| This thesis contains three essays dealing with the modeling of complex decision processes in marketing. Each of these deals with a different aspect of complex decision making, either at the individual or at the network level.;Essays 1 and 2 in this dissertation are studies using agent-based models. Essay 1 is an extension of Goldenberg, Libai, and Muller (2010), who use an agent-based model to demonstrate that contrary to intuition, products with network externalities tend to diffuse slower than those without (the "chilling" effect). In their study, they use a simple 2-dimensional Moore neighborhood as the underlying network substrate depicting the market for new product adoption. In keeping with other studies demonstrating that network structure affects diffusion dynamics, I adapt their simulations for real-world network data and find that while larger networks and networks with higher average degree tend to offset this chilling effect, clustering could enhance it. I also demonstrate that for the same high-level parameters, a cumulation of many local micro-level conditions could end up speeding diffusion with network externalities, actually making it faster than without network externalities.;Essay 2 deals with the controversy surrounding multilevel marketing (MLM) schemes and questions of their profitability to their freelance sales force. Building on the sparse literature in this field, I build an agent-based model of the growth of an MLM scheme on a social network.;Unlike extant work which neglects the role of recruits' business expenses on the decision to join, I include the same, and show that it has non-trivial effects on the proliferation of MLM schemes.;In essay 3, I build a new model of preferences based on the notion of anchoring. This vectorbased model is based on Lancaster's (1966) multiattribute utility model, but allows the weights to be shaped by context.;Context-dependent models are important in studying consumer choices, as for example, in explaining new product adoptions, new product takeoff, and market dynamics. Context dependent choice models can be used in conjoint analyses to provide calibrated input data to instantiate agent-based models that simulate new product growth. Thus, Essay 3 is a small but important piece in the overall jigsaw puzzle of complex decision processes. The proposed modeling approach can be used to simulate individual decision processes with what-if scenarios regarding options available to a single consumer, and thus be used to build an agent-based simulation of an entire market. |
