My sample includes only migrants that send remittances and non migrants.
I am trying to predict the household income if migration takes place for households with migrants and for households with no migrants.
To do so, I use the Heckman procedure to correct for selection bias. There is a self-selection because of migration and a self-selection about sending remittances, but I only take migrants that send remittances in my sample.
What bothers me is the choice of the selection equation. Does it make sense to only include a migration selection equation (since migration induce sending remittances and non-migration induce not sending remittances) or is it better to use a remittance selection equation?