Marketing Spending Models - Innovation and Marketing in the Pharmaceutical Industry - page 556

such as shareholders. At the individual level (physician, patient, or pharmacist), we
observe demand for ethical drugs in form of prescriptions. Aggregation across deci-
sion makers produces brand sales that may be further aggregated to reflect category
sales. In the following, I shall focus on brand demand models because they are
directly incorporated into the budgeting problem. For category demand effects of
pharmaceutical spend categories, I refer the interested reader to a recent in-depth
study of the US market by Fischer and Albers ( 2010 ).
Depending on the level of data analysis, different types of response models have
been suggested (for a general overview, see Hanssens et al. 2001 ; Leeflang et al.
2000 ). At the individual level, conditional logit models are available to model drug
choice. Double-log response models have been used to specify the impact of phar-
maceutical spending on aggregate brand sales. It is not only the type of data that
drives the choice of the model form, but also the marketing phenomenon the
researcher wants to model. For example, Bass-type diffusion models describe the
penetration of a new drug among a potential adopter population. All these model
types have been used in research on pharmaceutical brand demand.
Table 19.2 provides an overview of the rich literature on marketing spending
models for pharmaceuticals. While I do not claim this list to be complete, it should
represent the diversity and significance of contributions. In addition to the type of
demand model, the table classifies spending models according to their main focus
of application, the included pharmaceutical spending categories, and the consider-
ation of interactions and marketing dynamics. Spending models may be used for
description, prediction, or normative application such as optimizing the marketing
budget. They may include as many spending categories as are effectively used. It is
also important to understand to what extent the spending models do capture interac-
tions across decision variables, competitors, and countries. Because of the variety of
communication channels, the integration of international markets and the intense
competition, I suspect that such interactions do play a role.
The table is ordered chronologically. One of the first applications of a response
model to pharmaceutical products is by Montgomery and Silk ( 1972 ) who specify a
distributed lag model to capture the multiple dynamic effects of pharmaceutical
marketing efforts. These dynamics and the prediction of new product sales are also
the focus of diffusion models that have been suggested (e.g., Hahn et al. 1994 ;
Lilien et al. 1981 ). A distinctive feature of all pharmaceutical diffusion models is
that they model a complex diffusion process involving different segments of cus-
tomers including repeat buyers. Note that repeat buying (i.e., refills) are often a
multiple of first-time prescriptions. If marketing efforts successfully initiate trial
prescriptions, these efforts also impact future sales via repeat prescriptions.
Spending models thus need to account for these strong dynamic marketing effects.
In fact, the vast majority of models in Table 19.2 (87 %) do include marketing
dynamics, most often by specifying a marketing stock variable.
From Table 19.2 , it is obvious that detailing is part of virtually every demand
model reflecting the importance of this instrument. Since the abolishment of the ban
on patient-oriented marketing in the USA in 1997, studies with US data have
increasingly focused on this spending category. Models including expenditures on