MSC. THESIS DEFENSE
LUUK APPELS
Exploring the Use of a Dynamic Hierarchical Factor Model in a Marketing Context: Toward a New Way of Modeling in Marketing Science
THIS PRESENTATION
• Findings
• Contributions
• Limitations
FINDINGS
• Results of the Impulse Response and the price forecasts were not spectacular
• Variance decomposition revealed that
• There indeed exists some co-movement between prices/promo of different brands in a store.
• Some co-movement exists between prices/promo of different chains
• Very little co-movement exists between price and promo at the market level, except for maybe Albert Heijn.
• Concerning the FAVAR, we established that
• Adding factors = better results
CONTRIBUTIONS
• Pioneer the DHFM in a marketing context and prove the use of its tools. In this case, especially the variance decomposition.
LIMITATIONS & FUTURE RESEARCH
• Simple version of DHFM used for this research
• à select optimal number of lags and extract more factors.
• Generalizability is questionable
• à More similar studies should be done, covering more data.
• Too little regressions run for the FAVAR
• à run the full 1 million regressions
• Treating W endogenously in the FAVAR is not statistically correct
WHAT DID I LEARN?
• Independently work on a project of serious scale and time investment
• MATLAB, R + some additional statistics