Contributions published at Marketing and Market Research (René Algesheimer)

Valuing customers is essential to any firm and enables marketers to identify key customers. Customer lifetime value (CLV) is the central metric for valuing customers. It describes the long-term economic value of customers and gives managers an idea of how customers will evolve over time. With the Pareto/NBD model, modeling customer lifetime value for non-contractual businesses has become a straight-forward task, however this simplicity comes at a price. Individual-level predictions of customer lifetime value often lack precision. A possible explanation is that standard probabilistic customer attrition models do not consider important contextual factors, such as direct marketing or regularity purchase patterns. However, there is no generalization of the Pareto/NBD model that allows time-varying contextual factors to be considered. This study proposes a closed-form maximum likelihood extension to the Pareto/NBD model that allows both time-invariant and time-varying contextual factors to be modelled in continuous non-contractual settings. These contextual factors can influence either the purchase or the attrition process, or both. A benchmark using multiple retailing datasets shows a significant improvement in forecast accuracy for future customer activity when explicitly modeling time-varying contextual factors. Our findings have strong implications for both, marketing practice and research. Besides giving detailed recommendations on when to use which modeling approach, we also provide practical advices for applying probabilistic customer attrition models.

null

null

As new instances of nested organization—beyond ecological networks—are discovered, scholars are debating the coexistence of two apparently incompatible macroscale architectures: nestedness and modularity. The discussion is far from being solved, mainly for two reasons. First, nestedness and modularity appear to emerge from two contradictory dynamics, cooperation and competition. Second, existing methods to assess the presence of nestedness and modularity are flawed when it comes to the evaluation of concurrently nested and modular structures. In this work, we tackle the latter problem, presenting the concept of in-block nestedness, a structural property determining to what extent a network is composed of blocks whose internal connectivity exhibits nestedness. We then put forward a set of optimization methods that allow us to identify such organization successfully, in synthetic and in a large number of real networks. These findings challenge our understanding of the topology of ecological and social systems, calling for new models to explain how such patterns emerge.

Complex networks are often used to represent systems that are not static but grow with time: People make new friendships, new papers are published and refer to the existing ones, and so forth. To assess the statistical significance of measurements made on such networks, we propose a randomization methodology—a time-respecting null model—that preserves both the network's degree sequence and the time evolution of individual nodes' degree values. By preserving the temporal linking patterns of the analyzed system, the proposed model is able to factor out the effect of the system's temporal patterns on its structure. We apply the model to the citation network of Physical Review scholarly papers and the citation network of US movies. The model reveals that the two data sets are strikingly different with respect to their degree-degree correlations, and we discuss the important implications of this finding on the information provided by paradigmatic node centrality metrics such as indegree and Google's PageRank. The randomization methodology proposed here can be used to assess the significance of any structural property in growing networks, which could bring new insights into the problems where null models play a critical role, such as the detection of communities and network motifs.

null

The relevance of human values for the study of the motivational sources of interpersonal violent behavior was investigated in various fields of the social sciences. However, several past studies mixed up values with other dimensions like attitudes, norms, or beliefs, and only a few systematically assessed the effect of values on violent behavior relying on a value theory. Furthermore, in other studies, violence was often analyzed as a composite index of different forms of delinquent behavior rather than as violence per se. In the current study we address these gaps in the literature by building upon Schwartz’ theory of basic human values. We use it to explain attitudes toward interpersonal violence and interpersonal violent behavior. We analyze data of young people (n = 1,810) drawn from a German study in Duisburg, Germany, which assessed various types of self-reported violent behavior as well as values and attitudes toward violence. We test structural equation models in which we explain interpersonal violent behavior with basic human values, and where attitudes toward interpersonal violent behavior mediate this relation. Results show that self-transcendence and conservation values are associated negatively and power and stimulation values positively with interpersonal violent behavior. In addition, attitudes operate as a partial mediator for the former and as a full mediator for the latter in the relation between values and violent behavior. Despite a dominant association between attitudes and behavior, values themselves can significantly contribute to the explanation of violent behavior.

When estimating causal effects using observational data, researchers often apply matching methods to eliminate systematic differences between treatment and control groups. These methods only account for observed differences between groups, but observational data is likely to lack some important variables that influence treatment assignment and the out- come. This can lead to bias in the estimated treatment effect due to endogenous treatment assignment. This thesis compares the performance of different matching methods when only a subset of the confounding variables is observed. I find that (1) restrictive methods, such as exact and coarsened exact matching, achieve the best balance, but yield the most biased treatment effect estimate; and (2) genetic matching retrieves the least biased estimate, but fails at sufficiently balancing the treatment and control groups when the correlation between the observed and unobserved confounders is high. Overall, the higher the level of correlation between the omitted variable and the observed confounders the less biased is the estimate of the treatment effect. At the same time, it gets more difficult to sufficiently balance the observed variables for the treatment and control group.

null

Since the first round of the ESS (European Social Survey) in 2002 numerous comparative studies on changing attitudes towards immigrants have been published based on ESS data. The availability of two batteries measuring attitudes towards immigrants in the core module of each biennial survey enabled analysis of stability and change of these attitudes in Europe over the whole period 2002–2014. In these studies, variation can be studied from at least two perspectives, namely as differences between countries and/or as change over time. An exploration of published studies that used ESS data to study attitudes towards immigrants in Europe demonstrates that one can distinguish between five different research designs using the following criteria: the number of countries involved; inclusion of a time factor; treatment of context variables in the data analysed; inclusion of cross-level interactions; combination of cross-national and cross-temporal perspectives in the study of change. These designs are explained and illustrated with typical examples of analyses of ESS data. A dynamic version of the Group Conflict Theory (GCT) is the main theoretical background of these studies. Negative outgroup sentiments are seen as defensive reactions to perceived intergroup competition for scarce goods. After some theoretical reflections on this perspective, the main findings of our own studies are briefly presented and discussed.

