Spillovers and Other Externalities in Pharmaceutical Marketing - Innovation and Marketing in the Pharmaceutical Industry - page 678

To capture goodwill, we employ the standard Nerlove-Arrow exponential decay
model (see Nerlove and Arrow 1962 ; Lilien et al. 1992 , p. 280). Let det jt and
ome jt represent the category levels of detailing and other marketing efforts, respec-
tively, in period t . Then the goodwill stocks in period t (i.e., D jt , OME jt , and DTC jt )
are given by:

DD
jt

= q D, + det; (23.2)

j j t − 1 jt

OME = q OME + ome ; (23.3)

jt j O j t , − 1 jt

DTC = q DTC + dtc (23.4)

jt j DTC j t , 1 − jt

where, θ j D , θ j O , and θ j DTC are the country-specific carryover coefficients for detailing,
other marketing efforts, and DTCA combined expenditures, respectively. The
square root captures diminishing effects (Erickson 1992 ) and allows for zero values
for variables.
Detailing expenditures are deflated using the wage series for all workers. DTCA
spending was deflated using PPIs for advertising, while prices and other expendi-
tures were deflated using the Consumer Price Index (CPI) for all urban consumers.
The CPI and wage series were obtained from the Bureau of Labor Statistics (BLS)
for the USA, and from two websites (Statistics Canada and the Bank of Canada) for
Canada. The quarterly data in the latter series was transformed into monthly series
using the SAS CONVERT utility, in line with other studies (e.g., Ling et al. 2002 ).
Descriptive statistics for the data are in Table 23.2 . The average retail price in
Canada is about one-half of that in the USA for the same time period. Further, the
Canadian market is about a tenth of the size of the US market in terms of the total
number of prescriptions written in this category.
The carryover parameters θ j D , θ j O , and θ j DTC are determined via a grid search pro-
cedure (see Neslin 2001 and Rizzo 1999 for similar applications). We obtained the
following carryover parameters for ( 23.2 ), ( 23.3 ), and ( 23.4 ): 84, 0, and 55 % for
detailing, OME and combined DTCA, respectively. Even though we allowed for the
two countries to have different carryover parameters, the above values gave the best
results. For our estimation, we fixed the carryover rates at these values.
Our interest here is to explore whether DTCA spillover exists. It is possible, however,
that other types of (unobserved) spillovers may also be present; or, some common
demand shocks—such as due to the release of a new scientific study on drug efficacy—
may be present. Consequently, the error terms in the category demand functions for the
USA and Canada (given in ( 23.1 )) may be related. Therefore, we use the seemingly
unrelated regression estimation (SURE) procedure. We also obtained estimates from the
OLS procedure. The SURE and OLS results did not differ significantly.7 7

7 One potential concern is the endogeneity of marketing instruments, especially that of DTCA
(see e.g., Kremer et al. 2008 ).