Modelling firm heterogeneity with spatial ‘trends’

Abstract

The hypothesis underlying this article is that firm heterogeneity can be captured by spatial characteristics of the firm (similar to the inclusion of a time trend in time series models). The hypothesis is examined in the context of modelling electric generation by coal powered plants in the presence of firm heterogeneity.

Notes

¹ Examples of various applications that use time trends as a proxy for technology are in Griffiths et al. (1993), e.g. p. 491.

² A more refined pulverization increases combustion of coal in electric generation and, thus, increases efficiency.

³ Intuitively, the instrument based on ranks in Equation 6 reduces sources of feedback between the residual of the econometric model and the explanatory variable by using only qualitative information of the variable (ranking) suspicious of causing simultaneity biases. In particular, while the residual of the econometric model affects the ranking of the observations only if it also affects the variable in levels, the residual may explain the variable in levels without affecting the ranking of the variable.

⁴ The bandwidth (i.e. number of ones in the raw standardized matrix) is selected simultaneously with the spatial correlation coefficient accordingly with the criterion that maximizes the likelihood function. This criterion is valid since the same observations and parameters are defined for each bandwidth of the row standardized matrix.