APPENDIX 4
Memorandum from OXERA Environmental Ltd
PLANNING AND REGULATION OF MAINTENANCE EXPENDITURE
IN THE WATER INDUSTRY
1. INTRODUCTION
As part of the recent periodic review in water,
Ofwat placed great emphasis on the concept of "serviceability"
in predicting future capital maintenance needs. Broadly speaking,
this "top-down" approach involves examining whether,
given companies' historical level of capital maintenance spend,
service to customers has been stable. In cases where service to
customers—as measured by a selection of high-level indicators—is
deemed to have been stable (or has improved), the historical volume
of spend has been used to predict future funding requirements.
In Ofwat's view, over the past seven years,
serviceability for most companies has not deteriorated.
As such, for the majority of companies, Ofwat concluded it its
final determinations that "there is no need to increase capital
maintenance activity". Indeed, in the final determinations,
only three water companies were funded for increases in capital
maintenance activity to restore serviceability to customers.
Although a number of parties in the water industry
have expressed concerns that Ofwat's serviceability methodology
is unsatisfactory, Ofwat has maintained that its top-down approach
"provides a strong and coherent challenge to the detailed
asset management plans of the companies". Nonetheless, in
"MD161: Maintaining Serviceability to Customers", published
on 12 April, Ofwat acknowledged that "in advance of the next
price review, a better understanding is needed of the economic
case for the levels of capital maintenance charges that are to
be financed by customers through accounting charges". In
this context, Ofwat notes that, during the most recent price review,
few companies set out such economic analyses, and that these "should
have been used to justify the future levels of capital maintenance
included in the business plans."
Looking to the future, therefore, Ofwat has
stated that it would like companies to provide such appraisals,
which should include "systematic information" on the
following:
[the] cost of any potential loss of serviceability
to customers, including consideration of risk scenarios and their
probabilities as well as illustrations of how serviceability to
customers would decline, if the activity was not undertaken;
[the] impact on operating costs of capital maintenance
activity, before and after assets are renewed;
circumstances surrounding the timing of asset
replacement;
[the] impact of obsolescence and new lower cost
technology;
any terminal values and the discount rates assumed.
More recently, Ofwat has noted the Competition
Commission's recommendation that "more work be done in the
industry to broaden the understanding of the relationship between
serviceability and asset condition" (Ofwat press release
8 August 2000). The Commission's views on this matter will become
more apparent when its reports, on Mid Kent Water and Sutton and
East Surrey Water, are published in September.
The remainder of this note is organised as follows:
— section 2 examines the issues that
have emerged during the recent periodic review;
— section 3 presents a number of observations
on these matters;
— section 4 provides recommendations
for a more empirical approach to capital maintenance assessment.
2. ISSUES ARISING
DURING THE
PERIOD REVIEW
In the final determinations, the net level of
capital maintenance expenditure assumed by Ofwat for each company
was based on two main considerations:
— how much capital maintenance activity
need to be undertaken in the next five years—ie, the volume
of expenditure required; and
— the unit costs of undertaking this
given level of activity—ie, efficiency.
The methodologies used to determine the above
are examined in more detail below.
2.1 SERVICEABILITY
Ofwat's serviceability approach determined the
volume of expenditure included as part of the final determinations.
The trend in the levels of maintenance expenditure over a seven-year
period, from1992-93 to 1998-99, was examined and compared with
the trend in serviceability indicators. For example, for water
infrastructure (underground) expenditure, these indicators were:
— the extent of low-pressure problems;
— the scale of unplanned interruptions;
and
— quality-compliance parameters.
For most companies. Ofwat concluded that serviceability
was not deteriorating. However, as a secondary check, it also
examined companies' asset inventories to ascertain whether there
was any evidence that the underlying condition of their assets,
given past expenditure, had deteriorated. In the event, Ofwat
found that the percentage of assets classified in the "poor"
condition categories (Grades 4 and 5) had not increased materially.
As such, for all but three companies,[9]
Ofwat adopted the assumption that the historical annual average
of expenditure over the 1992-93 to 1997-98 period was sufficient
to maintain service to customers over the next five years.
However, the validity of the chosen serviceability
indicators has been disputed by a number of companies, and serviceability
indicators can be affected by circumstances outside companies'
control. Furthermore, with regard to non-infrastructure (overground)
expenditure, the number of indicators used was limited to two
for water and two for sewerage.
