The last RSS West Midlands local group meeting was hosted by the department of statistics, University of Warwick on 8 May 2014. Professor Giuliana Battisti from Warwick Business School gave a talk entitled 'Adoption-timing and causal directions in the diffusion of innovations'.
Giuliana began the talk by demonstrating that even when a cost-reducing innovation is readily available in the market, its diffusion within or across users can take several years and in some cases, decades. Examples include the case of Microsoft office and PCs which was adopted after about 35 years and the slow uptake of unleaded petrol in European countries. Diffusion is usually S-shaped, similar to epidemic models in epidemiology. Therefore, models for analysing epidemics have been used to model diffusions. These models involve defining link functions such as logistic and log-logistic. Economic modellers have extended these models to mixtures so that the models reflect uncertainty in the link function and also to let the diffusion rate parameter to be a function so that it is possible to include, for example, environmental factors and individual characteristics.
Another methodology issue that needs to be considered while fitting diffusion models is: why does it take long to adopt innovations? Is this rational? Giuliana argued uptake of innovations is not influenced by marketing (or advertising) strategies only. She demonstrated that diffusions of innovations happen in clusters, that is, there are synergies or complementarities (sequential or simultaneous) in diffusions of different innovations. For example, uptake of PCs coincided with uptake of certain softwares. Another example is the diffusion of technologies such as computer aided design (CAD) and organisational practices such as joint design (JOD) teams involving customers and suppliers.
Presence of complementarity means joint models are required to model diffusions processes. Before using joint models, one needs to test for the assumption of presence of complementarity. Giuliana presented a graph whose edges represented relationships and indicated directions of relationships for the different diffusions in innovations. Giuliana then demonstrated how the edges can be used to test for simultaneous and sequential complementarities within a discrete multistage Markov framework. For the CAD and JOD example, the model not only provided evidence of complementarities but also of technology-led organisational change in that JOD uptake was better if CAD was already in place (sequential complementarity) but CAD uptake did not improve if JOD was present (no sequential complementarity).