exception of Roder, Herrmann & Connor (2000), examine the relationship using cross-
section data.
We utilize an extensive annual panel data for the period 1983-2004 and account
for endogeneity of market concentration. These features distinguish our study from the
prior empirical work on the US food industry. Zellner (1989) accounted for endogeneity
of market concentration but used cross section analysis. Roder, Herrmann and Connor
(2000) study a panel of new product introductions in USA from 1988-1994. However, in
this study, the potential endogeneity of concentration has not been accounted for and
concentration measures for five years out of the seven-year study period are imputed
using data from two years of Census of Manufacturers.
In addition to addressing the methodological issues, we test the new theory of
product introduction in anticipation of future mergers (Innes, 2006) by analyzing the
impact of product introduction on subsequent mergers in the food industry.
3. Data
We use annual data on new product introductions (NPI hence forth) in the USA,
classified by various categories of processed food7. The data used in this study covers the
period 1983 to 2004. The number of new product introductions have been used to proxy
for product variety in the different segments of processed food industry in USA.
The food processing companies, like other manufacturing companies, are
classified by SIC codes to identify their industry segments8. Using annual firm level data
from Compustat, we compute annual industry level (identified by 4-digit SIC codes)
7 Data Source: Various issues of ‘The Food Institute Report’
8 The detailed description of the SIC codes were obtained from
http ://www.osha.gov/pls/imis/sic manual. display? id=13&tab=group