This paper shows that aggregation bias is generated by ignoring the parameter heterogeneity and using linearly aggregated data for nonlinear macro model with traditional NEIO approach. The aggregation bias can be minimized by imposing market share data to estimation process of conjectural elasticity.