HEFCE GMP 107 ASTON UNIVERSITY

1
Identifying Cost Efficient Practices In
Administrative Services In UK Universities
Prof. Emmanuel Thanassoulis Aston
University e.thanassoulis_at_aston.ac.uk
2
Aims of the Study
To identify good quality cost efficient
practices in the delivery of Central
Administrative Services in UK Universities.
3
The Unit of Assessment
Core CAS
4
The Input-Output Framework
Support Services To Students
Labour
Support Services To Staff
Central Administration
Liaisons with Other Bodies
Capital
Support to Technology Transfer
5
The Input-Output Set
6
Ratios Under Multiple Resource and/or Outcome
Measures

• The two ratios do not lead to the same benchmark
  operating unit.

• A benchmarking methodology taking multiple
  resource and
• outcome measures JOINTLY into account is
  required.

7
Data Envelopment Analysis (DEA)
8
The space of all possible production points in a
DEA model is specified as the feasible region of
a linear programming model.
The distance of a real production point (unit)
from the boundary of the space constructed is
determined by optimising an objective function on
the above linear programming model.
An introduction to Data Envelopment Analysis can
be found in
E. Thanassoulis (2001) Introduction to the Theory
and Application of Data Envelopment Analysis A
foundation text with integrated software. Kluwer
Academic Publishers, Boston, Hardbound, ISBN
0-7923-7429-0
9
TERMINOLOGY
Benchmark CAS unitsThese are units that
relative to the rest of the CAS units are found
to have the lowest level of spend when we control
for their mix and absolute levels of student
income, non-CAS staff spend and Technology
Transfer.
Non Benchmark CAS unitThese are CAS units which
are not benchmark in the foregoing sense.
Benchmark spendThis is the level of spend a CAS
unit is estimated would need to have to match the
benchmark CAS units when we control for mix and
absolute levels of student income, non-CAS staff
spend and Technology Transfer.
10
Three Related Measures of CAS Spend Have Been
Modelled
- Aggregate CAS staff and Operating expenditure
- CAS staff expenditure Only
- CAS OPerating EXpenditure (OPEX) only.
In each case we have controlled for mix and
absolute levels of student income, non-CAS staff
spend and Technology Transfer.
The data we have analysed relates to 1999/2000.
11

• The aggregate CAS expenditure model is a more
  appropriate benchmarking instrument the more
  mutually substitutable OPEX and CAS staff spend
  are in delivering CAS.

• The separate CAS staff and CAS OPEX models are
  more appropriate benchmarking instruments the
  more mutually NON - substitutable OPEX and CAS
  staff expenses are in securing the CAS
  deliverables.

12
CAS Devolved To Academic Departments
NO DEVOLVED ADMIN In this model the assumption
was made that non-academic staff costs in
academic departments ARE NOT part of CAS staff
expenditure.
DEVOLVED ADMIN In this model the assumption was
made that ALL non-academic staff costs in
academic departments ARE part of CAS staff
expenditure.
MEAN DEVOLVED ADMIN The average between the two
benchmark spends derived from the two purist
models give the mean devolved benchmark spend.
13
Summary of Benchmarking Computations
Models Used No
Devolvement Devolvement Mean DevolvementResource
being modelled Aggregate CAS
Staff and OPEX Spend ? ? ? CAS Staff
Spend Only ? ? ? CAS OPEX Spend Only
? ? ?
14
Interpreting Benchmark Spend Percentages

• Take as an example the Mean Devolvement Benchmark
  spend for I36 as Percent of Actual Spend

Aggregate CAS CAS Staff CAS OPEX
I36 91.790 78.036 94.550

• The 91.79 under Aggregate CAS applies if we
  assume CAS staff and OPEX spend are for the most
  part mutually substitutable. In that case the
  overall CAS expenditure of I36 can reduce by
  about 8.

• The percentages under CAS Staff and CAS OPEX
  apply if we assume CAS staff and OPEX spend are
  for the most part NOT mutually substitutable.

