How to compute the Wealth Lorenz curve with negative values? I have some data with asset holdings (wealth) of a large group of people. I want to compute the Lorenz curve and Gini coefficient of my population. The problem is that a large fraction of people have negative wealth. 
Say I have wealth that ranges from -1 to 4. About 50% of people have 0 or less wealth. The other remaining households have positive values. In this imaginary country, for every lender we have a borrower, so if we add all of the wealth of everybody we would get a total net wealth of zero. How can I compute the Lorenz curve?
My intuition, which I suspect is wrong, is to move the mass of people with negative wealth to zero. That is, the Lorenz curve only starts to move upwards from 0.5 on the x axis onward. Do you have any suggestions or example?  
 A: If others like me happen to stumble upon this question as I did I'd like to share my solution.
In On the Gini coefficient normalization when attributes with negative values are considered 2015 by Emanuela Raffinetti, Elena Siletti & Achille Vernizzi. There is proposal for a Gini coefficient based on a generalized Lorentz curve.
Consider an example with incomes [-0.3, 0.1, 0.5, 0.7]. Then according to their definition this generalized Gini coefficient becomes the difference in area between two polygons in the Generalized Lorentz curve, see Figure.

Here the x-values of the points points are simply 0, 1/N,...,1; we define y as the cumulative sum of the sorted incomes and y_lower will share the edges with y while the middle points all are min(y).
So for our example the polygons with vertices (x, y) and (x, y_lower) are defined by (0, 0), (0.25, -0.3), (0.5, -0.2), (0.75, 0.3), (1,1) (orange crosses in figure) and (0, 0), (0.25, -0.3), (0.5, -0.3), (0.75, -0.3), (1,1) (blue dots in figure).
Then the generalized Gini coefficient are the ratio between these two areas.
A few notes:

*

*There are at least a few other attempts at negative Gini coefficient, see referenced paper that compares against one other.

*This generalized Gini coefficient does not give quite the same values as the regular Gini coefficient (for maximal inequality this becomes 1 instead of 1-1/N).

A: You may find this reference interesting. In the end, it all depends on whether you are able to find an interpretation of a decreasing Lorenz curve. http://www.jstor.org/stable/2662589
