Estimation of a complete system of nonlinear Engel curves: further evidence from Box-Cox Engel curves for Sri Lanka

The specification and estimation of Engle curves has drawn much attention over many years since the earliest attempt of Working (1943). Among the many approaches, a complete system of Engel curves stood out as more appealing as it explicitly imposes the important budget constraint in the allocation of family budgets. Further, extended researches were undertaken on nonlinear specifications as they are likely to produce results that are closer to the reality than the linear systems. Beneito (2003) attempted Box-Cox form of nonlinear specification but ended up estimating a set of linear equations using seemingly unrelated regression equations method. The present research extends the previous work on Engle curves in two ways; first by incorporating key demographic characteristics into the determination of expenditures in the specification of the Box-Cox form of nonlinear system, and second by using primary micro data from large household surveys to estimate the system by nonlinear estimation methods. Our estimations have produced robust results; the expenditure patterns are significantly nonlinear and the expenditure patterns are significantly different for different demographic groups.