Leveraging Peer-to-Peer Networks in Pharmaceutical Marketing - Innovation and Marketing in the Pharmaceutical Industry - page 459

standard maximum likelihood. BW estimate their simultaneous equations model
using three-stage least squares. There are also several similarities in the models. All
three models assume that the social network is exogenous and static. All three models
assume that the physicians learn from the prescription experiences of other physicians
in their social network, and that the prescriptions written by the opinion leader will
influence the prescriptions written by the influenced physician/s.
NMB quantify the impact of social interactions and peer effects in the context of
prescription choices by physicians. They specify a parsimonious model for measuring
the effect of physician i i's self-reported opinion leader(s) (represented as j ( i )). The
dependent variable is physician i i's prescriptions y , and the independent variables are
detailing D , prescriptions x , and control variable z . The control variable z helps mitigate
the problem of correlated observables and is calculated as the mean prescription of all
other physicians in physician i i's zip-code. NMB control for endogenous group forma-
tion using physician fixed effects, and for simultaneity by using instrumental variables
for the endogenous variable x . The instrumental variables they use are detailing to the
opinion leader, D j ( i ), t , as well as the mean prescriptions of all other physicians in the
opinion leader's zip-code, z − j ( i ), t which influence x but not y .

y =+++++ = agbd D x g z e , i Nt 1 ,.., ; = 1 ,.., T (15.1)

it i t it ji t (), - i t , it

The corresponding specification for i i's opinion leader is:

x =++ at v D + V x e y + z +

( ), ,

t = 1 ,.., T (15.2)

ji t (), j i ( ) t ji t ( ), it − j it (), ji t

They estimate both specifications via fixed-effects panel data linear instrumental
variables regression. NMB find asymmetric peer effects where they find the effect
of opinion leader's prescriptions on physician i i's prescriptions ( δ ) to be positive and
significant while the effect of physician i i's prescriptions on the opinion leader's
prescription ( ς ) is not significant. The interesting finding in this paper is that these
effects are significant only when there in uncertainty in the system, which they
specify as the release of new guidelines for treatment of the disease.
NMB calculate a social multiplier for each opinion leader. The social multiplier
measures the ratio of the effect on the average action caused by a change in a param-
eter to the effect on the average action that would occur if individual agents ignored
the change in actions of their peers. In this case, for example, there is a direct effect
of targeting the opinion leader, which is the increase in the number of prescriptions
written by the opinion leader. There is also an indirect effect of targeting the opinion
leader. The opinion leader influences other physicians to prescribe more of the focal
drug. The social multiplier for each opinion leader will be the ratio of total revenue
(direct and indirect) received by targeting the opinion leader to the revenue received
directly from just the prescriptions written by the opinion leader. This gives an idea
of the true value of the opinion leader. So, an opinion leader with a social multiplier
of 1.25 generates 25 % more revenue from influenced physicians (indirectly) than
from his prescriptions alone (directly). The higher the social multiplier, the more the
influence of the opinion leader, and the higher the amount of indirect revenue from