Unification and generalization of inequality measures

Many inequality measures are functionals of the Lorenz curve; as far as I know they have not been unified as one LpL^p-type family.

At some point before writing about the Lorenz curvature based inequality index, I was thinking about the more general problem of unifying inequality indices under the family of LpL^p-norms. In this family of indices one could also weigh the Lorenz curves L\mathcal{L} by an arbitrary power (so not necessarily a pure norm but rather a generalized power mean if it comes down to that), as well as measure the distance of L\mathcal{L} with respect to any angle from the line of perfect equality. At least superficially, if the weight is non-zero, this could also make a difference.

I feel there are quite many indices around there which could easily be put under a single family with such a construction. For example, both Atkinson and Amato-Kakwani are different integrals of a power of an expression in L′\mathcal{L}'. Again, the generalization could easily go further by including any power of any derivative of L\mathcal{L}.

Such a generalized L\mathcal{L}-based index could include all powers, all functions of the population share in front, and the question would no longer be about the framework, but rather about the interpretation of parameters. For example, with the curvature index, the natural interpretation was due to curvature measuring the existence of different classes. Such family of parameter interpretations could be develop along the way. To some extent I feel a similarity here to parameters present in statistical mechanics, where the fluctuation-dissipation theorem gives a connection of a spontaneous fluctuation of a system to its response. What one would look for in the interpretation is what kind of perturbation a certain parameter would cause in the measured inequality.