Modeling the Effects of Promotional Efforts on Aggregate Pharmaceutical Demand: What We Know and Challenges for the Future - Innovation and Marketing in the Pharmaceutical Industry - page 602

accumulated prelaunch demand that is not infl uenced by marketing efforts, with the
diffusion process of new drugs in the late market infl uenced by marketing efforts.
They fi nd that medical journal advertising and DTCA are statistically signifi cant but
that detailing is not. They point out several possible explanations for the relative
ineffectiveness of detailing, such as higher saturation in detailing than medical jour-
nal advertising or the absence of serious side effects for the drugs in the analyzed
category.
Ruiz-Conde et al. ( 2011 ) extend the model of Hahn et al. ( 1994 ) by incorporating
the effect of company and competitor promotional efforts separately. In contrast to
existing studies on pharmaceuticals, their model accommodates heterogeneity in
the effects of different marketing instruments (detailing, medical journal advertis-
ing, physician meetings, and DTCA). The authors test several versions of their
model on 34 prescription drugs from three different categories. They fi nd that
longitudinal relationships exist both for “own” and competitor marketing efforts
with “own” (competing) marketing expenditures increasing (decreasing) the trial
rate. However, the most disaggregated version of their model (which accounts for
the separate effects of the four marketing instruments) suffers from severe
multicollinearity.
There are three important papers on diffusion of new drugs that are not included
in Table 20.2 because they incorporate neither marketing variables nor repeat sales
within the complete Bass theoretical framework (i.e., innovation/external infl uence
and imitation/internal infl uence).
Berndt et al. ( 2002 ) consider physician detailing and medical journal advertising
(similar to the variables studied by Hahn et al. ( 1994 )), but in addition they include
price (the industry average price per patient day of therapy). They propose a share
model as a semi-log specifi cation (by using the logistic diffusion expression) to
examine the impact of price, marketing efforts, and other variables (e.g., product
quality) on the demand of new antidepressant drugs in the United States. They fi nd
that marketing efforts have a positive and highly signifi cant impact on the rate of
diffusion. Regarding the role of price, they fi nd that the long-run industry demand
price elasticity is negative and signifi cant, but that within the therapeutic class,
market shares of individual products are not price sensitive.
Van den Bulte and Joshi ( 2007 ) study 33 data series with one of them consisting
of monthly sales of a prescription drug. They consider a two-segment structure with
asymmetric infl uence: the infl uential segment, consisting of physicians who are
more in touch with new developments, and the segment of imitators whose own
adoptions do not affect the members of the infl uential segment. Their model allows
diffusion researchers to operationalize the theories of asymmetric infl uence in the
absence of micro-level diffusion data and to estimate parameters from real data.
Guseo and Guidolin ( 2009 ) propose a model with a dynamic potential market
size that considers the Bass model as a nested special case. They assume that
the communication network is not observable. They test the performance of their
model with weekly data on the diffusion of a new drug in two geographical areas of
Italy. Their model yields higher values of R 2 and the F -test than the Bass model.
Their results show that the imitative adoption pattern is signifi cantly different