I have a data-set with N=131 and T=9. N are mostly rural regions with few big towns (50.000 - 180.000 inhabitants) and T are years from 2010 to 2018. The period can be considered as crisis that started in 2008 (rural areas are lagging).
Theoretical framework point into the SDM model and empirical evidence (Hausman test) points into the direction of FE SDM model. Also, Pesaran and Frees tests show cross-sectional dependence.
To analyze the spatial correlation I have to use the spatial FE SDM. That is because the error is time-invariant and it accounts for any regional-specific effect. Am I correct?
Code:
xsmle ln_Y ln_X1 ln_X2 ln_X3 ln_X4 ln_X5, wmat(W) model(sdm) fe type(ind) vce(cluster jls) effects vceffects(sim, nsim(5000)) hausman nolog
Code:
xsmle ln_Y ln_X1 ln_X2 ln_X3 ln_X4 ln_X5, wmat(W) model(sdm) fe type(time) vce(cluster jls) effects vceffects(sim, nsim(5000)) hausman nolog
Code:
xsmle ln_Y ln_X1 ln_X2 ln_X3 ln_X4 ln_X5, wmat(W) model(sdm) fe type(both) vce(cluster jls) effects vceffects(sim, nsim(5000)) hausman nolog
0 Response to Time and/or Spatial fixed SDM model estimation
Post a Comment