Popular forecasting models for long-term planning of industrial sites assume a strong linear relationship between employment and land use. New Dutch panel data for 1997 to 2008 were used to test this assumption. In a linear model, employment alone does not suffice to predict land use. Although non-linear models show relationships at municipal and industrial sites level, this is not the case for regions.
However, these relationships are strongly biased by unobserved heterogeneity. For planning purposes and prediction, additional explanatory variables are necessary to tackle this problem. The findings do not support the use of simple linear models for the planning of industrial sites.
PBL Working paper 9