2.2 EFFICIENCY
ADJUSTMENTS
Having established the levels of capital maintenance
activity required using the serviceability approach, Ofwat applied
a series of efficiency assumptions. For each company, this was
comprised of a "minimum efficiency" target (applied
equally across the industry), and a "catch-up" target.
The catch-up target was based on two comparative-efficiency exercises—the
cost-base and an econometric approach. Using the results of these
analyses, companies classed as being relatively inefficient were
set tougher catch-up targets.
The cost-base exercise involved companies submitting
unit costs to Ofwat for standard maintenance projects. The companies'
relative efficiencies were assessed according to the cost estimates
they submitted. The exercise was criticised by some in the industry
on comparability issues, and because some companies were not planning
to carry out all of the standard projects, but were still required
to submit estimates which contributed to their efficiency appraisal.
An alternative approach would have been to determine efficiency
levels based on a basket of projects drawn from the full set,
the basket being chosen to reflect the company's activities, with
weightings applied to the relative efficiencies to reflect the
volume of those activities planned by the companies.
Econometric modelling was also used to assess
the relative efficiency of companies in undertaking maintenance
expenditure. Previously, in the 1994 review, only the cost-base
approach was used to assess capital maintenance efficiency. The
new econometric models essentially extended the approach already
used to assess operating expenditure (OPEX) efficiency. However,
the relationship between "cost drivers" and costs is,
arguably, more complex in the case of capital maintenance than
for OPEX. Furthermore, the models arguably suffered from too few
degrees of freedom (data points) to capture all the drivers and
permit a satisfactory specification (choice of explanatory factors).
In addition, the models did not explicitly capture trade-offs
between OPEX and capital maintenance expenditure.
3. OBSERVATIONS
As recorded above, there have been a number
of contentious issues surrounding the determination of capital
maintenance spend during the recent water price review. To recap,
the economic regulator's role in determining the appropriate level
of spend on asset maintenance is twofold—to determine the
volume of expenditure, and to determine how cheaply this can be
carried out.
However, in parallel, the company could be asked
to demonstrate:
— an economic and engineering case
for its overall budgeted maintenance expenditure;
— the efficient prioritisation of maintenance
expenditure;
— the efficient procurement of maintenance
projects.
In order for the two players to deliver the
above, both the regulator and the company need to understand:
— a common definition of serviceability
and agreement about the objective to maintain the serviceability
of assets or to maintain levels of service to customers;
— a common definition of long-term
trends in the maintenance of serviceability;
— the relationship between the performance
of individual assets and levels of service experienced by customers;
— the relationship between the observable
indicators of asset condition and its expected performance;
— the relative costs and benefits of
proactive and reactive maintenance expenditure for different types
of assets.
With this information, the company could link
its expected maintenance expenditure to expected levels of service
performance, and the model could be audited by the economic regulator
and its reporters (engineering consultants). This approach allows
the model to apply economic tests to the choices available, rather
than building on the received wisdom of engineering precedents.
3.1 INFORMATION
AVAILABLE TO
THE REGULATOR
Ofwat has faced company submissions that could
have gone further to provide information of use in making the
determination. None of the business plans submitted to Ofwat has
been published (although most companies placed a cut-down version
of their plans in the public domain). Nonetheless, the following
are examples that are not thought to have been standard practice
across the industry:
— information on consistency of measurements
of asset condition and performance, survey sample sizes, the detail
of top-down and bottom-up modelling methods;
— an explanation of choices made in
management of maintenance programmes over the previous period
and the assessment of the outcomes, especially the management
of assets critical to the delivery of services and the mixture
of proactive and reactive expenditure;
— back-casting to test the reliability
of these models;
— expected uncertainties of estimates
(confidence limits); and
— the prioritisation of expenditure
expressed in terms of risk reduction or levels of service.
The regulator has to rely heavily on the judgement
of reporters (which all work as contractors for other companies
within the industry). Therefore, if there is any inherent bias,
for example towards gold-plating or caution within water services
engineering practice as a whole, then this may not be corrected.
This bias is most likely to enter through standards of performance,
which are common industry tools for identifying assets requiring
maintenance. The standards ought to be set in line with the overall
performance target. They should reflect the local criticality
or environmental factors, and be distanced from professional practices
that may contain a precautionary margin. However, if uniform component
standards are applied to the components individually, the overall
system performance might be far higher than optimal.
3.2 INFORMATION
AVAILABLE TO
THE COMPANIES
In general, companies were not able to estimate
the impact of individual asset failures on customer levels of
service, and could not therefore show what impact budget reductions
would have made on customer levels of service either in internal
negotiations of budget allocation or in negotiations with the
regulator.