15
Interpreting Benchmark Spend Percentages

• The CAS Staff percentage of 78.036 shows that
  the unit can save about 22 of actual CAS staff
  spend relative to its enchmarks on CAS staff.

• The CAS OPEX percentage of 94.55 shows the unit
  can save about 5 of actual OPEX spend relative
  to its benchmarks on OPEX.

• The three estimates of benchmark spend above need
  to be seen as broad brush indications given
  the caveats on data shortcomings to be made
  later.

16
Summary of Benchmarking Computations
Aggregate CAS Staff and OPEX Benchmarking
Mean Devolved Admin
BENCHMARKS
Devolved Admin
No Devolved Admin
Benchmark expenditure as percent of observed
expenditure
17
Aggregate CAS Staff and OPEX Spend
The sum of benchmark and scope for savings is the
observed level of spend. The scope for savings is
stated in m.
780m
737m
693m
18
Benchmark CAS Units On Aggregate Staff And OPEX
Inst No Devol Devolved Admin
Admin 1 100 100 2 100 100 3 100 1
00 4 100 100 5 100 100 6 100 100 7 100 100
8 97.822 100 9 100 100 10 100 98.468 11 10
0 100 12 100 98.572 13 100 100 15 99.66 100
16 100 89.106 17 100 100 18 100 82.459 24
100 100

• The benchmark spend as of the observed spend is
  shown. Where we have 100 we have a benchmark
  unit under the model concerned.

• Clearly benchmark CAS units are virtually
  identical under the devolved and not devolved
  administration models where aggregate spend is
  concerned.

19
CAS Staff Spend Benchmarking
Mean Devolved Admin
BENCHMARKS
Devolved Admin
No Devolved Admin
Benchmark expenditure as percent of observed
expenditure
20
Benchmark CAS units on Staff Spend
No Devld Admin
Devd Admin
HEI
1 93.24 100 4 100 58.55 8 99.58 100 9 97.11 1
00 10 100 98.27 11 100 81.06 12 100 100 15
86.91 100 17 100 100 22 88.67 100 23 100 84
.41 24 59.2 100 25 100 94.38 26 100 75.62
30 95.36 100 34 100 95.58 50 100 82.7 95 100
89.87

• Where we have 100 we have a benchmark unit under
  the model concerned.

• We have a large measure of agreement between the
  models on benchmarks but also some significant
  differences (highlighted).

• The differences arise mainly where a benchmark
  unit has a very large part of CAS devolved to
  academic departments.

21
CAS OPEX Benchmarking
BENCHMARKS
Mean Devolved Admin
Median 65
Median 67
Admin
Devolved Admin
Median 62
No Devolved Admin
Benchmark expenditure as percent of observed
expenditure
22
Using The Identified Benchmarks

• All our estimates of benchmark spend are based
  on SPECIFIC BENCHMARK units identified to match
  the activity volumes and mix of the unit whose
  benchmark spend we wish to estimate.

• Special sets of benchmarks are available for all
  non-benchmark CAS units from each one of the
  Devolved - No Devolved administration models and
  for each resource (Aggregate, OPEX or Staff )
  modelled.

• An illustration of how benchmarks specific to
  each non-benchmark unit can be useful follows.

23
Using The Identified Benchmarks

• The table below shows the specific benchmarks
  identified for I36 under the Devolved Admin
  model, when CAS staff is the resource modelled.
  Data on each variable is indexed for anonymity
  so that I36 100.

• I36 and I30 are post 1992 Universities while I46
  is an old university.

• All three units have about 50 of potential CAS
  staff spend devolved to academic departments and
  so we have in effect controlled for CAS staff
  devolvement.

24
Using The Identified Benchmarks

• Usually, but not always, one or more of the
  special benchmarks chosen by DEA can be used to
  see why the non-benchmark unit was found to have
  scope for efficiency savings.

• In the case of I36 its benchmark I46 can play
  this role.

• I46 administers 4 times the volume of Technology
  transfer of I36, nearly 30 larger non CAS staff
  spend and 84 of the student income of I36 .

