transport system. However, collaboration requires close alignment in the form of data exchange and integration of information systems, which creates dependencies and a risk of disruption through technical failure, cyber attacks, or organizational conflicts. Earlier works has shown that disruption in interdependent networks can propagate and lead to a cascade of failure, which casts doubt on the claim that more collaboration has a solely positive impact. This research aims at identifying and quantifying a trade-off in collaborative transport systems between synergies and vulnerabilities depending on the level of collaborative connectivity. Therefore, a multi-layer network model for collaborative transport systems is coupled with a model for propagation of intended or unintended data disruption, and the impact on network performance under varying levels of collaborative connectivity is observed for two classes of random networks as well as for the European intermodal transport service network. We find that increasing collaborative connectivity does not have a monotone effect on performance, but there is a threshold, beyond which additional collaborations have an adverse impact on the robustness of the system. Below this threshold, more collaborations have a mostly positive impact on performance, since unused synergy potential is high while the risk of disruption causing a cascade is low. Above it, failure cascades become larger and more likely while the marginal added synergies are diminishing. The formation of highly connected cliques of carriers in the collaboration network facilitates the propagation of failure and leads to a lower threshold beyond which vulnerabilities offset synergies.

Collaborative transport creates synergies as it enables efficient use of decentrally operated transport resources. However, extensive collaboration between carriers also comes with vulnerabilities, which can make disruptions very impactful. While the synergies of collaborative transport are widely addressed in the literature, existing transportation research models are not able to capture the vulnerabilities to disruption induced by collaboration. We aim to fill this gap by establishing a novel multi-layer network model capturing the functional dependence between the physical transport network, of which performance is impacted by the disruptions, and the network of carrier collaborations, which is being disrupted. The model is applicable to particular instances of collaborative transport networks generated from data or from random graphs. Moreover, by integrating the constraints imposed by collaborations into existing analytical methods for the analysis of network properties, the model allows for an analytical derivation of the impact of structural properties of characteristic random network populations on vulnerability. Using the model on a mix of probabilistic network populations and populations of networks generated through simulation, we show that market structure, represented by disparity in carrier sizes, has a non-trivial impact on the vulnerability of a collaborative transport network to targeted disruption at the collaborative level, resulting from the interplay between a system's dependence on collaboration and its susceptibility to targeted disruption. Networks are most vulnerable if they have intermediate disparity in carrier sizes.

The competitive position of sea ports depends on their capability to forward incoming cargo from overseas to its final destination. Such capability is associated with the hinterland connectivity of the sea port. Hinterland connectivity of a port is mostly treated as a local indicator, describing the number of different hinterland locations served by a port via a direct service. However, with intermodality being on the rise, transfer connections and connections including multimodal transshipment become more feasible and common. This turns hinterland transport into a complex multimodal network, whose connectivity cannot be captured by the existing local notion. Using methods from the science of complex networks, this work extends the notion of hinterland connectivity by non-local (network) and multimodal aspects, and uses this notion to analyze hinterland connectivity for the European hinterland transport network of scheduled rail and barge services. The results show that overall structural capability to perform hinterland transport assignments increases tremendously as transfer connections and multimodal routes are established. Moreover, non-local measures show that ports with poor local connectivity can still be well positioned within a network, an insight that would be overlooked in a purely local analysis. Last but not least, all ports benefit individually from multimodal integration, but some do more than others. For instance, 'Multimodal hubs' are the most important contributors to multimodal integration, but their relative accessibility does not improve much.

Propagating phenomena in networks have received significant amount of attention within various domains, ranging from contagion in epidemiology, to diffusion of innovations and social influence on behavior and communication. Often these studies attempt to model propagation processes in networks to create interventions that steer propagation dynamics towards desired or away from undesired outcomes. Traditionally, studies have used relatively simple models of the propagation mechanism. In most propagation models this mechanism is described as a monolithic process and a single parameter for the infection rate. Such a description of the propagation mechanism is a severe simplification of mechanisms described in various theoretical exchange theories and phenomena found in real world settings, and largely fails to capture the nuances present in such descriptions. Recent work has suggested that such a simplification may not be sufficient to explain observed propagation dynamics, as nuances of the mechanism of propagation can have a severe impact on its dynamics. This suggests a better understanding of the role of the propagation mechanism is desired. In this paper we put forward a novel framework and model for propagation, the RTR framework. This framework, based on communication theory, decomposes the propagation mechanism into three sub-processes; Radiation, Transmission and Reception (RTR). We show that the RTR framework provides a more detailed way for specifying and conceptually thinking about the process of propagation, aligns better with existing real world interventions, and allows for gaining new insights into effective intervention strategies. By decomposing the propagation mechanism, we show that the specifications of this mechanism can have significant impact on the effectiveness of network interventions. We show that for the same composite single-parameter specification, different decompositions in Radiation, Transmission and Reception yield very different effectiveness estimates for the same network intervention, from 30% less effective to 70% more effective. We find that the appropriate choice for intervention depends strongly on the decomposition of the propagation mechanism. Our findings highlight that a correct decomposition of the mechanism is a prerequisite for developing effective network intervention strategies, and that the use of monolithic models, which oversimplify the mechanism, can be problematic of supporting decisions related to network interventions. In contrast, by allowing more detailed specification of the propagation mechanism and enabling this mechanism to be linked to existing interventions, the RTR framework provides a valuable tool for those designing interventions and implementing interventions strategies.

