View this page on Reckon Online

Reckon home page

Data envelopment analysis

"When assumptions go unquestioned"

This page is for corrections, additions, comments, thoughts, and any other material relating to Reckon's November 2004 article on DEA (24 pages, PDF), which reviews the assumptions on which DEA rests.

Summary of Reckon's November 2004 article

Data Envelopment Analysis (DEA) is a technique used to estimate the production function on the basis of what is observed. DEA rests on a set of assumptions about the accuracy of the data used (are the data accurate reflection of the levels of inputs and outputs involved in the relevant production process) and on the technology to be estimated (given a set of observations what can be inferred about what input-output combinations are feasible).

The paper discusses the specific assumptions underlying the more familiar models of DEA and argues that, at the peril of drawing invalid inferences, the analyst must justify those assumptions before programmes are run and numbers cranked out.

Such good practice does not appear to be adhered to in DEA studies in the context of the UK water and electricity distribution. To the best of our knowledge, the bulk of DEA carried out in this context are based on one of the two "standard" models. The paper suggests that some of the assumptions on which these standard models rely are not likely to be satisfied in these sectors. At best, this suggests that the assumptions of the "standard" models should not be made, and that the analyst should look further afield into the sizeable DEA literature for models that do not rely on those assumptions found to be unjustifiable. At worst, it suggests that the application of DEA in these sectors is fundamentally misplaced and should not be pursued.

Back to top

Links

References in Reckoning article (those available online):

Data used in the article:

DNO
Network
length
Customer
number
Units
distributed
Composite
variable
Normalised
cost
Adjusted
normalised
cost
CN_Midlands
60.3
2.33
27.3
21.9
67
63.9
CN_East_Midlands
68.9
2.43
28.9
24.0
63
60.7
United_Utilities
59
2.3
25.4
21.2
70.4
67.1
CE_NEDL
39.9
1.5
17
14.2
40.5
38.2
CE_YEDL
51.1
2.15
24.3
19.2
54.2
52.1
WPD_South_West
48.1
1.44
15.4
15.1
54.2
51.3
WPD_South_Wales
33.5
1.055
12.6
11.1
38
36.1
EDF_LPN
30.7
2.1
27
15.2
62.4
59.4
EDF_SPN
49.5
2.14
21.2
18.3
68.5
66.1
EDF_EPN
92.1
3.385
36.3
32.0
86.9
84.4
SP_Distribution
67.3
1.94
22.3
21.0
62.7
57.5
SP_Manweb
45.5
1.44
16.8
15.0
52.6
51.4
SSE_Hydro
48.3
0.69
8.5
10.8
36.4
32.9
SSE_Southern
75
2.7
32.8
26.6
62.6
58.5
Sources:
  • Normalised and adjusted normalised costs: Table A3 in Ofgem (2004) Electricity Distribution Price Control Review, Update paper.
  • Network length, customer number, units distributed: miscellaneous public sources and Reckon estimates.
The formula for the composite variable is:
  • composite variable=(network length^.5)*(customer number^.25)*(units distributed^.25)

Back to top

This page is managed by Reckon LLP. Users are encouraged to contribute relevant views and links. Use the "Edit this page" or "Add a comment" button above to edit the text or to add comments and notes. Advice on making contributions (including trackback) is in the Guidelines.

Last changed by Richard at 10:39 AM on Thursday 11 August 2011.

Reference for this page:
Reckon Open "Data envelopment analysis" 2011-08-11T10:39:05
Link within Reckon Open: [[Data envelopment analysis]]