2006 Fiscal Year Final Research Report Summary
Making Model Building and Its Application with Integrating Heterogeneous Decision Makers
Project/Area Number |
16330083
|
Research Category |
Grant-in-Aid for Scientific Research (B)
|
Allocation Type | Single-year Grants |
Section | 一般 |
Research Field |
Commerce
|
Research Institution | Osaka University |
Principal Investigator |
NAKAJIMA Nozomi Osaka University, Graduate School of Economics, Professor, 大学院経済学研究科, 教授 (00095936)
|
Co-Investigator(Kenkyū-buntansha) |
TERUI Nobuhiko Tohoku University, Graduate School of Economics, Professor, 大学院経済学研究科, 教授 (50207495)
ABE Makoto University of Tokyo, Graduate School of Economics, Professor, 大学院経済学研究科, 教授 (70302677)
SATOMURA Takuya Keio University, Faculty of Business & Commerce, Associate Professor, 商学部, 助教授 (40324743)
KONDO Fumiyo Osaka University, Grad. School of Systems & Information Engineering, Assistant Professor, システム情報工学研究科, 講師 (40322010)
WIRAWAN Dony dahana Osaka University, Graduate School of Economics, Assistant Professor, 大学院経済学研究科, 講師 (90432426)
|
Project Period (FY) |
2004 – 2006
|
Keywords | customer heterogeneity / stochastic model / consumer behavior / market segmentation / nonlinear response / hierarchical Bayes |
Research Abstract |
We have developed several models in the Bayesian framework to cope with the various types of customer heterogeneity from a new perspective. The recent progress in the computational methods of Bayesian statistics has provided powerful model building approaches, together with the rapid improvement of technological infrastructures, such as the Internet and mobile phones, scanner data and home delivery systems as well as computerized customer management systems. A brand choice model with heterogeneous price-threshold parameters was proposed to investigate the relationships between the aspects of consumer price sensitivity and price thresholds using hierarchical Bayes modeling with the MCMC method. The empirical application revealed that the reference effect and loss aversion were more marked with the introduction of price thresholds into the response models. The similar approach was taken in constructing the advertising exposure model to investigate the effects of advertising stocks on cons
… More
umers' buying behavior processes. Another consumer behavior model that permits the estimation of customer lifetime value (CLV) from standard RFM data in "non-contractual" setting was proposed in the hierarchical Bayes framework. The model also relates customer characteristics to dropout, frequency, and spending behavior, which, in turn, is linked to CLV to provide useful insight into acquisition. Using frequent shopper program data, the model was shown to perform better than the benchmark Pareto/NBD-based model at both aggregate and disaggregate levels on calibration and holdout samples. The model enables to describe the interdependence of these three behavioral processes, and our data exhibited a highly negative correlation between purchase frequency and the dropout process, violating the assumption of the Pareto/NBD model. We also contributed to popularization of the Bayesian approach to marketing problems by providing monographs and organizing two international conferences held in Japan as well as presenting several papers there. Less
|