Pharmaceutical Detailing Elasticities: A Meta-Analysis - Innovation and Marketing in the Pharmaceutical Industry - page 537

elasticities between models which account for promotions vs. those that do not. We
fi nd that, holding all moderators and interaction effects at their average level, model
that include promotions report detailing elasticities to be lower by 0 . 16 . This could
be because the most common form of promotion, namely, free drug samples, are
heavily used in the introductory and early stages of prescription drug life cycles
(e.g., Joseph and Mantrala 2009 ), and once the drug has been launched, reps' detail-
ing may be adding little value beyond (increasing) sample drop-offs in the eyes of
physicians especially when they are in the “trying out” mode (see also Manchanda
and Chintagunta 2004 , p. 139).
As regards detailing elasticity estimates after accounting for advertising, the
result shown in Table 18.2 indicates pharma sales response models that include
advertising result in detailing elasticities that are on average lower by 0.106 than
those from models that exclude advertising ( b = −0.106, p < 0.05).

Inclusion of interactions of detailing with other marketing communication vari-
ables : Physician prescriptions can be affected by interactions of detailing and other
marketing communication variables such as journal advertising (e.g., Narayanan
et al. 2004 ). Since such interactions usually have positive effects, excluding them
from a model could result in a defl ated estimate of detailing elasticity. This is
because the marginal effect of detailing on sales will be underestimated as it will not
include the coeffi cient pertaining to the interaction term. Thus we expect that detail-
ing elasticities from response models that include interactions between detailing
and other marketing communication variables will be higher than those from
response models that exclude such interaction effects.
We dummy coded a variable for the interactions of detailing with other market-
ing communication variables (1 if included, 0 if not). The result shown in Table 18.3
indicates models that include such interactions result in detailing elasticities that are
higher than models that do not ( b = −0.196, p < 0.01).

18.4.3 Research Methodology Bias-Corrected Benchmark

Elasticity

Following the lead of AMS ( 2010 ) and Tellis ( 1988 , p. 337), a practically useful
generalized estimate to draw from this meta-analysis is the mean of detailing elas-
ticities after correcting each for the statistically signifi cant biases introduced by
researchers ' methodology choices . Specifi cally, we consider that the “unbiased”
reference response model is one that includes promotions, involves a product in the
early stage of the PLC, includes advertising, corrects for endogeneity especially if
the geographic setting is USA, and includes lagged effects. Correcting each of the
373 observed elasticities for their deviations from the above reference values of the
signifi cant independent variables, and taking the average of the corrected estimates,
we obtain the mean corrected elasticity of 0 . 178 , compared to 0 . 227 in AMS ( 2010 ).
As we discuss in the next section, this market-based benchmark value of detailing
elasticity can serve as a useful input for managers, and sales researchers.