This study provides a complete framework to define and classify the value that firms can attain by their closed-loop supply chains (CLSC) and analyzes the role of information systems (IS) in this process. We present a novel typology to identify the four value types a firm can generate in CLSCs, namely sourcing, environmental, customer, and informational value. Particularly the last two types have not been recognized in literature. We adopt a case methodology and analyze 8 cases to illustrate the generation of each value type through CLSCs and highlight the role that IS play in this value generation. We ground our analysis on the integrative model of IT business value in [30] which considers the role of external business partners, which is essential to CLSCs. We find 3 key results: (1) IS is an essential enabler for all value types, (2) while sourcing value and to some extent environmental value, can be created with IS internal to the firm, the novel value types (customer and informational) can only be created with extraorganizational IS, (3) the value created by extraorganizational systems can only be created if the appropriate intraorganizational systems are in place. Our findings show that substantial value can be gained from implementing IS in CLSCs but stakeholder collaboration is necessary to reap the full value. This is an important managerial insight for the firms who still consider CLSC activities as costly and do not notice all the benefits that they can obtain from CLSCs .

The increased transparency of online markets enables highly detailed insights into competitive processes between firms that were previously hard to observe. One of the most fundamental issues in such competitive processes is the reaction of a firm to an action of its competitor(s). In this paper we develop a model outlining the conditions for competitive reactions, using detailed data from a price-comparison in the online market for laptops. The analysis reveals that the probability for a reaction rises by a firm’s distance to its preferred position (rank and placement), by a lower price rank as well as if the firm reacted to the same sellers in the past, although the precise timing of the reaction is much more difficult to determine. Furthermore, when companies react, they react stronger if they are further away from their preferred position (rank and placement) and if they face a disadvantage on non-price dimensions.

We show that contacts in formal, informal and especially multiplex networks explain transfer of innovative knowledge in an organization. The contribution of informal contacts has been much acknowledged, while that of formal contacts did not receive much attention in the literature in recent decades. No study thus far has included both these different kinds of contacts in a firm, let alone considered their combined effect. The exact overlap between formal as well as informal contacts between individuals, forming multiplex or what we call rich ties because of their contribution, especially drives the transfer of new, innovative knowledge in a firm. Studying two cases in very different settings suggests these rich ties have a particularly strong effect on knowledge transfer in an organization, even when controlling for the strength of ties. Some of the effects on knowledge transfer in an organization previously ascribed to either the formal network or the informal network may actually be due to their combined effect in a rich tie.

Explanations of knowledge sharing in organizations emphasize either personality variables such as motivation or network-related structural variables such as centrality. Little empirical research examines how these two types of variables are in fact related: how do extrinsic and intrinsic motivation explain the position that an employee entertains in a knowledge sharing network within an organization? Much is to be gained from a better understanding of how, empirically, psychological variables and an organization's network interrelate (Burt et al., 1998; Kalish and Robins, 2006; Moch, 1980; Teigland and Wasko, 2009). Still, this line of enquiry is not pursued much (Foss et al., 2009). This paper integrates the structural characteristics known to be implicated in knowledge transfer typically focused on in the social network literature on the one hand, with the motivational perspective commonly identified in the organization literature. This study examines how motivation - extrinsic (expected organizational rewards, reciprocal benefits) and intrinsic (knowledge self-efficacy, enjoyment in helping others) - might explain how employees may be better connected in the full knowledge transfer network or might be engaged more in inter-unit knowledge transfer. Connectedness (closeness centrality) and inter-unit ties are well-known to contribute to knowledge transfer. Analyzing data from a survey at two large European organizations, this study, counterintuitively, shows that neither intrinsic nor extrinsic motivation explain an individual's favorable position in a knowledge transfer network. (C) 2013 Published by Elsevier B.V.

