Model Requirements
On the basis of these considerations, we maintain four criteria
for our model of dynamic interactions among marketing
and performance variables. First, the model should provide
a flexible treatment of both short- and long-term effects
(Dekimpe and Hanssens 1995). Second, the model should
be robust to deviations from stationarity, particularly the
presence of random walks in stock prices, which can lead to
spurious regression problems (Granger and Newbold
1986).2 Third, the model should provide a forecast and
expected baseline for each performance variable, so that we
can capture the impact of unexpected events as deviations
from the baseline. Both econometric models and survey
methods have been shown to perform well in generating
these expectations (Cheng, Hopwood, and McKeown 1992;
Fried and Givoly 1982). Consequently, our model uses forecasts
based on an econometric model and controls for
changes in analyst earnings expectations. Fourth, the model
should allow for various dynamic feedback loops between
marketing and business performance variables.
In summary, the study of the longitudinal impact of new
product introductions and promotional incentives requires a
carefully designed system of equations that accounts both
for the time-series properties of performance and marketing
variables and for their dynamic interactions