Some companies faced difficulties in upgrading
their maintenance planning assessments from the previous periodic
review because they had poor asset condition and maintenance records
(short time-series, uncertain consistency, and poor causality
information on historic incidents) especially for historical periods.
Data collection ought to be a high priority. The data-collection
exercises can be designed carefully to provide adequate sample
sizes, especially for underground assets, so that statistical
methods can be applied once the data has been collected.
With adequate data, models can then be used
to examine the causality of trends in service performance. Any
models based on extrapolation of historic trends without conclusive
evidence for drivers of trends may not recognise the potential
for future changes in trends. For example, in recent years, the
improved management of asset failure incidents, and the introduction
of automatic control has delivered improved levels of service
to customers without an improvement in the asset condition. This
is because the impact of individual failures on the customer has
been reduced. These factors ought to be considered explicitly
so that the significant changes in system management observed
since privatisation can be separated out from the effects in underlying
changes in asset stock and network systems.
Because of Ofwat's arm's-length interaction
via the reporters, it did not appear to be in a position to challenge
the understanding of the basis of maintenance budget with senior
management outside the operations management within the companies.
One example is the variation in policies between
companies on renovation and replacement, which never appeared
to be resolved satisfactorily.
There appears to have been a heavy reliance
on engineering protocols which have set maintenance schedules
and acceptable asset condition descriptions, and assume average
operating conditions that have been established historically.
These are typically only revised infrequently. They should be
driven by the maintenance planning process and include an explicit
economic appraisal.
The companies experienced difficulties in interpreting
Ofwat's condition and performance-grading criteria. The criteria
mix together concepts of physical state, ability to deliver function
(serviceability), and the criticality or consequences of component
failure. For example, the condition grades contain measures of
failure rate; and performance grades include damage on failure,
mixed with design capability and historic reliability. Using the
data collected according to these definitions, it is not possible
to take a pure physical condition or serviceability approach to
the assessment of the asset stock.
Finally, Ofwat's approach does not examine the
impact of OPEX on both serviceability and the need for capital
maintenance expenditure. Neither does it reflect the fact that
OPEX may be influenced by the level of asset failures, by proactive
maintenance and by changes in the management of individual assets.
The substitution of OPEX for capital maintenance ought to be recognised
as an output in the comparative-efficiency assessments for OPEX
and capital maintenance, and an adjustment made in the maintenance
budget. Cost allocation is notoriously difficult, but it is important
because a poor accounting rule can distort company behaviour.
4. RECOMMENDATION:
A MORE EMPIRICAL
APPROACH
In conclusion, there are several important economic
and engineering systems questions that could be built into the
periodic review, and methodologies developed to answer them.
— Data on asset failure and design
can, for ach type of asset, provide failure patterns (for example
predictable, unpredictable, constant probability throughout life,
time/usc-dependent probability) which can be used to determine
the optimal maintenance programme. A predictive model of failure
behaviour based on existing asset data would be useful, and could
be prepared from a pooling of industry-wide data, as has been
seen in the electricity distribution industry.
— The asset survey should be targeted
at proxies of failure probability. The ability of proxies to predict
failure (age, physical corrosion, wear, etc) could be measured
against empirical observations.
— A model could be built to link the
impact of asset failure on customer levels of service. This requires
analysis of system reliability. The analysis could take into account
the risks of multiple coincident failures, the mitigation by duty/standby
arrangements, and might even incorporate the impact of excess
loading on supporting assets during a failure episode.
— The overall budget for, and prioritisation
of, maintenance expenditure might consider the cost and benefit
(levels of service) trade-off of expenditure options. The business
plan could demonstrate a least-cost programme. The analysis might
include the costs of hot standby and duplicate capacity, demand-management
contracts, insurance and regulatory penalties. It could compare
replacement and renovation, reactive and proactive maintenance,
and could produce results in which the uncertainties or bounds
of expected outcome are described. All the modelling assumptions
could be listed.
— Key assets may merit a whole-life
maintenance profile for an asset in a particular operating role.
This might be part of a wider revision of the use of industry
standards currently in operation.
August 2000
9 Dwr Cymru was funded for a 10 per cent increase in
sewerage infrastructure maintenance, Southern for a 10 per cent
increase in sewerage maintenance, and Wessex for a 10 per cent
increase in water and sewerage infrastructure maintenance spend. Back
|
