Operations Research 2025
Abstract Submission

114. The value of consumer heterogeneity in simulating innovation diffusion: A computational experiment based on the Bass model

Invited abstract in session TC-7: Simulation & Analysis, stream Simulation and Quantum Computing.

Thursday, 11:45-13:15
Room: U2-205

Authors (first author is the speaker)

1. Theresa Elbracht
Department of Business Administration and Economics, Bielefeld University
2. Christian Stummer
Department of Business Administration and Economics, Bielefeld University

Abstract

Modeling the diffusion of innovations constitutes an established field of research. Its most prominent cornerstone, the Bass model, is based on strong assumptions, namely that (i) consumers are identical regarding their innovativeness and their propensity to conform with others and that (ii) they are embedded in a fully connected social network. Still, the model offers a reasonable estimation for the diffusion curve of many consumer products measured in terms of the R^2 value. Our research addresses these two limitations and investigates the effect of allowing for individual consumer traits or embedding consumers in a more realistic social network in which some are more strongly connected with their peers than others. To this end, we ran simulation experiments for four variants of an agent-based model. In the first, we use the same parameters as Bass and replicate the diffusion curves from his analytic model. In the second model, we stick with the fully connected network but assign individual consumer traits by drawing individual parameters for each consumer agent from a normal distribution with the original parameter as the mean. In the third model, in contrast, we stick with homogeneous consumer traits but employ a scale-free social network following the Barabasi-Albert approach. In the fourth model, we combine heterogeneity regarding consumer traits and heterogeneity resulting from the scale-free social network. The best results from the four models are then compared with respect to their fitting with sales data from the real-world application cases that Bass used. We find that heterogeneity makes a significant difference in most cases, which especially holds true for the network effect but also, although to a smaller extent, for heterogeneity in consumer traits.

Keywords

Status: accepted

Back to the list of papers