We performed the linear as well as negative binomial estimations for these models. The
results were qualitatively similar16. Table 6B presents the random effects estimates for
the panel data models as the Hausman statistic for FE vs RE failed to reject the null,
indicating that random effects estimates are consistent and more efficient.
The estimates indicate that the NPI with lag 3 and 4 have statistically significant
positive effect on mergers. This result is consistent across all the model specifications.
According to the aggregate analysis a one percent increase in NPIs can increase mergers
between 0.92 and 1.3 percent. The corresponding elasticity for the panel estimates ranges
between 0.06 and 0.16. Thus the mergers analysis reveals that new product introductions
positively affect subsequent mergers in the food industry. Hence it provides evidence in
support of the mechanism through which the anticipatory mergers theory predicts a
positive relationship between NPIs and market concentration.
5. Conclusion
This study provides important insight into the testing of theoretical predictions
about the relationship between market concentration and product variety. The analysis
highlights the fact that accounting for the endogeneity of market concentration plays a
vital role in estimating the effect of market concentration on new product introductions.
Our analysis based on the data of the food industry suggests that market concentration
increases product introductions17. Number of firms and R&D intensity are also important
factors that positively affect new product introductions. We also find that product
16 In case of negative binomial, convergence could not be achieved in some of the model specifications and
in case of the linear estimations the NPI lags were not significant for the last two models.
17 The number of new product introductions cannot distinguish between the quality of innovation i.e.
improvement of a base product vis -a- vis a new base product. Hence in order to infer about the quality of
innovation from this study, one will need to assume that there is one to one mapping between number of
new product introductions and quality of innovations.
14