Intended to bridge the gap between the latest methodological developments and cross-cultural research, this interdisciplinary resource presents the latest strategies for analyzing cross-cultural data. Techniques are demonstrated through the use of applications that employ cross-national data sets such as the latest European Social Survey. With an emphasis on the generalized latent variable approach, internationally prominent researchers from a variety of fields explain how the methods work, how to apply them, and how they relate to other methods presented in the book. Syntax and graphical and verbal explanations of the techniques are included. Online resources, available at www.routledge.com/9781138690271, include some of the data sets and syntax commands used in the book. Applications from the behavioral and social sciences that use real data-sets demonstrate: - The use of samples from 17 countries to validate the resistance to change scale across these nations - How to test the cross-national invariance properties of social trust - The interplay between social structure, religiosity, values, and social attitudes - A comparison of anti-immigrant attitudes and patterns of religious orientations across European countries. The second edition includes six new chapters and two revised ones presenting exciting developments in the literature of cross-cultural analysis including topics such as approximate measurement invariance, alignment optimization, sensitivity analyses, a mixed-methods approach to test for measurement invariance, and a multilevel structural equation modeling approach to explain noninvariance.　 This book is intended for researchers, practitioners, and advanced students interested in cross-cultural research. Because the applications span a variety of disciplines, the book will appeal to researchers and students in: psychology, political science, sociology, education, marketing and economics, geography, criminology, psychometrics, epidemiology, and public health, as well as those interested in methodology. It is also appropriate for an advanced methods course in cross-cultural analysis.

This article examines the structure of character strengths (Peterson and Seligman, 2004) following both variable-centered and person-centered approaches. We used the International Personality Item Pool-Values in Action (IPIP-VIA) questionnaire. The IPIP-VIA measures 24 character strengths and consists of 213 direct and reversed items. The present study was conducted in a heterogeneous group of N = 908 Poles (aged 18–78, M = 28.58). It was part of a validation project of a Polish version of the IPIP-VIA questionnaire. The variable-centered approach was used to examine the structure of character strengths on both the scale and item levels. The scale-level results indicated a four-factor structure that can be interpreted based on four of the five personality traits from the Big Five theory (excluding neuroticism). The item-level analysis suggested a slightly different and limited set of character strengths (17 not 24). After conducting a second-order analysis, a four-factor structure emerged, and three of the factors could be interpreted as being consistent with the scale-level factors. Three character strength profiles were found using the person-centered approach. Two of them were consistent with alpha and beta personality metatraits. The structure of character strengths can be described by using categories from the Five Factor Model of personality and metatraits. They form factors similar to some personality traits and occur in similar constellations as metatraits. The main contributions of this paper are: (1) the validation of IPIP-VIA conducted in variable-centered approach in a new research group (Poles) using a different measurement instrument; (2) introducing the person-centered approach to the study of the structure of character strengths.

The current study examines the extent to which religiosity account for ideological orientations in 16 countries from five continents (Australia, Brazil, Chile, Germany, Greece, Finland, Israel, Italy, Japan, Poland, Slovakia, Spain, Turkey, Ukraine, the United Kingdom, and the United States). Results showed that religiosity was consistently related to right and conservative ideologies in all countries, except Australia. This relation held across different religions, and did not vary across participant’s demographic conditions (i.e., gender, age, income, and education). After controlling for basic personal values, the contribution of religiosity on ideology was still significant. However, the effect was substantial only in countries where religion has played a prominent role in the public sphere, such as Spain, Poland, Greece, Italy, Slovakia, and Turkey. In the other countries, the unique contribution of religiosity was marginal or small.

Identifying individuals who are influential in diffusing information, ideas or products in a population remains a challenging problem. Most extant work can be abstracted by a process in which researchers first decide which features describe an influencer and then identify them as the individuals with the highest values of these features. This makes the identification dependent on the relevance of the selected features and it still remains uncertain if triggering the identified influencers leads to a behavioral change in others. Furthermore, most work was developed for cross-sectional or time-aggregated datasets, where the time-evolution of influence processes cannot be observed. We show that mapping the influencer identification to a wisdom of crowds problem overcomes these limitations. We present a framework in which the individuals in a social group repeatedly evaluate the contribution of other members according to what they perceive as valuable and not according to predefined features. We propose a method to aggregate the behavioral reactions of the members of the social group into a collective judgment that considers the temporal variation of influence processes. Using data from three large news providers, we show that the members of the group surprisingly agree on who are the influential individuals. The aggregation method addresses different sources of heterogeneity encountered in social systems and leads to results that are easily interpretable and comparable within and across systems. The approach we propose is computationally scalable and can be applied to any social systems where behavioral reactions are observable.