• Thus even if we ignore the higher volumes of non
  CAS staff and Technology Transfer at I46 we would
  expect its CAS spend to be of the order of 84
  of that of I36.

25
Using The Identified Benchmarks

• There may be factors outside the model (such as
  mix of students administered or the quality of
  service at I36 being better and costlier than at
  I46 ) that explain the apparent difference
  between I36 and I46 on CAS staff spend. However,
  it is also possible that there is a genuine
  difference in operating practices between the two
  that explains in part the lower CAS staff spend
  at I46 when we control for activity levels.

• Comparisons of this type can be made for all
  non-benchmark institutions on each model used
  relative to one or more of their specific
  benchmarks.

26
Contracting Two Benchmarks CAS Units
Percent CAS staff spend attributable to each
activity

• For unit 73 diseconomies of scale justify about
  1.22 of CAS staff spend. Over 80 of CAS
  staff spend needs to be attributed to TT activity
  for unit 73 to justify in full its CAS staff
  spend.

• For unit 46 there are no dis- or economies of
  scale. Further, it does not specifically need to
  play up the importance of any one of the three
  surrogate measures of activity we are using in
  order to justify its CAS staff spend.

27
Contrasting Two Benchmarks CAS Units

• Both I46 and I73 are collegiate old
  research-intensive universities. It does appear
  though that in either student or staff volume
  administration or both I46 may have better
  practices which would benefit I73 despite I73
  itself being a benchmark.

• The DEA analysis reveals information of this type
  which could benefit the benchmark CAS units to
  adopt the best practices from other benchmark
  units.

28
Summary

• We have used three surrogate measures of CAS
  activity

- Student income (000)
- Total non CAS staff costs (000)
- Technology Transfer (000) (Research grants,
other services rendered etc.)

• Controlling by means of DEA simultaneously for
  the three measures above we have benchmarked CAS
  units in turn on three measures of spend

- Aggregate CAS staff and Operating expenditure
- CAS staff expenditure only
- CAS OPerating EXpenditure (OPEX) only
29
Summary

• The data we have analysed relates to 1999/2000.

• The spend modelled was alternately computed
  assuming non-academic staff costs in academic
  departments are and are NOT part of CAS spend.

• We have found large measure of agreement in the
  benchmarks identified for each type of spend
  modelled, under the two alternate assumptions
  above on treating non academic staff spend in
  academic departments. This is less true when CAS
  staff spend is modelled.

30
Summary

• On balance, using the mean devolved
  administration result is likely to be a better
  estimate of the relative performance of a unit on
  each spend modelled.

• If we assume CAS OPEX and CAS staff spend are
  in large measure mutually substitutable then the
  aggregate CAS and OPEX spend model applies.

• Under this model we estimate that the median CAS
  unit can reduce total spend by some 25,
  amounting for the sector to potential savings of
  some 737m.

31
Summary

• DEA clusters each non-benchmark unit within a
  small subset of the benchmark units, those most
  closely matching it on scale size and mix of
  activities.

• We have indicated how such small groups of
  specially identified units may readily compare
  their data and generally share operating
  practices found at benchmark units.

• We have also indicated how benchmark units can
  themselves identify other benchmark units which
  will offer complementary best practice to their
  own.

32
Caveats

• Our findings could be biased for a number of
  reasons

- The surrogate variables we have used (student
income, non-CAS staff spend and Technology
Transfer income) may not reflect with similar
accuracy volumes of CAS activities across HEIs.
- We have not reflected in the modelling any
variation in quality of service offered by CAS
units across HEIs. This part of the project is in
the process of being carried out.
- Data may not be consistently returned by HEIs
as there is latitude in interpreting data
headings on HESA returns.
33
Caveats
- We have raised but not resolved the question
as to whether CAS staff and CAS OPEX spends are
mutually substitutable and if so to what degree.
- We have been unable to disentangle staff and
OPEX spend on CAS from other non academic staff
and OPEX spend in academic departments.