The Impact of Negativity on Separability Testing

During separability testing, it has not been common practice to verify if the data satisfy the negativity condition. The impacts of imposing negativity on separability test results are examined for two groups of commodities using Barten's generalized demand system and with Monte Carlo simulations. Results indicate that imposing negativity did not alter model estimates when negativity was spontaneously satisfied. Also, imposing negativity rarely altered separability test results when the data did not spontaneously satisfy negativity. Copyright 2001, Oxford University Press.