In types of flows such as migration where family units move, the third assumption is unlikely to be true.However, in such instances, interaction models could be calibrated with data on the flows of family units simply by dividing the number of individuals moving between i and j by the average size of a family unit.This would not alter the parameter estimates obtained from the interaction model, but it would make the assumption of independent outcomes more acceptable.
No takes yet. Share an insight, caveat, or question.
Fotheringham et al. (1983) studied this question.
Synapse has enriched 5 related papers. Consider them for comparative context: