Pharmaceutical Detailing Elasticities: A Meta-Analysis - Innovation and Marketing in the Pharmaceutical Industry - page 530

Table 18.2 List of results

Geographic setting
Europe USA

Panel A: Sample size by cell
Stage in product life cycle Late 32 206

Early 74 61

Panel B: Detailing elasticity means unadjusted for biases (grand mean = 0.210)
Stage in product life cycle Late 0.169 0.140

Early 0.410 0.230

of these estimates are positive. We fi nd the “raw” mean in our database (unadjusted
for any methodology-induced biases) to be 0.21, compared to 0.245 in AMS ( 2010 ).
In keeping with our research questions, we also present the means of the elasticities
by PLC stage and geographic setting. Panels A and B of Table 18.2 list some key
descriptive data of the subsamples. We have 32 elasticities from the late stages of
the PLC from Europe, and 206 from the late stage of the PLC from USA. 3 These
elasticities have a mean of 0.17 and 0.14 respectively. Also, we have 74 elasticities
from the early stages of the PLC from Europe, and 61 from the early stage of the
PLC from USA. These elasticities have means of 0.41 and 0.23 respectively.
Note that the means presented in Panel B of Table 18.2 are unadjusted for any
methodology-induced biases (e.g., the elasticities could have stemmed from a
model that omitted other marketing-mix variables, thereby biasing the estimate).
Therefore, to obtain a better understanding of the means by each category, we fi rst
need to identify any systematic biases that might be driving these means. For this
purpose, we build a meta-analytic model that can help identify the factors that
systematically explain upward or downward movement in the elasticities. Next, we
present the independent variables considered in our analysis.

18.3.4 Independent Variables (Determinants) and Coding

We group our independent variables into three categories: (1) variables capturing
market characteristics (stage in the product life cycle, geographic setting), (2) vari-
ables capturing dataset characteristics (years spanned by a dataset, absolute or rela-
tive form of sales output measure, and temporal aggregation of the data) and (3)
variables capturing researcher choices with respect to response model specifi cation
& estimation (consideration of dynamics, capturing endogeneity and heterogeneity
in sales response, including other marketing-mix instruments such as promotions

3 Se Footnote 2 for our defi nition of “early” vs. “late” PLC stages; the coding of elasticity measure-
ments into early or late PLC stage categories is described below.