In this paper, we examine the existence of agglomeration effect on production in the Japanese food industry from 1985 to 2000 using plant-level 4-digit subclassification, panel dataset and new agglomeration index in Akune and Tokunaga (2005), and Tokunaga, Kageyama, and Akune (2005), based on Ellison and Glaeser (1997). This is an improvement on the the conventional indices such as Location Quotient (LQ) or Location Gini Coefficient (L).When we apply a flexible translog production function and cost share equation as suggested by Kim (1992), we find that around 2% of positive agglomeration effect exists in absence of any restriction on homotheticity in the the case of employment based agglomeration (Gamma EG ).