Partial least squares (PLS) estimation of path models has become very popular in IS research, as an alternative to covariance-based methods. PLS path modeling is often referred to as being useful for “predictive” applications. In this work, we investigate the predictive aspects of PLS path modeling and its relation to predictive analytics and predictive assessment. In particular, we compare it to neural networks, which share several similarities. We conclude that PLS path modeling (the dominant form of usage in IS) is primarily causal-explanatory in nature. Introduction Partial Least Squares (PLS) estimation has become a popular method in information systems research, especially when it comes to analyzing adoption and usage of new electronic commerce applications (e.g., Pavlou and Fygenson (2006)). It is most commonly used to estimate structural equation models in place of covariance-based approaches (e.g., LISREL) in order to overcome the latter’s challenges (“technically demanding, often resulting in analytical errors, and, even if modeled correctly, is not a complete solution”, Chin et al. 2003, p.191). In structural equation models, one or more of the variables is assumed to be unobserved (latent). Theory is used to generate a causal diagram that relates the different variables (observable and latent) to each other. Then, data are used to quantify the hypothesized relationships, and the estimated relationships are used to test causal hypotheses. Hence, at their core, latent variable models are causal-explanatory. There have been claims that PLS path modeling is “causal-predictive” rather than causal-explanatory (Jöreskog and Wold, 1982; Anderson and Garbing, 1988). In this work we investigate the predictive aspects of PLS in light of the distinction between causal explanation and prediction by Shmueli & Koppius (2009). In particular, we look at the three steps in PLS path modeling: model estimation, validation, and use. PLS and Neural Networks To highlight the predictive nature of PLS path modeling, we examine it against a classic predictive model that appears somewhat similar: neural networks (NN) (Hackl and Westlund, 2000). NN are popular in data mining for predicting an outcome from a set of predictors. They are considered a “blackbox" in terms of interpretability and have been highly successful in terms of predictive accuracy. Similar to the use of latent variables (LVs) in PLS, in NN there is a diagram that connects the observable inputs and outputs via a network of unobservable (hidden) nodes and the NN algorithm then estimates the weights on the arrows connecting the different nodes. However, unlike path models, there is no underlying causal theoretical model, and the hidden nodes are conceptually unimportant. Hsu et al. (2006) compared PLS and NN numerically on simulated and real data. The simulations yielded similar results, whereas the real data yielded similar LV scores but different path coefficients. Model estimation: Two sets of models are estimated: outer models relate each latent variable to its observed variables (the measurement model), and inner models relate the latent variables to each other (the structural model). Each model is estimated using OLS regression (thereby assuming linear relationships). The path coefficients are then obtained via correlation or OLS regression between subsets of the estimated latent variables and/or observed variables. Estimation is done iteratively, until the various parameters are stable, by minimizing the various sums of squared errors. In other words, the final model best fits each of the regression models. Neural net estimation is similarly iterative. However, the algorithm tries to directly minimize the error of the observed output variables. NN also allow non-linear relationships between nodes. In light of these differences, PLS estimation is more similar to causal-explanatory modeling than to predictive modeling. Model validation: Validation is achieved by comparing the resulting estimates to the theoretical expectations. In particular, if coefficients do not have the right sign, then the corresponding observed variables are typically removed (Tenenhaus et al., 2005). Popular model validation measures include communality and redundancy. When computed by cross-validation (blindfolding), they measure how well the latent variables predict the observed variables (input or output). In contrast, in NN and in general predictive assessment, one examines how well the observed input variables predict the observed output variables. Hence, blindfolding serves more as a verification of the integrity of the estimated model in fitting the theoretical relationship between the observed and latent variables, rather than evaluates practical predictive power. Model use: The estimated PLS path model is used for testing causal hypotheses related to the structural diagram. Software output includes estimated weights, loadings, path coefficients, and correlations between latent variables. These are then examined carefully and used for inference. Due to the lack of an explicit overall statistical model and the nature of PLS estimation, statistical inference is achieved by using bootstrapping. Finally, path models are typically used only for causal testing and not for predicting new observations. In contrast, typical NN output does not include any inference-related information beyond the weights, and the focus is on its predictive accuracy on a holdout set (and comparing the training and holdout performance to detect over-fitting). The model is then used to predict new data. In summary, the final use of the estimated path model differs markedly from that of a predictive model. Conclusions In line with the differentiation between explanatory-causal statistical models and predictive analytics by Shmueli & Koppius (2009), our preliminary conclusion is that PLS path modeling is a causal-explanatory approach. Although PLS has some similarities with the predictive neural networks algorithm, its heavy reliance on a causal theoretical model, the conceptual importance of the latent variables, its method for model validation and its final use deem it very different from predictive models and from practical predictive use. If predictive power is important to PLS modelers then more emphasis should be given to the relationship between the observed input and output variables and the ability of the former to accurately predict the latter. Changes to the estimation procedure can also put more emphasis on prediction: Tenenhaus et al. (2005) suggested using PLS regression in place of OLS regression for outer model estimation. PLS regression is a predictive, shrinkage-based method which is indeed geared towards prediction. Finally, this is work in progress, and we plan to illustrate the various aspects mentioned above via simulation, the results of which will be presented at SCECR. References Anderson J C and Gerbing D W (1988), “Structural Equation Modeling in Practice: A Review and Recommended Two-Step Approach”, Psychological Bulletin, 103(3), pp. 411-423. Chin W W, Marcolin B L, and Newsted P R (2003), “A Partial Least Squares Latent Variable Modeling Approach for Measuring Interaction Effects: Results from a Monte Carlo Simulation Study and an Electronic-Mail Emotion/Adoption Study”, Information Systems Research, 14(2), pp. 189-217. Hackl P and Westlund A H (2000) “On Structural Equation Modelling for Customer Satisfaction Measurement”, Total Quality Management, 11, pp. S820-S825. Hsu S-H, Chen W-H, and Hsieh M-J (2006), “Robustness Testing of PLS, LISREL, EQS and ANN-based SEM for Measuring Customer Satisfaction”, Total Quality Management, 17(3), pp. 355–371. Jöreskog K G and Wold H (1982), “The ML and PLS Techniques For Modeling with Latent Variables: Historical and Comparative Aspects”, in Systems Under Indirect Observation: Causality, Structure, Prediction (Vol. I), Amsterdam: North-Holland, pp. 263-270. Pavlou P A and Fygenson M (2006), “Understanding and predicting electronic commerce adoption: An extension of the theory of planned behavior”, MIS Quarterly, 30(1), pp. 115-143. Shmueli G and Koppius O (2009), “The Challenge of Prediction in Information Systems Research”, Working Paper RHS 06-058, School of Business, UMD (http://ssrn.com/abstract=1112893) Tenenhaus M, Vinzi V E, Chatelin Y-M, and Lauro C (2005), “PLS Path Modeling”, Computational Statistics & Data Analysis, 48, pp. 159-205.

