Non-parametric efficiency, progress and regress measures for panel data: Methodological aspects

This purely methodological paper deals with the role of time in non-parametric efficiency analysis. Using both FDH and DEA technologies, it first shows how each observation in a panel can be characterized in efficiency terms vis-a-vis three different kinds of frontiers: (i) 'contemporaneous', (ii) 'sequential', and (iii) 'intertemporal'. These are then compared with window analysis. Next, frontier shifts 'outward' and 'inward', interpreted as progress or regress are considered for the two kinds of technologies, and computational methods are described in detail for evaluating such shifts in either case. These are also contrasted with what is measured by the 'Malmquist' productivity index. Finally, an alternative way of identifying progress and regress, independent of the frontier notion and referring instead to some 'benchmark' notion, is extended here to panel data.