aggregate panel data analysis: the scale of these analyses does not allow controlling for farm agro-ecological characteristics and implicitly assumes that the underlying theoretical model can be scaled up at a macro level.
farm level analysis using cross section data : This approach has the obvious shortcoming of neglecting dynamics.
The impact of weather related factors in the productive role of biodiversity is not taken into account.
We frame our analysis of impact of crop biodiversity on farm household productivity under the standard theory of portfolio choice.
The problem facing a representative risk averse farm household is to
1)Determine a production function that includes diversity
2) chose an optimal level of crop diversity to maximize her expected utility from final wealth at the end of the production period, given the production function and her land, labor and other resource constraints.
As diversity is measured as count index, a Poisson estimation is used in the diversity regression.
To estimate the productivity model, we employed a pseudo-fixed effect model.
A standard fixed effect model has an obvious advantage over random effect and other linear models : controlling unobserved heterogeneity.
However, standard fixed effect models rely on data transformation that removes the individual effect.
Pseudo-fixed effect model ( which is run as a random effect model but at the same enables controlling for unobserved heterogenity by including mean values of time varying explanatory variables) -Mundlak’s approach ( Wooldridge 2002).