We study the role of information systems in enabling closed-loop supply chains. Past research in green IS and closed-loop supply chains has shown that it can result in substantial cost savings and waste reduction. We complement this research by showing that the effects are more than that: using information systems can also create business value for a firm in closed-loop supply chains. We make a novel distinction between four types of value: sourcing value, environmental value, customer value and informational value. Particularly the last two types have not been recognized in past research. We then analyze 8 cases (2 for each of the 4 value types) to highlight the role that information systems play in enabling this value creation and find three key results. First, we find that IS is an essential enabler for all four value types. Second, while sourcing value and to some extent environmental value, can be created with IS that are internal to the firm, the novel types of value (customer value and informational value) can only be created with information systems that are extraorganizational, i.e. aimed at customers and supply chain partners. Third, the value created by extraorganizational systems can only be created if the appropriate intraorganizational systems are in place. Overall, our results show that substantial value can be gained from implementing green IS in closed-loop supply chains, but that collaboration between all stakeholders in the supply chain is necessary in order to reap the full value.

Considerable attention has been paid to the network determinants of knowledge sharing. However, most, if not all, of the studies investigating the determinants of knowledge sharing are either focused on knowledge-intensive organizations such as consultancy firms or R&D organizations, or knowledge workers in regular organizations, while lesser knowledge intensive organizations or non-knowledge workers are rarely explored. This is a gap in the literature on social networks and knowledge sharing. In this paper, the relations between network determinants and actor determinants of knowledge sharing are empirically tested by means of a network survey in a less knowledge intensive organization, specifically employees of a Dutch department store chain. The results show that individual-level variables such as departmental commitment and enjoyment in helping others are the major determinants of individuals’ knowledge sharing behavior, but none of the social network variables play a role. The results thus present an important boundary condition to social networks effects on knowledge sharing: social networks only seem to play a role in knowledge sharing for knowledge workers, not for blue-collar workers.

The ongoing evolution of interorganizational information systems (IOS) into industry-wide systems (IIOS) or digital ecosystem platforms raises new questions for researchers and practitioners alike, in particular the question of how organizations can gain advantage of these new ecosystems by orchestrating their business network. In an empirical study of the Dutch insurance industry, we use network theory and interorganizational systems theory to show that participation in a digital ecosystem leads to organizations orchestrating their network position into a larger portfolio of suppliers as well as a more differentiated portfolio of suppliers. A more differentiated supplier portfolio in turn leads to superior organizational performance.

Business networks have proliferated over the past three decades as many firms enabled by ICT have embraced the digital ecosystem form of organization. This implies that in addition to thinking of performance at the level of the individual firm, it becomes necessary to think in terms of performance at the level of the ecosystem, i.e. the business network as a whole, yet is not clear how to best conceptualize and measure business network performance. In this article, we show that business network performance can be conceptualized in three complementary perspectives: the firm perspective, the complex systems perspective and the customer perspective. We employ the customer perspective and develop two specific business network performance metrics, namely network effectiveness and network efficiency. Based on the information integration and visibility benefits that interorganizational systems (IOS) provide, we then develop a construct called network transparency. Using a series of network experiments with students, we find that an increase in network transparency increases business network performance in terms of network effectiveness and network efficiency. We thus show that the value of network transparency is measurable through clear metrics at the level of the business network as a whole.

Online product buzz refers to an online expression of interest in a product, such as online product reviews, blog posts and search trends. We study how well online buzz predicts actual sales across different phases in the product lifecycle. Using data from smartphone sales of a Dutch online retailer, we demonstrate a 28% overall increase in forecasting accuracy when measures of online product buzz variables are take into account. The value of online product buzz shows especially in early sales forecasting, an area in which traditional forecasting models have substantial difficulties. In early sales, local search trends, subscriptions for stock notifications and pageviews were important predictors and forecasts were on average improved by 44%. For mature sales, accuracy improved by 10%, with the most important predictors being on- and offsite reviews and, again, pageviews. These results also suggest different drivers of sales across phases in the product lifecycle.

The last few years have seen the rise of a new breed of interorganizational systems, built around web services and business process standards that allows for new and efficient ways of cooperation with new business partners in the network. This quick-connect capability in turn may affect how organizations in a business network structure their own network of relationships. In this paper, using a survey among organizations in the Dutch Graphimedia industry, we develop and validate a measure of the quick connect capability as consisting of four subcomponents: quick connect, quick complexity, quick disconnect and low switching costs, and we show that this quick connect capability can exist on both the supplier side as well as the customer side. We furthermore show that investing in interorganiza-tional systems that enable the usage of communication standards and business process standards leads organizations to develop such a quick connect capability, particularly on the supplier side. These results suggest that it is valuable for organizations to invest in the quick connect capability in order to achieve the flexibility that is necessary to compete in a dynamic business network.

Empirical research in Information Systems (IS) is dominated by the use of explanatory statistical models for testing causal hypotheses, and by a focus on explanatory power. Predictive statistical models, which are aimed at predicting out-of-sample observations with high accuracy, are rare, and so is attention to predictive power. The distinction between explanatory and predictive statistical models is key, as both types of models play a different, yet essential, role in advancing scientific research. Similarly, explanatory power and predictive accuracy are two distinct qualities of a statistical model, and are measured in different ways. A literature review of MISQ and ISR shows that predictive goals, predictive claims, and predictive statistical models are scarce in mainstream empirical IS research. In addition, we find three questionable common practices: First, even when the stated goal of modeling is predictive, explanatory statistical modeling is often employed. Second, the predictive power of a model is often inferred from its explanatory power. And third, the vast majority of explanatory statistical models lack proper predictive assessment, which is a key scientific requirement. In light of the distinction between explanatory and predictive statistical modeling and power, and current practice in IS, we highlight the main differences between them, focusing on practical issues that confront an empirical researcher in the data analysis process.

Most studies on the role of IT for economic exchange predicted that under a given set of exchange attributes buyers would choose a certain mode of relationship with suppliers. Our study of an online IT services marketplace revealed that buyers do not have a single, uniformly preferred type of relationship, but rather maintain a portfolio of relationships. Furthermore, different buyers arrange their portfolios of exchange relationships in different ways. We found four clusters of buyers' portfolios of relationships labeled Transactional buyers, Recurrent buyers, Small diversifiers and Large diversifiers, that differ in their usage of auction or negotiation mechanism, their supplier relations as well as their usage of preferred suppliers. Our results thus paint a richer picture of how buyers organize their supplier networks online.